A 45‐year‐old man with HIV infection (CD4 count of 6 cells per cubic millimeter) presented after 2 days of diminishing visual acuity and pain in his right eye. Examination revealed a corneal ulceration and hypopyon (Fig. 1, white arrow). Chest radiograph demonstrated right lower lobe pneumonia. Cultures of the hypopyon, sputum, and blood grew Streptococcus pneumoniae.

Figure 1

The patient was treated with IV ceftriaxone as well as fortified tobramycin, vancomycin, and doxycycline eye drops with intravitreal vancomycin. The patient's vision and eye pain gradually improved, and he was discharged home.

Infectious ulcerative keratitis is a rare entity, most often resulting from direct corneal invasion by bacterial or fungal organisms. This case appears to involve hematogenous spread. Streptococcus pneumoniae, Staphylococcus, and Pseudomonas are the most common bacterial pathogens. Broad‐spectrum topical antibiotics are the cornerstone of therapy. Topical steroids may be administered once the infection is under control.

A 45‐year‐old man with HIV infection (CD4 count of 6 cells per cubic millimeter) presented after 2 days of diminishing visual acuity and pain in his right eye. Examination revealed a corneal ulceration and hypopyon (Fig. 1, white arrow). Chest radiograph demonstrated right lower lobe pneumonia. Cultures of the hypopyon, sputum, and blood grew Streptococcus pneumoniae.

Figure 1

The patient was treated with IV ceftriaxone as well as fortified tobramycin, vancomycin, and doxycycline eye drops with intravitreal vancomycin. The patient's vision and eye pain gradually improved, and he was discharged home.

Infectious ulcerative keratitis is a rare entity, most often resulting from direct corneal invasion by bacterial or fungal organisms. This case appears to involve hematogenous spread. Streptococcus pneumoniae, Staphylococcus, and Pseudomonas are the most common bacterial pathogens. Broad‐spectrum topical antibiotics are the cornerstone of therapy. Topical steroids may be administered once the infection is under control.

A 45‐year‐old man with HIV infection (CD4 count of 6 cells per cubic millimeter) presented after 2 days of diminishing visual acuity and pain in his right eye. Examination revealed a corneal ulceration and hypopyon (Fig. 1, white arrow). Chest radiograph demonstrated right lower lobe pneumonia. Cultures of the hypopyon, sputum, and blood grew Streptococcus pneumoniae.

Figure 1

The patient was treated with IV ceftriaxone as well as fortified tobramycin, vancomycin, and doxycycline eye drops with intravitreal vancomycin. The patient's vision and eye pain gradually improved, and he was discharged home.

Infectious ulcerative keratitis is a rare entity, most often resulting from direct corneal invasion by bacterial or fungal organisms. This case appears to involve hematogenous spread. Streptococcus pneumoniae, Staphylococcus, and Pseudomonas are the most common bacterial pathogens. Broad‐spectrum topical antibiotics are the cornerstone of therapy. Topical steroids may be administered once the infection is under control.

This update reviews key clinical articles for hospitalists published over the past year. Selection criteria include high methodological quality, pertinence to hospital medicine, and likelihood that a change in practice is warranted. Table 1 summarizes practice changes.

Enoxaparin Dosing in Acute Coronary Syndromes

Allen La Pointe NM, Chen AY, Alexander KP, et al. Enoxaparin dosing and associated risk of in‐hospital bleeding and death in patients with non‐ST‐segment elevation acute coronary syndromes. Arch Intern Med. 2007;167:15391544.

Question: Among patients with non‐ST‐elevation acute coronary syndromes, how common and harmful is excess enoxaparin dosing?

Sponsors: Schering‐Plough Corp., Bristol‐Myers Squibb/Sanofi‐Aventis Pharmaceuticals Partnership, Millennium Pharmaceuticals, and the National Institutes of Health and National Institute on Aging.

Study Design: Observational study of prospective cohort data from the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA Guidelines (CRUSADE) National Quality Improvement Initiative.

Patients: A total of 10,687 patients receiving enoxaparin for non‐ST‐elevation acute coronary syndromes.

Setting: Three hundred thirty‐two US hospitals.

Outcomes: Rate of excess enoxaparin dose, defined as greater than 10 mg/day above the recommended dose of 1 mg/kg every 12 hours for creatinine clearance (CrCl) 30 mL/minute or 1 mg/kg every 24 hours for CrCl < 30 mL/minute; rates of in‐hospital major bleeding and death; and rate of lower than recommended enoxaparin dose.

Results: Excess enoxaparin dosing occurred in 18.7% of the cohort (2002/10,687). Of these, 57.8% (1157/2002) had CrCl < 30 mL/minute. Excess‐dose patients were more likely to be older and female and have a low body mass index (P < 0.001 for all comparisons). In‐hospital major bleeding (14.2% versus 7.3%, P< 0.001) and in‐hospital death (5.6% versus 2.4%, P < 0.001) were more common among excess‐dose patients. Enoxaparin underdosing occurred in 29.2% (3116/10 687) and was not associated with excess harm. Controlling for baseline characteristics, the authors found that the adjusted odds ratio for in‐hospital major bleeding in the excess‐dose cohort was 1.43 (1.181.75, P < 0.001) and the adjusted odds ratio for death was 1.35 (1.031.77, P = 0.03).

Conclusions: Excess enoxaparin dosing in non‐ST‐elevation acute coronary syndromes occurred in about 1 of every 5 patients treated in this prospective multihospital registry. Excess dosing was associated with substantially higher rates of major in‐hospital bleeding and death, with a number needed to harm of 78 for major bleeding and a number needed to harm of 167 for in‐hospital death. In comparison, the number needed to treat with another low‐molecular‐weight heparin (dalteparin) was 34 to prevent 1 death or myocardial infarction in the first 6 days, with a nonsignificant trend toward decreased mortality.1

Commentary: Providers likely underestimate the degree of renal impairment when looking solely at serum creatinine instead of estimates of CrCl. Excess dosing was more common among elderly, thin, and female patients. Clinicians must calculate the enoxaparin dose on the basis of careful estimates of CrCl to limit this risk. The Modification of Diet in Renal Disease (MDRD) equation is commonly used for this purpose.

Clinical Bottom Line: Enoxaparin excess dosing is common and harmful. Clinicians can mitigate this risk by more carefully estimating renal function when selecting the proper enoxaparin dose of 1 mg/kg twice daily for CrCl 30 mL/minute and 1 mg/kg once daily for CrCl < 30 mL/minute.

Venous Thromboembolism Prevention

Wein L, Wein S, Haas SJ, et al. Pharmacological venous thromboembolism prophylaxis in hospitalized medical patients. Arch Intern Med. 2007;167:14761486.

Question: What is the relative safety and efficacy of various pharmacological agents for preventing venous thromboembolism among hospitalized medical patients?

Sponsor: National Health and Medical Council of Australia.

Study Design: Meta‐analysis of 36 prospective randomized controlled trials involving about 48,000 patients.

Study Selection: Prospective randomized controlled trials enrolling at least 30 patients comparing 1 of 4 regimens: (1) unfractionated heparin (UFH) versus control, (2) low‐molecular‐weight heparin (LMWH) versus control, (3) LMWH versus UFH, or (4) Factor Xa inhibitor versus placebo. Trials of surgical, trauma, and critical care patients were excluded. Only 1 Factor Xa trial (fondaparinux) was located,2 and thus it was not eligible for meta‐analysis.

Outcomes: Pooled relative risks with 95% confidence intervals for deep venous thrombosis (DVT), pulmonary embolism (PE), mortality, and total bleeding. The authors also compared 2 UFH regimens: 5000 units twice daily versus 5000 units thrice daily.

Results: UFH (all doses, compared with control): The relative risk was 0.33 (95% CI 0.260.42) for DVT and 0.64 (95% CI 0.500.82) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 3.11 (95% CI 2.443.96, P = 0.001).

LMWH (compared with control): The relative risk was 0.56 (95% CI 0.450.70) for DVT and 0.37 (95% CI 0.210.64) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 1.92 (95% CI 1.322.78, P = 0.001).

LMWH (compared with UFH, all doses): The relative risk for DVT was 0.68 (95% CI 0.520.88, P = 0.004), but the risk was not different for PE, mortality, or major bleeding.

UFH (5000 units twice daily, compared with control): The relative risk for DVT was 0.52 (95% CI 0.280.96, P = 0.04). When the random‐effects model was used, this difference became statistically nonsignificant (relative risk = 0.41, 95% CI 0.101.73, P = 0.23).

UFH (5000 units 3 times daily, compared with control): The relative risk for DVT was 0.27 (95% CI 0.200.36, P = 0.001). This difference remained when the random‐effects model was applied (relative risk = 0.28, 95% confidence interval = 0.210.38, P = 0.001).

Conclusions: Both UFH and LMWH reduce DVT and PE in hospitalized medical patients. Neither affects mortality. Both increase the risk of major bleeding. LMWH reduces the risk of DVT but not the risk of PE in comparison with UFH (all doses). When adjusted for random effects, UFH at a dose of 5000 units twice daily does not appear to be different than the control.

Commentary: This well‐conducted meta‐analysis demonstrates the efficacy of heparin, whether unfractionated or low‐molecular‐weight, in the prevention of venous thromboembolism. Of note, the UFH dose of 5000 units twice daily did not appear to be different than placebo. The UFH dose of 5000 units 3 times daily, by contrast, was effective in both the fixed‐effects and random‐effects models. Mortality was unaffected by any of the regimens studied. All regimens were associated with increased risks of major bleeding.

Clinical Bottom Line: Pharmacological prophylaxis with UFH 3 times daily or LMWH reduces the risk for venous thromboembolism. Twice daily UFH is not clearly different from placebo. Overall mortality was unaffected by any of the regimens for prophylaxis.

Contrast Nephropathy Prevention

Briguori C, Airoldi F, D'Andrea D, et al. Renal insufficiency following contrast media administration trial (REMEDIAL): a randomized comparison of 3 preventive strategies. Circulation. 2007;115:12111217.

Question: What is the efficacy of saline versus bicarbonate for the prevention of contrast mediainduced nephropathy?

Sponsor: Institutional funding (C. Briguori, personal communication, January 2008).

Study Design: Randomized trial.

Patients: Three hundred twenty‐six consecutive patients with serum creatinine 2.0 mg/dL and/or an estimated glomerular filtration rate < 40 mL/minute/1.73 m² undergoing elective coronary and/or peripheral angiography.

Setting: Two interventional cardiology laboratories in Italy.

Intervention: Patients were randomized to 1 of 3 preventive regimens: (1) intravenous saline (0.9%) given at a rate of 1 mL/kg of body weight/hour 12 hours prior to the procedure and continuing for 12 hours afterward (reduced to 0.5 mL/kg/hour for patients with a left ventricular ejection fraction < 40%) plus N‐acetylcysteine (NAC; 1200 mg orally twice daily) on the day before the procedure and the day of the procedure; (2) intravenous sodium bicarbonate (154 mEq/L in dextrose and water) given as an initial bolus of 3 mL/kg over 1 hour prior to the procedure and continuing at a rate of 1 mL/kg/hour for 6 hours more plus NAC as above; or (3) intravenous saline as above plus intravenous ascorbic acid (3 g) 2 hours prior to the procedure followed by 2 g on the night and morning after the procedure plus NAC as above.

Outcomes: Rate of contrast‐induced nephropathy (CIN), which was defined as an increase in serum creatinine 25% from the baseline value at 48 hours after the administration of contrast or the need for hemodialysis.

Follow‐Up: Forty‐eight hours.

Results: The baseline serum creatinine was about 2.0 mg/dL and did not differ among the 3 groups. The rate of CIN was 9.9% (11/111) in the saline plus NAC group, 1.9% (2/108) in the bicarbonate plus NAC group, and 10.3% (11/107) in the saline plus ascorbic acid plus NAC group. The bicarbonate plus NAC regimen was superior to saline plus NAC (P = 0.019). The absolute risk reduction for bicarbonate plus NAC versus saline plus NAC was 8% (a number needed to treat of 13 to prevent 1 case of CIN). The saline plus NAC and saline plus ascorbic acid plus NAC groups did not differ in outcome.

Conclusions: Sodium bicarbonate plus NAC is superior to saline plus NAC for the prevention of CIN among patients with baseline chronic kidney disease.

Commentary: This trial confirms the results of the initial study by Merten et al.3 showing the superiority of bicarbonate versus saline in the prevention of CIN. That trial, published in 2004, did not use NAC. Also in 2007, 3 other single‐center randomized trials of saline versus bicarbonate in the prevention of CIN were published.46 All concluded that bicarbonate is superior to saline. Whether NAC is effective for CIN prevention remains unclear.7 Given its low side‐effect profile, it is not unreasonable to continue using NAC until further data are available. At‐risk patients receiving intravenous contrast for other indications (eg, computed tomography) would likely show similar benefit. Although there are now 5 prospective blinded controlled trials showing the superiority of bicarbonate, a recently published large retrospective cohort found that the use of sodium bicarbonate was associated with increased incidence of CIN.8 The concordant results of all 5 prospective randomized trials of sodium bicarbonate, along with the risk for unmeasured confounding variables with retrospective cohort analysis, suggest that bicarbonate is superior to saline in the prevention of CIN.

Clinical Bottom Line: Clinicians should consider selecting intravenous bicarbonate rather than saline for the prevention of CIN.

Acute Decompensated Heart Failure Treatment

Gheorghiade M, Konstam MA, Burnett JC, et al. Short‐term clinical effects of tolvaptan, an oral vasopressin antagonist, in patients hospitalized for heart failure: the EVEREST clinical status trials. JAMA. 2007;297:13321343.

Question: What is the efficacy and safety of short‐term tolvaptan added to standard therapy in the treatment of acute decompensated heart failure?

Sponsor: Otsuka America, Inc.

Study Design: Two concurrent randomized, double‐blind, placebo‐controlled trials. Two trials (each with different sites) were conducted to fulfill regulatory requirements for establishing efficacy from at least 2 independent, adequately powered, and well‐controlled trials.

Patients: Two thousand forty‐eight adults (trial A) and 2085 adults (trial B) hospitalized with heart failure. Eligibility criteria included a history of chronic heart failure requiring treatment for at least 30 days prior to admission, an ejection fraction 40% at any point in the prior year, dyspnea at rest or with minimal exertion, and 2 or more signs of congestion (dyspnea, jugular vein distension, or peripheral edema). Selected exclusionary criteria included active myocardial ischemia, recent cardiac surgery, systolic blood pressure < 90 mm Hg, serum creatinine > 3.5 mg/dL, serum potassium > 5.5 mg/dL, or hemoglobin < 9 g/dL.

Setting: Three hundred fifty‐nine sites across North America, South America, and Europe. Trial A patients were assigned from 179 of these sites. Trial B patients were assigned from 180 of these sites.

Intervention: Tolvaptan, a vasopressin antagonist (30 mg orally daily), versus matching placebo, in addition to standard therapy. Treatment was started within 48 hours of admission and was continued through discharge for a minimum of 60 days.

Outcomes: Composite of global clinical status and body weight at day 7 or at discharge if earlier. Additional secondary endpoints were dyspnea (day 1) and peripheral edema (day 7).

Follow‐Up: Seven days.

Results: Tolvaptan improved the composite primary endpoint compared with placebo, and this was primarily related to greater overall net diuresis: 3.35 kg of diuresis at day 7 or discharge with tolvaptan versus 2.73 kg with placebo (trial A) and 3.77 kg of diuresis at day 7 or discharge with tolvaptan versus 2.79 kg with placebo (trial B; P < 0.001 for both trials). Net diuresis at day 1 was also greater with tolvaptan. More patients reported improved dyspnea at day 1 with tolvaptan: 76.74% versus 70.61% (trial A) and 72.06% versus 65.32% (trial B; P < 0.001 for both comparisons). Edema scores at day 7 favored tolvaptan in trial B (P = 0.02) but in not trial A (P = 0.07). Hypernatremia was more common with tolvaptan in trial A (1.4% versus 0%, P < 0.001) but not in trial B (0.5% versus 0%, P = 0.06). Tolvaptan‐treated patients had lower average furosemide doses than placebo‐treated patients. Patient‐assessed global clinical status at day 7, as measured by a visual analog scale, was no different.

Conclusions: Tolvaptan, added to standard care for acute heart failure, safely improved many but not all short‐term heart failure signs and symptoms.

Commentary: The accompanying Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) outcomes trial demonstrated that longer term use of tolvaptan for 60 days was not associated with changes in cardiovascular morbidity and mortality.9 Concerns have been raised about the safety of nesiritide10 and inotropes11 in the treatment of acute decompensated heart failure. With the completion of this 2‐part trial, we have a safe addition to the current armamentarium of treatments for acute decompensated heart failure. Clinicians should exercise caution in adding tolvaptan only to patients whose characteristics mirror those in this trial.

Clinical Bottom Line: Tolvaptan represents an effective and safe addition to therapies for acute decompensated heart failure.

Cardiovascular Risk Reduction

Dentali F, Douketis JD, Lim W, Crowther M. Combined aspirin‐oral anticoagulant therapy compared with oral anticoagulant therapy alone among patients at risk for cardiovascular disease: a meta‐analysis of randomized trials. Arch Intern Med. 2007;167:117124.

Question: For patients receiving oral anticoagulant therapy (OAC), does the addition of aspirin reduce major adverse cardiovascular endpoints?

Sponsor: Heart and Stroke Foundation of Canada.

Study Design: Meta‐analysis of 10 randomized controlled trials.

Study Selection: From MEDLINE (to June 2005), EMBASE (to June 2005), and Cochrane (to 2005, issue 2) reviews, including manual reference list reviews, 10 studies were identified that satisfied 4 criteria: (1) a randomized controlled trial in patients requiring OAC therapy, (2) a comparison of combined aspirinOAC therapy with OAC alone (the same target international normalized ratio in both arms), (3) follow‐up of at least 3 months, and (4) at least 1 prespecified outcome that was objectively documented. The 10 trials meeting these criteria studied 4180 patients. The target international normalized ratio varied across the trials on the basis of the population studied. The aspirin dose was at least 75 mg/day in all studies.

Outcomes: Arterial thromboembolism, all‐cause mortality, and major bleeding. Secondary outcomes included fatal arterial thromboembolism and fatal major bleeding.

Results: Arterial thromboembolism was lower with combined aspirinOAC therapy (6.3%) versus OAC therapy alone (8.8%; absolute risk reduction = 2.5%, number needed to treat = 40, P < 0.001). In subgroup analysis, this difference was found only among patients with mechanical heart valves (odds ratio = 0.27, 95% CI 0.150.49). There was no benefit among patients with atrial fibrillation (odds ratio = 0.99, 95% CI 0.472.07) or coronary artery disease (odds ratio = 0.69, 95% CI 0.351.3). Mortality was no different. Major bleeding was more common with combined therapy (3.8%) versus OAC therapy alone (2.8%; absolute risk reduction = 1.0%, number needed to harm = 100, P = 0.05). Secondary outcomes were not different.

Conclusions: Combined aspirinOAC therapy does not protect against future arterial thromboembolism in comparison with OAC therapy alone, except among patients with mechanical heart valves. Combined therapy, however, is associated with higher rates of major bleeding.

Commentary: These findings question the current practice of combining OAC with aspirin in patients with separate indications for each. Looking in more detail at the analyzed trials, the researchers found that there were relatively few patients with proven coronary artery disease. There may have been insufficient power to show a benefit for combined therapy among these patients. Patients with mechanical heart valves, however, clearly showed benefit. A recently published retrospective study of more than 4000 patients also concluded that the hemorrhagic risk of combined aspirinOAC therapy versus OAC therapy alone appeared to outweigh the benefit.12

Clinical Bottom Line: Except among patients with mechanical heart valves, combined aspirinOAC increases bleeding risk without proven benefit. Until further data are available, clinicians should individualize antithrombotic therapy on the basis of a careful assessment of risk and benefit.

Cardioembolic Stroke Treatment

Paciaroni M, Agnelli G, Micheli S, Caso V. Efficacy and safety of anticoagulant treatment in acute cardioembolic stroke: a meta‐analysis of randomized controlled trials. Stroke. 2007;38:423430.

Question: What are the safety and efficacy of anticoagulation in the treatment of acute cardioembolic stroke?

Sponsor: None.

Study Design: Meta‐analysis of 7 randomized controlled trials.

Study Selection: Trials randomizing patients within 48 hours from stroke onset with objectively diagnosed stroke of presumed cardioembolic origin that compared full‐dose anticoagulants (unfractionated heparin, low‐molecular‐weight heparin, and heparinoid) to other treatments (aspirin or placebo) for initial therapy and used objective methods to assess study outcomes.

Outcomes: A composite of death or disability at final follow‐up (at least 3 months), all new strokes (ischemic and hemorrhagic) at 14 days, and pulmonary embolism. The safety outcome was symptomatic intracranial bleeding.

Results: The odds ratio (95% CI) for death or disability with anticoagulation versus aspirin or placebo was 1.01 (95% CI 0.821.24); the odds ratio for all new strokes with anticoagulation versus aspirin or placebo was 1.18 (95% CI 0.741.88). The odds ratio for pulmonary embolism with anticoagulation versus aspirin was 0.94 (95% CI 0.442.00). None of these were statistically significant. However, the odds ratio for symptomatic intracranial hemorrhage with anticoagulation versus aspirin or placebo was 2.89 (95% CI 1.197.01, P = 0.02). The absolute increase in symptomatic intracranial bleeding with anticoagulation was 1.8% (number needed to harm = 55). Of the 7 trials analyzed, 1 trial did show a reduction in overall death or disability with anticoagulation, in which therapy was started within 3 hours of symptom onset (odds ratio = 0.49, 95% CI 0.260.93). This trial was small, and subgroup analysis in the other, larger trials failed to confirm this finding.

Conclusion: Anticoagulation for acute stroke of suspected cardioembolic origin does not improve outcomes but is associated with higher rates of symptomatic intracranial hemorrhage.

Commentary: Long‐term anticoagulation with sodium warfarin clearly lowers cardioembolic stroke risk for patients with chronic atrial fibrillation. This meta‐analysis demonstrates that acute anticoagulation does not reduce the composite endpoint of death or disability, recurrent stroke, or pulmonary embolism. The risk of symptomatic intracranial hemorrhage is substantially increased and argues against the use of anticoagulants during the acute phase of suspected cardioembolic stroke.

Clinical Bottom Line: Anticoagulation is harmful and does not reduce death or disability in the acute phase of suspected cardioembolic stroke.

Clostridium Difficile Associated Diarrhea

Zar FA, Bakkanagari SR, Moorthi KM, Davis MB. A comparison of vancomycin and metronidazole for the treatment of Clostridium difficile‐associated diarrhea, stratified by disease severity. Clin Infect Dis. 2007;45:302307.

Question: What is the best first‐line treatment for Clostridium difficileassociated diarrhea (CDAD)?

Sponsor: None.

Study Design: Randomized, double‐blind, placebo‐controlled trial.

Patients: One hundred fifty patients with 3 or more nonformed stools in 24 hours with a positive stool C. difficile toxin A test or the presence of pseudomembranous colitis on endoscopy.

Setting: A 200‐bed community teaching hospital affiliated with an academic medical center.

Intervention: Metronidazole (250 mg 4 times daily) plus vancomycin liquid placebo versus metronidazole placebo plus vancomycin liquid (125 mg 4 times daily), both for 10 days.

Outcomes: The primary outcomes were cure (resolution of diarrhea by day 6 of treatment and a negative stool toxin at both 6 and 10 days post‐treatment), treatment failure (persistent diarrhea and/or an inability to clear the toxin at 6 days, the need for colectomy, or death after 5 days of treatment), and relapse (recurrence of toxin‐positive CDAD by day 21 after the initial cure). Disease was categorized as mild (<2 points) or severe ( 2 points), with 1 point each for age > 60 years, temperature > 38.3C, albumin < 2.5 mg/dL, and a peripheral white blood count > 15,000 cells/mm³ within 48 hours of enrollment. Two points were allotted for endoscopic findings of pseudomembranous colitis.

Follow‐Up: Patients were monitored for 21 days for resolution of diarrhea (2 formed stools in 24 hours). Stool toxin was measured at days 6 and 10 of treatment and at day 21 if diarrhea was still present.

Results: One hundred fifty patients (81 patients with mild disease and 69 patients with severe disease) finished the trial, with no significant differences in patients categorized into the 2 treatment arms. Overall, 84% (66/79) of patients receiving metronidazole were cured versus 97% (69/71) of patients receiving vancomycin (P = 0.006). In patients with mild disease, 90% (37/41) and 98% (39/40) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.36). In patients with severe disease, 76% (29/38) and 97% (30/31) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.02). After the initial cure, relapse occurred in 7% (5/76), 15% (9/59), 14% (9/66), and 7% (5/69) of patients with mild disease, severe disease (P = 0.15 for mild versus severe), metronidazole treatment, and vancomycin treatment (P = 0.27 between treatments), respectively. In patients with severe CDAD, low albumin, intensive care, and presence of pseudomembranous colitis were associated with metronidazole treatment failure.

Conclusion: Metronidazole is equally effective as vancomycin in treating mild CDAD; however, vancomycin appears superior to metronidazole in treating patients with severe CDAD.

Commentary: Two prior studies evaluating metronidazole and vancomycin for CDAD revealed no significant difference between the 2 therapies.13, 14 However, these studies had serious methodological flaws, including a lack of blinding and too little power to show a difference. This randomized, double‐blind, placebo‐controlled trial provides convincing evidence that oral vancomycin is superior to metronidazole in patients with severe CDAD. This is an especially important finding as the recently described hypervirulent epidemic strain of C. difficile becomes more prevalent.

A single‐center retrospective study of 102 veterans with metronidazole‐treated CDAD showed analogous findings with a slightly different scoring system.15 In 94% of metronidazole responders, the score was 2 or less. In 67% of true failures, the score was greater than 2. Taken together, these studies suggest that higher scores predict metronidazole failure.

Clinical Bottom Line: Vancomycin appears to be more effective than metronidazole in treating more severe forms of CDAD.

Consultative Medicine: Orthopedics

Lyles KW, Coln‐Emeric CS, Magaziner JS, et al. Zoledronic acid and clinical fractures and mortality after hip fracture. N Engl J Med. 2007;357:17991809.

Question: Does an annual dose of zoledronic acid reduce the rate of subsequent fractures and mortality in patients with a recent hip fracture?

Sponsor: Novartis.

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: A total of 2127 men and women 50 years old or older with a surgically repaired low‐impact hip fracture (eg, fall from a standing height) within 90 days of study entry who were unwilling or unable to take an oral bisphosphonate.

Setting: International and multicenter.

Intervention: A single 5‐mg intravenous dose of zoledronic acid within 90 days of a hip fracture repair versus an intravenous placebo, given annually. All patients with documented vitamin D deficiency or no documentation of a serum 25‐hydroxyvitamin D level received a loading dose of vitamin D₃ or D₂ 14 days prior to the first infusion. All patients received oral calcium and vitamin D daily after the first infusion.

Outcomes: The primary outcome was a new clinical fracture excluding facial, digital, or abnormal bone (eg, bone with metastases) fractures. Secondary outcomes included changes in the bone mineral density in the nonfractured hip, the number of new vertebral, nonvertebral, and hip fractures, and predetermined safety outcomes.

Follow‐Up: Quarterly phone calls and annual clinic visits for up to 5 years.

Results: The trial was stopped early after prespecified efficacy objectives were met. At an average follow‐up of 1.9 years, 8.6% of subjects receiving zoledronic acid and 13.9% of those receiving placebo suffered subsequent fractures (P = 0.001). Statistically significant improvements in bone mineral density were seen at both the total hip and femoral neck sites in the zoledronic acid group versus the placebo group. Approximately 80% of patients experienced an adverse event in each group, with statistically significantly more pyrexia, myalgias, and bone pain in the zoledronic acid cohort and higher mortality in the placebo group, that is, 9.6% versus 13.3% (hazard ratio = 0.72, 95% CI 0.560.72, P = 0.01).

Conclusion: Annual treatment with 5 mg of intravenous zoledronic acid reduces clinical fractures and mortality when it is dosed within 90 days of a hip fracture repair.

Commentary: Patients who suffer a hip fracture are at high risk for successive fractures, with a considerable morbid and financial burden on the patient and the healthcare system. Additionally, as many as 1 in 4 of these patients will die in the subsequent year. Poor adherence to oral bisphosphonates and prescriber nonadherence to fracture guidelines are common sources of noncompliance and have been associated with increased fracture burden. The findings that an annual infusion can achieve reductions in the fracture rate and mortality are notable and offer options for patients who otherwise could not comply with therapy because of side effects or an inability to take a more frequently dosed medication.

Clinical Bottom Line: An annual dose of zoledronic acid reduces the rate of subsequent fractures and death in patients with a recent hip fracture.

Critical Care Medicine

Francois B, Bellissant E, Gissot V, et al. 12‐h pretreatment with methylprednisolone versus placebo for prevention of postextubation laryngeal edema: a randomized double blind trial. Lancet. 2007;369:10831089.

Question: Do pre‐extubation steroids prior to planned extubation prevent postextubation laryngeal edema?

Sponsor: Institutional funding (P. Vignon, personal communication, March 2008).

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: Seven hundred sixty‐one adult patients with at least 36 hours of mechanical ventilation and planned extubation.

Setting: Fifteen intensive care units in France.

Intervention: Intravenous methylprednisolone (20 mg) starting 12 hours before extubation and continuing every 4 hours until extubation, including the time of extubation (total dose = 80 mg), or a placebo identical in appearance and delivery.

Outcomes: The primary outcome was the development of minor (inspiratory stridor associated with respiratory distress requiring intervention) or major (reintubation secondary to laryngoscopically visualized upper airway obstruction) laryngeal edema within 24 hours of extubation.

Follow‐Up: Clinical assessments were performed 10 minutes and 1, 1.5, 3, 6, 12, and 24 hours after extubation.

Results: Six hundred ninety‐eight patients completed the trial. The median duration of intubation prior to extubation was 6 days. Any laryngeal edema occurred in 22% (76/343) and 3% (11/355) of patients in the placebo and treatment groups, respectively (P < 0.0001). When edema was present, the severity and timing of the onset of edema did not differ between the 2 groups. Reintubation was reduced from 8% (26/343) in the placebo group to 4% (13/355) in the treatment groups (P = 0.02). When necessary, reintubation was deemed secondary to major edema in 54% (14/26) of the placebo group and 8% (1/13) of the treatment group, respectively. An intention‐to‐treat analysis did not alter the study findings. One patient in each group suffered a serious adverse event: respiratory failure and death 23 hours after extubation in the placebo group and septic shock and death 26 hours after extubation in the treatment group. Rates of hyperglycemia and infections were not reported.

Conclusion: The use of 20‐mg intravenous doses of methylprednisolone spaced 4 hours apart and starting 12 hours prior to planned extubation is associated with significant reductions in the rates of tracheal edema and reintubation.

Commentary: Postextubation laryngeal edema is common (2%22% incidence) and results in reintubation for 0.74.7% of extubations. This work shows that a simple pretreatment with intravenous steroids 12 hours before planned extubation can reduce the rate of postextubation edema 7‐fold, including a 2‐fold reduction in the reintubation rate. Prior trials using shorter periods of treatment (<6 hours) have not shown benefit, so achieving this study's results likely requires the full 12‐hour protocol.

Clinical Bottom Line: Intravenous methylprednisolone dosed 12 hours before and every 4 hours until planned extubation reduces the rate of reintubation due to tracheal edema.

This update reviews key clinical articles for hospitalists published over the past year. Selection criteria include high methodological quality, pertinence to hospital medicine, and likelihood that a change in practice is warranted. Table 1 summarizes practice changes.

Enoxaparin Dosing in Acute Coronary Syndromes

Allen La Pointe NM, Chen AY, Alexander KP, et al. Enoxaparin dosing and associated risk of in‐hospital bleeding and death in patients with non‐ST‐segment elevation acute coronary syndromes. Arch Intern Med. 2007;167:15391544.

Question: Among patients with non‐ST‐elevation acute coronary syndromes, how common and harmful is excess enoxaparin dosing?

Sponsors: Schering‐Plough Corp., Bristol‐Myers Squibb/Sanofi‐Aventis Pharmaceuticals Partnership, Millennium Pharmaceuticals, and the National Institutes of Health and National Institute on Aging.

Study Design: Observational study of prospective cohort data from the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA Guidelines (CRUSADE) National Quality Improvement Initiative.

Patients: A total of 10,687 patients receiving enoxaparin for non‐ST‐elevation acute coronary syndromes.

Setting: Three hundred thirty‐two US hospitals.

Outcomes: Rate of excess enoxaparin dose, defined as greater than 10 mg/day above the recommended dose of 1 mg/kg every 12 hours for creatinine clearance (CrCl) 30 mL/minute or 1 mg/kg every 24 hours for CrCl < 30 mL/minute; rates of in‐hospital major bleeding and death; and rate of lower than recommended enoxaparin dose.

Results: Excess enoxaparin dosing occurred in 18.7% of the cohort (2002/10,687). Of these, 57.8% (1157/2002) had CrCl < 30 mL/minute. Excess‐dose patients were more likely to be older and female and have a low body mass index (P < 0.001 for all comparisons). In‐hospital major bleeding (14.2% versus 7.3%, P< 0.001) and in‐hospital death (5.6% versus 2.4%, P < 0.001) were more common among excess‐dose patients. Enoxaparin underdosing occurred in 29.2% (3116/10 687) and was not associated with excess harm. Controlling for baseline characteristics, the authors found that the adjusted odds ratio for in‐hospital major bleeding in the excess‐dose cohort was 1.43 (1.181.75, P < 0.001) and the adjusted odds ratio for death was 1.35 (1.031.77, P = 0.03).

Conclusions: Excess enoxaparin dosing in non‐ST‐elevation acute coronary syndromes occurred in about 1 of every 5 patients treated in this prospective multihospital registry. Excess dosing was associated with substantially higher rates of major in‐hospital bleeding and death, with a number needed to harm of 78 for major bleeding and a number needed to harm of 167 for in‐hospital death. In comparison, the number needed to treat with another low‐molecular‐weight heparin (dalteparin) was 34 to prevent 1 death or myocardial infarction in the first 6 days, with a nonsignificant trend toward decreased mortality.1

Commentary: Providers likely underestimate the degree of renal impairment when looking solely at serum creatinine instead of estimates of CrCl. Excess dosing was more common among elderly, thin, and female patients. Clinicians must calculate the enoxaparin dose on the basis of careful estimates of CrCl to limit this risk. The Modification of Diet in Renal Disease (MDRD) equation is commonly used for this purpose.

Clinical Bottom Line: Enoxaparin excess dosing is common and harmful. Clinicians can mitigate this risk by more carefully estimating renal function when selecting the proper enoxaparin dose of 1 mg/kg twice daily for CrCl 30 mL/minute and 1 mg/kg once daily for CrCl < 30 mL/minute.

Venous Thromboembolism Prevention

Wein L, Wein S, Haas SJ, et al. Pharmacological venous thromboembolism prophylaxis in hospitalized medical patients. Arch Intern Med. 2007;167:14761486.

Question: What is the relative safety and efficacy of various pharmacological agents for preventing venous thromboembolism among hospitalized medical patients?

Sponsor: National Health and Medical Council of Australia.

Study Design: Meta‐analysis of 36 prospective randomized controlled trials involving about 48,000 patients.

Study Selection: Prospective randomized controlled trials enrolling at least 30 patients comparing 1 of 4 regimens: (1) unfractionated heparin (UFH) versus control, (2) low‐molecular‐weight heparin (LMWH) versus control, (3) LMWH versus UFH, or (4) Factor Xa inhibitor versus placebo. Trials of surgical, trauma, and critical care patients were excluded. Only 1 Factor Xa trial (fondaparinux) was located,2 and thus it was not eligible for meta‐analysis.

Outcomes: Pooled relative risks with 95% confidence intervals for deep venous thrombosis (DVT), pulmonary embolism (PE), mortality, and total bleeding. The authors also compared 2 UFH regimens: 5000 units twice daily versus 5000 units thrice daily.

Results: UFH (all doses, compared with control): The relative risk was 0.33 (95% CI 0.260.42) for DVT and 0.64 (95% CI 0.500.82) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 3.11 (95% CI 2.443.96, P = 0.001).

LMWH (compared with control): The relative risk was 0.56 (95% CI 0.450.70) for DVT and 0.37 (95% CI 0.210.64) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 1.92 (95% CI 1.322.78, P = 0.001).

LMWH (compared with UFH, all doses): The relative risk for DVT was 0.68 (95% CI 0.520.88, P = 0.004), but the risk was not different for PE, mortality, or major bleeding.

UFH (5000 units twice daily, compared with control): The relative risk for DVT was 0.52 (95% CI 0.280.96, P = 0.04). When the random‐effects model was used, this difference became statistically nonsignificant (relative risk = 0.41, 95% CI 0.101.73, P = 0.23).

UFH (5000 units 3 times daily, compared with control): The relative risk for DVT was 0.27 (95% CI 0.200.36, P = 0.001). This difference remained when the random‐effects model was applied (relative risk = 0.28, 95% confidence interval = 0.210.38, P = 0.001).

Conclusions: Both UFH and LMWH reduce DVT and PE in hospitalized medical patients. Neither affects mortality. Both increase the risk of major bleeding. LMWH reduces the risk of DVT but not the risk of PE in comparison with UFH (all doses). When adjusted for random effects, UFH at a dose of 5000 units twice daily does not appear to be different than the control.

Commentary: This well‐conducted meta‐analysis demonstrates the efficacy of heparin, whether unfractionated or low‐molecular‐weight, in the prevention of venous thromboembolism. Of note, the UFH dose of 5000 units twice daily did not appear to be different than placebo. The UFH dose of 5000 units 3 times daily, by contrast, was effective in both the fixed‐effects and random‐effects models. Mortality was unaffected by any of the regimens studied. All regimens were associated with increased risks of major bleeding.

Clinical Bottom Line: Pharmacological prophylaxis with UFH 3 times daily or LMWH reduces the risk for venous thromboembolism. Twice daily UFH is not clearly different from placebo. Overall mortality was unaffected by any of the regimens for prophylaxis.

Contrast Nephropathy Prevention

Briguori C, Airoldi F, D'Andrea D, et al. Renal insufficiency following contrast media administration trial (REMEDIAL): a randomized comparison of 3 preventive strategies. Circulation. 2007;115:12111217.

Question: What is the efficacy of saline versus bicarbonate for the prevention of contrast mediainduced nephropathy?

Sponsor: Institutional funding (C. Briguori, personal communication, January 2008).

Study Design: Randomized trial.

Patients: Three hundred twenty‐six consecutive patients with serum creatinine 2.0 mg/dL and/or an estimated glomerular filtration rate < 40 mL/minute/1.73 m² undergoing elective coronary and/or peripheral angiography.

Setting: Two interventional cardiology laboratories in Italy.

Intervention: Patients were randomized to 1 of 3 preventive regimens: (1) intravenous saline (0.9%) given at a rate of 1 mL/kg of body weight/hour 12 hours prior to the procedure and continuing for 12 hours afterward (reduced to 0.5 mL/kg/hour for patients with a left ventricular ejection fraction < 40%) plus N‐acetylcysteine (NAC; 1200 mg orally twice daily) on the day before the procedure and the day of the procedure; (2) intravenous sodium bicarbonate (154 mEq/L in dextrose and water) given as an initial bolus of 3 mL/kg over 1 hour prior to the procedure and continuing at a rate of 1 mL/kg/hour for 6 hours more plus NAC as above; or (3) intravenous saline as above plus intravenous ascorbic acid (3 g) 2 hours prior to the procedure followed by 2 g on the night and morning after the procedure plus NAC as above.

Outcomes: Rate of contrast‐induced nephropathy (CIN), which was defined as an increase in serum creatinine 25% from the baseline value at 48 hours after the administration of contrast or the need for hemodialysis.

Follow‐Up: Forty‐eight hours.

Results: The baseline serum creatinine was about 2.0 mg/dL and did not differ among the 3 groups. The rate of CIN was 9.9% (11/111) in the saline plus NAC group, 1.9% (2/108) in the bicarbonate plus NAC group, and 10.3% (11/107) in the saline plus ascorbic acid plus NAC group. The bicarbonate plus NAC regimen was superior to saline plus NAC (P = 0.019). The absolute risk reduction for bicarbonate plus NAC versus saline plus NAC was 8% (a number needed to treat of 13 to prevent 1 case of CIN). The saline plus NAC and saline plus ascorbic acid plus NAC groups did not differ in outcome.

Conclusions: Sodium bicarbonate plus NAC is superior to saline plus NAC for the prevention of CIN among patients with baseline chronic kidney disease.

Commentary: This trial confirms the results of the initial study by Merten et al.3 showing the superiority of bicarbonate versus saline in the prevention of CIN. That trial, published in 2004, did not use NAC. Also in 2007, 3 other single‐center randomized trials of saline versus bicarbonate in the prevention of CIN were published.46 All concluded that bicarbonate is superior to saline. Whether NAC is effective for CIN prevention remains unclear.7 Given its low side‐effect profile, it is not unreasonable to continue using NAC until further data are available. At‐risk patients receiving intravenous contrast for other indications (eg, computed tomography) would likely show similar benefit. Although there are now 5 prospective blinded controlled trials showing the superiority of bicarbonate, a recently published large retrospective cohort found that the use of sodium bicarbonate was associated with increased incidence of CIN.8 The concordant results of all 5 prospective randomized trials of sodium bicarbonate, along with the risk for unmeasured confounding variables with retrospective cohort analysis, suggest that bicarbonate is superior to saline in the prevention of CIN.

Clinical Bottom Line: Clinicians should consider selecting intravenous bicarbonate rather than saline for the prevention of CIN.

Acute Decompensated Heart Failure Treatment

Gheorghiade M, Konstam MA, Burnett JC, et al. Short‐term clinical effects of tolvaptan, an oral vasopressin antagonist, in patients hospitalized for heart failure: the EVEREST clinical status trials. JAMA. 2007;297:13321343.

Question: What is the efficacy and safety of short‐term tolvaptan added to standard therapy in the treatment of acute decompensated heart failure?

Sponsor: Otsuka America, Inc.

Study Design: Two concurrent randomized, double‐blind, placebo‐controlled trials. Two trials (each with different sites) were conducted to fulfill regulatory requirements for establishing efficacy from at least 2 independent, adequately powered, and well‐controlled trials.

Patients: Two thousand forty‐eight adults (trial A) and 2085 adults (trial B) hospitalized with heart failure. Eligibility criteria included a history of chronic heart failure requiring treatment for at least 30 days prior to admission, an ejection fraction 40% at any point in the prior year, dyspnea at rest or with minimal exertion, and 2 or more signs of congestion (dyspnea, jugular vein distension, or peripheral edema). Selected exclusionary criteria included active myocardial ischemia, recent cardiac surgery, systolic blood pressure < 90 mm Hg, serum creatinine > 3.5 mg/dL, serum potassium > 5.5 mg/dL, or hemoglobin < 9 g/dL.

Setting: Three hundred fifty‐nine sites across North America, South America, and Europe. Trial A patients were assigned from 179 of these sites. Trial B patients were assigned from 180 of these sites.

Intervention: Tolvaptan, a vasopressin antagonist (30 mg orally daily), versus matching placebo, in addition to standard therapy. Treatment was started within 48 hours of admission and was continued through discharge for a minimum of 60 days.

Outcomes: Composite of global clinical status and body weight at day 7 or at discharge if earlier. Additional secondary endpoints were dyspnea (day 1) and peripheral edema (day 7).

Follow‐Up: Seven days.

Results: Tolvaptan improved the composite primary endpoint compared with placebo, and this was primarily related to greater overall net diuresis: 3.35 kg of diuresis at day 7 or discharge with tolvaptan versus 2.73 kg with placebo (trial A) and 3.77 kg of diuresis at day 7 or discharge with tolvaptan versus 2.79 kg with placebo (trial B; P < 0.001 for both trials). Net diuresis at day 1 was also greater with tolvaptan. More patients reported improved dyspnea at day 1 with tolvaptan: 76.74% versus 70.61% (trial A) and 72.06% versus 65.32% (trial B; P < 0.001 for both comparisons). Edema scores at day 7 favored tolvaptan in trial B (P = 0.02) but in not trial A (P = 0.07). Hypernatremia was more common with tolvaptan in trial A (1.4% versus 0%, P < 0.001) but not in trial B (0.5% versus 0%, P = 0.06). Tolvaptan‐treated patients had lower average furosemide doses than placebo‐treated patients. Patient‐assessed global clinical status at day 7, as measured by a visual analog scale, was no different.

Conclusions: Tolvaptan, added to standard care for acute heart failure, safely improved many but not all short‐term heart failure signs and symptoms.

Commentary: The accompanying Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) outcomes trial demonstrated that longer term use of tolvaptan for 60 days was not associated with changes in cardiovascular morbidity and mortality.9 Concerns have been raised about the safety of nesiritide10 and inotropes11 in the treatment of acute decompensated heart failure. With the completion of this 2‐part trial, we have a safe addition to the current armamentarium of treatments for acute decompensated heart failure. Clinicians should exercise caution in adding tolvaptan only to patients whose characteristics mirror those in this trial.

Clinical Bottom Line: Tolvaptan represents an effective and safe addition to therapies for acute decompensated heart failure.

Cardiovascular Risk Reduction

Dentali F, Douketis JD, Lim W, Crowther M. Combined aspirin‐oral anticoagulant therapy compared with oral anticoagulant therapy alone among patients at risk for cardiovascular disease: a meta‐analysis of randomized trials. Arch Intern Med. 2007;167:117124.

Question: For patients receiving oral anticoagulant therapy (OAC), does the addition of aspirin reduce major adverse cardiovascular endpoints?

Sponsor: Heart and Stroke Foundation of Canada.

Study Design: Meta‐analysis of 10 randomized controlled trials.

Study Selection: From MEDLINE (to June 2005), EMBASE (to June 2005), and Cochrane (to 2005, issue 2) reviews, including manual reference list reviews, 10 studies were identified that satisfied 4 criteria: (1) a randomized controlled trial in patients requiring OAC therapy, (2) a comparison of combined aspirinOAC therapy with OAC alone (the same target international normalized ratio in both arms), (3) follow‐up of at least 3 months, and (4) at least 1 prespecified outcome that was objectively documented. The 10 trials meeting these criteria studied 4180 patients. The target international normalized ratio varied across the trials on the basis of the population studied. The aspirin dose was at least 75 mg/day in all studies.

Outcomes: Arterial thromboembolism, all‐cause mortality, and major bleeding. Secondary outcomes included fatal arterial thromboembolism and fatal major bleeding.

Results: Arterial thromboembolism was lower with combined aspirinOAC therapy (6.3%) versus OAC therapy alone (8.8%; absolute risk reduction = 2.5%, number needed to treat = 40, P < 0.001). In subgroup analysis, this difference was found only among patients with mechanical heart valves (odds ratio = 0.27, 95% CI 0.150.49). There was no benefit among patients with atrial fibrillation (odds ratio = 0.99, 95% CI 0.472.07) or coronary artery disease (odds ratio = 0.69, 95% CI 0.351.3). Mortality was no different. Major bleeding was more common with combined therapy (3.8%) versus OAC therapy alone (2.8%; absolute risk reduction = 1.0%, number needed to harm = 100, P = 0.05). Secondary outcomes were not different.

Conclusions: Combined aspirinOAC therapy does not protect against future arterial thromboembolism in comparison with OAC therapy alone, except among patients with mechanical heart valves. Combined therapy, however, is associated with higher rates of major bleeding.

Commentary: These findings question the current practice of combining OAC with aspirin in patients with separate indications for each. Looking in more detail at the analyzed trials, the researchers found that there were relatively few patients with proven coronary artery disease. There may have been insufficient power to show a benefit for combined therapy among these patients. Patients with mechanical heart valves, however, clearly showed benefit. A recently published retrospective study of more than 4000 patients also concluded that the hemorrhagic risk of combined aspirinOAC therapy versus OAC therapy alone appeared to outweigh the benefit.12

Clinical Bottom Line: Except among patients with mechanical heart valves, combined aspirinOAC increases bleeding risk without proven benefit. Until further data are available, clinicians should individualize antithrombotic therapy on the basis of a careful assessment of risk and benefit.

Cardioembolic Stroke Treatment

Paciaroni M, Agnelli G, Micheli S, Caso V. Efficacy and safety of anticoagulant treatment in acute cardioembolic stroke: a meta‐analysis of randomized controlled trials. Stroke. 2007;38:423430.

Question: What are the safety and efficacy of anticoagulation in the treatment of acute cardioembolic stroke?

Sponsor: None.

Study Design: Meta‐analysis of 7 randomized controlled trials.

Study Selection: Trials randomizing patients within 48 hours from stroke onset with objectively diagnosed stroke of presumed cardioembolic origin that compared full‐dose anticoagulants (unfractionated heparin, low‐molecular‐weight heparin, and heparinoid) to other treatments (aspirin or placebo) for initial therapy and used objective methods to assess study outcomes.

Outcomes: A composite of death or disability at final follow‐up (at least 3 months), all new strokes (ischemic and hemorrhagic) at 14 days, and pulmonary embolism. The safety outcome was symptomatic intracranial bleeding.

Results: The odds ratio (95% CI) for death or disability with anticoagulation versus aspirin or placebo was 1.01 (95% CI 0.821.24); the odds ratio for all new strokes with anticoagulation versus aspirin or placebo was 1.18 (95% CI 0.741.88). The odds ratio for pulmonary embolism with anticoagulation versus aspirin was 0.94 (95% CI 0.442.00). None of these were statistically significant. However, the odds ratio for symptomatic intracranial hemorrhage with anticoagulation versus aspirin or placebo was 2.89 (95% CI 1.197.01, P = 0.02). The absolute increase in symptomatic intracranial bleeding with anticoagulation was 1.8% (number needed to harm = 55). Of the 7 trials analyzed, 1 trial did show a reduction in overall death or disability with anticoagulation, in which therapy was started within 3 hours of symptom onset (odds ratio = 0.49, 95% CI 0.260.93). This trial was small, and subgroup analysis in the other, larger trials failed to confirm this finding.

Conclusion: Anticoagulation for acute stroke of suspected cardioembolic origin does not improve outcomes but is associated with higher rates of symptomatic intracranial hemorrhage.

Commentary: Long‐term anticoagulation with sodium warfarin clearly lowers cardioembolic stroke risk for patients with chronic atrial fibrillation. This meta‐analysis demonstrates that acute anticoagulation does not reduce the composite endpoint of death or disability, recurrent stroke, or pulmonary embolism. The risk of symptomatic intracranial hemorrhage is substantially increased and argues against the use of anticoagulants during the acute phase of suspected cardioembolic stroke.

Clinical Bottom Line: Anticoagulation is harmful and does not reduce death or disability in the acute phase of suspected cardioembolic stroke.

Clostridium Difficile Associated Diarrhea

Zar FA, Bakkanagari SR, Moorthi KM, Davis MB. A comparison of vancomycin and metronidazole for the treatment of Clostridium difficile‐associated diarrhea, stratified by disease severity. Clin Infect Dis. 2007;45:302307.

Question: What is the best first‐line treatment for Clostridium difficileassociated diarrhea (CDAD)?

Sponsor: None.

Study Design: Randomized, double‐blind, placebo‐controlled trial.

Patients: One hundred fifty patients with 3 or more nonformed stools in 24 hours with a positive stool C. difficile toxin A test or the presence of pseudomembranous colitis on endoscopy.

Setting: A 200‐bed community teaching hospital affiliated with an academic medical center.

Intervention: Metronidazole (250 mg 4 times daily) plus vancomycin liquid placebo versus metronidazole placebo plus vancomycin liquid (125 mg 4 times daily), both for 10 days.

Outcomes: The primary outcomes were cure (resolution of diarrhea by day 6 of treatment and a negative stool toxin at both 6 and 10 days post‐treatment), treatment failure (persistent diarrhea and/or an inability to clear the toxin at 6 days, the need for colectomy, or death after 5 days of treatment), and relapse (recurrence of toxin‐positive CDAD by day 21 after the initial cure). Disease was categorized as mild (<2 points) or severe ( 2 points), with 1 point each for age > 60 years, temperature > 38.3C, albumin < 2.5 mg/dL, and a peripheral white blood count > 15,000 cells/mm³ within 48 hours of enrollment. Two points were allotted for endoscopic findings of pseudomembranous colitis.

Follow‐Up: Patients were monitored for 21 days for resolution of diarrhea (2 formed stools in 24 hours). Stool toxin was measured at days 6 and 10 of treatment and at day 21 if diarrhea was still present.

Results: One hundred fifty patients (81 patients with mild disease and 69 patients with severe disease) finished the trial, with no significant differences in patients categorized into the 2 treatment arms. Overall, 84% (66/79) of patients receiving metronidazole were cured versus 97% (69/71) of patients receiving vancomycin (P = 0.006). In patients with mild disease, 90% (37/41) and 98% (39/40) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.36). In patients with severe disease, 76% (29/38) and 97% (30/31) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.02). After the initial cure, relapse occurred in 7% (5/76), 15% (9/59), 14% (9/66), and 7% (5/69) of patients with mild disease, severe disease (P = 0.15 for mild versus severe), metronidazole treatment, and vancomycin treatment (P = 0.27 between treatments), respectively. In patients with severe CDAD, low albumin, intensive care, and presence of pseudomembranous colitis were associated with metronidazole treatment failure.

Conclusion: Metronidazole is equally effective as vancomycin in treating mild CDAD; however, vancomycin appears superior to metronidazole in treating patients with severe CDAD.

Commentary: Two prior studies evaluating metronidazole and vancomycin for CDAD revealed no significant difference between the 2 therapies.13, 14 However, these studies had serious methodological flaws, including a lack of blinding and too little power to show a difference. This randomized, double‐blind, placebo‐controlled trial provides convincing evidence that oral vancomycin is superior to metronidazole in patients with severe CDAD. This is an especially important finding as the recently described hypervirulent epidemic strain of C. difficile becomes more prevalent.

A single‐center retrospective study of 102 veterans with metronidazole‐treated CDAD showed analogous findings with a slightly different scoring system.15 In 94% of metronidazole responders, the score was 2 or less. In 67% of true failures, the score was greater than 2. Taken together, these studies suggest that higher scores predict metronidazole failure.

Clinical Bottom Line: Vancomycin appears to be more effective than metronidazole in treating more severe forms of CDAD.

Consultative Medicine: Orthopedics

Lyles KW, Coln‐Emeric CS, Magaziner JS, et al. Zoledronic acid and clinical fractures and mortality after hip fracture. N Engl J Med. 2007;357:17991809.

Question: Does an annual dose of zoledronic acid reduce the rate of subsequent fractures and mortality in patients with a recent hip fracture?

Sponsor: Novartis.

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: A total of 2127 men and women 50 years old or older with a surgically repaired low‐impact hip fracture (eg, fall from a standing height) within 90 days of study entry who were unwilling or unable to take an oral bisphosphonate.

Setting: International and multicenter.

Intervention: A single 5‐mg intravenous dose of zoledronic acid within 90 days of a hip fracture repair versus an intravenous placebo, given annually. All patients with documented vitamin D deficiency or no documentation of a serum 25‐hydroxyvitamin D level received a loading dose of vitamin D₃ or D₂ 14 days prior to the first infusion. All patients received oral calcium and vitamin D daily after the first infusion.

Outcomes: The primary outcome was a new clinical fracture excluding facial, digital, or abnormal bone (eg, bone with metastases) fractures. Secondary outcomes included changes in the bone mineral density in the nonfractured hip, the number of new vertebral, nonvertebral, and hip fractures, and predetermined safety outcomes.

Follow‐Up: Quarterly phone calls and annual clinic visits for up to 5 years.

Results: The trial was stopped early after prespecified efficacy objectives were met. At an average follow‐up of 1.9 years, 8.6% of subjects receiving zoledronic acid and 13.9% of those receiving placebo suffered subsequent fractures (P = 0.001). Statistically significant improvements in bone mineral density were seen at both the total hip and femoral neck sites in the zoledronic acid group versus the placebo group. Approximately 80% of patients experienced an adverse event in each group, with statistically significantly more pyrexia, myalgias, and bone pain in the zoledronic acid cohort and higher mortality in the placebo group, that is, 9.6% versus 13.3% (hazard ratio = 0.72, 95% CI 0.560.72, P = 0.01).

Conclusion: Annual treatment with 5 mg of intravenous zoledronic acid reduces clinical fractures and mortality when it is dosed within 90 days of a hip fracture repair.

Commentary: Patients who suffer a hip fracture are at high risk for successive fractures, with a considerable morbid and financial burden on the patient and the healthcare system. Additionally, as many as 1 in 4 of these patients will die in the subsequent year. Poor adherence to oral bisphosphonates and prescriber nonadherence to fracture guidelines are common sources of noncompliance and have been associated with increased fracture burden. The findings that an annual infusion can achieve reductions in the fracture rate and mortality are notable and offer options for patients who otherwise could not comply with therapy because of side effects or an inability to take a more frequently dosed medication.

Clinical Bottom Line: An annual dose of zoledronic acid reduces the rate of subsequent fractures and death in patients with a recent hip fracture.

Critical Care Medicine

Francois B, Bellissant E, Gissot V, et al. 12‐h pretreatment with methylprednisolone versus placebo for prevention of postextubation laryngeal edema: a randomized double blind trial. Lancet. 2007;369:10831089.

Question: Do pre‐extubation steroids prior to planned extubation prevent postextubation laryngeal edema?

Sponsor: Institutional funding (P. Vignon, personal communication, March 2008).

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: Seven hundred sixty‐one adult patients with at least 36 hours of mechanical ventilation and planned extubation.

Setting: Fifteen intensive care units in France.

Intervention: Intravenous methylprednisolone (20 mg) starting 12 hours before extubation and continuing every 4 hours until extubation, including the time of extubation (total dose = 80 mg), or a placebo identical in appearance and delivery.

Outcomes: The primary outcome was the development of minor (inspiratory stridor associated with respiratory distress requiring intervention) or major (reintubation secondary to laryngoscopically visualized upper airway obstruction) laryngeal edema within 24 hours of extubation.

Follow‐Up: Clinical assessments were performed 10 minutes and 1, 1.5, 3, 6, 12, and 24 hours after extubation.

Results: Six hundred ninety‐eight patients completed the trial. The median duration of intubation prior to extubation was 6 days. Any laryngeal edema occurred in 22% (76/343) and 3% (11/355) of patients in the placebo and treatment groups, respectively (P < 0.0001). When edema was present, the severity and timing of the onset of edema did not differ between the 2 groups. Reintubation was reduced from 8% (26/343) in the placebo group to 4% (13/355) in the treatment groups (P = 0.02). When necessary, reintubation was deemed secondary to major edema in 54% (14/26) of the placebo group and 8% (1/13) of the treatment group, respectively. An intention‐to‐treat analysis did not alter the study findings. One patient in each group suffered a serious adverse event: respiratory failure and death 23 hours after extubation in the placebo group and septic shock and death 26 hours after extubation in the treatment group. Rates of hyperglycemia and infections were not reported.

Conclusion: The use of 20‐mg intravenous doses of methylprednisolone spaced 4 hours apart and starting 12 hours prior to planned extubation is associated with significant reductions in the rates of tracheal edema and reintubation.

Commentary: Postextubation laryngeal edema is common (2%22% incidence) and results in reintubation for 0.74.7% of extubations. This work shows that a simple pretreatment with intravenous steroids 12 hours before planned extubation can reduce the rate of postextubation edema 7‐fold, including a 2‐fold reduction in the reintubation rate. Prior trials using shorter periods of treatment (<6 hours) have not shown benefit, so achieving this study's results likely requires the full 12‐hour protocol.

Clinical Bottom Line: Intravenous methylprednisolone dosed 12 hours before and every 4 hours until planned extubation reduces the rate of reintubation due to tracheal edema.

This update reviews key clinical articles for hospitalists published over the past year. Selection criteria include high methodological quality, pertinence to hospital medicine, and likelihood that a change in practice is warranted. Table 1 summarizes practice changes.

Enoxaparin Dosing in Acute Coronary Syndromes

Allen La Pointe NM, Chen AY, Alexander KP, et al. Enoxaparin dosing and associated risk of in‐hospital bleeding and death in patients with non‐ST‐segment elevation acute coronary syndromes. Arch Intern Med. 2007;167:15391544.

Question: Among patients with non‐ST‐elevation acute coronary syndromes, how common and harmful is excess enoxaparin dosing?

Sponsors: Schering‐Plough Corp., Bristol‐Myers Squibb/Sanofi‐Aventis Pharmaceuticals Partnership, Millennium Pharmaceuticals, and the National Institutes of Health and National Institute on Aging.

Study Design: Observational study of prospective cohort data from the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA Guidelines (CRUSADE) National Quality Improvement Initiative.

Patients: A total of 10,687 patients receiving enoxaparin for non‐ST‐elevation acute coronary syndromes.

Setting: Three hundred thirty‐two US hospitals.

Outcomes: Rate of excess enoxaparin dose, defined as greater than 10 mg/day above the recommended dose of 1 mg/kg every 12 hours for creatinine clearance (CrCl) 30 mL/minute or 1 mg/kg every 24 hours for CrCl < 30 mL/minute; rates of in‐hospital major bleeding and death; and rate of lower than recommended enoxaparin dose.

Results: Excess enoxaparin dosing occurred in 18.7% of the cohort (2002/10,687). Of these, 57.8% (1157/2002) had CrCl < 30 mL/minute. Excess‐dose patients were more likely to be older and female and have a low body mass index (P < 0.001 for all comparisons). In‐hospital major bleeding (14.2% versus 7.3%, P< 0.001) and in‐hospital death (5.6% versus 2.4%, P < 0.001) were more common among excess‐dose patients. Enoxaparin underdosing occurred in 29.2% (3116/10 687) and was not associated with excess harm. Controlling for baseline characteristics, the authors found that the adjusted odds ratio for in‐hospital major bleeding in the excess‐dose cohort was 1.43 (1.181.75, P < 0.001) and the adjusted odds ratio for death was 1.35 (1.031.77, P = 0.03).

Conclusions: Excess enoxaparin dosing in non‐ST‐elevation acute coronary syndromes occurred in about 1 of every 5 patients treated in this prospective multihospital registry. Excess dosing was associated with substantially higher rates of major in‐hospital bleeding and death, with a number needed to harm of 78 for major bleeding and a number needed to harm of 167 for in‐hospital death. In comparison, the number needed to treat with another low‐molecular‐weight heparin (dalteparin) was 34 to prevent 1 death or myocardial infarction in the first 6 days, with a nonsignificant trend toward decreased mortality.1

Commentary: Providers likely underestimate the degree of renal impairment when looking solely at serum creatinine instead of estimates of CrCl. Excess dosing was more common among elderly, thin, and female patients. Clinicians must calculate the enoxaparin dose on the basis of careful estimates of CrCl to limit this risk. The Modification of Diet in Renal Disease (MDRD) equation is commonly used for this purpose.

Clinical Bottom Line: Enoxaparin excess dosing is common and harmful. Clinicians can mitigate this risk by more carefully estimating renal function when selecting the proper enoxaparin dose of 1 mg/kg twice daily for CrCl 30 mL/minute and 1 mg/kg once daily for CrCl < 30 mL/minute.

Venous Thromboembolism Prevention

Wein L, Wein S, Haas SJ, et al. Pharmacological venous thromboembolism prophylaxis in hospitalized medical patients. Arch Intern Med. 2007;167:14761486.

Question: What is the relative safety and efficacy of various pharmacological agents for preventing venous thromboembolism among hospitalized medical patients?

Sponsor: National Health and Medical Council of Australia.

Study Design: Meta‐analysis of 36 prospective randomized controlled trials involving about 48,000 patients.

Study Selection: Prospective randomized controlled trials enrolling at least 30 patients comparing 1 of 4 regimens: (1) unfractionated heparin (UFH) versus control, (2) low‐molecular‐weight heparin (LMWH) versus control, (3) LMWH versus UFH, or (4) Factor Xa inhibitor versus placebo. Trials of surgical, trauma, and critical care patients were excluded. Only 1 Factor Xa trial (fondaparinux) was located,2 and thus it was not eligible for meta‐analysis.

Outcomes: Pooled relative risks with 95% confidence intervals for deep venous thrombosis (DVT), pulmonary embolism (PE), mortality, and total bleeding. The authors also compared 2 UFH regimens: 5000 units twice daily versus 5000 units thrice daily.

Results: UFH (all doses, compared with control): The relative risk was 0.33 (95% CI 0.260.42) for DVT and 0.64 (95% CI 0.500.82) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 3.11 (95% CI 2.443.96, P = 0.001).

LMWH (compared with control): The relative risk was 0.56 (95% CI 0.450.70) for DVT and 0.37 (95% CI 0.210.64) for PE (P = 0.001 for both). Mortality was not different. The relative risk for major bleeding was 1.92 (95% CI 1.322.78, P = 0.001).

LMWH (compared with UFH, all doses): The relative risk for DVT was 0.68 (95% CI 0.520.88, P = 0.004), but the risk was not different for PE, mortality, or major bleeding.

UFH (5000 units twice daily, compared with control): The relative risk for DVT was 0.52 (95% CI 0.280.96, P = 0.04). When the random‐effects model was used, this difference became statistically nonsignificant (relative risk = 0.41, 95% CI 0.101.73, P = 0.23).

UFH (5000 units 3 times daily, compared with control): The relative risk for DVT was 0.27 (95% CI 0.200.36, P = 0.001). This difference remained when the random‐effects model was applied (relative risk = 0.28, 95% confidence interval = 0.210.38, P = 0.001).

Conclusions: Both UFH and LMWH reduce DVT and PE in hospitalized medical patients. Neither affects mortality. Both increase the risk of major bleeding. LMWH reduces the risk of DVT but not the risk of PE in comparison with UFH (all doses). When adjusted for random effects, UFH at a dose of 5000 units twice daily does not appear to be different than the control.

Commentary: This well‐conducted meta‐analysis demonstrates the efficacy of heparin, whether unfractionated or low‐molecular‐weight, in the prevention of venous thromboembolism. Of note, the UFH dose of 5000 units twice daily did not appear to be different than placebo. The UFH dose of 5000 units 3 times daily, by contrast, was effective in both the fixed‐effects and random‐effects models. Mortality was unaffected by any of the regimens studied. All regimens were associated with increased risks of major bleeding.

Clinical Bottom Line: Pharmacological prophylaxis with UFH 3 times daily or LMWH reduces the risk for venous thromboembolism. Twice daily UFH is not clearly different from placebo. Overall mortality was unaffected by any of the regimens for prophylaxis.

Contrast Nephropathy Prevention

Briguori C, Airoldi F, D'Andrea D, et al. Renal insufficiency following contrast media administration trial (REMEDIAL): a randomized comparison of 3 preventive strategies. Circulation. 2007;115:12111217.

Question: What is the efficacy of saline versus bicarbonate for the prevention of contrast mediainduced nephropathy?

Sponsor: Institutional funding (C. Briguori, personal communication, January 2008).

Study Design: Randomized trial.

Patients: Three hundred twenty‐six consecutive patients with serum creatinine 2.0 mg/dL and/or an estimated glomerular filtration rate < 40 mL/minute/1.73 m² undergoing elective coronary and/or peripheral angiography.

Setting: Two interventional cardiology laboratories in Italy.

Intervention: Patients were randomized to 1 of 3 preventive regimens: (1) intravenous saline (0.9%) given at a rate of 1 mL/kg of body weight/hour 12 hours prior to the procedure and continuing for 12 hours afterward (reduced to 0.5 mL/kg/hour for patients with a left ventricular ejection fraction < 40%) plus N‐acetylcysteine (NAC; 1200 mg orally twice daily) on the day before the procedure and the day of the procedure; (2) intravenous sodium bicarbonate (154 mEq/L in dextrose and water) given as an initial bolus of 3 mL/kg over 1 hour prior to the procedure and continuing at a rate of 1 mL/kg/hour for 6 hours more plus NAC as above; or (3) intravenous saline as above plus intravenous ascorbic acid (3 g) 2 hours prior to the procedure followed by 2 g on the night and morning after the procedure plus NAC as above.

Outcomes: Rate of contrast‐induced nephropathy (CIN), which was defined as an increase in serum creatinine 25% from the baseline value at 48 hours after the administration of contrast or the need for hemodialysis.

Follow‐Up: Forty‐eight hours.

Results: The baseline serum creatinine was about 2.0 mg/dL and did not differ among the 3 groups. The rate of CIN was 9.9% (11/111) in the saline plus NAC group, 1.9% (2/108) in the bicarbonate plus NAC group, and 10.3% (11/107) in the saline plus ascorbic acid plus NAC group. The bicarbonate plus NAC regimen was superior to saline plus NAC (P = 0.019). The absolute risk reduction for bicarbonate plus NAC versus saline plus NAC was 8% (a number needed to treat of 13 to prevent 1 case of CIN). The saline plus NAC and saline plus ascorbic acid plus NAC groups did not differ in outcome.

Conclusions: Sodium bicarbonate plus NAC is superior to saline plus NAC for the prevention of CIN among patients with baseline chronic kidney disease.

Commentary: This trial confirms the results of the initial study by Merten et al.3 showing the superiority of bicarbonate versus saline in the prevention of CIN. That trial, published in 2004, did not use NAC. Also in 2007, 3 other single‐center randomized trials of saline versus bicarbonate in the prevention of CIN were published.46 All concluded that bicarbonate is superior to saline. Whether NAC is effective for CIN prevention remains unclear.7 Given its low side‐effect profile, it is not unreasonable to continue using NAC until further data are available. At‐risk patients receiving intravenous contrast for other indications (eg, computed tomography) would likely show similar benefit. Although there are now 5 prospective blinded controlled trials showing the superiority of bicarbonate, a recently published large retrospective cohort found that the use of sodium bicarbonate was associated with increased incidence of CIN.8 The concordant results of all 5 prospective randomized trials of sodium bicarbonate, along with the risk for unmeasured confounding variables with retrospective cohort analysis, suggest that bicarbonate is superior to saline in the prevention of CIN.

Clinical Bottom Line: Clinicians should consider selecting intravenous bicarbonate rather than saline for the prevention of CIN.

Acute Decompensated Heart Failure Treatment

Gheorghiade M, Konstam MA, Burnett JC, et al. Short‐term clinical effects of tolvaptan, an oral vasopressin antagonist, in patients hospitalized for heart failure: the EVEREST clinical status trials. JAMA. 2007;297:13321343.

Question: What is the efficacy and safety of short‐term tolvaptan added to standard therapy in the treatment of acute decompensated heart failure?

Sponsor: Otsuka America, Inc.

Study Design: Two concurrent randomized, double‐blind, placebo‐controlled trials. Two trials (each with different sites) were conducted to fulfill regulatory requirements for establishing efficacy from at least 2 independent, adequately powered, and well‐controlled trials.

Patients: Two thousand forty‐eight adults (trial A) and 2085 adults (trial B) hospitalized with heart failure. Eligibility criteria included a history of chronic heart failure requiring treatment for at least 30 days prior to admission, an ejection fraction 40% at any point in the prior year, dyspnea at rest or with minimal exertion, and 2 or more signs of congestion (dyspnea, jugular vein distension, or peripheral edema). Selected exclusionary criteria included active myocardial ischemia, recent cardiac surgery, systolic blood pressure < 90 mm Hg, serum creatinine > 3.5 mg/dL, serum potassium > 5.5 mg/dL, or hemoglobin < 9 g/dL.

Setting: Three hundred fifty‐nine sites across North America, South America, and Europe. Trial A patients were assigned from 179 of these sites. Trial B patients were assigned from 180 of these sites.

Intervention: Tolvaptan, a vasopressin antagonist (30 mg orally daily), versus matching placebo, in addition to standard therapy. Treatment was started within 48 hours of admission and was continued through discharge for a minimum of 60 days.

Outcomes: Composite of global clinical status and body weight at day 7 or at discharge if earlier. Additional secondary endpoints were dyspnea (day 1) and peripheral edema (day 7).

Follow‐Up: Seven days.

Results: Tolvaptan improved the composite primary endpoint compared with placebo, and this was primarily related to greater overall net diuresis: 3.35 kg of diuresis at day 7 or discharge with tolvaptan versus 2.73 kg with placebo (trial A) and 3.77 kg of diuresis at day 7 or discharge with tolvaptan versus 2.79 kg with placebo (trial B; P < 0.001 for both trials). Net diuresis at day 1 was also greater with tolvaptan. More patients reported improved dyspnea at day 1 with tolvaptan: 76.74% versus 70.61% (trial A) and 72.06% versus 65.32% (trial B; P < 0.001 for both comparisons). Edema scores at day 7 favored tolvaptan in trial B (P = 0.02) but in not trial A (P = 0.07). Hypernatremia was more common with tolvaptan in trial A (1.4% versus 0%, P < 0.001) but not in trial B (0.5% versus 0%, P = 0.06). Tolvaptan‐treated patients had lower average furosemide doses than placebo‐treated patients. Patient‐assessed global clinical status at day 7, as measured by a visual analog scale, was no different.

Conclusions: Tolvaptan, added to standard care for acute heart failure, safely improved many but not all short‐term heart failure signs and symptoms.

Commentary: The accompanying Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) outcomes trial demonstrated that longer term use of tolvaptan for 60 days was not associated with changes in cardiovascular morbidity and mortality.9 Concerns have been raised about the safety of nesiritide10 and inotropes11 in the treatment of acute decompensated heart failure. With the completion of this 2‐part trial, we have a safe addition to the current armamentarium of treatments for acute decompensated heart failure. Clinicians should exercise caution in adding tolvaptan only to patients whose characteristics mirror those in this trial.

Clinical Bottom Line: Tolvaptan represents an effective and safe addition to therapies for acute decompensated heart failure.

Cardiovascular Risk Reduction

Dentali F, Douketis JD, Lim W, Crowther M. Combined aspirin‐oral anticoagulant therapy compared with oral anticoagulant therapy alone among patients at risk for cardiovascular disease: a meta‐analysis of randomized trials. Arch Intern Med. 2007;167:117124.

Question: For patients receiving oral anticoagulant therapy (OAC), does the addition of aspirin reduce major adverse cardiovascular endpoints?

Sponsor: Heart and Stroke Foundation of Canada.

Study Design: Meta‐analysis of 10 randomized controlled trials.

Study Selection: From MEDLINE (to June 2005), EMBASE (to June 2005), and Cochrane (to 2005, issue 2) reviews, including manual reference list reviews, 10 studies were identified that satisfied 4 criteria: (1) a randomized controlled trial in patients requiring OAC therapy, (2) a comparison of combined aspirinOAC therapy with OAC alone (the same target international normalized ratio in both arms), (3) follow‐up of at least 3 months, and (4) at least 1 prespecified outcome that was objectively documented. The 10 trials meeting these criteria studied 4180 patients. The target international normalized ratio varied across the trials on the basis of the population studied. The aspirin dose was at least 75 mg/day in all studies.

Outcomes: Arterial thromboembolism, all‐cause mortality, and major bleeding. Secondary outcomes included fatal arterial thromboembolism and fatal major bleeding.

Results: Arterial thromboembolism was lower with combined aspirinOAC therapy (6.3%) versus OAC therapy alone (8.8%; absolute risk reduction = 2.5%, number needed to treat = 40, P < 0.001). In subgroup analysis, this difference was found only among patients with mechanical heart valves (odds ratio = 0.27, 95% CI 0.150.49). There was no benefit among patients with atrial fibrillation (odds ratio = 0.99, 95% CI 0.472.07) or coronary artery disease (odds ratio = 0.69, 95% CI 0.351.3). Mortality was no different. Major bleeding was more common with combined therapy (3.8%) versus OAC therapy alone (2.8%; absolute risk reduction = 1.0%, number needed to harm = 100, P = 0.05). Secondary outcomes were not different.

Conclusions: Combined aspirinOAC therapy does not protect against future arterial thromboembolism in comparison with OAC therapy alone, except among patients with mechanical heart valves. Combined therapy, however, is associated with higher rates of major bleeding.

Commentary: These findings question the current practice of combining OAC with aspirin in patients with separate indications for each. Looking in more detail at the analyzed trials, the researchers found that there were relatively few patients with proven coronary artery disease. There may have been insufficient power to show a benefit for combined therapy among these patients. Patients with mechanical heart valves, however, clearly showed benefit. A recently published retrospective study of more than 4000 patients also concluded that the hemorrhagic risk of combined aspirinOAC therapy versus OAC therapy alone appeared to outweigh the benefit.12

Clinical Bottom Line: Except among patients with mechanical heart valves, combined aspirinOAC increases bleeding risk without proven benefit. Until further data are available, clinicians should individualize antithrombotic therapy on the basis of a careful assessment of risk and benefit.

Cardioembolic Stroke Treatment

Paciaroni M, Agnelli G, Micheli S, Caso V. Efficacy and safety of anticoagulant treatment in acute cardioembolic stroke: a meta‐analysis of randomized controlled trials. Stroke. 2007;38:423430.

Question: What are the safety and efficacy of anticoagulation in the treatment of acute cardioembolic stroke?

Sponsor: None.

Study Design: Meta‐analysis of 7 randomized controlled trials.

Study Selection: Trials randomizing patients within 48 hours from stroke onset with objectively diagnosed stroke of presumed cardioembolic origin that compared full‐dose anticoagulants (unfractionated heparin, low‐molecular‐weight heparin, and heparinoid) to other treatments (aspirin or placebo) for initial therapy and used objective methods to assess study outcomes.

Outcomes: A composite of death or disability at final follow‐up (at least 3 months), all new strokes (ischemic and hemorrhagic) at 14 days, and pulmonary embolism. The safety outcome was symptomatic intracranial bleeding.

Results: The odds ratio (95% CI) for death or disability with anticoagulation versus aspirin or placebo was 1.01 (95% CI 0.821.24); the odds ratio for all new strokes with anticoagulation versus aspirin or placebo was 1.18 (95% CI 0.741.88). The odds ratio for pulmonary embolism with anticoagulation versus aspirin was 0.94 (95% CI 0.442.00). None of these were statistically significant. However, the odds ratio for symptomatic intracranial hemorrhage with anticoagulation versus aspirin or placebo was 2.89 (95% CI 1.197.01, P = 0.02). The absolute increase in symptomatic intracranial bleeding with anticoagulation was 1.8% (number needed to harm = 55). Of the 7 trials analyzed, 1 trial did show a reduction in overall death or disability with anticoagulation, in which therapy was started within 3 hours of symptom onset (odds ratio = 0.49, 95% CI 0.260.93). This trial was small, and subgroup analysis in the other, larger trials failed to confirm this finding.

Conclusion: Anticoagulation for acute stroke of suspected cardioembolic origin does not improve outcomes but is associated with higher rates of symptomatic intracranial hemorrhage.

Commentary: Long‐term anticoagulation with sodium warfarin clearly lowers cardioembolic stroke risk for patients with chronic atrial fibrillation. This meta‐analysis demonstrates that acute anticoagulation does not reduce the composite endpoint of death or disability, recurrent stroke, or pulmonary embolism. The risk of symptomatic intracranial hemorrhage is substantially increased and argues against the use of anticoagulants during the acute phase of suspected cardioembolic stroke.

Clinical Bottom Line: Anticoagulation is harmful and does not reduce death or disability in the acute phase of suspected cardioembolic stroke.

Clostridium Difficile Associated Diarrhea

Zar FA, Bakkanagari SR, Moorthi KM, Davis MB. A comparison of vancomycin and metronidazole for the treatment of Clostridium difficile‐associated diarrhea, stratified by disease severity. Clin Infect Dis. 2007;45:302307.

Question: What is the best first‐line treatment for Clostridium difficileassociated diarrhea (CDAD)?

Sponsor: None.

Study Design: Randomized, double‐blind, placebo‐controlled trial.

Patients: One hundred fifty patients with 3 or more nonformed stools in 24 hours with a positive stool C. difficile toxin A test or the presence of pseudomembranous colitis on endoscopy.

Setting: A 200‐bed community teaching hospital affiliated with an academic medical center.

Intervention: Metronidazole (250 mg 4 times daily) plus vancomycin liquid placebo versus metronidazole placebo plus vancomycin liquid (125 mg 4 times daily), both for 10 days.

Outcomes: The primary outcomes were cure (resolution of diarrhea by day 6 of treatment and a negative stool toxin at both 6 and 10 days post‐treatment), treatment failure (persistent diarrhea and/or an inability to clear the toxin at 6 days, the need for colectomy, or death after 5 days of treatment), and relapse (recurrence of toxin‐positive CDAD by day 21 after the initial cure). Disease was categorized as mild (<2 points) or severe ( 2 points), with 1 point each for age > 60 years, temperature > 38.3C, albumin < 2.5 mg/dL, and a peripheral white blood count > 15,000 cells/mm³ within 48 hours of enrollment. Two points were allotted for endoscopic findings of pseudomembranous colitis.

Follow‐Up: Patients were monitored for 21 days for resolution of diarrhea (2 formed stools in 24 hours). Stool toxin was measured at days 6 and 10 of treatment and at day 21 if diarrhea was still present.

Results: One hundred fifty patients (81 patients with mild disease and 69 patients with severe disease) finished the trial, with no significant differences in patients categorized into the 2 treatment arms. Overall, 84% (66/79) of patients receiving metronidazole were cured versus 97% (69/71) of patients receiving vancomycin (P = 0.006). In patients with mild disease, 90% (37/41) and 98% (39/40) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.36). In patients with severe disease, 76% (29/38) and 97% (30/31) were cured in the metronidazole‐treated and vancomycin‐treated groups, respectively (P = 0.02). After the initial cure, relapse occurred in 7% (5/76), 15% (9/59), 14% (9/66), and 7% (5/69) of patients with mild disease, severe disease (P = 0.15 for mild versus severe), metronidazole treatment, and vancomycin treatment (P = 0.27 between treatments), respectively. In patients with severe CDAD, low albumin, intensive care, and presence of pseudomembranous colitis were associated with metronidazole treatment failure.

Conclusion: Metronidazole is equally effective as vancomycin in treating mild CDAD; however, vancomycin appears superior to metronidazole in treating patients with severe CDAD.

Commentary: Two prior studies evaluating metronidazole and vancomycin for CDAD revealed no significant difference between the 2 therapies.13, 14 However, these studies had serious methodological flaws, including a lack of blinding and too little power to show a difference. This randomized, double‐blind, placebo‐controlled trial provides convincing evidence that oral vancomycin is superior to metronidazole in patients with severe CDAD. This is an especially important finding as the recently described hypervirulent epidemic strain of C. difficile becomes more prevalent.

A single‐center retrospective study of 102 veterans with metronidazole‐treated CDAD showed analogous findings with a slightly different scoring system.15 In 94% of metronidazole responders, the score was 2 or less. In 67% of true failures, the score was greater than 2. Taken together, these studies suggest that higher scores predict metronidazole failure.

Clinical Bottom Line: Vancomycin appears to be more effective than metronidazole in treating more severe forms of CDAD.

Consultative Medicine: Orthopedics

Lyles KW, Coln‐Emeric CS, Magaziner JS, et al. Zoledronic acid and clinical fractures and mortality after hip fracture. N Engl J Med. 2007;357:17991809.

Question: Does an annual dose of zoledronic acid reduce the rate of subsequent fractures and mortality in patients with a recent hip fracture?

Sponsor: Novartis.

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: A total of 2127 men and women 50 years old or older with a surgically repaired low‐impact hip fracture (eg, fall from a standing height) within 90 days of study entry who were unwilling or unable to take an oral bisphosphonate.

Setting: International and multicenter.

Intervention: A single 5‐mg intravenous dose of zoledronic acid within 90 days of a hip fracture repair versus an intravenous placebo, given annually. All patients with documented vitamin D deficiency or no documentation of a serum 25‐hydroxyvitamin D level received a loading dose of vitamin D₃ or D₂ 14 days prior to the first infusion. All patients received oral calcium and vitamin D daily after the first infusion.

Outcomes: The primary outcome was a new clinical fracture excluding facial, digital, or abnormal bone (eg, bone with metastases) fractures. Secondary outcomes included changes in the bone mineral density in the nonfractured hip, the number of new vertebral, nonvertebral, and hip fractures, and predetermined safety outcomes.

Follow‐Up: Quarterly phone calls and annual clinic visits for up to 5 years.

Results: The trial was stopped early after prespecified efficacy objectives were met. At an average follow‐up of 1.9 years, 8.6% of subjects receiving zoledronic acid and 13.9% of those receiving placebo suffered subsequent fractures (P = 0.001). Statistically significant improvements in bone mineral density were seen at both the total hip and femoral neck sites in the zoledronic acid group versus the placebo group. Approximately 80% of patients experienced an adverse event in each group, with statistically significantly more pyrexia, myalgias, and bone pain in the zoledronic acid cohort and higher mortality in the placebo group, that is, 9.6% versus 13.3% (hazard ratio = 0.72, 95% CI 0.560.72, P = 0.01).

Conclusion: Annual treatment with 5 mg of intravenous zoledronic acid reduces clinical fractures and mortality when it is dosed within 90 days of a hip fracture repair.

Commentary: Patients who suffer a hip fracture are at high risk for successive fractures, with a considerable morbid and financial burden on the patient and the healthcare system. Additionally, as many as 1 in 4 of these patients will die in the subsequent year. Poor adherence to oral bisphosphonates and prescriber nonadherence to fracture guidelines are common sources of noncompliance and have been associated with increased fracture burden. The findings that an annual infusion can achieve reductions in the fracture rate and mortality are notable and offer options for patients who otherwise could not comply with therapy because of side effects or an inability to take a more frequently dosed medication.

Clinical Bottom Line: An annual dose of zoledronic acid reduces the rate of subsequent fractures and death in patients with a recent hip fracture.

Critical Care Medicine

Francois B, Bellissant E, Gissot V, et al. 12‐h pretreatment with methylprednisolone versus placebo for prevention of postextubation laryngeal edema: a randomized double blind trial. Lancet. 2007;369:10831089.

Question: Do pre‐extubation steroids prior to planned extubation prevent postextubation laryngeal edema?

Sponsor: Institutional funding (P. Vignon, personal communication, March 2008).

Study Design: Placebo‐controlled, double‐blinded, randomized controlled trial.

Patients: Seven hundred sixty‐one adult patients with at least 36 hours of mechanical ventilation and planned extubation.

Setting: Fifteen intensive care units in France.

Intervention: Intravenous methylprednisolone (20 mg) starting 12 hours before extubation and continuing every 4 hours until extubation, including the time of extubation (total dose = 80 mg), or a placebo identical in appearance and delivery.

Outcomes: The primary outcome was the development of minor (inspiratory stridor associated with respiratory distress requiring intervention) or major (reintubation secondary to laryngoscopically visualized upper airway obstruction) laryngeal edema within 24 hours of extubation.

Follow‐Up: Clinical assessments were performed 10 minutes and 1, 1.5, 3, 6, 12, and 24 hours after extubation.

Results: Six hundred ninety‐eight patients completed the trial. The median duration of intubation prior to extubation was 6 days. Any laryngeal edema occurred in 22% (76/343) and 3% (11/355) of patients in the placebo and treatment groups, respectively (P < 0.0001). When edema was present, the severity and timing of the onset of edema did not differ between the 2 groups. Reintubation was reduced from 8% (26/343) in the placebo group to 4% (13/355) in the treatment groups (P = 0.02). When necessary, reintubation was deemed secondary to major edema in 54% (14/26) of the placebo group and 8% (1/13) of the treatment group, respectively. An intention‐to‐treat analysis did not alter the study findings. One patient in each group suffered a serious adverse event: respiratory failure and death 23 hours after extubation in the placebo group and septic shock and death 26 hours after extubation in the treatment group. Rates of hyperglycemia and infections were not reported.

Conclusion: The use of 20‐mg intravenous doses of methylprednisolone spaced 4 hours apart and starting 12 hours prior to planned extubation is associated with significant reductions in the rates of tracheal edema and reintubation.

Commentary: Postextubation laryngeal edema is common (2%22% incidence) and results in reintubation for 0.74.7% of extubations. This work shows that a simple pretreatment with intravenous steroids 12 hours before planned extubation can reduce the rate of postextubation edema 7‐fold, including a 2‐fold reduction in the reintubation rate. Prior trials using shorter periods of treatment (<6 hours) have not shown benefit, so achieving this study's results likely requires the full 12‐hour protocol.

Clinical Bottom Line: Intravenous methylprednisolone dosed 12 hours before and every 4 hours until planned extubation reduces the rate of reintubation due to tracheal edema.

The prevalence of intimate partner violence (IPV; defined as mental and/or physical violence directed from 1 person in an intimate relationship to the other) varies widely, depending on the population sampled and method of data collection. In the United States, IPV against women, occurring within the year prior to contact with a healthcare professional, ranges from 2% to 15% in surveys done by telephone, in primary care clinics, or in face‐to‐face home interviews19 and from 10% to 30% in surveys of patients visiting urgent care or emergency departments.1012 The prevalence of IPV occurring at any time during the life of the patient ranges from 18% in the aforementioned settings to as high as 88% in women applying for welfare.1, 2, 4, 5, 10, 1214

Although reports indicate that victims of IPV are more likely to be hospitalized,1517 the only study assessing the prevalence of IPV in hospitalized patients included women on medical, surgical, and obstetrical services and reported 1‐year and lifetime prevalences of only 5% and 23%, respectively.18

We hypothesized that the prevalence of IPV in hospitalized patients would be at least as high as that reported from emergency departments and sought to measure the 1‐year and lifetime prevalences of IPV in women admitted to a general internal medicine service. In addition, because studies done in various outpatient settings have reported that victims of IPV have a variety of somatic complaints and an increased prevalence of chronic and functional illnesses,1923 we also sought to determine whether women with a history of IPV and women without a history of IPV had different numbers or types of positive responses to questions asked on the review of systems.

PATIENTS AND METHODS

This study was approved by the Colorado Multiple Institution Review Board, and informed consent was obtained from all participants.

Women between the ages of 18 and 60 who were admitted to the internal medicine floor service of Denver Health Medical Center (a university‐affiliated public safety‐net hospital) between January 1 and February 28, 2004 and between October 1 and October 30, 2004 were approached to participate. These dates were selected on the basis of the availability of our interviewers. Patients older than 60 were excluded to avoid overlap between IPV and the problem of elder abuse. Women were excluded if they were unable to give informed consent, were pregnant, were incarcerated, were on contact precautions, or spoke a language other than English or Spanish. Although IPV is common in pregnant women and may occur in women who are incarcerated, these are considered vulnerable populations with respect to obtaining approval from internal review boards.

The questionnaire consisted of 23 review‐of‐systems questions,24 4 questions adapted from a previously validated screen for IPV11 (Table 1), and 1 question about attempts to seek help (Table 1). Women were considered to have experienced IPV if they gave positive responses to any of the 4 questions targeting IPV. According to patient preference, the combined questionnaire was either read and filled out by each subject independently or was read to her by a female interviewer who then recorded the subject's verbal responses. All interviewers were women with a shared common concern about, and interest in, IPV. Although none had advanced training in psychology, social work, or other formal discipline that involved interviewing skills, all interviews were scripted so that interactions with subjects and completion of the questionnaires would be uniform. Responses indicating sometimes were considered to be positive. Responses that were not answered, left blank, or marked as not applicable were considered to be negative.

Each patient's medical record was reviewed to determine her age, race, number of previous hospital admissions, visits to the emergency department and walk‐in clinic, visits to primary care and subspecialty physicians, and whether the patient had been screened for IPV as recorded on the admission history and physical template. Admission diagnosis was obtained from the history and physical template, and the discharge diagnosis was obtained from the discharge paperwork. Functional diagnoses were considered to be symptoms (eg, shortness of breath) or problems (eg, constipation) that could not clearly be linked to a specific disease process. All participants were offered a card containing a list of resources for victims of IPV.

Data were analyzed with SAS 8.1 (SAS Institute, Cary, NC) and SPSS 11.5 (SPSS, Chicago, IL). The Student t test was used to compare continuous variables. Data are reported as means standard deviation. Chi‐square analysis was used to test associations between race, primary language, level of education, insurance status, admitting diagnosis, discharge diagnosis, number of previous hospital admissions, visit type, and the presence of IPV. For these, P < 0.05 was considered to be significant. The association of positive review‐of‐systems responses with the presence of IPV was also tested by chi‐square analysis, but P < 0.002 was considered to be significant on the basis of a Bonferroni adjustment for multiple comparisons. A receiver operating characteristic curve was used to assess the relationship between the number of positive responses to the questions included in the review of systems and a history of IPV. The odds ratio and confidence intervals were calculated to test the association between the number of positive responses to the review‐of‐systems questions and a lifetime history of IPV.

RESULTS

Throughout the dates of the study, 245 women were admitted to the internal medicine service, and 106 were excluded (Figure 1). Of the 139 eligible women, 78 were available to the interviewers and asked to participate, and 72 (92%) agreed. IPV occurring within the year prior to the interview or at any point in the patient's lifetime was reported by 16 (22%) and 44 (61%) subjects, respectively. No significant differences were seen in women who did or did not experience IPV at anytime in their life with respect to age, race, insurance status, education, number of scheduled outpatient, urgent, or emergent visits, or admission or discharge diagnosis even when the diagnoses were grouped into a functional category (although at best our study was powered to detect only >35% differences in prevalences; Tables 2 and 3). Of women reporting a lifetime history of IPV, 26 of 44 (59%) had previously sought help, and 9 of those 26 (35%) said that they sought help from a physician.

Figure 1

Women with a 1‐year history of IPV and women without a 1‐year history of IPV had 11.4 4.7 and 7.7 5.4 positive responses to the review of systems (P < 0.01), respectively. Women with a lifetime history of IPV and women without a lifetime history of IPV had 10.9 4.4 and 7.7 5.4 positive responses (P < 0.01), respectively. The receiver operating characteristic curve of the number of positive responses versus a lifetime history of IPV is presented in Figure 2. Subjects with 10 or more positive responses were 4.8 times more likely to report a lifetime history of IPV than subjects with 9 or fewer positive responses (confidence interval = 1.614.2, P = 0.003). The c‐statistic indicating the ability of the review of systems to properly classify cases when there were 10 or more positive responses was 0.692.

Figure 2

No differences were observed in the responses to the individual review of systems questions in women who did or did not have a lifetime history of IPV, with the exception that those with a positive history more commonly complained of difficulty sleeping and numbness and tingling in their hands or feet (although at best our study was sufficiently powered to detect only >20% differences in prevalences; Table 4). Although the sensitivity of having problems sleeping or experiencing numbness or tingling in patients with IPV was high, the specificity and positive and negative predictive values were not (Table 5).

The admission history forms filled out by first‐year admitting residents showed that only 18 (25%) of the women were screened for IPV, even though the history and physical examination template used at Denver Health Medical Center includes a prompt in the social history section pertaining to a history of violence as a reminder.

DISCUSSION

The important findings of this study were that women admitted to the internal medicine service of a university‐affiliated public safety‐net hospital had a high prevalence of IPV (22% and 61% 1‐year and lifetime prevalences, respectively), that most women with a history of IPV had previously sought help for the problem, many from physicians, that women were more likely to have a history of IPV if they had >10 positive responses to questions asked in a routine review of systems (particularly problems sleeping and experiencing numbness or tingling in their extremities), and that routine screening for IPV was uncommon at the time of admission.

These conclusions should be interpreted with respect to a number of limitations in our study. First, although our study was designed to be a consecutive series, the interviewers did not have sufficient time to meet with and interview every woman admitted before they were discharged. This occurred in part because the interviewers were available only for a portion of each day, some patients were discharged within 24 hours of admission, and many were out of their rooms for ancillary testing. Within the interviewers' time constraints, however, all hospitalized women meeting entry criteria who were available were approached. Our data could, however, overrepresent the prevalence of IPV if hospitalized women with a history of IPV had longer hospital stays than those who did not or if those experiencing IPV were out of their rooms less frequently (eg, for diagnostic tests). On the other hand, our data could underrepresent the true prevalence of IPV if patients with a history of IPV had shorter hospital stays or if they received more ancillary testing that caused them to be out of their rooms more frequently. Second, none of our interviewers had specific training in interviewing techniques. Accordingly, our data could have underestimated the true prevalence of IPV if interviewers with advanced training in probing sensitive topics had more success in eliciting positive responses. Third, the relationship between a history of IPV and multiple positive responses to the review of systems may be confounded if some of these patients also had a history of adverse childhood experiences or other experiences resulting in posttraumatic stress disorder as these patients also have an increased prevalence of chronic and functional disorders.2527 Finally, as our numbers were small, we were not powered to detect clinically important differences in demographics or specific positive answers on the review of systems.

To the best of our knowledge, the only study presenting IPV prevalence data in patients hospitalized for other than psychiatric problems was performed by McKenzie and colleagues18 in 1997. In their group of 130 patients (61 on internal medicine, 59 on surgery, 7 on obstetrics, and 3 on psychiatry), the 1‐year and lifetime prevalences of IPV were only 5% and 26%, respectively. McKenzie and colleagues used only 1 question to screen for IPV, but that single question incorporated 2 of the 4 questions used in our survey. Forty‐three of our 44 patients (98%) with a history of IPV were discovered on the basis of these 2 questions. The hospitals in which the 2 studies were done were similar, as were the ages and levels of education of the 2 populations studied and the percentage of eligible patients who agreed to participate. The patients in the 2 studies were different with respect to race, language mix, and the percentage who were insured, but neither study found differences in the prevalence of IPV as a function of race or insurance (although others have found an association of IPV with being uninsured1, 3, 4, 12, 23). Our study was conducted in women admitted exclusively to an internal medicine service, whereas nearly half of the patients studied by McKenzie and colleagues were admitted to surgical, gynecologic, or psychiatric services. Although McKenzie and colleagues found no difference in the prevalence of IPV as a function of admitting service, others have suggested that the prevalence of IPV is higher in patients admitted for trauma or psychiatric problems.1517, 28 The percentage of patients who self‐administered the questionnaires was 57% in our study and 77% in the study by McKenzie and colleagues. Neither study, however, found a difference in the percentage of IPV in patients who self‐administered the survey versus those who were interviewed. Women may have become more comfortable discussing this issue in the 10‐year interval between these 2 studies, or the prevalence of IPV may have increased. The only other study of IPV in hospitalized patients of which we are aware reported a 90% 1‐year prevalence in suicidal women admitted to a psychiatric service.28

Several studies have reported that victims of IPV have multiple somatic complaints and an increased prevalence of chronic and functional illnesses.1923 We confirmed that women experiencing IPV have more positive responses to questions posed in a review of systems, but the low specificity and positive and negative predictive values of the responses make this association of little clinical utility.

For only 18 of the 72 patients (25%) in our study was there evidence that they were screened for a history of IPV by the admitting resident. If more women were screened without a response being recorded, or if women were screened only for a current history of violence, our data may not accurately reflect the true rate at which screening occurred; however, the rate of screening that we observed is consistent with a number of other studies.12, 22, 2931 Fourteen of 18 patients who were screened for IPV by the resident gave negative responses. Ten of these, however, gave positive responses to our interviewers. Accordingly, the sensitivity, specificity, and positive and negative predictive values of the information recorded by the admitting resident were 40%, 100%, 100%, and 57%, respectively (assuming that the responses given to the IPV survey represent the gold standard), and this confirms that routine screening underestimates the prevalence of this problem. Accordingly, we identified 2 problems pertaining to screening for IPV: (1) it is not routinely done at the time of hospital admission, and (2) responses reported during routine screening are frequently incorrect. A number of barriers to routine screening have been previously identified, as have interventions designed to increase screening.32 Providing specific screening questions increases the identification of victims of IPV, but simply educating healthcare providers does not.32 Our history and physical templates have a prompt for violence victim to facilitate the screening, but as a result of this study, we are changing our prompting question and indicating what should be done if the response is positive.

The US Preventive Services Task Force and the Canadian Task Force on Preventive Health Care both concluded that there was insufficient evidence to recommend for or against routine screening for IPV.3335 Their rationale was that trials assessing the effectiveness of screening have not been published, that studies designed to assess the effectiveness of any resulting intervention are few in number, focused on pregnant women, and limited by problems in study design, that no studies have determined the accuracy of the screening tools, and that none have addressed the potential harm of screening.3335 The US Preventive Services Task Force did recommend screening if providers were concerned about IPV.34 Our data would suggest that there is little in the admission history that distinguishes women who might be victims of IPV from those who might not. Guidelines published by the American Medical Association, the American Academy of Family Physicians, and the American College of Obstetricians and Gynecologists promote routine screening of all patients.3638 Janssen and colleagues39 support the importance of screening on the basis that IPV is associated with numerous physical and mental health problems (eg, arthritis, migraines and other types of headaches, vaginal bleeding, ulcers, spastic colon, chronic pain, substance abuse, depression, and suicide ideation) and that establishing the link between these conditions and IPV could be important with respect to developing appropriate diagnostic and therapeutic approaches to patients' complaints. Screening also allows physicians to become more knowledgeable about their patients' lives, facilitating their ability to provide a supportive relationship that, in turn, increases women's likelihood of using an intervention method.39 We did not confirm an increased prevalence of any of the complaints noted by Janssen and colleagues in the women experiencing a history of IPV, but we did find an increased prevalence of insomnia and extremity numbness in women admitting to IPV as well as an overall increase in the number of positive responses to the review of systems. Screening identifies women who should receive information about reporting IPV, obtaining available assistance, planning for personal safety, and formal counseling as these have all been shown to reduce the severity of IPV and to improve the quality of life in rather large, randomized controlled trials.4043

As previously observed by others,13, 22, 29, 4446 the large majority of women that we approached welcomed screening for IPV. Over half of those with a history of IPV had previously sought help for the problem, over one‐third of these sought help from physicians, and most took the resource card that we offered, regardless of whether they did or did not have a history of IPV (this suggests either that our data may actually underestimate the true prevalence of IPV or that patients taking the information knew of others experiencing this problem). Accordingly, regardless of whether physicians believe that routine screening is warranted, patients see physicians and other healthcare workers as a resource for this problem.

We have confirmed that a history of IPV is very common in women admitted to an internal medicine service of a university‐affiliated public hospital and that female victims of IPV have more positive responses on the review of systems (particularly difficulty sleeping and extremity numbness or tingling) than those who have not. Although we initially hypothesized that finding numerous somatic complaints might serve as a marker for IPV, thereby identifying patients for whom more careful screening should occur, finding such a high prevalence of IPV argues that screening should be a routine part of the history for all women admitted to internal medicine inpatient services.

The prevalence of intimate partner violence (IPV; defined as mental and/or physical violence directed from 1 person in an intimate relationship to the other) varies widely, depending on the population sampled and method of data collection. In the United States, IPV against women, occurring within the year prior to contact with a healthcare professional, ranges from 2% to 15% in surveys done by telephone, in primary care clinics, or in face‐to‐face home interviews19 and from 10% to 30% in surveys of patients visiting urgent care or emergency departments.1012 The prevalence of IPV occurring at any time during the life of the patient ranges from 18% in the aforementioned settings to as high as 88% in women applying for welfare.1, 2, 4, 5, 10, 1214

Although reports indicate that victims of IPV are more likely to be hospitalized,1517 the only study assessing the prevalence of IPV in hospitalized patients included women on medical, surgical, and obstetrical services and reported 1‐year and lifetime prevalences of only 5% and 23%, respectively.18

We hypothesized that the prevalence of IPV in hospitalized patients would be at least as high as that reported from emergency departments and sought to measure the 1‐year and lifetime prevalences of IPV in women admitted to a general internal medicine service. In addition, because studies done in various outpatient settings have reported that victims of IPV have a variety of somatic complaints and an increased prevalence of chronic and functional illnesses,1923 we also sought to determine whether women with a history of IPV and women without a history of IPV had different numbers or types of positive responses to questions asked on the review of systems.

PATIENTS AND METHODS

This study was approved by the Colorado Multiple Institution Review Board, and informed consent was obtained from all participants.

Women between the ages of 18 and 60 who were admitted to the internal medicine floor service of Denver Health Medical Center (a university‐affiliated public safety‐net hospital) between January 1 and February 28, 2004 and between October 1 and October 30, 2004 were approached to participate. These dates were selected on the basis of the availability of our interviewers. Patients older than 60 were excluded to avoid overlap between IPV and the problem of elder abuse. Women were excluded if they were unable to give informed consent, were pregnant, were incarcerated, were on contact precautions, or spoke a language other than English or Spanish. Although IPV is common in pregnant women and may occur in women who are incarcerated, these are considered vulnerable populations with respect to obtaining approval from internal review boards.

The questionnaire consisted of 23 review‐of‐systems questions,24 4 questions adapted from a previously validated screen for IPV11 (Table 1), and 1 question about attempts to seek help (Table 1). Women were considered to have experienced IPV if they gave positive responses to any of the 4 questions targeting IPV. According to patient preference, the combined questionnaire was either read and filled out by each subject independently or was read to her by a female interviewer who then recorded the subject's verbal responses. All interviewers were women with a shared common concern about, and interest in, IPV. Although none had advanced training in psychology, social work, or other formal discipline that involved interviewing skills, all interviews were scripted so that interactions with subjects and completion of the questionnaires would be uniform. Responses indicating sometimes were considered to be positive. Responses that were not answered, left blank, or marked as not applicable were considered to be negative.

Each patient's medical record was reviewed to determine her age, race, number of previous hospital admissions, visits to the emergency department and walk‐in clinic, visits to primary care and subspecialty physicians, and whether the patient had been screened for IPV as recorded on the admission history and physical template. Admission diagnosis was obtained from the history and physical template, and the discharge diagnosis was obtained from the discharge paperwork. Functional diagnoses were considered to be symptoms (eg, shortness of breath) or problems (eg, constipation) that could not clearly be linked to a specific disease process. All participants were offered a card containing a list of resources for victims of IPV.

Data were analyzed with SAS 8.1 (SAS Institute, Cary, NC) and SPSS 11.5 (SPSS, Chicago, IL). The Student t test was used to compare continuous variables. Data are reported as means standard deviation. Chi‐square analysis was used to test associations between race, primary language, level of education, insurance status, admitting diagnosis, discharge diagnosis, number of previous hospital admissions, visit type, and the presence of IPV. For these, P < 0.05 was considered to be significant. The association of positive review‐of‐systems responses with the presence of IPV was also tested by chi‐square analysis, but P < 0.002 was considered to be significant on the basis of a Bonferroni adjustment for multiple comparisons. A receiver operating characteristic curve was used to assess the relationship between the number of positive responses to the questions included in the review of systems and a history of IPV. The odds ratio and confidence intervals were calculated to test the association between the number of positive responses to the review‐of‐systems questions and a lifetime history of IPV.

RESULTS

Throughout the dates of the study, 245 women were admitted to the internal medicine service, and 106 were excluded (Figure 1). Of the 139 eligible women, 78 were available to the interviewers and asked to participate, and 72 (92%) agreed. IPV occurring within the year prior to the interview or at any point in the patient's lifetime was reported by 16 (22%) and 44 (61%) subjects, respectively. No significant differences were seen in women who did or did not experience IPV at anytime in their life with respect to age, race, insurance status, education, number of scheduled outpatient, urgent, or emergent visits, or admission or discharge diagnosis even when the diagnoses were grouped into a functional category (although at best our study was powered to detect only >35% differences in prevalences; Tables 2 and 3). Of women reporting a lifetime history of IPV, 26 of 44 (59%) had previously sought help, and 9 of those 26 (35%) said that they sought help from a physician.

Figure 1

Women with a 1‐year history of IPV and women without a 1‐year history of IPV had 11.4 4.7 and 7.7 5.4 positive responses to the review of systems (P < 0.01), respectively. Women with a lifetime history of IPV and women without a lifetime history of IPV had 10.9 4.4 and 7.7 5.4 positive responses (P < 0.01), respectively. The receiver operating characteristic curve of the number of positive responses versus a lifetime history of IPV is presented in Figure 2. Subjects with 10 or more positive responses were 4.8 times more likely to report a lifetime history of IPV than subjects with 9 or fewer positive responses (confidence interval = 1.614.2, P = 0.003). The c‐statistic indicating the ability of the review of systems to properly classify cases when there were 10 or more positive responses was 0.692.

Figure 2

No differences were observed in the responses to the individual review of systems questions in women who did or did not have a lifetime history of IPV, with the exception that those with a positive history more commonly complained of difficulty sleeping and numbness and tingling in their hands or feet (although at best our study was sufficiently powered to detect only >20% differences in prevalences; Table 4). Although the sensitivity of having problems sleeping or experiencing numbness or tingling in patients with IPV was high, the specificity and positive and negative predictive values were not (Table 5).

The admission history forms filled out by first‐year admitting residents showed that only 18 (25%) of the women were screened for IPV, even though the history and physical examination template used at Denver Health Medical Center includes a prompt in the social history section pertaining to a history of violence as a reminder.

DISCUSSION

The important findings of this study were that women admitted to the internal medicine service of a university‐affiliated public safety‐net hospital had a high prevalence of IPV (22% and 61% 1‐year and lifetime prevalences, respectively), that most women with a history of IPV had previously sought help for the problem, many from physicians, that women were more likely to have a history of IPV if they had >10 positive responses to questions asked in a routine review of systems (particularly problems sleeping and experiencing numbness or tingling in their extremities), and that routine screening for IPV was uncommon at the time of admission.

These conclusions should be interpreted with respect to a number of limitations in our study. First, although our study was designed to be a consecutive series, the interviewers did not have sufficient time to meet with and interview every woman admitted before they were discharged. This occurred in part because the interviewers were available only for a portion of each day, some patients were discharged within 24 hours of admission, and many were out of their rooms for ancillary testing. Within the interviewers' time constraints, however, all hospitalized women meeting entry criteria who were available were approached. Our data could, however, overrepresent the prevalence of IPV if hospitalized women with a history of IPV had longer hospital stays than those who did not or if those experiencing IPV were out of their rooms less frequently (eg, for diagnostic tests). On the other hand, our data could underrepresent the true prevalence of IPV if patients with a history of IPV had shorter hospital stays or if they received more ancillary testing that caused them to be out of their rooms more frequently. Second, none of our interviewers had specific training in interviewing techniques. Accordingly, our data could have underestimated the true prevalence of IPV if interviewers with advanced training in probing sensitive topics had more success in eliciting positive responses. Third, the relationship between a history of IPV and multiple positive responses to the review of systems may be confounded if some of these patients also had a history of adverse childhood experiences or other experiences resulting in posttraumatic stress disorder as these patients also have an increased prevalence of chronic and functional disorders.2527 Finally, as our numbers were small, we were not powered to detect clinically important differences in demographics or specific positive answers on the review of systems.

To the best of our knowledge, the only study presenting IPV prevalence data in patients hospitalized for other than psychiatric problems was performed by McKenzie and colleagues18 in 1997. In their group of 130 patients (61 on internal medicine, 59 on surgery, 7 on obstetrics, and 3 on psychiatry), the 1‐year and lifetime prevalences of IPV were only 5% and 26%, respectively. McKenzie and colleagues used only 1 question to screen for IPV, but that single question incorporated 2 of the 4 questions used in our survey. Forty‐three of our 44 patients (98%) with a history of IPV were discovered on the basis of these 2 questions. The hospitals in which the 2 studies were done were similar, as were the ages and levels of education of the 2 populations studied and the percentage of eligible patients who agreed to participate. The patients in the 2 studies were different with respect to race, language mix, and the percentage who were insured, but neither study found differences in the prevalence of IPV as a function of race or insurance (although others have found an association of IPV with being uninsured1, 3, 4, 12, 23). Our study was conducted in women admitted exclusively to an internal medicine service, whereas nearly half of the patients studied by McKenzie and colleagues were admitted to surgical, gynecologic, or psychiatric services. Although McKenzie and colleagues found no difference in the prevalence of IPV as a function of admitting service, others have suggested that the prevalence of IPV is higher in patients admitted for trauma or psychiatric problems.1517, 28 The percentage of patients who self‐administered the questionnaires was 57% in our study and 77% in the study by McKenzie and colleagues. Neither study, however, found a difference in the percentage of IPV in patients who self‐administered the survey versus those who were interviewed. Women may have become more comfortable discussing this issue in the 10‐year interval between these 2 studies, or the prevalence of IPV may have increased. The only other study of IPV in hospitalized patients of which we are aware reported a 90% 1‐year prevalence in suicidal women admitted to a psychiatric service.28

Several studies have reported that victims of IPV have multiple somatic complaints and an increased prevalence of chronic and functional illnesses.1923 We confirmed that women experiencing IPV have more positive responses to questions posed in a review of systems, but the low specificity and positive and negative predictive values of the responses make this association of little clinical utility.

For only 18 of the 72 patients (25%) in our study was there evidence that they were screened for a history of IPV by the admitting resident. If more women were screened without a response being recorded, or if women were screened only for a current history of violence, our data may not accurately reflect the true rate at which screening occurred; however, the rate of screening that we observed is consistent with a number of other studies.12, 22, 2931 Fourteen of 18 patients who were screened for IPV by the resident gave negative responses. Ten of these, however, gave positive responses to our interviewers. Accordingly, the sensitivity, specificity, and positive and negative predictive values of the information recorded by the admitting resident were 40%, 100%, 100%, and 57%, respectively (assuming that the responses given to the IPV survey represent the gold standard), and this confirms that routine screening underestimates the prevalence of this problem. Accordingly, we identified 2 problems pertaining to screening for IPV: (1) it is not routinely done at the time of hospital admission, and (2) responses reported during routine screening are frequently incorrect. A number of barriers to routine screening have been previously identified, as have interventions designed to increase screening.32 Providing specific screening questions increases the identification of victims of IPV, but simply educating healthcare providers does not.32 Our history and physical templates have a prompt for violence victim to facilitate the screening, but as a result of this study, we are changing our prompting question and indicating what should be done if the response is positive.

The US Preventive Services Task Force and the Canadian Task Force on Preventive Health Care both concluded that there was insufficient evidence to recommend for or against routine screening for IPV.3335 Their rationale was that trials assessing the effectiveness of screening have not been published, that studies designed to assess the effectiveness of any resulting intervention are few in number, focused on pregnant women, and limited by problems in study design, that no studies have determined the accuracy of the screening tools, and that none have addressed the potential harm of screening.3335 The US Preventive Services Task Force did recommend screening if providers were concerned about IPV.34 Our data would suggest that there is little in the admission history that distinguishes women who might be victims of IPV from those who might not. Guidelines published by the American Medical Association, the American Academy of Family Physicians, and the American College of Obstetricians and Gynecologists promote routine screening of all patients.3638 Janssen and colleagues39 support the importance of screening on the basis that IPV is associated with numerous physical and mental health problems (eg, arthritis, migraines and other types of headaches, vaginal bleeding, ulcers, spastic colon, chronic pain, substance abuse, depression, and suicide ideation) and that establishing the link between these conditions and IPV could be important with respect to developing appropriate diagnostic and therapeutic approaches to patients' complaints. Screening also allows physicians to become more knowledgeable about their patients' lives, facilitating their ability to provide a supportive relationship that, in turn, increases women's likelihood of using an intervention method.39 We did not confirm an increased prevalence of any of the complaints noted by Janssen and colleagues in the women experiencing a history of IPV, but we did find an increased prevalence of insomnia and extremity numbness in women admitting to IPV as well as an overall increase in the number of positive responses to the review of systems. Screening identifies women who should receive information about reporting IPV, obtaining available assistance, planning for personal safety, and formal counseling as these have all been shown to reduce the severity of IPV and to improve the quality of life in rather large, randomized controlled trials.4043

As previously observed by others,13, 22, 29, 4446 the large majority of women that we approached welcomed screening for IPV. Over half of those with a history of IPV had previously sought help for the problem, over one‐third of these sought help from physicians, and most took the resource card that we offered, regardless of whether they did or did not have a history of IPV (this suggests either that our data may actually underestimate the true prevalence of IPV or that patients taking the information knew of others experiencing this problem). Accordingly, regardless of whether physicians believe that routine screening is warranted, patients see physicians and other healthcare workers as a resource for this problem.

We have confirmed that a history of IPV is very common in women admitted to an internal medicine service of a university‐affiliated public hospital and that female victims of IPV have more positive responses on the review of systems (particularly difficulty sleeping and extremity numbness or tingling) than those who have not. Although we initially hypothesized that finding numerous somatic complaints might serve as a marker for IPV, thereby identifying patients for whom more careful screening should occur, finding such a high prevalence of IPV argues that screening should be a routine part of the history for all women admitted to internal medicine inpatient services.

The prevalence of intimate partner violence (IPV; defined as mental and/or physical violence directed from 1 person in an intimate relationship to the other) varies widely, depending on the population sampled and method of data collection. In the United States, IPV against women, occurring within the year prior to contact with a healthcare professional, ranges from 2% to 15% in surveys done by telephone, in primary care clinics, or in face‐to‐face home interviews19 and from 10% to 30% in surveys of patients visiting urgent care or emergency departments.1012 The prevalence of IPV occurring at any time during the life of the patient ranges from 18% in the aforementioned settings to as high as 88% in women applying for welfare.1, 2, 4, 5, 10, 1214

Although reports indicate that victims of IPV are more likely to be hospitalized,1517 the only study assessing the prevalence of IPV in hospitalized patients included women on medical, surgical, and obstetrical services and reported 1‐year and lifetime prevalences of only 5% and 23%, respectively.18

We hypothesized that the prevalence of IPV in hospitalized patients would be at least as high as that reported from emergency departments and sought to measure the 1‐year and lifetime prevalences of IPV in women admitted to a general internal medicine service. In addition, because studies done in various outpatient settings have reported that victims of IPV have a variety of somatic complaints and an increased prevalence of chronic and functional illnesses,1923 we also sought to determine whether women with a history of IPV and women without a history of IPV had different numbers or types of positive responses to questions asked on the review of systems.

PATIENTS AND METHODS

This study was approved by the Colorado Multiple Institution Review Board, and informed consent was obtained from all participants.

Women between the ages of 18 and 60 who were admitted to the internal medicine floor service of Denver Health Medical Center (a university‐affiliated public safety‐net hospital) between January 1 and February 28, 2004 and between October 1 and October 30, 2004 were approached to participate. These dates were selected on the basis of the availability of our interviewers. Patients older than 60 were excluded to avoid overlap between IPV and the problem of elder abuse. Women were excluded if they were unable to give informed consent, were pregnant, were incarcerated, were on contact precautions, or spoke a language other than English or Spanish. Although IPV is common in pregnant women and may occur in women who are incarcerated, these are considered vulnerable populations with respect to obtaining approval from internal review boards.

The questionnaire consisted of 23 review‐of‐systems questions,24 4 questions adapted from a previously validated screen for IPV11 (Table 1), and 1 question about attempts to seek help (Table 1). Women were considered to have experienced IPV if they gave positive responses to any of the 4 questions targeting IPV. According to patient preference, the combined questionnaire was either read and filled out by each subject independently or was read to her by a female interviewer who then recorded the subject's verbal responses. All interviewers were women with a shared common concern about, and interest in, IPV. Although none had advanced training in psychology, social work, or other formal discipline that involved interviewing skills, all interviews were scripted so that interactions with subjects and completion of the questionnaires would be uniform. Responses indicating sometimes were considered to be positive. Responses that were not answered, left blank, or marked as not applicable were considered to be negative.

Each patient's medical record was reviewed to determine her age, race, number of previous hospital admissions, visits to the emergency department and walk‐in clinic, visits to primary care and subspecialty physicians, and whether the patient had been screened for IPV as recorded on the admission history and physical template. Admission diagnosis was obtained from the history and physical template, and the discharge diagnosis was obtained from the discharge paperwork. Functional diagnoses were considered to be symptoms (eg, shortness of breath) or problems (eg, constipation) that could not clearly be linked to a specific disease process. All participants were offered a card containing a list of resources for victims of IPV.

Data were analyzed with SAS 8.1 (SAS Institute, Cary, NC) and SPSS 11.5 (SPSS, Chicago, IL). The Student t test was used to compare continuous variables. Data are reported as means standard deviation. Chi‐square analysis was used to test associations between race, primary language, level of education, insurance status, admitting diagnosis, discharge diagnosis, number of previous hospital admissions, visit type, and the presence of IPV. For these, P < 0.05 was considered to be significant. The association of positive review‐of‐systems responses with the presence of IPV was also tested by chi‐square analysis, but P < 0.002 was considered to be significant on the basis of a Bonferroni adjustment for multiple comparisons. A receiver operating characteristic curve was used to assess the relationship between the number of positive responses to the questions included in the review of systems and a history of IPV. The odds ratio and confidence intervals were calculated to test the association between the number of positive responses to the review‐of‐systems questions and a lifetime history of IPV.

RESULTS

Throughout the dates of the study, 245 women were admitted to the internal medicine service, and 106 were excluded (Figure 1). Of the 139 eligible women, 78 were available to the interviewers and asked to participate, and 72 (92%) agreed. IPV occurring within the year prior to the interview or at any point in the patient's lifetime was reported by 16 (22%) and 44 (61%) subjects, respectively. No significant differences were seen in women who did or did not experience IPV at anytime in their life with respect to age, race, insurance status, education, number of scheduled outpatient, urgent, or emergent visits, or admission or discharge diagnosis even when the diagnoses were grouped into a functional category (although at best our study was powered to detect only >35% differences in prevalences; Tables 2 and 3). Of women reporting a lifetime history of IPV, 26 of 44 (59%) had previously sought help, and 9 of those 26 (35%) said that they sought help from a physician.

Figure 1

Women with a 1‐year history of IPV and women without a 1‐year history of IPV had 11.4 4.7 and 7.7 5.4 positive responses to the review of systems (P < 0.01), respectively. Women with a lifetime history of IPV and women without a lifetime history of IPV had 10.9 4.4 and 7.7 5.4 positive responses (P < 0.01), respectively. The receiver operating characteristic curve of the number of positive responses versus a lifetime history of IPV is presented in Figure 2. Subjects with 10 or more positive responses were 4.8 times more likely to report a lifetime history of IPV than subjects with 9 or fewer positive responses (confidence interval = 1.614.2, P = 0.003). The c‐statistic indicating the ability of the review of systems to properly classify cases when there were 10 or more positive responses was 0.692.

Figure 2

No differences were observed in the responses to the individual review of systems questions in women who did or did not have a lifetime history of IPV, with the exception that those with a positive history more commonly complained of difficulty sleeping and numbness and tingling in their hands or feet (although at best our study was sufficiently powered to detect only >20% differences in prevalences; Table 4). Although the sensitivity of having problems sleeping or experiencing numbness or tingling in patients with IPV was high, the specificity and positive and negative predictive values were not (Table 5).

The admission history forms filled out by first‐year admitting residents showed that only 18 (25%) of the women were screened for IPV, even though the history and physical examination template used at Denver Health Medical Center includes a prompt in the social history section pertaining to a history of violence as a reminder.

DISCUSSION

The important findings of this study were that women admitted to the internal medicine service of a university‐affiliated public safety‐net hospital had a high prevalence of IPV (22% and 61% 1‐year and lifetime prevalences, respectively), that most women with a history of IPV had previously sought help for the problem, many from physicians, that women were more likely to have a history of IPV if they had >10 positive responses to questions asked in a routine review of systems (particularly problems sleeping and experiencing numbness or tingling in their extremities), and that routine screening for IPV was uncommon at the time of admission.

These conclusions should be interpreted with respect to a number of limitations in our study. First, although our study was designed to be a consecutive series, the interviewers did not have sufficient time to meet with and interview every woman admitted before they were discharged. This occurred in part because the interviewers were available only for a portion of each day, some patients were discharged within 24 hours of admission, and many were out of their rooms for ancillary testing. Within the interviewers' time constraints, however, all hospitalized women meeting entry criteria who were available were approached. Our data could, however, overrepresent the prevalence of IPV if hospitalized women with a history of IPV had longer hospital stays than those who did not or if those experiencing IPV were out of their rooms less frequently (eg, for diagnostic tests). On the other hand, our data could underrepresent the true prevalence of IPV if patients with a history of IPV had shorter hospital stays or if they received more ancillary testing that caused them to be out of their rooms more frequently. Second, none of our interviewers had specific training in interviewing techniques. Accordingly, our data could have underestimated the true prevalence of IPV if interviewers with advanced training in probing sensitive topics had more success in eliciting positive responses. Third, the relationship between a history of IPV and multiple positive responses to the review of systems may be confounded if some of these patients also had a history of adverse childhood experiences or other experiences resulting in posttraumatic stress disorder as these patients also have an increased prevalence of chronic and functional disorders.2527 Finally, as our numbers were small, we were not powered to detect clinically important differences in demographics or specific positive answers on the review of systems.

To the best of our knowledge, the only study presenting IPV prevalence data in patients hospitalized for other than psychiatric problems was performed by McKenzie and colleagues18 in 1997. In their group of 130 patients (61 on internal medicine, 59 on surgery, 7 on obstetrics, and 3 on psychiatry), the 1‐year and lifetime prevalences of IPV were only 5% and 26%, respectively. McKenzie and colleagues used only 1 question to screen for IPV, but that single question incorporated 2 of the 4 questions used in our survey. Forty‐three of our 44 patients (98%) with a history of IPV were discovered on the basis of these 2 questions. The hospitals in which the 2 studies were done were similar, as were the ages and levels of education of the 2 populations studied and the percentage of eligible patients who agreed to participate. The patients in the 2 studies were different with respect to race, language mix, and the percentage who were insured, but neither study found differences in the prevalence of IPV as a function of race or insurance (although others have found an association of IPV with being uninsured1, 3, 4, 12, 23). Our study was conducted in women admitted exclusively to an internal medicine service, whereas nearly half of the patients studied by McKenzie and colleagues were admitted to surgical, gynecologic, or psychiatric services. Although McKenzie and colleagues found no difference in the prevalence of IPV as a function of admitting service, others have suggested that the prevalence of IPV is higher in patients admitted for trauma or psychiatric problems.1517, 28 The percentage of patients who self‐administered the questionnaires was 57% in our study and 77% in the study by McKenzie and colleagues. Neither study, however, found a difference in the percentage of IPV in patients who self‐administered the survey versus those who were interviewed. Women may have become more comfortable discussing this issue in the 10‐year interval between these 2 studies, or the prevalence of IPV may have increased. The only other study of IPV in hospitalized patients of which we are aware reported a 90% 1‐year prevalence in suicidal women admitted to a psychiatric service.28

Several studies have reported that victims of IPV have multiple somatic complaints and an increased prevalence of chronic and functional illnesses.1923 We confirmed that women experiencing IPV have more positive responses to questions posed in a review of systems, but the low specificity and positive and negative predictive values of the responses make this association of little clinical utility.

For only 18 of the 72 patients (25%) in our study was there evidence that they were screened for a history of IPV by the admitting resident. If more women were screened without a response being recorded, or if women were screened only for a current history of violence, our data may not accurately reflect the true rate at which screening occurred; however, the rate of screening that we observed is consistent with a number of other studies.12, 22, 2931 Fourteen of 18 patients who were screened for IPV by the resident gave negative responses. Ten of these, however, gave positive responses to our interviewers. Accordingly, the sensitivity, specificity, and positive and negative predictive values of the information recorded by the admitting resident were 40%, 100%, 100%, and 57%, respectively (assuming that the responses given to the IPV survey represent the gold standard), and this confirms that routine screening underestimates the prevalence of this problem. Accordingly, we identified 2 problems pertaining to screening for IPV: (1) it is not routinely done at the time of hospital admission, and (2) responses reported during routine screening are frequently incorrect. A number of barriers to routine screening have been previously identified, as have interventions designed to increase screening.32 Providing specific screening questions increases the identification of victims of IPV, but simply educating healthcare providers does not.32 Our history and physical templates have a prompt for violence victim to facilitate the screening, but as a result of this study, we are changing our prompting question and indicating what should be done if the response is positive.

The US Preventive Services Task Force and the Canadian Task Force on Preventive Health Care both concluded that there was insufficient evidence to recommend for or against routine screening for IPV.3335 Their rationale was that trials assessing the effectiveness of screening have not been published, that studies designed to assess the effectiveness of any resulting intervention are few in number, focused on pregnant women, and limited by problems in study design, that no studies have determined the accuracy of the screening tools, and that none have addressed the potential harm of screening.3335 The US Preventive Services Task Force did recommend screening if providers were concerned about IPV.34 Our data would suggest that there is little in the admission history that distinguishes women who might be victims of IPV from those who might not. Guidelines published by the American Medical Association, the American Academy of Family Physicians, and the American College of Obstetricians and Gynecologists promote routine screening of all patients.3638 Janssen and colleagues39 support the importance of screening on the basis that IPV is associated with numerous physical and mental health problems (eg, arthritis, migraines and other types of headaches, vaginal bleeding, ulcers, spastic colon, chronic pain, substance abuse, depression, and suicide ideation) and that establishing the link between these conditions and IPV could be important with respect to developing appropriate diagnostic and therapeutic approaches to patients' complaints. Screening also allows physicians to become more knowledgeable about their patients' lives, facilitating their ability to provide a supportive relationship that, in turn, increases women's likelihood of using an intervention method.39 We did not confirm an increased prevalence of any of the complaints noted by Janssen and colleagues in the women experiencing a history of IPV, but we did find an increased prevalence of insomnia and extremity numbness in women admitting to IPV as well as an overall increase in the number of positive responses to the review of systems. Screening identifies women who should receive information about reporting IPV, obtaining available assistance, planning for personal safety, and formal counseling as these have all been shown to reduce the severity of IPV and to improve the quality of life in rather large, randomized controlled trials.4043

As previously observed by others,13, 22, 29, 4446 the large majority of women that we approached welcomed screening for IPV. Over half of those with a history of IPV had previously sought help for the problem, over one‐third of these sought help from physicians, and most took the resource card that we offered, regardless of whether they did or did not have a history of IPV (this suggests either that our data may actually underestimate the true prevalence of IPV or that patients taking the information knew of others experiencing this problem). Accordingly, regardless of whether physicians believe that routine screening is warranted, patients see physicians and other healthcare workers as a resource for this problem.

We have confirmed that a history of IPV is very common in women admitted to an internal medicine service of a university‐affiliated public hospital and that female victims of IPV have more positive responses on the review of systems (particularly difficulty sleeping and extremity numbness or tingling) than those who have not. Although we initially hypothesized that finding numerous somatic complaints might serve as a marker for IPV, thereby identifying patients for whom more careful screening should occur, finding such a high prevalence of IPV argues that screening should be a routine part of the history for all women admitted to internal medicine inpatient services.

As a hospitalist of about 6 years, I enjoy hospital medicine and hope, over the course of my career, to see it develop into an increasingly respected, diverse, and influential specialty. There is abundant evidence that this is occurring, primarily through the praiseworthy efforts of the leadership and members of the Society of Hospital Medicine (SHM). Efforts to prove our value to inpatient care and align ourselves with quality improvement, as promoted early in the hospitalist movement,2 are coming to fruition. However, I would like to raise a flag of concern; and this is based on my experience working as a hospitalist in 10 community hospitals in 5 states, including positions as a locum tenens hospitalist, staff hospitalist, medical director of a hospitalist group, and full‐time teaching hospitalist for a community hospital residency program. I believe that hospitalists, particularly those working in community hospitals (approximately 80% of all hospitalists),3 are currently at a critical crossroad, with the option of either actively expanding their clinical, administrative, and quality improvement roles or allowing these roles to stagnate or atrophy. As in any career, we are, like Alice, perched on a mushroom, one side of which will make us grow taller and the other side of which will make us grow shorter. Which side are we choosing in our careers as hospitalists?

Hospitalists currently have numerous opportunities to expand their clinical, administrative, and quality improvement roles and responsibilities (Table 1), and these opportunities are in full alignment with the mission statement of SHM: to promote the highest quality of care for all hospitalized patients.4 My concern is that, for one reason or another, hospitalists in some settings are shrinking away from roles that they could or should fill, and this is a trend that I believe could affect our specialty adversely over time and that we, as an organization, should find ways to prevent. Although family medicine and traditional internal medicine physicians who work in the hospital face similar challenges, if we as hospitalists wish to qualify one day as board‐certified hospital medicine specialists, we are obligated to develop knowledge and skill sets that are truly unique to our profession.5 Holding to this goal, we cannot settle into a narrow comfort zone. I believe that the development of the hospital medicine core competencies by SHM6 was an important step in helping us define our intended reach, but even so, what are the specific growth factors or inhibitors that are influencing the expansion or shrinking of hospitalists and hospital medicine groups?

On the basis of my observations, I believe that this problem is due in large part to a misalignment of incentives. Specifically, I believe that the expansion of hospitalist roles and responsibilities is often counteraligned with the bottom‐line productivity goals of the group. That is, to maintain high productivity, a hospitalist has a tendency to minimize his or her role in ways that save time. For example, there may be a tendency to overuse subspecialty consultations, which can take away some of the burden of complex clinical decision making, or to quickly transfer patients that are sicker and require more time to a higher level of care (if available). There may also be a tendency to avoid performing inpatient procedures (a significant part of the core competencies) because of time constraints and the demands of a higher census. Excessively rapid rounding results, and this diminishes other claimed benefits of the hospitalist model of care: patient satisfaction, safety, quality, and communication. Length‐of‐stay measures also suffer as productivity exceeds the limits of efficient care. Moreover, in such a productivity‐based environment, there is certainly no incentive for hospitalists to become enthusiastically involved in hospital committees, education, or quality improvement efforts, all of which are critical to the development of hospital medicine as a unique subspecialty. In essence, the incentive to expand one's role as a hospitalist in such a setting is almost completely absent, and I believe that this puts the future influence and reach of our specialty at significant risk.

Particularly as hospitals face increasing scrutiny about their quality and safety, and especially as the costs of hospital care increase and reimbursements threaten to decline, the value of hospitalists to the hospital has become different from that of all other physicians. Their value lies not in sheer productivity but in their ability to improve the cost, quality, efficiency, and safety of inpatient care simultaneously. If hospitalists settle into or are forced into a lesser role, hospital medicine will not be worthy of consideration as a unique subspecialty. Some of the remaining roles of the shrunken hospitalist may, at some point and in some settings, shift to nonphysicians,7 with a decline in the ratio of physicians to mid‐level providers in hospital medicine programs, and the jobs of some hospitalists will be effectively eliminated. Market forces will lead to improved training of mid‐level providers, allowing hospitals to fill inpatient care needs in a more cost‐effective way.

Having worked with some very capable nurse practitioners in 4 different community hospitals, I believe that a well‐trained mid‐level provider, with appropriate physician backup, can effectively manage many of the typical general medical admissions and surgical consultations seen in a community hospital setting. I admit that this may not be the case in larger referral centers or academic medical centers.

In developing and defining this new specialty and also in training new physicians for the field, we do not want to lose this transient opportunity to define ourselves as broadly as possible, pushing beyond traditional internal medicine to new areas of inpatient care and management and managing more complex conditions than a traditional primary care physician would typically manage, conditions that have always fallen within the broad spectrum of inpatient internal medicine (Table 2). If we instead develop a tendency to admit, consult, and walk away and do not have the time or appropriate incentives to expand our roles in other important ways (noted in Table 1) because of a focus on productivity, what is our specialty destined to become?

That said, how can incentives be restructured to encourage hospitalists to expand their universe? Perhaps the simplest way of influencing the incentive structure of hospital medicine programs is more selectivity in the choice of jobs: seeking out jobs that offer us clear incentives (typically financial) to expand our universe by rewarding efforts to improve the quality, safety, and efficiency of inpatient care. According to the SHM 20052006 survey, about two‐thirds of responding hospital medicine programs reimbursed their physicians with a mix of salary and productivity/performance bonuses, with productivity being the dominant incentive (more than 80%). However, bonuses based on quality/efficiency measures were also being rewarded (about 60%), as well as bonuses for committee or project work (about 25%). Of all responding groups, that leaves about 60% of programs with no financial incentives for quality/efficiency measures. There is certainly room for progress in this area, and we can influence the process positively by requesting that such incentives be added to our contract before making a final commitment to a job or by negotiating changes to our current incentive structure at the time of contract renewal. This would be in the best interest of our individual careers as well as our specialty.

As we consider different job opportunities, we may also wish to consider the possible effect of the employment model on the incentive structure. Although it may seem logical that hospital‐employed groups would have broader goals than independent groups and thus might be more motivated to provide proper incentives, I do not believe that this is the case universally. Conversely, private groups who might be expected to focus more on productivity measures may actually offer excellent growth‐promoting incentives. In either case, careful consideration of the incentive structure is warranted when we choose to work in a given employment model.

Perhaps another way of encouraging hospitalists to expand their role would be through a program of national recognition, potentially established by SHM, that would allow individual hospitalists to formally claim specialization in a particular area of hospital medicine and benefit from such distinctions. For example, a hospitalist that was particularly proficient with inpatient procedures could submit documentation of procedures completed in a given time period and subsequently receive a formal designation as a certified procedural hospitalist or something similar. Alternatively, a hospitalist who preferred to focus on quality improvement efforts could submit information regarding his involvement with quality improvement initiatives and results and, on the basis of defined criteria, receive a formal designation as a quality improvement hospitalist. This approach could apply to any area of focus, and more than one designation could be achieved by each hospitalist. As the specialty of hospital medicine matures, these designations (similar to academic rank) could eventually correlate with salary ranges or incentive bonuses as hospitals learned to value the diverse skills of individual hospitalists.

Discouraging overconsultation of subspecialists while concurrently encouraging the broadening of our clinical skills is particularly difficult to address. The only solution to this issue that I can imagine would be to somehow align physician reimbursement more closely to the actual complexity of and time spent in managing patients with multiple comorbidities. Currently, the actual hospitalist physician reimbursement for subsequent visits of patients, with or without subspecialists involved, likely does not vary much. However, if hospitalists knew their extra effort in managing more complex conditions would be reimbursed differently (ie, billing for critical care time), they would certainly tend to broaden their practice to the benefit of their careers and the future of the specialty.

In summary, I believe that misaligned incentives are causing some hospitalists to underestimate their potential; this has the potential to adversely affect the future of the specialty of hospital medicine. I hope that this opinion will serve to generate discussion on the potential origins of and solutions to this problem and ultimately promote the future expansion of our hospital medicine universe, so that we do not find ourselves in Alice's predicament:

As a hospitalist of about 6 years, I enjoy hospital medicine and hope, over the course of my career, to see it develop into an increasingly respected, diverse, and influential specialty. There is abundant evidence that this is occurring, primarily through the praiseworthy efforts of the leadership and members of the Society of Hospital Medicine (SHM). Efforts to prove our value to inpatient care and align ourselves with quality improvement, as promoted early in the hospitalist movement,2 are coming to fruition. However, I would like to raise a flag of concern; and this is based on my experience working as a hospitalist in 10 community hospitals in 5 states, including positions as a locum tenens hospitalist, staff hospitalist, medical director of a hospitalist group, and full‐time teaching hospitalist for a community hospital residency program. I believe that hospitalists, particularly those working in community hospitals (approximately 80% of all hospitalists),3 are currently at a critical crossroad, with the option of either actively expanding their clinical, administrative, and quality improvement roles or allowing these roles to stagnate or atrophy. As in any career, we are, like Alice, perched on a mushroom, one side of which will make us grow taller and the other side of which will make us grow shorter. Which side are we choosing in our careers as hospitalists?

Hospitalists currently have numerous opportunities to expand their clinical, administrative, and quality improvement roles and responsibilities (Table 1), and these opportunities are in full alignment with the mission statement of SHM: to promote the highest quality of care for all hospitalized patients.4 My concern is that, for one reason or another, hospitalists in some settings are shrinking away from roles that they could or should fill, and this is a trend that I believe could affect our specialty adversely over time and that we, as an organization, should find ways to prevent. Although family medicine and traditional internal medicine physicians who work in the hospital face similar challenges, if we as hospitalists wish to qualify one day as board‐certified hospital medicine specialists, we are obligated to develop knowledge and skill sets that are truly unique to our profession.5 Holding to this goal, we cannot settle into a narrow comfort zone. I believe that the development of the hospital medicine core competencies by SHM6 was an important step in helping us define our intended reach, but even so, what are the specific growth factors or inhibitors that are influencing the expansion or shrinking of hospitalists and hospital medicine groups?

On the basis of my observations, I believe that this problem is due in large part to a misalignment of incentives. Specifically, I believe that the expansion of hospitalist roles and responsibilities is often counteraligned with the bottom‐line productivity goals of the group. That is, to maintain high productivity, a hospitalist has a tendency to minimize his or her role in ways that save time. For example, there may be a tendency to overuse subspecialty consultations, which can take away some of the burden of complex clinical decision making, or to quickly transfer patients that are sicker and require more time to a higher level of care (if available). There may also be a tendency to avoid performing inpatient procedures (a significant part of the core competencies) because of time constraints and the demands of a higher census. Excessively rapid rounding results, and this diminishes other claimed benefits of the hospitalist model of care: patient satisfaction, safety, quality, and communication. Length‐of‐stay measures also suffer as productivity exceeds the limits of efficient care. Moreover, in such a productivity‐based environment, there is certainly no incentive for hospitalists to become enthusiastically involved in hospital committees, education, or quality improvement efforts, all of which are critical to the development of hospital medicine as a unique subspecialty. In essence, the incentive to expand one's role as a hospitalist in such a setting is almost completely absent, and I believe that this puts the future influence and reach of our specialty at significant risk.

Particularly as hospitals face increasing scrutiny about their quality and safety, and especially as the costs of hospital care increase and reimbursements threaten to decline, the value of hospitalists to the hospital has become different from that of all other physicians. Their value lies not in sheer productivity but in their ability to improve the cost, quality, efficiency, and safety of inpatient care simultaneously. If hospitalists settle into or are forced into a lesser role, hospital medicine will not be worthy of consideration as a unique subspecialty. Some of the remaining roles of the shrunken hospitalist may, at some point and in some settings, shift to nonphysicians,7 with a decline in the ratio of physicians to mid‐level providers in hospital medicine programs, and the jobs of some hospitalists will be effectively eliminated. Market forces will lead to improved training of mid‐level providers, allowing hospitals to fill inpatient care needs in a more cost‐effective way.

Having worked with some very capable nurse practitioners in 4 different community hospitals, I believe that a well‐trained mid‐level provider, with appropriate physician backup, can effectively manage many of the typical general medical admissions and surgical consultations seen in a community hospital setting. I admit that this may not be the case in larger referral centers or academic medical centers.

In developing and defining this new specialty and also in training new physicians for the field, we do not want to lose this transient opportunity to define ourselves as broadly as possible, pushing beyond traditional internal medicine to new areas of inpatient care and management and managing more complex conditions than a traditional primary care physician would typically manage, conditions that have always fallen within the broad spectrum of inpatient internal medicine (Table 2). If we instead develop a tendency to admit, consult, and walk away and do not have the time or appropriate incentives to expand our roles in other important ways (noted in Table 1) because of a focus on productivity, what is our specialty destined to become?

That said, how can incentives be restructured to encourage hospitalists to expand their universe? Perhaps the simplest way of influencing the incentive structure of hospital medicine programs is more selectivity in the choice of jobs: seeking out jobs that offer us clear incentives (typically financial) to expand our universe by rewarding efforts to improve the quality, safety, and efficiency of inpatient care. According to the SHM 20052006 survey, about two‐thirds of responding hospital medicine programs reimbursed their physicians with a mix of salary and productivity/performance bonuses, with productivity being the dominant incentive (more than 80%). However, bonuses based on quality/efficiency measures were also being rewarded (about 60%), as well as bonuses for committee or project work (about 25%). Of all responding groups, that leaves about 60% of programs with no financial incentives for quality/efficiency measures. There is certainly room for progress in this area, and we can influence the process positively by requesting that such incentives be added to our contract before making a final commitment to a job or by negotiating changes to our current incentive structure at the time of contract renewal. This would be in the best interest of our individual careers as well as our specialty.

As we consider different job opportunities, we may also wish to consider the possible effect of the employment model on the incentive structure. Although it may seem logical that hospital‐employed groups would have broader goals than independent groups and thus might be more motivated to provide proper incentives, I do not believe that this is the case universally. Conversely, private groups who might be expected to focus more on productivity measures may actually offer excellent growth‐promoting incentives. In either case, careful consideration of the incentive structure is warranted when we choose to work in a given employment model.

Perhaps another way of encouraging hospitalists to expand their role would be through a program of national recognition, potentially established by SHM, that would allow individual hospitalists to formally claim specialization in a particular area of hospital medicine and benefit from such distinctions. For example, a hospitalist that was particularly proficient with inpatient procedures could submit documentation of procedures completed in a given time period and subsequently receive a formal designation as a certified procedural hospitalist or something similar. Alternatively, a hospitalist who preferred to focus on quality improvement efforts could submit information regarding his involvement with quality improvement initiatives and results and, on the basis of defined criteria, receive a formal designation as a quality improvement hospitalist. This approach could apply to any area of focus, and more than one designation could be achieved by each hospitalist. As the specialty of hospital medicine matures, these designations (similar to academic rank) could eventually correlate with salary ranges or incentive bonuses as hospitals learned to value the diverse skills of individual hospitalists.

Discouraging overconsultation of subspecialists while concurrently encouraging the broadening of our clinical skills is particularly difficult to address. The only solution to this issue that I can imagine would be to somehow align physician reimbursement more closely to the actual complexity of and time spent in managing patients with multiple comorbidities. Currently, the actual hospitalist physician reimbursement for subsequent visits of patients, with or without subspecialists involved, likely does not vary much. However, if hospitalists knew their extra effort in managing more complex conditions would be reimbursed differently (ie, billing for critical care time), they would certainly tend to broaden their practice to the benefit of their careers and the future of the specialty.

In summary, I believe that misaligned incentives are causing some hospitalists to underestimate their potential; this has the potential to adversely affect the future of the specialty of hospital medicine. I hope that this opinion will serve to generate discussion on the potential origins of and solutions to this problem and ultimately promote the future expansion of our hospital medicine universe, so that we do not find ourselves in Alice's predicament:

As a hospitalist of about 6 years, I enjoy hospital medicine and hope, over the course of my career, to see it develop into an increasingly respected, diverse, and influential specialty. There is abundant evidence that this is occurring, primarily through the praiseworthy efforts of the leadership and members of the Society of Hospital Medicine (SHM). Efforts to prove our value to inpatient care and align ourselves with quality improvement, as promoted early in the hospitalist movement,2 are coming to fruition. However, I would like to raise a flag of concern; and this is based on my experience working as a hospitalist in 10 community hospitals in 5 states, including positions as a locum tenens hospitalist, staff hospitalist, medical director of a hospitalist group, and full‐time teaching hospitalist for a community hospital residency program. I believe that hospitalists, particularly those working in community hospitals (approximately 80% of all hospitalists),3 are currently at a critical crossroad, with the option of either actively expanding their clinical, administrative, and quality improvement roles or allowing these roles to stagnate or atrophy. As in any career, we are, like Alice, perched on a mushroom, one side of which will make us grow taller and the other side of which will make us grow shorter. Which side are we choosing in our careers as hospitalists?

Hospitalists currently have numerous opportunities to expand their clinical, administrative, and quality improvement roles and responsibilities (Table 1), and these opportunities are in full alignment with the mission statement of SHM: to promote the highest quality of care for all hospitalized patients.4 My concern is that, for one reason or another, hospitalists in some settings are shrinking away from roles that they could or should fill, and this is a trend that I believe could affect our specialty adversely over time and that we, as an organization, should find ways to prevent. Although family medicine and traditional internal medicine physicians who work in the hospital face similar challenges, if we as hospitalists wish to qualify one day as board‐certified hospital medicine specialists, we are obligated to develop knowledge and skill sets that are truly unique to our profession.5 Holding to this goal, we cannot settle into a narrow comfort zone. I believe that the development of the hospital medicine core competencies by SHM6 was an important step in helping us define our intended reach, but even so, what are the specific growth factors or inhibitors that are influencing the expansion or shrinking of hospitalists and hospital medicine groups?

On the basis of my observations, I believe that this problem is due in large part to a misalignment of incentives. Specifically, I believe that the expansion of hospitalist roles and responsibilities is often counteraligned with the bottom‐line productivity goals of the group. That is, to maintain high productivity, a hospitalist has a tendency to minimize his or her role in ways that save time. For example, there may be a tendency to overuse subspecialty consultations, which can take away some of the burden of complex clinical decision making, or to quickly transfer patients that are sicker and require more time to a higher level of care (if available). There may also be a tendency to avoid performing inpatient procedures (a significant part of the core competencies) because of time constraints and the demands of a higher census. Excessively rapid rounding results, and this diminishes other claimed benefits of the hospitalist model of care: patient satisfaction, safety, quality, and communication. Length‐of‐stay measures also suffer as productivity exceeds the limits of efficient care. Moreover, in such a productivity‐based environment, there is certainly no incentive for hospitalists to become enthusiastically involved in hospital committees, education, or quality improvement efforts, all of which are critical to the development of hospital medicine as a unique subspecialty. In essence, the incentive to expand one's role as a hospitalist in such a setting is almost completely absent, and I believe that this puts the future influence and reach of our specialty at significant risk.

Particularly as hospitals face increasing scrutiny about their quality and safety, and especially as the costs of hospital care increase and reimbursements threaten to decline, the value of hospitalists to the hospital has become different from that of all other physicians. Their value lies not in sheer productivity but in their ability to improve the cost, quality, efficiency, and safety of inpatient care simultaneously. If hospitalists settle into or are forced into a lesser role, hospital medicine will not be worthy of consideration as a unique subspecialty. Some of the remaining roles of the shrunken hospitalist may, at some point and in some settings, shift to nonphysicians,7 with a decline in the ratio of physicians to mid‐level providers in hospital medicine programs, and the jobs of some hospitalists will be effectively eliminated. Market forces will lead to improved training of mid‐level providers, allowing hospitals to fill inpatient care needs in a more cost‐effective way.

Having worked with some very capable nurse practitioners in 4 different community hospitals, I believe that a well‐trained mid‐level provider, with appropriate physician backup, can effectively manage many of the typical general medical admissions and surgical consultations seen in a community hospital setting. I admit that this may not be the case in larger referral centers or academic medical centers.

In developing and defining this new specialty and also in training new physicians for the field, we do not want to lose this transient opportunity to define ourselves as broadly as possible, pushing beyond traditional internal medicine to new areas of inpatient care and management and managing more complex conditions than a traditional primary care physician would typically manage, conditions that have always fallen within the broad spectrum of inpatient internal medicine (Table 2). If we instead develop a tendency to admit, consult, and walk away and do not have the time or appropriate incentives to expand our roles in other important ways (noted in Table 1) because of a focus on productivity, what is our specialty destined to become?

That said, how can incentives be restructured to encourage hospitalists to expand their universe? Perhaps the simplest way of influencing the incentive structure of hospital medicine programs is more selectivity in the choice of jobs: seeking out jobs that offer us clear incentives (typically financial) to expand our universe by rewarding efforts to improve the quality, safety, and efficiency of inpatient care. According to the SHM 20052006 survey, about two‐thirds of responding hospital medicine programs reimbursed their physicians with a mix of salary and productivity/performance bonuses, with productivity being the dominant incentive (more than 80%). However, bonuses based on quality/efficiency measures were also being rewarded (about 60%), as well as bonuses for committee or project work (about 25%). Of all responding groups, that leaves about 60% of programs with no financial incentives for quality/efficiency measures. There is certainly room for progress in this area, and we can influence the process positively by requesting that such incentives be added to our contract before making a final commitment to a job or by negotiating changes to our current incentive structure at the time of contract renewal. This would be in the best interest of our individual careers as well as our specialty.

As we consider different job opportunities, we may also wish to consider the possible effect of the employment model on the incentive structure. Although it may seem logical that hospital‐employed groups would have broader goals than independent groups and thus might be more motivated to provide proper incentives, I do not believe that this is the case universally. Conversely, private groups who might be expected to focus more on productivity measures may actually offer excellent growth‐promoting incentives. In either case, careful consideration of the incentive structure is warranted when we choose to work in a given employment model.

Perhaps another way of encouraging hospitalists to expand their role would be through a program of national recognition, potentially established by SHM, that would allow individual hospitalists to formally claim specialization in a particular area of hospital medicine and benefit from such distinctions. For example, a hospitalist that was particularly proficient with inpatient procedures could submit documentation of procedures completed in a given time period and subsequently receive a formal designation as a certified procedural hospitalist or something similar. Alternatively, a hospitalist who preferred to focus on quality improvement efforts could submit information regarding his involvement with quality improvement initiatives and results and, on the basis of defined criteria, receive a formal designation as a quality improvement hospitalist. This approach could apply to any area of focus, and more than one designation could be achieved by each hospitalist. As the specialty of hospital medicine matures, these designations (similar to academic rank) could eventually correlate with salary ranges or incentive bonuses as hospitals learned to value the diverse skills of individual hospitalists.

Discouraging overconsultation of subspecialists while concurrently encouraging the broadening of our clinical skills is particularly difficult to address. The only solution to this issue that I can imagine would be to somehow align physician reimbursement more closely to the actual complexity of and time spent in managing patients with multiple comorbidities. Currently, the actual hospitalist physician reimbursement for subsequent visits of patients, with or without subspecialists involved, likely does not vary much. However, if hospitalists knew their extra effort in managing more complex conditions would be reimbursed differently (ie, billing for critical care time), they would certainly tend to broaden their practice to the benefit of their careers and the future of the specialty.

In summary, I believe that misaligned incentives are causing some hospitalists to underestimate their potential; this has the potential to adversely affect the future of the specialty of hospital medicine. I hope that this opinion will serve to generate discussion on the potential origins of and solutions to this problem and ultimately promote the future expansion of our hospital medicine universe, so that we do not find ourselves in Alice's predicament:

Accelerating the development of clinical research in academic hospitalist programs is a worthwhile goal if pursued with clarity, objectivity, and a thorough understanding of the process and its implications. In their articles, Flanders et al.1 and Wright et al.2 identify major barriers to growing academic hospitalist programs. These barriers include the need for protected time, the shortage of trained research faculty, the lack of infrastructure, and the limited availability of senior mentors. Both Flanders et al. and Wright et al. offer smart and innovative ways of addressing these issues. However, building an academic program from the ground up is more complex and challenging than it may seem at first glance. It takes time, patience, creativity, diplomacy, and the ability to recruit collaborators and advocates who are willing to share infrastructure and resources.

Although both articles add significantly to the discussion of strategies for creating an academic hospitalist program, they are unclear about the definition of academic in this context. The term academic is often misunderstood to be synonymous with research. However, research is just one component of an academic program, which also includes education, quality improvement (QI), administration, and program development. It may be helpful, therefore, to replace academic with scholarship, which can be defined as a process that involves peer review and dissemination of ideas at local, regional, and national levels. Scholarship also goes beyond research, encompassing education and other areas such as QI. Although academic programs are not necessarily involved with funded research, there is usually an expectation of peer review, through either presentations at regional and national meetings or publication. For the purposes of this discussion, the term academic hospitalist program will be defined broadly to include any program affiliated with a university that is involved in the teaching of residents and medical students and whose faculty is required to participate in a promotions process.

All members of an academic division should be expected to participate in scholarship, whether it is education, QI projects, or research. If there is a strong expectation that traditional National Institute of Health (NIH) funded research will take place, this expectation must come with sufficient resources. Without infrastructure for research and investment in research faculty, procuring NIH funds for research is not a reasonable expectation. Organizers of hospitalist programs currently within academic divisions of general internal medicine should consider ways to better integrate programs into the existing research infrastructure in their divisions. For either freestanding hospitalist programs or programs within academic divisions of general internal medicine, investments in infrastructure and faculty are needed to nurture this area of research and build an academic focus in hospital medicine. However, if obtaining NIH research funds is not the expectation and resources are not available for hospitalists or for any other division or department at that institution, then academic expectations should focus on other pursuits. Examples include participation in the education and QI initiatives.

For programs with expectations of both funded research and other scholarship, a successful program will most likely include a small core of skilled clinical researchers working closely with well‐trained clinical educators, all of whom are involved in scholarship. Both clinical educators and researchers need to be continuously developing, and to reach their full potential, all should have access to infrastructure that supports these activities, including resources such as MPH‐level project managers, research assistants, database managers, and, most importantly, appropriate mentors.

Clinician educators must be both proficient clinicians and dedicated teachers. Ideally, they should have strong familiarity with educational theory in addition to skills in hands‐on teaching. Their responsibilities include mastering the skills that students need, staying up to the minute in their areas of expertise, and serving as role models in their attitudes toward patients, colleagues, and their work.3 Many hospitalists may not have these skills when they begin, often fresh out of residency, and will need help developing them.

PROTECTED TIME

Protected time is crucial to the success of any academic program. Finding this time presents a challenge for any clinical group, but the challenge is exacerbated for hospitalists, who face tremendous pressure to serve full‐time clinical jobs with little emphasis on academic elements such as education, QI, and participation in funded research. As both Wright et al.2 and Flanders et al.1 point out, to build a group of well‐developed academic clinician educators, academic hospitalists not on track for funding need to be given adequate protected time to participate in committees, sharpen their teaching skills, develop QI projects that can be converted to scholarships, participate in research, and present at national and regional meetings.

The requirements of protected time for researchers are more challenging than those for educators. Building a newly funded research unit within a hospitalist group, as with any group, will entail hiring fellowship trained faculty with significant protected time (approximately 80%) to give them time to obtain funding such as a K award and eventually become independent investigators with RO1 grant funding. Building research units requires support from collaborators with infrastructure and mentors already in place that can be tapped during the incubation stage of the academic program. For most hospitalist programs, infrastructure and mentors will be found in their divisions of general internal medicine.

Protected time should be considered an integral element of academic hospitalist positionsnot a perkas long as the time is used responsibly and productively. As both Wright et al.2 and Flanders et al.1 correctly point out, herein lies the major challenge of creating any kind of academic program: How will the program support protected time for both educators and researchers? In most instances, significant seed money will be needed to support junior faculty over the first few years of their careers. It is noteworthy that building a federally funded hospital medicine research program will be particularly difficult in today's economy because funding levels at the NIH, Agency for Healthcare Research and Quality, Health Resources and Services Administration, and other traditional funders of clinical research either are flat or have been reduced dramatically.

To many Academic Medical Centers (AMC), it may not be immediately obvious why a strong Academic Hospitalist program is in their economic best interest. Hospitalist programs may confront consistently high levels of turnover and a shrinking supply of general internists. The associated high costs of hiring new, junior faculty include the time and effort needed to interview, credential, train, and most importantly build familiarity with the complex systems encountered in maneuvering through a hospital, especially one with widespread dissemination of electronic medical records for documentation and order entry. However, hospitalists provided with opportunities for academic development are more likely to stay on the job longer and perform at a higher level, providing convincing motivation for hospitals to invest in their academic hospitalist programs. Retaining high‐quality hospitalists may be one of the most cost‐efficient methods for an AMC to support a hospital medicine program.

STRATEGIES IN ACTION

Flanders et al.1 point to a shortage of well‐trained clinician investigators with a focus on inpatient research as a barrier to the development of academic hospitalist programming. They describe a strategy of collaboration with specialty groups. This highlights the importance of collaboration with more well‐established research units as a key ingredient to building a new academic unit. Wright et al.2 describe their mentorship program and how they created protected time for scholarship for hospitalists, including supporting mentors' time. These examples highlight another key benchmark of a viable academic program, mentoring, and the importance of ensuring that mentors have time for this essential effort.

However, it is critically important to remember that all politics is local, to quote the late Tip O'Neill, long‐time speaker of the US House of Representatives. What works in one setting may not work as well in a different contexthence the need for creativity and political acumen.

For example, although the specialty‐group collaboration described by Flanders et al.1 may be helpful in one setting, other strategic alliances may work on a larger scale and over a longer time period. Most hospital medicine groups are currently within academic divisions of general internal medicine, where infrastructure and mentoring may already exist for both research and educational scholarship. In those cases, fostering interaction between the division's hospitalists and its researchers would be a critical first step. The programming and growth developed in this way can be leveraged to support ongoing academic activities by hospitalists rather than being limited to a single project.

In the Division of General Internal Medicine at Mount Sinai, which houses the academic hospitalists, building research entailed collaboration with the well‐established Departments of Geriatrics and Health Policy, which had preexisting research infrastructure and mentors. At the same time, we developed a research fellowship program by applying jointly with the Division of General Pediatrics for federal grant support. Such diversity of collaboration enhanced our application.

LOOKING AHEAD

Putting the academic into academic hospitalist programs is the key to the future of hospital medicine. To be successful, one must consider all the issues described in light of available resources and the local and federal political landscape. As Flanders et al.1 and Wright et al.2 emphasize, collaboration will be the main component for success in the current academic landscape.

Accelerating the development of clinical research in academic hospitalist programs is a worthwhile goal if pursued with clarity, objectivity, and a thorough understanding of the process and its implications. In their articles, Flanders et al.1 and Wright et al.2 identify major barriers to growing academic hospitalist programs. These barriers include the need for protected time, the shortage of trained research faculty, the lack of infrastructure, and the limited availability of senior mentors. Both Flanders et al. and Wright et al. offer smart and innovative ways of addressing these issues. However, building an academic program from the ground up is more complex and challenging than it may seem at first glance. It takes time, patience, creativity, diplomacy, and the ability to recruit collaborators and advocates who are willing to share infrastructure and resources.

Although both articles add significantly to the discussion of strategies for creating an academic hospitalist program, they are unclear about the definition of academic in this context. The term academic is often misunderstood to be synonymous with research. However, research is just one component of an academic program, which also includes education, quality improvement (QI), administration, and program development. It may be helpful, therefore, to replace academic with scholarship, which can be defined as a process that involves peer review and dissemination of ideas at local, regional, and national levels. Scholarship also goes beyond research, encompassing education and other areas such as QI. Although academic programs are not necessarily involved with funded research, there is usually an expectation of peer review, through either presentations at regional and national meetings or publication. For the purposes of this discussion, the term academic hospitalist program will be defined broadly to include any program affiliated with a university that is involved in the teaching of residents and medical students and whose faculty is required to participate in a promotions process.

All members of an academic division should be expected to participate in scholarship, whether it is education, QI projects, or research. If there is a strong expectation that traditional National Institute of Health (NIH) funded research will take place, this expectation must come with sufficient resources. Without infrastructure for research and investment in research faculty, procuring NIH funds for research is not a reasonable expectation. Organizers of hospitalist programs currently within academic divisions of general internal medicine should consider ways to better integrate programs into the existing research infrastructure in their divisions. For either freestanding hospitalist programs or programs within academic divisions of general internal medicine, investments in infrastructure and faculty are needed to nurture this area of research and build an academic focus in hospital medicine. However, if obtaining NIH research funds is not the expectation and resources are not available for hospitalists or for any other division or department at that institution, then academic expectations should focus on other pursuits. Examples include participation in the education and QI initiatives.

For programs with expectations of both funded research and other scholarship, a successful program will most likely include a small core of skilled clinical researchers working closely with well‐trained clinical educators, all of whom are involved in scholarship. Both clinical educators and researchers need to be continuously developing, and to reach their full potential, all should have access to infrastructure that supports these activities, including resources such as MPH‐level project managers, research assistants, database managers, and, most importantly, appropriate mentors.

Clinician educators must be both proficient clinicians and dedicated teachers. Ideally, they should have strong familiarity with educational theory in addition to skills in hands‐on teaching. Their responsibilities include mastering the skills that students need, staying up to the minute in their areas of expertise, and serving as role models in their attitudes toward patients, colleagues, and their work.3 Many hospitalists may not have these skills when they begin, often fresh out of residency, and will need help developing them.

PROTECTED TIME

Protected time is crucial to the success of any academic program. Finding this time presents a challenge for any clinical group, but the challenge is exacerbated for hospitalists, who face tremendous pressure to serve full‐time clinical jobs with little emphasis on academic elements such as education, QI, and participation in funded research. As both Wright et al.2 and Flanders et al.1 point out, to build a group of well‐developed academic clinician educators, academic hospitalists not on track for funding need to be given adequate protected time to participate in committees, sharpen their teaching skills, develop QI projects that can be converted to scholarships, participate in research, and present at national and regional meetings.

The requirements of protected time for researchers are more challenging than those for educators. Building a newly funded research unit within a hospitalist group, as with any group, will entail hiring fellowship trained faculty with significant protected time (approximately 80%) to give them time to obtain funding such as a K award and eventually become independent investigators with RO1 grant funding. Building research units requires support from collaborators with infrastructure and mentors already in place that can be tapped during the incubation stage of the academic program. For most hospitalist programs, infrastructure and mentors will be found in their divisions of general internal medicine.

Protected time should be considered an integral element of academic hospitalist positionsnot a perkas long as the time is used responsibly and productively. As both Wright et al.2 and Flanders et al.1 correctly point out, herein lies the major challenge of creating any kind of academic program: How will the program support protected time for both educators and researchers? In most instances, significant seed money will be needed to support junior faculty over the first few years of their careers. It is noteworthy that building a federally funded hospital medicine research program will be particularly difficult in today's economy because funding levels at the NIH, Agency for Healthcare Research and Quality, Health Resources and Services Administration, and other traditional funders of clinical research either are flat or have been reduced dramatically.

To many Academic Medical Centers (AMC), it may not be immediately obvious why a strong Academic Hospitalist program is in their economic best interest. Hospitalist programs may confront consistently high levels of turnover and a shrinking supply of general internists. The associated high costs of hiring new, junior faculty include the time and effort needed to interview, credential, train, and most importantly build familiarity with the complex systems encountered in maneuvering through a hospital, especially one with widespread dissemination of electronic medical records for documentation and order entry. However, hospitalists provided with opportunities for academic development are more likely to stay on the job longer and perform at a higher level, providing convincing motivation for hospitals to invest in their academic hospitalist programs. Retaining high‐quality hospitalists may be one of the most cost‐efficient methods for an AMC to support a hospital medicine program.

STRATEGIES IN ACTION

Flanders et al.1 point to a shortage of well‐trained clinician investigators with a focus on inpatient research as a barrier to the development of academic hospitalist programming. They describe a strategy of collaboration with specialty groups. This highlights the importance of collaboration with more well‐established research units as a key ingredient to building a new academic unit. Wright et al.2 describe their mentorship program and how they created protected time for scholarship for hospitalists, including supporting mentors' time. These examples highlight another key benchmark of a viable academic program, mentoring, and the importance of ensuring that mentors have time for this essential effort.

However, it is critically important to remember that all politics is local, to quote the late Tip O'Neill, long‐time speaker of the US House of Representatives. What works in one setting may not work as well in a different contexthence the need for creativity and political acumen.

For example, although the specialty‐group collaboration described by Flanders et al.1 may be helpful in one setting, other strategic alliances may work on a larger scale and over a longer time period. Most hospital medicine groups are currently within academic divisions of general internal medicine, where infrastructure and mentoring may already exist for both research and educational scholarship. In those cases, fostering interaction between the division's hospitalists and its researchers would be a critical first step. The programming and growth developed in this way can be leveraged to support ongoing academic activities by hospitalists rather than being limited to a single project.

In the Division of General Internal Medicine at Mount Sinai, which houses the academic hospitalists, building research entailed collaboration with the well‐established Departments of Geriatrics and Health Policy, which had preexisting research infrastructure and mentors. At the same time, we developed a research fellowship program by applying jointly with the Division of General Pediatrics for federal grant support. Such diversity of collaboration enhanced our application.

LOOKING AHEAD

Putting the academic into academic hospitalist programs is the key to the future of hospital medicine. To be successful, one must consider all the issues described in light of available resources and the local and federal political landscape. As Flanders et al.1 and Wright et al.2 emphasize, collaboration will be the main component for success in the current academic landscape.

Accelerating the development of clinical research in academic hospitalist programs is a worthwhile goal if pursued with clarity, objectivity, and a thorough understanding of the process and its implications. In their articles, Flanders et al.1 and Wright et al.2 identify major barriers to growing academic hospitalist programs. These barriers include the need for protected time, the shortage of trained research faculty, the lack of infrastructure, and the limited availability of senior mentors. Both Flanders et al. and Wright et al. offer smart and innovative ways of addressing these issues. However, building an academic program from the ground up is more complex and challenging than it may seem at first glance. It takes time, patience, creativity, diplomacy, and the ability to recruit collaborators and advocates who are willing to share infrastructure and resources.

Although both articles add significantly to the discussion of strategies for creating an academic hospitalist program, they are unclear about the definition of academic in this context. The term academic is often misunderstood to be synonymous with research. However, research is just one component of an academic program, which also includes education, quality improvement (QI), administration, and program development. It may be helpful, therefore, to replace academic with scholarship, which can be defined as a process that involves peer review and dissemination of ideas at local, regional, and national levels. Scholarship also goes beyond research, encompassing education and other areas such as QI. Although academic programs are not necessarily involved with funded research, there is usually an expectation of peer review, through either presentations at regional and national meetings or publication. For the purposes of this discussion, the term academic hospitalist program will be defined broadly to include any program affiliated with a university that is involved in the teaching of residents and medical students and whose faculty is required to participate in a promotions process.

All members of an academic division should be expected to participate in scholarship, whether it is education, QI projects, or research. If there is a strong expectation that traditional National Institute of Health (NIH) funded research will take place, this expectation must come with sufficient resources. Without infrastructure for research and investment in research faculty, procuring NIH funds for research is not a reasonable expectation. Organizers of hospitalist programs currently within academic divisions of general internal medicine should consider ways to better integrate programs into the existing research infrastructure in their divisions. For either freestanding hospitalist programs or programs within academic divisions of general internal medicine, investments in infrastructure and faculty are needed to nurture this area of research and build an academic focus in hospital medicine. However, if obtaining NIH research funds is not the expectation and resources are not available for hospitalists or for any other division or department at that institution, then academic expectations should focus on other pursuits. Examples include participation in the education and QI initiatives.

For programs with expectations of both funded research and other scholarship, a successful program will most likely include a small core of skilled clinical researchers working closely with well‐trained clinical educators, all of whom are involved in scholarship. Both clinical educators and researchers need to be continuously developing, and to reach their full potential, all should have access to infrastructure that supports these activities, including resources such as MPH‐level project managers, research assistants, database managers, and, most importantly, appropriate mentors.

Clinician educators must be both proficient clinicians and dedicated teachers. Ideally, they should have strong familiarity with educational theory in addition to skills in hands‐on teaching. Their responsibilities include mastering the skills that students need, staying up to the minute in their areas of expertise, and serving as role models in their attitudes toward patients, colleagues, and their work.3 Many hospitalists may not have these skills when they begin, often fresh out of residency, and will need help developing them.

PROTECTED TIME

Protected time is crucial to the success of any academic program. Finding this time presents a challenge for any clinical group, but the challenge is exacerbated for hospitalists, who face tremendous pressure to serve full‐time clinical jobs with little emphasis on academic elements such as education, QI, and participation in funded research. As both Wright et al.2 and Flanders et al.1 point out, to build a group of well‐developed academic clinician educators, academic hospitalists not on track for funding need to be given adequate protected time to participate in committees, sharpen their teaching skills, develop QI projects that can be converted to scholarships, participate in research, and present at national and regional meetings.

The requirements of protected time for researchers are more challenging than those for educators. Building a newly funded research unit within a hospitalist group, as with any group, will entail hiring fellowship trained faculty with significant protected time (approximately 80%) to give them time to obtain funding such as a K award and eventually become independent investigators with RO1 grant funding. Building research units requires support from collaborators with infrastructure and mentors already in place that can be tapped during the incubation stage of the academic program. For most hospitalist programs, infrastructure and mentors will be found in their divisions of general internal medicine.

Protected time should be considered an integral element of academic hospitalist positionsnot a perkas long as the time is used responsibly and productively. As both Wright et al.2 and Flanders et al.1 correctly point out, herein lies the major challenge of creating any kind of academic program: How will the program support protected time for both educators and researchers? In most instances, significant seed money will be needed to support junior faculty over the first few years of their careers. It is noteworthy that building a federally funded hospital medicine research program will be particularly difficult in today's economy because funding levels at the NIH, Agency for Healthcare Research and Quality, Health Resources and Services Administration, and other traditional funders of clinical research either are flat or have been reduced dramatically.

To many Academic Medical Centers (AMC), it may not be immediately obvious why a strong Academic Hospitalist program is in their economic best interest. Hospitalist programs may confront consistently high levels of turnover and a shrinking supply of general internists. The associated high costs of hiring new, junior faculty include the time and effort needed to interview, credential, train, and most importantly build familiarity with the complex systems encountered in maneuvering through a hospital, especially one with widespread dissemination of electronic medical records for documentation and order entry. However, hospitalists provided with opportunities for academic development are more likely to stay on the job longer and perform at a higher level, providing convincing motivation for hospitals to invest in their academic hospitalist programs. Retaining high‐quality hospitalists may be one of the most cost‐efficient methods for an AMC to support a hospital medicine program.

STRATEGIES IN ACTION

Flanders et al.1 point to a shortage of well‐trained clinician investigators with a focus on inpatient research as a barrier to the development of academic hospitalist programming. They describe a strategy of collaboration with specialty groups. This highlights the importance of collaboration with more well‐established research units as a key ingredient to building a new academic unit. Wright et al.2 describe their mentorship program and how they created protected time for scholarship for hospitalists, including supporting mentors' time. These examples highlight another key benchmark of a viable academic program, mentoring, and the importance of ensuring that mentors have time for this essential effort.

However, it is critically important to remember that all politics is local, to quote the late Tip O'Neill, long‐time speaker of the US House of Representatives. What works in one setting may not work as well in a different contexthence the need for creativity and political acumen.

For example, although the specialty‐group collaboration described by Flanders et al.1 may be helpful in one setting, other strategic alliances may work on a larger scale and over a longer time period. Most hospital medicine groups are currently within academic divisions of general internal medicine, where infrastructure and mentoring may already exist for both research and educational scholarship. In those cases, fostering interaction between the division's hospitalists and its researchers would be a critical first step. The programming and growth developed in this way can be leveraged to support ongoing academic activities by hospitalists rather than being limited to a single project.

In the Division of General Internal Medicine at Mount Sinai, which houses the academic hospitalists, building research entailed collaboration with the well‐established Departments of Geriatrics and Health Policy, which had preexisting research infrastructure and mentors. At the same time, we developed a research fellowship program by applying jointly with the Division of General Pediatrics for federal grant support. Such diversity of collaboration enhanced our application.

LOOKING AHEAD

Putting the academic into academic hospitalist programs is the key to the future of hospital medicine. To be successful, one must consider all the issues described in light of available resources and the local and federal political landscape. As Flanders et al.1 and Wright et al.2 emphasize, collaboration will be the main component for success in the current academic landscape.

Promotion through the ranks is the hallmark of success in academia. The support and infrastructure necessary to develop junior faculty members at academic medical centers may be inadequate.1, 2 Academic hospitalists are particularly vulnerable and at high risk for failure because of their heavy clinical commitment and limited time to pursue scholarly interests. Further, relatively few have pursued fellowship training, which means that many hospitalists must learn research‐related skills and the nuances of academia after joining the faculty.

Top‐notch mentors are believed to be integral to the success of the academic physician.36 Among other responsibilities, mentors (1) direct mentees toward promising opportunities, (2) serve as advocates for mentees, and (3) lend expertise to mentees' studies and scholarship. In general, there is concern that the cadre of talented, committed, and capable mentors is dwindling such that they are insufficient in number to satisfy and support the needs of the faculty.7, 8 In hospital medicine, experienced mentorship is particularly in short supply because the field is relatively new and there has been tremendous growth in the number of academic hospitalists, producing a large demand.

Like many hospitalist groups, our hospitalist division, the Collaborative Inpatient Medicine Service (CIMS), has experienced significant growth. It became apparent that the faculty needed and deserved a well‐designed academic support program to foster the development of skills necessary for academic success. The remainder of this article discusses our approach toward fulfilling these needs and the results to date.

DEVELOPING THE HOSPITALIST ACADEMIC SUPPORT PROGRAM

EVALUATION DATA

In the 2 years since implementation of the scholarly support program, individual faculty in the CIMS have been meeting the above‐mentioned goals. Specifically, with respect to acquiring knowledge and skills, 2 faculty members have completed their master's degrees, and 6 others have made use of select courses to augment their knowledge and skills. All faculty members (100%) have a scholarly project they are leading, and most have reached out to a colleague in the CIMS to assist them, such that nearly all are team members on at least 1 other scholarly project. Through informal mentoring sessions and a once‐yearly formal meeting related to academic promotion, all members (100%) of the faculty are aware of the expectations and requirements for promotion.

Table 1 shows the accomplishment of the 5 faculty members in the academic track who have been division members for 3 years or more. Among the 5 faculty in the academic track, publications and extramural funding are improving. In the 5 years before the initiative, CIMS faculty averaged approximately 0.5 publications per person per year; in the first 2 years of this initiative, that number has increased to 1.3 publications per person per year. The 1 physician who has not yet been published has completed projects and has several article in process. External funding (largely in the form of 3 extramural grants from private foundations) has increased dramatically from an average of 4% per FTE before the intervention to approximately 15% per FTE afterward. In addition, all faculty members have secured a source of additional funding to reduce their clinical efforts since the implementation of this program. One foundation funded project that involved all division members, whose goal was to develop mechanisms to improve the discharge process of elderly patients to their homes, won the award at the SGIM 2007 National Meeting for the best clinical innovation. As illustrated in Table 1, 1 of the founding CIMS members transferred out of the academic track in 2003 in alignment with this physician's personal and professional goals and preferences. Two faculty members have moved up an academic rank, and several others are poised to do so.

Thus, the divisional objectives (increasing number of publications, securing funding to increase the time devoted to scholarship, new leadership roles, and progression toward promotion) are being met as well.

CONCLUSIONS

Our rapidly growing hospitalist division recognized that several factors threatened the ability of the division and individuals to succeed academically. Divisional, departmental, and medical center leadership was committed to creating a supportive structure that would be available to all hospitalists as opposed to expecting each individual to unearth the necessary resources on their own. The innovative approach to foster individual, and therefore divisional, academic and scholarly success was designed around the following strategies: retention of an expert mentor (who is a not a hospitalist) and securing 20% of his time, investing in scholarship by protecting 30% nonclinical time for academic pursuits, and attempting to seek out fellowship‐trained hospitalists when hiring.

Although quality mentorship, protected time, and recruiting the best‐available talent to fill needs may not seem all that innovative, we believe the systematic approach to the problem and our steadfast application of the strategic plan is unique, innovative, and may present a model to be emulated by other divisions. Some may contend that it is impossible to protect 30% FTE of academic hospitalists indefinitely. Our group has made substantial investment in supporting the academic pursuits of our physicians, and we believe this is essential to maintaining their satisfaction and commitment to scholarship. This amount of protected time is offered to the entire physician faculty and continues even as our division has almost tripled in size. This initiative represents a carefully calculated investment that has influenced our ability to recruit and retain excellent people. Ongoing prospective study of this intervention over time will provide additional perspective on its value and shortcomings. Nonetheless, early data suggest that the plan is indeed working and that our group is satisfied with the return on investment to date.