A 67‐year‐old male presented to an outside hospital with a 1‐day history of fevers up to 39.4C, bilateral upper extremity weakness, and confusion. Forty‐eight hours prior to his presentation he had undergone uncomplicated bilateral carpal tunnel release surgery for the complaint of bilateral upper extremity paresthesias.

Bilateral carpal tunnel syndrome should prompt consideration of systemic diseases that infiltrate or impinge both canals (eg, rheumatoid arthritis, acromegaly, hypothyroidism, amyloidosis), although it is most frequently explained by a bilateral repetitive stress (eg, workplace typing). The development of upper extremity weakness suggests that an alternative condition such as cervical myelopathy, bilateral radiculopathy, or a rapidly progressive peripheral neuropathy may be responsible for his paresthesias. It would be unusual for a central nervous system process to selectively cause bilateral upper extremity weakness. Occasionally, patients emerge from surgery with limb weakness caused by peripheral nerve injury sustained from malpositioning of the extremity, but this would have been evident immediately following the operation.

Postoperative fevers are frequently unexplained, but require a search for common healthcare‐associated infections, such as pneumonia, urinary tract infection, intravenous catheter thrombophlebitis, wound infection, or Clostridium difficile colitis. However, such complications are unlikely following an ambulatory procedure. Confusion and fever together point to a central nervous system infection (meningoencephalitis or brain abscess) or a systemic infection that has impaired cognition. Malignancies can cause fever and altered mental status, but these are typically asynchronous events.

His past medical history was notable for hypertension, dyslipidemia, gout, actinic keratosis, and gastroesophageal reflux. His surgical history included bilateral knee replacements, repair of a left rotator cuff injury, and a herniorrhaphy. He was a nonsmoker who consumed 4 to 6 beers daily. His medications included clonidine, colchicine, atorvastatin, extended release metoprolol, triamterene‐hydrochlorothiazide, probenecid, and as‐needed ibuprofen and omeprazole.

Upon presentation he was cooperative and in no distress. Temperature was 38.9C, pulse 119 beats per minute, blood pressure 140/90 mm Hg, and oxygen saturation 94% on room air. He was noted to have logical thinking but impaired concentration. His upper extremity movement was restricted because of postoperative discomfort and swelling rather than true weakness. The rest of the exam was normal.

Metabolic, infectious, structural (intracranial), and toxic disorders can cause altered mental status. His heavy alcohol use puts him at risk for alcohol withdrawal and infections (such as Listeria meningitis), both of which may explain his fever and altered mental status. Signs and symptoms of meningitis are absent at this time. His knee prostheses could have harbored an infection preoperatively and therefore warrant close examination. Patients sometimes have adverse reactions to medications they have been prescribed but are not exposed to until hospitalization, although his surgical procedure was likely done on an outpatient basis. Empiric thiamine should be administered early given his confusion and alcohol habits.

Basic laboratories revealed a hemoglobin of 11.2 g/dL, white blood cell (WBC) count of 6,900/mm³ with 75% neutrophils, platelets of 206,000/mm³. Mean corpuscular volume was 97 mm³. Serum albumin was 2.4 g/dl, sodium 134 mmol/L, potassium 3.9 mmol/L, blood urea nitrogen 12 mg/dL, and creatinine 0.9 mg/dL. The aspartate aminotransferase was 93 U/L, alanine aminotransferase 73 U/L, alkaline phosphatase 254 U/L, and total bilirubin 1.0 mg/dL. Urinalysis was normal. Over the next 16 days fevers and waxing and waning mentation continued. The following studies were normal or negative: blood and urine cultures; transthoracic echocardiogram, antinuclear antibodies, hepatitis B surface antigen, hepatitis C antibody, and human immunodeficiency virus antibody; magnetic resonance imaging of the brain, electroencephalogram, and lower extremity venous ultrasound.

Hypoalbuminemia may signal chronic illness, hypoproduction from liver disease (caused by his heavy alcohol use), or losses from the kidney or gastrointestinal tract. His anemia may reflect chronic disease or point toward a specific underlying disorder. For example, fever and anemia could arise from hemolytic processes such as thrombotic thrombocytopenic purpura or clostridial infections.

An extensive workup has not revealed a cause for his prolonged fever (eg, infection, malignancy, autoimmune condition, or toxin). Likewise, an explanation for confusion is lacking. Because systemic illness and structural brain disease have not been uncovered, a lumbar puncture is indicated.

A lumbar puncture under fluoroscopic guidance revealed a cerebrospinal fluid (CSF) WBC count of 6/mm³, red blood cell count (RBC) 2255/mm³, protein 49 mg/dL, and glucose 54 mg/dL. The WBC differential was not reported. No growth was reported on bacterial cultures. Polymerase chain reactions for enterovirus and herpes simplex viruses 1 and 2 were negative. Cryptococcal antigen and Venereal Disease Research Laboratory serologies were also negative.

A CSF WBC count of 6 is out of the normal range, but could be explained by a traumatic tap given the elevated RBC; the protein and glucose are likewise at the border of normal. Collectively, these are nonspecific findings that could point to an infectious or noninfectious cause of intrathecal or paraspinous inflammation, but are not suggestive of bacterial meningitis.

The patient developed pneumonia, for which he received ertapenem. On hospital day 17 he was intubated for hypoxia and respiratory distress and was extubated after 4 days of mechanical ventilation. Increasing weakness in all extremities prompted magnetic resonance imaging of the spine, which revealed fluid and enhancement involving the soft tissues around C3‐C4 and C5‐C6, raising concerns for discitis and osteomyelitis. Possible septic arthritis at the C3‐C4 and C4‐C5 facets was noted. Ring enhancing fluid collections from T2‐T8 compatible with an epidural abscess with cord compression at T4‐T5 and T6‐T7 were seen. Enhancement and fluid involving the facet joints between T2‐T7 was also consistent with septic arthritis (Figure 1).

Figure 1

His pneumonia appears to have developed many days into his hospitalization, and therefore is unlikely to account for his initial fever and confusion. Blood cultures and echocardiogram have not suggested an endovascular infection that could account for such widespread vertebral and epidural deposition. A wide number of bacteria can cause epidural abscesses and septic arthritis, most commonly Staphylococcus aureus. Less common pathogens with a predilection for osteoarticular involvement, such as Brucella species, warrant consideration when there is appropriate epidemiologic risk.

Systemic bacterial infection remains a concern with his alcoholism rendering him partially immunosuppressed. However, a large number of adjacent spinal joints harboring a bacterial infection is unusual, and a working diagnosis of multilevel spinal infection, therefore, should prompt consideration of noninfectious processes. When a patient develops a swollen peripheral joint and fever in the postoperative setting, gout or pseudogout is a leading consideration. That same thinking should be applied to the vertebrae, where spinal gout can manifest. Surgery itself or associated changes in alcohol consumption patterns or changes in medications (at least 4 of which are relevant to goutcolchicine, hydrochlorothiazide, probenecid, and ibuprofen) could predispose him to a flare.

Aspiration of the epidural collection yielded a negative Gram stain and culture. He developed swelling in the bilateral proximal interphalangeal joints and was treated with steroids and colchicine for suspected gout flare. Vancomycin and piperacillin‐tazobactam were initiated, and on hospital day 22 the patient was transferred to another hospital for further evaluation by neurosurgery.

The negative Gram stain and culture argues against septic arthritis, but these are imperfect tests and will not detect atypical pathogens (eg, spinal tuberculosis). Reexamination of the aspirate for urate and calcium pyrophosphate crystals would be useful. Initiation of steroids in the setting of potentially undiagnosed infection requires a careful risk/benefit analysis. It may be reasonable to treat the patient with colchicine alone while withholding steroids and avoiding nonsteroidal agents in case invasive procedures are planned.

On exam his temperature was 36C, blood pressure 156/92 mm Hg, pulse 100 beats per minute, respirations 21 per minute, and oxygenation 97% on room air. He was not in acute distress and was only oriented to self. Bilateral 2+ lower extremity pitting edema up to the knees was noted. Examination of the heart and lungs was unremarkable. Gouty tophi were noted over both elbows. His joints were normal.

Cranial nerves IIXII were normal. Motor exam revealed normal muscle tone and bulk. Muscle strength was approximately 3/5 in the right upper extremity and 4+/5 in the left upper extremity. Bilateral lower extremity strength was 3/5 in hip flexion, knee flexion, and knee extension. Dorsiflexion and plantar flexion were approximately 2/5 bilaterally. Sensation was intact to light touch and pinprick, and proprioception was normal. Gait was not tested. A Foley catheter was in place.

This examination confirms ongoing encephalopathy and incomplete quadriplegia. The lower extremity weakness is nearly equal proximally and distally, which can be seen with an advanced peripheral neuropathy but is more characteristic of myelopathy. The expected concomitant sensory deficit of myelopathy is not present, although this may be difficult to detect in a confused patient. Reflex testing would help in distinguishing myelopathy (favored because of the imaging findings) from a rapid progressive peripheral motor neuropathy (eg, acute inflammatory demyelinating polyneuropathy or acute intermittent porphyria).

The pitting edema likely represents fluid overload, which can be nonspecific after prolonged immobility during hospitalization; hypoalbuminemia is oftentimes speculated to play a role when this develops. His alcohol use puts him at risk for heart failure (although there is no evidence of this on exam) and liver disease (which his liver function tests suggest). The tophi speak to the extent and chronicity of his hyperuricemia.

On arrival he reported recent onset diarrhea. Medications at transfer included metoprolol, omeprazole, prednisone, piperacillin/tazobactam, vancomycin, and colchicine; acetaminophen, bisacodyl, diphenhydramine, fentanyl, subcutaneous insulin, and labetalol were administered as needed. Laboratory studies included a hemoglobin of 9.5 g/dL, WBC count of 7,300/mm³ with 95% neutrophils, platelets 301,000/mm³, sodium 151 mmol/L, potassium 2.9 mmol/L, blood urea nitrogen 76 mg/dL, creatinine 2.0 mg/dL, aspartate aminotransferase 171 U/L, and alanine aminotransferase 127 U/L. Serum albumin was 1.7 g/dL.

At least 3 of his medicationsdiphenhydramine, fentanyl, and prednisonemay be contributing to his ongoing altered mental status, which may be further compounded by hypernatremia. Although his liver disease remains uncharacterized, hepatic encephalopathy may be contributing to his confusion as well.

Colchicine is likely responsible for his diarrhea, which would be the most readily available explanation for his hypernatremia, hypokalemia, and acute kidney injury (AKI). Acute kidney injury could result from progressive liver disease (hepatorenal syndrome), decreased arterial perfusion (suggested by third spacing or his diarrhea), acute tubular necrosis (from infection or medication), or urinary retention secondary to catheter obstruction. Acute hyperuricemia can also cause AKI (urate nephropathy).

Anemia has progressed and requires evaluation for blood loss as well as hemolysis. Hepatotoxicity from any of his medications (eg, acetaminophen) must be considered. Coagulation studies and review of the previous abdominal computed tomography would help determine the extent of his liver disease.

Neurosurgical consultation was obtained and the patient and his family elected to proceed with a thoracic laminectomy. Cheesy fluid was identified at the facet joints at T6‐T7, which was found to contain rare deposits of monosodium urate crystals. Surgical specimen cultures were sterile. His mental status and strength slowly improved to baseline following the surgery. He was discharged on postoperative day 7 to a rehabilitation facility. On the telephone follow‐up he reported that he has regained his strength completely.

The fluid analysis and clinical course confirms spinal gout. The presenting encephalopathy remains unexplained; I am unaware of gout leading to altered mental status.

COMMENTARY

Gout is an inflammatory condition triggered by the deposition of monosodium urate crystals in tissues in association with hyperuricemia.[1] Based on the 20072008 National Health and Nutrition Examination Survey, the prevalence of gout among US adults was 3.9% (8.3 million individuals).[2] These rates are increasing and are thought to be spurred by the aging population, increasing rates of obesity, and changing dietary habits including increases in the consumption of soft drinks and red meat.[3, 4, 5] The development of gout during hospitalization can prolong length of stay, and the implementation of a management protocol appears to help decrease treatment delays and the inappropriate discontinuation of gout prophylaxis.[6, 7] Surgery, with its associated physiologic stressors, can trigger gout, which is often polyarticular and presents with fever leading to testing and consultations for the febrile episode.[8]

Gout is an ancient disease that is familiar to most clinicians. In 1666, Daniel Sennert, a German physician, described gout as the physician's shame because of its infrequent recognition.[9] Clinical gout spans 3 stages: asymptomatic hyperuricemia, acute and intercritical gout, and chronic gouty arthritis. The typical acute presentation is monoarticular with the abrupt onset of pain, swelling, warmth, and erythema in a peripheral joint. It manifests most characteristically in the first metatarsophalangeal joint (podagra), but also frequently involves the midfoot, ankle, knee, and wrist and sometimes affects multiple joints simultaneously (polyarticular gout).[1, 10] The visualization of monosodium urate crystals either in synovial fluid or from a tophus is diagnostic of gout; however, guidelines recognize that a classic presentation of gout may be diagnosed based on clinical criteria alone.[11] Dual energy computerized tomography and ultrasonography are emerging as techniques for the visualization of monosodium urate crystals; however, they are not currently routinely recommended.[12]

There are many unusual presentations of gout, with an increase in such reports paralleling both the overall increase in the prevalence of gout and improvements in available imaging techniques.[13] Atypical presentations present diagnostic challenges and are often caused by tophaceous deposits in unusual locations. Reports of atypical gout have described entrapment neuropathies (eg, gouty deposits inducing carpal tunnel syndrome), ocular gout manifested as conjunctival deposits and uveitis, pancreatic gout presenting as a mass, and dermatologic manifestations including panniculitis.[13, 14]

Spinal gout (also known as axial gout) manifests when crystal‐induced inflammation, erosive arthritis, and tophaceous deposits occur along the spinal column. A cross‐sectional study of patients with poorly controlled gout reported the prevalence of spinal gout diagnosed by computerized tomography to be 35%. These radiographic findings were not consistently correlated with back pain.[15] Imaging features that are suggestive of spinal gout include intra‐articular and juxta‐articular erosions with sclerotic margins and density greater than the surrounding muscle. Periosteal new bone formation adjacent to bony destruction can form overhanging edges.[16] When retrospectively presented with the final diagnosis, the radiologist at our institution noted that the appearance was typical gout in an atypical location.

Spinal gout can be confused with spinal metastasis, infection, and stenosis. It can remain asymptomatic or present with back pain, radiculopathy, or cord compression. The lumbar spine is the most frequently affected site.[17, 18] Many patients with spinal gout have had chronic tophaceous gout with radiologic evidence of erosions in the peripheral joints.[15] Patients with spinal gout also have elevated urate levels and markers of inflammation.[18] Surgical decompression and stabilization is recommended when there is frank cord compression, progressive neurologic compromise, or lack of improvement with gout therapy alone.[18]

This patient's male gender, history of gout, hypertension, alcohol consumption, and thiazide diuretic use placed him at an increased risk of a gout attack.[19, 20] The possible interruption of urate‐lowering therapy for the surgical procedure and surgery itself further heightened his risk of suffering acute gouty arthritis in the perioperative period.[21] The patient's encephalopathy may have masked back pain and precluded an accurate neurologic exam. There is one case report to our knowledge describing encephalopathy that improved with colchicine and was possibly related to gout.[22] This patient's encephalopathy was deemed multifactorial and attributed to alcohol withdrawal, medications (including opioids and steroids), and infection (pneumonia).

Gout is best known for its peripheral arthritis and is rarely invoked in the consideration of spinal and myelopathic processes where more pressing competing diagnoses, such as infection and malignancy, are typically considered. In addition, when surgical specimens are submitted for examination for pathology in formaldehyde (rather than alcohol), monosodium urate crystals are dissolved and are thus difficult to identify in the specimen.

This case reminds us that gout remains a diagnostic challenge and should be considered in the differential of an inflammatory process. Recognition of the multifaceted nature of gout can allow for the earlier recognition and treatment of the less typical presentations of this ancient malady.

KEY TEACHING POINTS

1. Crystalline disease is a common cause of postoperative arthritis.
2. Gout (and pseudogout) should be considered in cases of focal inflammation (detected by examination or imaging) when the evidence or predisposition for infection is limited or nonexistent.
3. Spinal gout presents with back pain, radiculopathy, or cord compression and may be confused with spinal metastasis, infection, and stenosis.

A 67‐year‐old male presented to an outside hospital with a 1‐day history of fevers up to 39.4C, bilateral upper extremity weakness, and confusion. Forty‐eight hours prior to his presentation he had undergone uncomplicated bilateral carpal tunnel release surgery for the complaint of bilateral upper extremity paresthesias.

Bilateral carpal tunnel syndrome should prompt consideration of systemic diseases that infiltrate or impinge both canals (eg, rheumatoid arthritis, acromegaly, hypothyroidism, amyloidosis), although it is most frequently explained by a bilateral repetitive stress (eg, workplace typing). The development of upper extremity weakness suggests that an alternative condition such as cervical myelopathy, bilateral radiculopathy, or a rapidly progressive peripheral neuropathy may be responsible for his paresthesias. It would be unusual for a central nervous system process to selectively cause bilateral upper extremity weakness. Occasionally, patients emerge from surgery with limb weakness caused by peripheral nerve injury sustained from malpositioning of the extremity, but this would have been evident immediately following the operation.

Postoperative fevers are frequently unexplained, but require a search for common healthcare‐associated infections, such as pneumonia, urinary tract infection, intravenous catheter thrombophlebitis, wound infection, or Clostridium difficile colitis. However, such complications are unlikely following an ambulatory procedure. Confusion and fever together point to a central nervous system infection (meningoencephalitis or brain abscess) or a systemic infection that has impaired cognition. Malignancies can cause fever and altered mental status, but these are typically asynchronous events.

His past medical history was notable for hypertension, dyslipidemia, gout, actinic keratosis, and gastroesophageal reflux. His surgical history included bilateral knee replacements, repair of a left rotator cuff injury, and a herniorrhaphy. He was a nonsmoker who consumed 4 to 6 beers daily. His medications included clonidine, colchicine, atorvastatin, extended release metoprolol, triamterene‐hydrochlorothiazide, probenecid, and as‐needed ibuprofen and omeprazole.

Upon presentation he was cooperative and in no distress. Temperature was 38.9C, pulse 119 beats per minute, blood pressure 140/90 mm Hg, and oxygen saturation 94% on room air. He was noted to have logical thinking but impaired concentration. His upper extremity movement was restricted because of postoperative discomfort and swelling rather than true weakness. The rest of the exam was normal.

Metabolic, infectious, structural (intracranial), and toxic disorders can cause altered mental status. His heavy alcohol use puts him at risk for alcohol withdrawal and infections (such as Listeria meningitis), both of which may explain his fever and altered mental status. Signs and symptoms of meningitis are absent at this time. His knee prostheses could have harbored an infection preoperatively and therefore warrant close examination. Patients sometimes have adverse reactions to medications they have been prescribed but are not exposed to until hospitalization, although his surgical procedure was likely done on an outpatient basis. Empiric thiamine should be administered early given his confusion and alcohol habits.

Basic laboratories revealed a hemoglobin of 11.2 g/dL, white blood cell (WBC) count of 6,900/mm³ with 75% neutrophils, platelets of 206,000/mm³. Mean corpuscular volume was 97 mm³. Serum albumin was 2.4 g/dl, sodium 134 mmol/L, potassium 3.9 mmol/L, blood urea nitrogen 12 mg/dL, and creatinine 0.9 mg/dL. The aspartate aminotransferase was 93 U/L, alanine aminotransferase 73 U/L, alkaline phosphatase 254 U/L, and total bilirubin 1.0 mg/dL. Urinalysis was normal. Over the next 16 days fevers and waxing and waning mentation continued. The following studies were normal or negative: blood and urine cultures; transthoracic echocardiogram, antinuclear antibodies, hepatitis B surface antigen, hepatitis C antibody, and human immunodeficiency virus antibody; magnetic resonance imaging of the brain, electroencephalogram, and lower extremity venous ultrasound.

Hypoalbuminemia may signal chronic illness, hypoproduction from liver disease (caused by his heavy alcohol use), or losses from the kidney or gastrointestinal tract. His anemia may reflect chronic disease or point toward a specific underlying disorder. For example, fever and anemia could arise from hemolytic processes such as thrombotic thrombocytopenic purpura or clostridial infections.

An extensive workup has not revealed a cause for his prolonged fever (eg, infection, malignancy, autoimmune condition, or toxin). Likewise, an explanation for confusion is lacking. Because systemic illness and structural brain disease have not been uncovered, a lumbar puncture is indicated.

A lumbar puncture under fluoroscopic guidance revealed a cerebrospinal fluid (CSF) WBC count of 6/mm³, red blood cell count (RBC) 2255/mm³, protein 49 mg/dL, and glucose 54 mg/dL. The WBC differential was not reported. No growth was reported on bacterial cultures. Polymerase chain reactions for enterovirus and herpes simplex viruses 1 and 2 were negative. Cryptococcal antigen and Venereal Disease Research Laboratory serologies were also negative.

A CSF WBC count of 6 is out of the normal range, but could be explained by a traumatic tap given the elevated RBC; the protein and glucose are likewise at the border of normal. Collectively, these are nonspecific findings that could point to an infectious or noninfectious cause of intrathecal or paraspinous inflammation, but are not suggestive of bacterial meningitis.

The patient developed pneumonia, for which he received ertapenem. On hospital day 17 he was intubated for hypoxia and respiratory distress and was extubated after 4 days of mechanical ventilation. Increasing weakness in all extremities prompted magnetic resonance imaging of the spine, which revealed fluid and enhancement involving the soft tissues around C3‐C4 and C5‐C6, raising concerns for discitis and osteomyelitis. Possible septic arthritis at the C3‐C4 and C4‐C5 facets was noted. Ring enhancing fluid collections from T2‐T8 compatible with an epidural abscess with cord compression at T4‐T5 and T6‐T7 were seen. Enhancement and fluid involving the facet joints between T2‐T7 was also consistent with septic arthritis (Figure 1).

Figure 1

His pneumonia appears to have developed many days into his hospitalization, and therefore is unlikely to account for his initial fever and confusion. Blood cultures and echocardiogram have not suggested an endovascular infection that could account for such widespread vertebral and epidural deposition. A wide number of bacteria can cause epidural abscesses and septic arthritis, most commonly Staphylococcus aureus. Less common pathogens with a predilection for osteoarticular involvement, such as Brucella species, warrant consideration when there is appropriate epidemiologic risk.

Systemic bacterial infection remains a concern with his alcoholism rendering him partially immunosuppressed. However, a large number of adjacent spinal joints harboring a bacterial infection is unusual, and a working diagnosis of multilevel spinal infection, therefore, should prompt consideration of noninfectious processes. When a patient develops a swollen peripheral joint and fever in the postoperative setting, gout or pseudogout is a leading consideration. That same thinking should be applied to the vertebrae, where spinal gout can manifest. Surgery itself or associated changes in alcohol consumption patterns or changes in medications (at least 4 of which are relevant to goutcolchicine, hydrochlorothiazide, probenecid, and ibuprofen) could predispose him to a flare.

Aspiration of the epidural collection yielded a negative Gram stain and culture. He developed swelling in the bilateral proximal interphalangeal joints and was treated with steroids and colchicine for suspected gout flare. Vancomycin and piperacillin‐tazobactam were initiated, and on hospital day 22 the patient was transferred to another hospital for further evaluation by neurosurgery.

The negative Gram stain and culture argues against septic arthritis, but these are imperfect tests and will not detect atypical pathogens (eg, spinal tuberculosis). Reexamination of the aspirate for urate and calcium pyrophosphate crystals would be useful. Initiation of steroids in the setting of potentially undiagnosed infection requires a careful risk/benefit analysis. It may be reasonable to treat the patient with colchicine alone while withholding steroids and avoiding nonsteroidal agents in case invasive procedures are planned.

On exam his temperature was 36C, blood pressure 156/92 mm Hg, pulse 100 beats per minute, respirations 21 per minute, and oxygenation 97% on room air. He was not in acute distress and was only oriented to self. Bilateral 2+ lower extremity pitting edema up to the knees was noted. Examination of the heart and lungs was unremarkable. Gouty tophi were noted over both elbows. His joints were normal.

Cranial nerves IIXII were normal. Motor exam revealed normal muscle tone and bulk. Muscle strength was approximately 3/5 in the right upper extremity and 4+/5 in the left upper extremity. Bilateral lower extremity strength was 3/5 in hip flexion, knee flexion, and knee extension. Dorsiflexion and plantar flexion were approximately 2/5 bilaterally. Sensation was intact to light touch and pinprick, and proprioception was normal. Gait was not tested. A Foley catheter was in place.

This examination confirms ongoing encephalopathy and incomplete quadriplegia. The lower extremity weakness is nearly equal proximally and distally, which can be seen with an advanced peripheral neuropathy but is more characteristic of myelopathy. The expected concomitant sensory deficit of myelopathy is not present, although this may be difficult to detect in a confused patient. Reflex testing would help in distinguishing myelopathy (favored because of the imaging findings) from a rapid progressive peripheral motor neuropathy (eg, acute inflammatory demyelinating polyneuropathy or acute intermittent porphyria).

The pitting edema likely represents fluid overload, which can be nonspecific after prolonged immobility during hospitalization; hypoalbuminemia is oftentimes speculated to play a role when this develops. His alcohol use puts him at risk for heart failure (although there is no evidence of this on exam) and liver disease (which his liver function tests suggest). The tophi speak to the extent and chronicity of his hyperuricemia.

On arrival he reported recent onset diarrhea. Medications at transfer included metoprolol, omeprazole, prednisone, piperacillin/tazobactam, vancomycin, and colchicine; acetaminophen, bisacodyl, diphenhydramine, fentanyl, subcutaneous insulin, and labetalol were administered as needed. Laboratory studies included a hemoglobin of 9.5 g/dL, WBC count of 7,300/mm³ with 95% neutrophils, platelets 301,000/mm³, sodium 151 mmol/L, potassium 2.9 mmol/L, blood urea nitrogen 76 mg/dL, creatinine 2.0 mg/dL, aspartate aminotransferase 171 U/L, and alanine aminotransferase 127 U/L. Serum albumin was 1.7 g/dL.

At least 3 of his medicationsdiphenhydramine, fentanyl, and prednisonemay be contributing to his ongoing altered mental status, which may be further compounded by hypernatremia. Although his liver disease remains uncharacterized, hepatic encephalopathy may be contributing to his confusion as well.

Colchicine is likely responsible for his diarrhea, which would be the most readily available explanation for his hypernatremia, hypokalemia, and acute kidney injury (AKI). Acute kidney injury could result from progressive liver disease (hepatorenal syndrome), decreased arterial perfusion (suggested by third spacing or his diarrhea), acute tubular necrosis (from infection or medication), or urinary retention secondary to catheter obstruction. Acute hyperuricemia can also cause AKI (urate nephropathy).

Anemia has progressed and requires evaluation for blood loss as well as hemolysis. Hepatotoxicity from any of his medications (eg, acetaminophen) must be considered. Coagulation studies and review of the previous abdominal computed tomography would help determine the extent of his liver disease.

Neurosurgical consultation was obtained and the patient and his family elected to proceed with a thoracic laminectomy. Cheesy fluid was identified at the facet joints at T6‐T7, which was found to contain rare deposits of monosodium urate crystals. Surgical specimen cultures were sterile. His mental status and strength slowly improved to baseline following the surgery. He was discharged on postoperative day 7 to a rehabilitation facility. On the telephone follow‐up he reported that he has regained his strength completely.

The fluid analysis and clinical course confirms spinal gout. The presenting encephalopathy remains unexplained; I am unaware of gout leading to altered mental status.

COMMENTARY

Gout is an inflammatory condition triggered by the deposition of monosodium urate crystals in tissues in association with hyperuricemia.[1] Based on the 20072008 National Health and Nutrition Examination Survey, the prevalence of gout among US adults was 3.9% (8.3 million individuals).[2] These rates are increasing and are thought to be spurred by the aging population, increasing rates of obesity, and changing dietary habits including increases in the consumption of soft drinks and red meat.[3, 4, 5] The development of gout during hospitalization can prolong length of stay, and the implementation of a management protocol appears to help decrease treatment delays and the inappropriate discontinuation of gout prophylaxis.[6, 7] Surgery, with its associated physiologic stressors, can trigger gout, which is often polyarticular and presents with fever leading to testing and consultations for the febrile episode.[8]

Gout is an ancient disease that is familiar to most clinicians. In 1666, Daniel Sennert, a German physician, described gout as the physician's shame because of its infrequent recognition.[9] Clinical gout spans 3 stages: asymptomatic hyperuricemia, acute and intercritical gout, and chronic gouty arthritis. The typical acute presentation is monoarticular with the abrupt onset of pain, swelling, warmth, and erythema in a peripheral joint. It manifests most characteristically in the first metatarsophalangeal joint (podagra), but also frequently involves the midfoot, ankle, knee, and wrist and sometimes affects multiple joints simultaneously (polyarticular gout).[1, 10] The visualization of monosodium urate crystals either in synovial fluid or from a tophus is diagnostic of gout; however, guidelines recognize that a classic presentation of gout may be diagnosed based on clinical criteria alone.[11] Dual energy computerized tomography and ultrasonography are emerging as techniques for the visualization of monosodium urate crystals; however, they are not currently routinely recommended.[12]

There are many unusual presentations of gout, with an increase in such reports paralleling both the overall increase in the prevalence of gout and improvements in available imaging techniques.[13] Atypical presentations present diagnostic challenges and are often caused by tophaceous deposits in unusual locations. Reports of atypical gout have described entrapment neuropathies (eg, gouty deposits inducing carpal tunnel syndrome), ocular gout manifested as conjunctival deposits and uveitis, pancreatic gout presenting as a mass, and dermatologic manifestations including panniculitis.[13, 14]

Spinal gout (also known as axial gout) manifests when crystal‐induced inflammation, erosive arthritis, and tophaceous deposits occur along the spinal column. A cross‐sectional study of patients with poorly controlled gout reported the prevalence of spinal gout diagnosed by computerized tomography to be 35%. These radiographic findings were not consistently correlated with back pain.[15] Imaging features that are suggestive of spinal gout include intra‐articular and juxta‐articular erosions with sclerotic margins and density greater than the surrounding muscle. Periosteal new bone formation adjacent to bony destruction can form overhanging edges.[16] When retrospectively presented with the final diagnosis, the radiologist at our institution noted that the appearance was typical gout in an atypical location.

Spinal gout can be confused with spinal metastasis, infection, and stenosis. It can remain asymptomatic or present with back pain, radiculopathy, or cord compression. The lumbar spine is the most frequently affected site.[17, 18] Many patients with spinal gout have had chronic tophaceous gout with radiologic evidence of erosions in the peripheral joints.[15] Patients with spinal gout also have elevated urate levels and markers of inflammation.[18] Surgical decompression and stabilization is recommended when there is frank cord compression, progressive neurologic compromise, or lack of improvement with gout therapy alone.[18]

This patient's male gender, history of gout, hypertension, alcohol consumption, and thiazide diuretic use placed him at an increased risk of a gout attack.[19, 20] The possible interruption of urate‐lowering therapy for the surgical procedure and surgery itself further heightened his risk of suffering acute gouty arthritis in the perioperative period.[21] The patient's encephalopathy may have masked back pain and precluded an accurate neurologic exam. There is one case report to our knowledge describing encephalopathy that improved with colchicine and was possibly related to gout.[22] This patient's encephalopathy was deemed multifactorial and attributed to alcohol withdrawal, medications (including opioids and steroids), and infection (pneumonia).

Gout is best known for its peripheral arthritis and is rarely invoked in the consideration of spinal and myelopathic processes where more pressing competing diagnoses, such as infection and malignancy, are typically considered. In addition, when surgical specimens are submitted for examination for pathology in formaldehyde (rather than alcohol), monosodium urate crystals are dissolved and are thus difficult to identify in the specimen.

This case reminds us that gout remains a diagnostic challenge and should be considered in the differential of an inflammatory process. Recognition of the multifaceted nature of gout can allow for the earlier recognition and treatment of the less typical presentations of this ancient malady.

KEY TEACHING POINTS

1. Crystalline disease is a common cause of postoperative arthritis.
2. Gout (and pseudogout) should be considered in cases of focal inflammation (detected by examination or imaging) when the evidence or predisposition for infection is limited or nonexistent.
3. Spinal gout presents with back pain, radiculopathy, or cord compression and may be confused with spinal metastasis, infection, and stenosis.

A 67‐year‐old male presented to an outside hospital with a 1‐day history of fevers up to 39.4C, bilateral upper extremity weakness, and confusion. Forty‐eight hours prior to his presentation he had undergone uncomplicated bilateral carpal tunnel release surgery for the complaint of bilateral upper extremity paresthesias.

Bilateral carpal tunnel syndrome should prompt consideration of systemic diseases that infiltrate or impinge both canals (eg, rheumatoid arthritis, acromegaly, hypothyroidism, amyloidosis), although it is most frequently explained by a bilateral repetitive stress (eg, workplace typing). The development of upper extremity weakness suggests that an alternative condition such as cervical myelopathy, bilateral radiculopathy, or a rapidly progressive peripheral neuropathy may be responsible for his paresthesias. It would be unusual for a central nervous system process to selectively cause bilateral upper extremity weakness. Occasionally, patients emerge from surgery with limb weakness caused by peripheral nerve injury sustained from malpositioning of the extremity, but this would have been evident immediately following the operation.

Postoperative fevers are frequently unexplained, but require a search for common healthcare‐associated infections, such as pneumonia, urinary tract infection, intravenous catheter thrombophlebitis, wound infection, or Clostridium difficile colitis. However, such complications are unlikely following an ambulatory procedure. Confusion and fever together point to a central nervous system infection (meningoencephalitis or brain abscess) or a systemic infection that has impaired cognition. Malignancies can cause fever and altered mental status, but these are typically asynchronous events.

His past medical history was notable for hypertension, dyslipidemia, gout, actinic keratosis, and gastroesophageal reflux. His surgical history included bilateral knee replacements, repair of a left rotator cuff injury, and a herniorrhaphy. He was a nonsmoker who consumed 4 to 6 beers daily. His medications included clonidine, colchicine, atorvastatin, extended release metoprolol, triamterene‐hydrochlorothiazide, probenecid, and as‐needed ibuprofen and omeprazole.

Upon presentation he was cooperative and in no distress. Temperature was 38.9C, pulse 119 beats per minute, blood pressure 140/90 mm Hg, and oxygen saturation 94% on room air. He was noted to have logical thinking but impaired concentration. His upper extremity movement was restricted because of postoperative discomfort and swelling rather than true weakness. The rest of the exam was normal.

Metabolic, infectious, structural (intracranial), and toxic disorders can cause altered mental status. His heavy alcohol use puts him at risk for alcohol withdrawal and infections (such as Listeria meningitis), both of which may explain his fever and altered mental status. Signs and symptoms of meningitis are absent at this time. His knee prostheses could have harbored an infection preoperatively and therefore warrant close examination. Patients sometimes have adverse reactions to medications they have been prescribed but are not exposed to until hospitalization, although his surgical procedure was likely done on an outpatient basis. Empiric thiamine should be administered early given his confusion and alcohol habits.

Basic laboratories revealed a hemoglobin of 11.2 g/dL, white blood cell (WBC) count of 6,900/mm³ with 75% neutrophils, platelets of 206,000/mm³. Mean corpuscular volume was 97 mm³. Serum albumin was 2.4 g/dl, sodium 134 mmol/L, potassium 3.9 mmol/L, blood urea nitrogen 12 mg/dL, and creatinine 0.9 mg/dL. The aspartate aminotransferase was 93 U/L, alanine aminotransferase 73 U/L, alkaline phosphatase 254 U/L, and total bilirubin 1.0 mg/dL. Urinalysis was normal. Over the next 16 days fevers and waxing and waning mentation continued. The following studies were normal or negative: blood and urine cultures; transthoracic echocardiogram, antinuclear antibodies, hepatitis B surface antigen, hepatitis C antibody, and human immunodeficiency virus antibody; magnetic resonance imaging of the brain, electroencephalogram, and lower extremity venous ultrasound.

Hypoalbuminemia may signal chronic illness, hypoproduction from liver disease (caused by his heavy alcohol use), or losses from the kidney or gastrointestinal tract. His anemia may reflect chronic disease or point toward a specific underlying disorder. For example, fever and anemia could arise from hemolytic processes such as thrombotic thrombocytopenic purpura or clostridial infections.

An extensive workup has not revealed a cause for his prolonged fever (eg, infection, malignancy, autoimmune condition, or toxin). Likewise, an explanation for confusion is lacking. Because systemic illness and structural brain disease have not been uncovered, a lumbar puncture is indicated.

A lumbar puncture under fluoroscopic guidance revealed a cerebrospinal fluid (CSF) WBC count of 6/mm³, red blood cell count (RBC) 2255/mm³, protein 49 mg/dL, and glucose 54 mg/dL. The WBC differential was not reported. No growth was reported on bacterial cultures. Polymerase chain reactions for enterovirus and herpes simplex viruses 1 and 2 were negative. Cryptococcal antigen and Venereal Disease Research Laboratory serologies were also negative.

A CSF WBC count of 6 is out of the normal range, but could be explained by a traumatic tap given the elevated RBC; the protein and glucose are likewise at the border of normal. Collectively, these are nonspecific findings that could point to an infectious or noninfectious cause of intrathecal or paraspinous inflammation, but are not suggestive of bacterial meningitis.

The patient developed pneumonia, for which he received ertapenem. On hospital day 17 he was intubated for hypoxia and respiratory distress and was extubated after 4 days of mechanical ventilation. Increasing weakness in all extremities prompted magnetic resonance imaging of the spine, which revealed fluid and enhancement involving the soft tissues around C3‐C4 and C5‐C6, raising concerns for discitis and osteomyelitis. Possible septic arthritis at the C3‐C4 and C4‐C5 facets was noted. Ring enhancing fluid collections from T2‐T8 compatible with an epidural abscess with cord compression at T4‐T5 and T6‐T7 were seen. Enhancement and fluid involving the facet joints between T2‐T7 was also consistent with septic arthritis (Figure 1).

Figure 1

His pneumonia appears to have developed many days into his hospitalization, and therefore is unlikely to account for his initial fever and confusion. Blood cultures and echocardiogram have not suggested an endovascular infection that could account for such widespread vertebral and epidural deposition. A wide number of bacteria can cause epidural abscesses and septic arthritis, most commonly Staphylococcus aureus. Less common pathogens with a predilection for osteoarticular involvement, such as Brucella species, warrant consideration when there is appropriate epidemiologic risk.

Systemic bacterial infection remains a concern with his alcoholism rendering him partially immunosuppressed. However, a large number of adjacent spinal joints harboring a bacterial infection is unusual, and a working diagnosis of multilevel spinal infection, therefore, should prompt consideration of noninfectious processes. When a patient develops a swollen peripheral joint and fever in the postoperative setting, gout or pseudogout is a leading consideration. That same thinking should be applied to the vertebrae, where spinal gout can manifest. Surgery itself or associated changes in alcohol consumption patterns or changes in medications (at least 4 of which are relevant to goutcolchicine, hydrochlorothiazide, probenecid, and ibuprofen) could predispose him to a flare.

Aspiration of the epidural collection yielded a negative Gram stain and culture. He developed swelling in the bilateral proximal interphalangeal joints and was treated with steroids and colchicine for suspected gout flare. Vancomycin and piperacillin‐tazobactam were initiated, and on hospital day 22 the patient was transferred to another hospital for further evaluation by neurosurgery.

The negative Gram stain and culture argues against septic arthritis, but these are imperfect tests and will not detect atypical pathogens (eg, spinal tuberculosis). Reexamination of the aspirate for urate and calcium pyrophosphate crystals would be useful. Initiation of steroids in the setting of potentially undiagnosed infection requires a careful risk/benefit analysis. It may be reasonable to treat the patient with colchicine alone while withholding steroids and avoiding nonsteroidal agents in case invasive procedures are planned.

On exam his temperature was 36C, blood pressure 156/92 mm Hg, pulse 100 beats per minute, respirations 21 per minute, and oxygenation 97% on room air. He was not in acute distress and was only oriented to self. Bilateral 2+ lower extremity pitting edema up to the knees was noted. Examination of the heart and lungs was unremarkable. Gouty tophi were noted over both elbows. His joints were normal.

Cranial nerves IIXII were normal. Motor exam revealed normal muscle tone and bulk. Muscle strength was approximately 3/5 in the right upper extremity and 4+/5 in the left upper extremity. Bilateral lower extremity strength was 3/5 in hip flexion, knee flexion, and knee extension. Dorsiflexion and plantar flexion were approximately 2/5 bilaterally. Sensation was intact to light touch and pinprick, and proprioception was normal. Gait was not tested. A Foley catheter was in place.

This examination confirms ongoing encephalopathy and incomplete quadriplegia. The lower extremity weakness is nearly equal proximally and distally, which can be seen with an advanced peripheral neuropathy but is more characteristic of myelopathy. The expected concomitant sensory deficit of myelopathy is not present, although this may be difficult to detect in a confused patient. Reflex testing would help in distinguishing myelopathy (favored because of the imaging findings) from a rapid progressive peripheral motor neuropathy (eg, acute inflammatory demyelinating polyneuropathy or acute intermittent porphyria).

The pitting edema likely represents fluid overload, which can be nonspecific after prolonged immobility during hospitalization; hypoalbuminemia is oftentimes speculated to play a role when this develops. His alcohol use puts him at risk for heart failure (although there is no evidence of this on exam) and liver disease (which his liver function tests suggest). The tophi speak to the extent and chronicity of his hyperuricemia.

On arrival he reported recent onset diarrhea. Medications at transfer included metoprolol, omeprazole, prednisone, piperacillin/tazobactam, vancomycin, and colchicine; acetaminophen, bisacodyl, diphenhydramine, fentanyl, subcutaneous insulin, and labetalol were administered as needed. Laboratory studies included a hemoglobin of 9.5 g/dL, WBC count of 7,300/mm³ with 95% neutrophils, platelets 301,000/mm³, sodium 151 mmol/L, potassium 2.9 mmol/L, blood urea nitrogen 76 mg/dL, creatinine 2.0 mg/dL, aspartate aminotransferase 171 U/L, and alanine aminotransferase 127 U/L. Serum albumin was 1.7 g/dL.

At least 3 of his medicationsdiphenhydramine, fentanyl, and prednisonemay be contributing to his ongoing altered mental status, which may be further compounded by hypernatremia. Although his liver disease remains uncharacterized, hepatic encephalopathy may be contributing to his confusion as well.

Colchicine is likely responsible for his diarrhea, which would be the most readily available explanation for his hypernatremia, hypokalemia, and acute kidney injury (AKI). Acute kidney injury could result from progressive liver disease (hepatorenal syndrome), decreased arterial perfusion (suggested by third spacing or his diarrhea), acute tubular necrosis (from infection or medication), or urinary retention secondary to catheter obstruction. Acute hyperuricemia can also cause AKI (urate nephropathy).

Anemia has progressed and requires evaluation for blood loss as well as hemolysis. Hepatotoxicity from any of his medications (eg, acetaminophen) must be considered. Coagulation studies and review of the previous abdominal computed tomography would help determine the extent of his liver disease.

Neurosurgical consultation was obtained and the patient and his family elected to proceed with a thoracic laminectomy. Cheesy fluid was identified at the facet joints at T6‐T7, which was found to contain rare deposits of monosodium urate crystals. Surgical specimen cultures were sterile. His mental status and strength slowly improved to baseline following the surgery. He was discharged on postoperative day 7 to a rehabilitation facility. On the telephone follow‐up he reported that he has regained his strength completely.

The fluid analysis and clinical course confirms spinal gout. The presenting encephalopathy remains unexplained; I am unaware of gout leading to altered mental status.

COMMENTARY

Gout is an inflammatory condition triggered by the deposition of monosodium urate crystals in tissues in association with hyperuricemia.[1] Based on the 20072008 National Health and Nutrition Examination Survey, the prevalence of gout among US adults was 3.9% (8.3 million individuals).[2] These rates are increasing and are thought to be spurred by the aging population, increasing rates of obesity, and changing dietary habits including increases in the consumption of soft drinks and red meat.[3, 4, 5] The development of gout during hospitalization can prolong length of stay, and the implementation of a management protocol appears to help decrease treatment delays and the inappropriate discontinuation of gout prophylaxis.[6, 7] Surgery, with its associated physiologic stressors, can trigger gout, which is often polyarticular and presents with fever leading to testing and consultations for the febrile episode.[8]

Gout is an ancient disease that is familiar to most clinicians. In 1666, Daniel Sennert, a German physician, described gout as the physician's shame because of its infrequent recognition.[9] Clinical gout spans 3 stages: asymptomatic hyperuricemia, acute and intercritical gout, and chronic gouty arthritis. The typical acute presentation is monoarticular with the abrupt onset of pain, swelling, warmth, and erythema in a peripheral joint. It manifests most characteristically in the first metatarsophalangeal joint (podagra), but also frequently involves the midfoot, ankle, knee, and wrist and sometimes affects multiple joints simultaneously (polyarticular gout).[1, 10] The visualization of monosodium urate crystals either in synovial fluid or from a tophus is diagnostic of gout; however, guidelines recognize that a classic presentation of gout may be diagnosed based on clinical criteria alone.[11] Dual energy computerized tomography and ultrasonography are emerging as techniques for the visualization of monosodium urate crystals; however, they are not currently routinely recommended.[12]

There are many unusual presentations of gout, with an increase in such reports paralleling both the overall increase in the prevalence of gout and improvements in available imaging techniques.[13] Atypical presentations present diagnostic challenges and are often caused by tophaceous deposits in unusual locations. Reports of atypical gout have described entrapment neuropathies (eg, gouty deposits inducing carpal tunnel syndrome), ocular gout manifested as conjunctival deposits and uveitis, pancreatic gout presenting as a mass, and dermatologic manifestations including panniculitis.[13, 14]

Spinal gout (also known as axial gout) manifests when crystal‐induced inflammation, erosive arthritis, and tophaceous deposits occur along the spinal column. A cross‐sectional study of patients with poorly controlled gout reported the prevalence of spinal gout diagnosed by computerized tomography to be 35%. These radiographic findings were not consistently correlated with back pain.[15] Imaging features that are suggestive of spinal gout include intra‐articular and juxta‐articular erosions with sclerotic margins and density greater than the surrounding muscle. Periosteal new bone formation adjacent to bony destruction can form overhanging edges.[16] When retrospectively presented with the final diagnosis, the radiologist at our institution noted that the appearance was typical gout in an atypical location.

Spinal gout can be confused with spinal metastasis, infection, and stenosis. It can remain asymptomatic or present with back pain, radiculopathy, or cord compression. The lumbar spine is the most frequently affected site.[17, 18] Many patients with spinal gout have had chronic tophaceous gout with radiologic evidence of erosions in the peripheral joints.[15] Patients with spinal gout also have elevated urate levels and markers of inflammation.[18] Surgical decompression and stabilization is recommended when there is frank cord compression, progressive neurologic compromise, or lack of improvement with gout therapy alone.[18]

This patient's male gender, history of gout, hypertension, alcohol consumption, and thiazide diuretic use placed him at an increased risk of a gout attack.[19, 20] The possible interruption of urate‐lowering therapy for the surgical procedure and surgery itself further heightened his risk of suffering acute gouty arthritis in the perioperative period.[21] The patient's encephalopathy may have masked back pain and precluded an accurate neurologic exam. There is one case report to our knowledge describing encephalopathy that improved with colchicine and was possibly related to gout.[22] This patient's encephalopathy was deemed multifactorial and attributed to alcohol withdrawal, medications (including opioids and steroids), and infection (pneumonia).

Gout is best known for its peripheral arthritis and is rarely invoked in the consideration of spinal and myelopathic processes where more pressing competing diagnoses, such as infection and malignancy, are typically considered. In addition, when surgical specimens are submitted for examination for pathology in formaldehyde (rather than alcohol), monosodium urate crystals are dissolved and are thus difficult to identify in the specimen.

This case reminds us that gout remains a diagnostic challenge and should be considered in the differential of an inflammatory process. Recognition of the multifaceted nature of gout can allow for the earlier recognition and treatment of the less typical presentations of this ancient malady.

KEY TEACHING POINTS

1. Crystalline disease is a common cause of postoperative arthritis.
2. Gout (and pseudogout) should be considered in cases of focal inflammation (detected by examination or imaging) when the evidence or predisposition for infection is limited or nonexistent.
3. Spinal gout presents with back pain, radiculopathy, or cord compression and may be confused with spinal metastasis, infection, and stenosis.

In 2006,[1] we questioned whether rapid response systems (RRSs) were an effective strategy for detecting and managing deteriorating general ward patients. Since then, the implementation of RRSs has flourished, especially in the United States where accreditors (Joint Commission)[2] and patient‐safety organizations (Institute for Healthcare Improvement 100,000 Live Campaign)[3] have strongly supported RRSs. Decades of evidence show that general ward patients often experience unrecognized deterioration and cardiorespiratory arrest (CA). The low sensitivity and accuracy of periodic assessments by staff are thought to be a major reason for these lapses, as are imbalances between patient needs and clinician (primarily nursing) resources. Additionally, a medical culture that punishes speaking up or bypassing the chain of command are also likely contributors to the problem. A system that effectively recognizes the early signs of deterioration and quickly responds should catch problems before they become life threatening. Over the last decade, RRSs have been the primary intervention implemented to do this. The potential for RRSs to improve outcomes has strong face validity, but researchers have struggled to demonstrate consistent improvements in outcomes across institutions. Given this, are RRSs the best intervention to prevent this failure to rescue? In this editorial we examine the progress of RRSs, how they compare to other options, and we consider whether we should continue to question their implementation.

In our 2007 systematic review,[4] we concluded there was weak to moderate evidence supporting RRSs. Since then, 6 other systematic reviews of the effectiveness or implementation of RRSs have been published. One high‐quality review of effectiveness studies published through 2008 by Chan et al.[5] found that RRSs significantly reduced non‐intensive care unit (ICU) CA (relative risk [RR], 0.66; 95% confidence interval [CI], 0.54‐0.80), but not total hospital mortality (RR, 0.96; 95% CI, 0.84‐1.09) in adult inpatients. In pediatric inpatients, RRSs led to significant improvements in both non‐ICU CA (RR, 0.62; 95% CI, 0.46 to 0.84) and total hospital mortality (RR, 0.79; 95% CI, 0.63 to 0.98). Subsequent to 2008, a structured search[6] finds 26 additional studies.[7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30] Although the benefit for CA in both adults and children has remained robust, even more so since Chan's review, mortality reductions in adult patients appear to have had the most notable shift. In aggregate, the point estimate (for those studies providing analyzable data), for adult mortality has strengthened to 0.88, with a confidence interval of 0.82‐0.96 in favor of the RRS strategy.

This change has occurred as the analyzable studies since 2008 have all had favorable point estimates, and 4 have had statistically significant confidence intervals. Prior to 2008, 5 had unfavorable point estimates, and only 2 had favorable confidence intervals. As RRSs expand, the benefits, although not universal (some hospitals still experience no improvement in outcomes), seem to be getting stronger and more consistent. This may be secondary to maturation of the intervention and implementation strategies, or it may be the result of secular trends outside of the RRS intervention, although studies controlling for this found it not to be the case.[10] The factors associated with successful implementation of the RRS or improved outcomes include knowledge of activation criteria, communication, teamwork, lack of criticism for activating the RRS, and better attitudes about the team's positive effect on nurses and patients. Many of these factors relate to an improved safety culture in general. Additionally, activation rates may have increased in more recent studies, as greater utilization is associated with improved outcomes.[31] Finally, RRSs, like other patient‐safety and quality interventions, mature with time, often taking several years before they have a full effect on outcomes.[31, 32]

Despite these more favorable results for RRSs, we still see a large discrepancy between the magnitude of benefit for CA and mortality. This may partly be because the exposure groups are different; most studies examined non‐ICU CA, yet studies reporting mortality used total hospital mortality (ICU and non‐ICU). Additionally, although RRSs may effectively prevent CA, this intervention may have a more limited effect in preventing the patient's ultimate demise (particularly in the ICU).

We also still see that effectiveness reports for RRSs continue to be of low to moderate quality. Many reports give no statistics or denominator data or have missing data. Few control for secular trends in providers, outcomes, and confounders. Outcome measures vary widely, and none conducted blinded outcome assessments. Most studies use a pre‐post design without concurrent controls, substantially increasing the risk of bias. The better‐designed studies that use concurrent controls or cluster randomization (Priestley,[33] Bristow,[34] and the MERIT trial[35]) tend to show lower treatment effects, although interestingly in the MERIT trial, while the cluster‐randomized data showed no benefit, the pre‐post data showed significant improvement in the RRS intervention hospitals. These results have been attributed to the control hospitals using their code teams for RRS activities,[36] negating a comparative improvement in the intervention hospitals.

Can we improve RRS research? Likely, yes. We can begin by being more careful about defining the exposure group. Ideally, studies should not include data from the ICU or the emergency department because these patient populations are not part of the exposure group. Although most studies removed ICU and emergency department data for CA, they did not do so for hospital mortality. ICU mortality is likely biased, because only a small proportion of ICU patients have been exposed to an RRS. Definitions also need to be stringent and uniform. For example, CA may be defined in a variety of ways such as calling the code team versus documented cardiopulmonary resuscitation. Unexpected hospital mortality is often defined as excluding patients with do not resuscitate (DNR) orders, but this may or may not accurately exclude expected deaths. We also need to better attempt to control for confounders and secular trends. Outcomes such as CA and mortality are strongly influenced by changes in patient case‐mix over time, the frequency of care limitation/DNR orders, or by poor triage decisions.[37] Outcomes such as unanticipated ICU admission are indirect and may be heavily influenced by local cultural factors. Finally, authors need to provide robust statistical data and clear numerators and denominators to support their conclusions.

Although we need to do our best to improve the quality of the RRS literature, the near ubiquitous presence of this patient‐safety intervention in North American hospitals raises a crucial question, Do we even need more effectiveness studies and if so what kind? Randomized controlled trials are not likely. It is hard to argue that we still sit at a position of equipoise, and randomizing patients who are deteriorating to standard care versus an RRS is neither practical nor ethical. Finding appropriate concurrent control hospitals that have not implemented some type of RRS would also be very difficult.

We should, however, continue to test the effectiveness of RRSs but in a more diverse manner. RRSs should be more directly compared to other interventions that can improve the problem of failure to rescue such as increased nurse staffing[38, 39, 40] and hospitalist staffing.[41] The low sensitivity and accuracy of monitoring vital signs on general wards by staff is also an area strongly deserving of investigation, as it is likely central to the problem. Researchers have sought to use various combinations of vital signs, including aggregated or weighted scoring systems, and recent data suggest some approaches may be superior to others.[42] Many have advocated for continuous monitoring of a limited set of vital signs similar to the ICU, and there are some recent data indicating that this might be effective.[43, 44] This work is in the early stages, and we do not yet know whether this strategy will affect outcomes. It is conceivable that if the false alarm rate can be kept very low and we can minimize the failure to recognize deteriorating patients (good sensitivity, specificity, and positive predictive value), the need for the RRS response team may be reduced or even eliminated. Additionally, as electronic medical records (EMRs) have expanded, there has been growing interest in leveraging these systems to improve the effectiveness of RRSs.[45] There is a tremendous amount of information within the EMRs that can be used to complement vital‐sign monitoring (manual or continuous), because baseline medical problems, laboratory values, and recent history may have a strong impact on the predictive value of changes in vital signs.

Research should also focus on the possible unintended consequences, costs, and the cost‐effectiveness of RRSs compared with other interventions that can or may reduce the rate of failure to rescue. Certainly, establishing RRSs has costs including staff time and the need to pull staff from other clinical duties to respond. Unintended harm, such as diversion of ICU staff from their usual care, are often mentioned but never rigorously evaluated. Increasing nurse staffing has very substantial costs, but how these costs compare to the costs of the RRS are unclear, although likely the comparison would be very favorable to the RRS, because staffing typically relies on existing employees with expertise in caring for the critically ill as opposed to workforce expansion. Given the current healthcare economic climate, any model that relies on additional employees is not likely to gain support. Establishing continuous monitoring systems have up‐front capital costs, although they may reduce other costs in the long run (eg, staff, medical liability). They also have intangible costs for provider workload if the false alarm rates are too high. Again, this strategy is too new to know the answers to these concerns. As we move forward, such evaluations are needed to guide policy decisions.

We also need more evaluation of RRS implementation science. The optimal way to organize, train, and staff RRSs is unknown. Most programs use physician‐led teams, although some use nurse‐led teams. Few studies have compared the various models, although 1 study that compared a resident‐led to an attending‐led team found no difference.[17] Education is ubiquitous, although actual staff training (simulation for example) is not commonly described. In addition, there is wide variation in the frequency of RRS activation. We know nurses and residents often feel pressured not to activate RRSs, and much of the success of the RRS relies on nurses identifying deteriorating patients and calling the response team. The use of continuous monitoring combined with automatic notification of staff may reduce the barriers to activating RRSs, increasing activation rates, but until then we need more understanding of how to break down these barriers. Family/patient access to activation has also gained ground (1 program demonstrated outcome improvement only after this was established[13]), but is not yet widespread.

The role of the RRS in improving processes of care, such as the appropriate institution of DNR orders, end of life/palliative care discussions, and early goal‐directed therapy for sepsis, have been presented in several studies[46, 47] but remain inadequately evaluated. Here too, there is much to learn about how we might realize the full effectiveness of this patient‐safety strategy beyond outcomes such as CA and hospital mortality. Ideally, if all appropriate patients had DNR orders and we stopped failing to recognize and respond to deteriorating ward patients, CAs on general hospital wards could be nearly eliminated.

RRSs have been described as a band‐aid for a failed model of general ward care.[37] What is clear is that many patients suffer preventable harm from unrecognized deterioration. This needs to be challenged, but are RRSs the best intervention? Despite the Joint Commission's Patient Safety Goal 16, should we still question their implementation? Should we (and the Joint Commission) reconsider our approach and prioritize our efforts elsewhere or should we feel comfortable with the investment that we have made in these systems? Even though there are many unknowns, and the quality of RRS studies needs improvement, the literature is accumulating that RRSs do reduce non‐ICU CA and improve hospital mortality. Without direct comparison studies demonstrating superiority of other expensive strategies, there is little reason to reconsider the RRS concept or question their implementation and our investment. We should instead invest further in this foundational patient‐safety strategy to make it as effective as it can be.

Disclosures: Dr. Pronovost reports the following potential conflicts of interest: grant or contract support from the Agency for Healthcare Research and Quality, and the Gordon and Betty Moore Foundation (research related to patient safety and quality of care), and the National Institutes of Health (acute lung injury research); consulting fees from the Association of Professionals in Infection Control and Epidemiology, Inc.; honoraria from various hospitals, health systems, and the Leigh Bureau to speak on quality and patient safety; book royalties from the Penguin Group; and board membership for the Cantel Medical Group. Dr. Winters reports the following potential conflicts of interest: contract or grant support from Masimo Corporation, honoraria from 3M Corporation and various hospitals and health systems, royalties from Lippincott Williams &Wilkins (UptoDate), and consulting fees from several legal firms for medical legal consulting.

In 2006,[1] we questioned whether rapid response systems (RRSs) were an effective strategy for detecting and managing deteriorating general ward patients. Since then, the implementation of RRSs has flourished, especially in the United States where accreditors (Joint Commission)[2] and patient‐safety organizations (Institute for Healthcare Improvement 100,000 Live Campaign)[3] have strongly supported RRSs. Decades of evidence show that general ward patients often experience unrecognized deterioration and cardiorespiratory arrest (CA). The low sensitivity and accuracy of periodic assessments by staff are thought to be a major reason for these lapses, as are imbalances between patient needs and clinician (primarily nursing) resources. Additionally, a medical culture that punishes speaking up or bypassing the chain of command are also likely contributors to the problem. A system that effectively recognizes the early signs of deterioration and quickly responds should catch problems before they become life threatening. Over the last decade, RRSs have been the primary intervention implemented to do this. The potential for RRSs to improve outcomes has strong face validity, but researchers have struggled to demonstrate consistent improvements in outcomes across institutions. Given this, are RRSs the best intervention to prevent this failure to rescue? In this editorial we examine the progress of RRSs, how they compare to other options, and we consider whether we should continue to question their implementation.

In our 2007 systematic review,[4] we concluded there was weak to moderate evidence supporting RRSs. Since then, 6 other systematic reviews of the effectiveness or implementation of RRSs have been published. One high‐quality review of effectiveness studies published through 2008 by Chan et al.[5] found that RRSs significantly reduced non‐intensive care unit (ICU) CA (relative risk [RR], 0.66; 95% confidence interval [CI], 0.54‐0.80), but not total hospital mortality (RR, 0.96; 95% CI, 0.84‐1.09) in adult inpatients. In pediatric inpatients, RRSs led to significant improvements in both non‐ICU CA (RR, 0.62; 95% CI, 0.46 to 0.84) and total hospital mortality (RR, 0.79; 95% CI, 0.63 to 0.98). Subsequent to 2008, a structured search[6] finds 26 additional studies.[7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30] Although the benefit for CA in both adults and children has remained robust, even more so since Chan's review, mortality reductions in adult patients appear to have had the most notable shift. In aggregate, the point estimate (for those studies providing analyzable data), for adult mortality has strengthened to 0.88, with a confidence interval of 0.82‐0.96 in favor of the RRS strategy.

This change has occurred as the analyzable studies since 2008 have all had favorable point estimates, and 4 have had statistically significant confidence intervals. Prior to 2008, 5 had unfavorable point estimates, and only 2 had favorable confidence intervals. As RRSs expand, the benefits, although not universal (some hospitals still experience no improvement in outcomes), seem to be getting stronger and more consistent. This may be secondary to maturation of the intervention and implementation strategies, or it may be the result of secular trends outside of the RRS intervention, although studies controlling for this found it not to be the case.[10] The factors associated with successful implementation of the RRS or improved outcomes include knowledge of activation criteria, communication, teamwork, lack of criticism for activating the RRS, and better attitudes about the team's positive effect on nurses and patients. Many of these factors relate to an improved safety culture in general. Additionally, activation rates may have increased in more recent studies, as greater utilization is associated with improved outcomes.[31] Finally, RRSs, like other patient‐safety and quality interventions, mature with time, often taking several years before they have a full effect on outcomes.[31, 32]

Despite these more favorable results for RRSs, we still see a large discrepancy between the magnitude of benefit for CA and mortality. This may partly be because the exposure groups are different; most studies examined non‐ICU CA, yet studies reporting mortality used total hospital mortality (ICU and non‐ICU). Additionally, although RRSs may effectively prevent CA, this intervention may have a more limited effect in preventing the patient's ultimate demise (particularly in the ICU).

We also still see that effectiveness reports for RRSs continue to be of low to moderate quality. Many reports give no statistics or denominator data or have missing data. Few control for secular trends in providers, outcomes, and confounders. Outcome measures vary widely, and none conducted blinded outcome assessments. Most studies use a pre‐post design without concurrent controls, substantially increasing the risk of bias. The better‐designed studies that use concurrent controls or cluster randomization (Priestley,[33] Bristow,[34] and the MERIT trial[35]) tend to show lower treatment effects, although interestingly in the MERIT trial, while the cluster‐randomized data showed no benefit, the pre‐post data showed significant improvement in the RRS intervention hospitals. These results have been attributed to the control hospitals using their code teams for RRS activities,[36] negating a comparative improvement in the intervention hospitals.

Can we improve RRS research? Likely, yes. We can begin by being more careful about defining the exposure group. Ideally, studies should not include data from the ICU or the emergency department because these patient populations are not part of the exposure group. Although most studies removed ICU and emergency department data for CA, they did not do so for hospital mortality. ICU mortality is likely biased, because only a small proportion of ICU patients have been exposed to an RRS. Definitions also need to be stringent and uniform. For example, CA may be defined in a variety of ways such as calling the code team versus documented cardiopulmonary resuscitation. Unexpected hospital mortality is often defined as excluding patients with do not resuscitate (DNR) orders, but this may or may not accurately exclude expected deaths. We also need to better attempt to control for confounders and secular trends. Outcomes such as CA and mortality are strongly influenced by changes in patient case‐mix over time, the frequency of care limitation/DNR orders, or by poor triage decisions.[37] Outcomes such as unanticipated ICU admission are indirect and may be heavily influenced by local cultural factors. Finally, authors need to provide robust statistical data and clear numerators and denominators to support their conclusions.

Although we need to do our best to improve the quality of the RRS literature, the near ubiquitous presence of this patient‐safety intervention in North American hospitals raises a crucial question, Do we even need more effectiveness studies and if so what kind? Randomized controlled trials are not likely. It is hard to argue that we still sit at a position of equipoise, and randomizing patients who are deteriorating to standard care versus an RRS is neither practical nor ethical. Finding appropriate concurrent control hospitals that have not implemented some type of RRS would also be very difficult.

We should, however, continue to test the effectiveness of RRSs but in a more diverse manner. RRSs should be more directly compared to other interventions that can improve the problem of failure to rescue such as increased nurse staffing[38, 39, 40] and hospitalist staffing.[41] The low sensitivity and accuracy of monitoring vital signs on general wards by staff is also an area strongly deserving of investigation, as it is likely central to the problem. Researchers have sought to use various combinations of vital signs, including aggregated or weighted scoring systems, and recent data suggest some approaches may be superior to others.[42] Many have advocated for continuous monitoring of a limited set of vital signs similar to the ICU, and there are some recent data indicating that this might be effective.[43, 44] This work is in the early stages, and we do not yet know whether this strategy will affect outcomes. It is conceivable that if the false alarm rate can be kept very low and we can minimize the failure to recognize deteriorating patients (good sensitivity, specificity, and positive predictive value), the need for the RRS response team may be reduced or even eliminated. Additionally, as electronic medical records (EMRs) have expanded, there has been growing interest in leveraging these systems to improve the effectiveness of RRSs.[45] There is a tremendous amount of information within the EMRs that can be used to complement vital‐sign monitoring (manual or continuous), because baseline medical problems, laboratory values, and recent history may have a strong impact on the predictive value of changes in vital signs.

Research should also focus on the possible unintended consequences, costs, and the cost‐effectiveness of RRSs compared with other interventions that can or may reduce the rate of failure to rescue. Certainly, establishing RRSs has costs including staff time and the need to pull staff from other clinical duties to respond. Unintended harm, such as diversion of ICU staff from their usual care, are often mentioned but never rigorously evaluated. Increasing nurse staffing has very substantial costs, but how these costs compare to the costs of the RRS are unclear, although likely the comparison would be very favorable to the RRS, because staffing typically relies on existing employees with expertise in caring for the critically ill as opposed to workforce expansion. Given the current healthcare economic climate, any model that relies on additional employees is not likely to gain support. Establishing continuous monitoring systems have up‐front capital costs, although they may reduce other costs in the long run (eg, staff, medical liability). They also have intangible costs for provider workload if the false alarm rates are too high. Again, this strategy is too new to know the answers to these concerns. As we move forward, such evaluations are needed to guide policy decisions.

We also need more evaluation of RRS implementation science. The optimal way to organize, train, and staff RRSs is unknown. Most programs use physician‐led teams, although some use nurse‐led teams. Few studies have compared the various models, although 1 study that compared a resident‐led to an attending‐led team found no difference.[17] Education is ubiquitous, although actual staff training (simulation for example) is not commonly described. In addition, there is wide variation in the frequency of RRS activation. We know nurses and residents often feel pressured not to activate RRSs, and much of the success of the RRS relies on nurses identifying deteriorating patients and calling the response team. The use of continuous monitoring combined with automatic notification of staff may reduce the barriers to activating RRSs, increasing activation rates, but until then we need more understanding of how to break down these barriers. Family/patient access to activation has also gained ground (1 program demonstrated outcome improvement only after this was established[13]), but is not yet widespread.

The role of the RRS in improving processes of care, such as the appropriate institution of DNR orders, end of life/palliative care discussions, and early goal‐directed therapy for sepsis, have been presented in several studies[46, 47] but remain inadequately evaluated. Here too, there is much to learn about how we might realize the full effectiveness of this patient‐safety strategy beyond outcomes such as CA and hospital mortality. Ideally, if all appropriate patients had DNR orders and we stopped failing to recognize and respond to deteriorating ward patients, CAs on general hospital wards could be nearly eliminated.

RRSs have been described as a band‐aid for a failed model of general ward care.[37] What is clear is that many patients suffer preventable harm from unrecognized deterioration. This needs to be challenged, but are RRSs the best intervention? Despite the Joint Commission's Patient Safety Goal 16, should we still question their implementation? Should we (and the Joint Commission) reconsider our approach and prioritize our efforts elsewhere or should we feel comfortable with the investment that we have made in these systems? Even though there are many unknowns, and the quality of RRS studies needs improvement, the literature is accumulating that RRSs do reduce non‐ICU CA and improve hospital mortality. Without direct comparison studies demonstrating superiority of other expensive strategies, there is little reason to reconsider the RRS concept or question their implementation and our investment. We should instead invest further in this foundational patient‐safety strategy to make it as effective as it can be.

Disclosures: Dr. Pronovost reports the following potential conflicts of interest: grant or contract support from the Agency for Healthcare Research and Quality, and the Gordon and Betty Moore Foundation (research related to patient safety and quality of care), and the National Institutes of Health (acute lung injury research); consulting fees from the Association of Professionals in Infection Control and Epidemiology, Inc.; honoraria from various hospitals, health systems, and the Leigh Bureau to speak on quality and patient safety; book royalties from the Penguin Group; and board membership for the Cantel Medical Group. Dr. Winters reports the following potential conflicts of interest: contract or grant support from Masimo Corporation, honoraria from 3M Corporation and various hospitals and health systems, royalties from Lippincott Williams &Wilkins (UptoDate), and consulting fees from several legal firms for medical legal consulting.

In 2006,[1] we questioned whether rapid response systems (RRSs) were an effective strategy for detecting and managing deteriorating general ward patients. Since then, the implementation of RRSs has flourished, especially in the United States where accreditors (Joint Commission)[2] and patient‐safety organizations (Institute for Healthcare Improvement 100,000 Live Campaign)[3] have strongly supported RRSs. Decades of evidence show that general ward patients often experience unrecognized deterioration and cardiorespiratory arrest (CA). The low sensitivity and accuracy of periodic assessments by staff are thought to be a major reason for these lapses, as are imbalances between patient needs and clinician (primarily nursing) resources. Additionally, a medical culture that punishes speaking up or bypassing the chain of command are also likely contributors to the problem. A system that effectively recognizes the early signs of deterioration and quickly responds should catch problems before they become life threatening. Over the last decade, RRSs have been the primary intervention implemented to do this. The potential for RRSs to improve outcomes has strong face validity, but researchers have struggled to demonstrate consistent improvements in outcomes across institutions. Given this, are RRSs the best intervention to prevent this failure to rescue? In this editorial we examine the progress of RRSs, how they compare to other options, and we consider whether we should continue to question their implementation.

In our 2007 systematic review,[4] we concluded there was weak to moderate evidence supporting RRSs. Since then, 6 other systematic reviews of the effectiveness or implementation of RRSs have been published. One high‐quality review of effectiveness studies published through 2008 by Chan et al.[5] found that RRSs significantly reduced non‐intensive care unit (ICU) CA (relative risk [RR], 0.66; 95% confidence interval [CI], 0.54‐0.80), but not total hospital mortality (RR, 0.96; 95% CI, 0.84‐1.09) in adult inpatients. In pediatric inpatients, RRSs led to significant improvements in both non‐ICU CA (RR, 0.62; 95% CI, 0.46 to 0.84) and total hospital mortality (RR, 0.79; 95% CI, 0.63 to 0.98). Subsequent to 2008, a structured search[6] finds 26 additional studies.[7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30] Although the benefit for CA in both adults and children has remained robust, even more so since Chan's review, mortality reductions in adult patients appear to have had the most notable shift. In aggregate, the point estimate (for those studies providing analyzable data), for adult mortality has strengthened to 0.88, with a confidence interval of 0.82‐0.96 in favor of the RRS strategy.

This change has occurred as the analyzable studies since 2008 have all had favorable point estimates, and 4 have had statistically significant confidence intervals. Prior to 2008, 5 had unfavorable point estimates, and only 2 had favorable confidence intervals. As RRSs expand, the benefits, although not universal (some hospitals still experience no improvement in outcomes), seem to be getting stronger and more consistent. This may be secondary to maturation of the intervention and implementation strategies, or it may be the result of secular trends outside of the RRS intervention, although studies controlling for this found it not to be the case.[10] The factors associated with successful implementation of the RRS or improved outcomes include knowledge of activation criteria, communication, teamwork, lack of criticism for activating the RRS, and better attitudes about the team's positive effect on nurses and patients. Many of these factors relate to an improved safety culture in general. Additionally, activation rates may have increased in more recent studies, as greater utilization is associated with improved outcomes.[31] Finally, RRSs, like other patient‐safety and quality interventions, mature with time, often taking several years before they have a full effect on outcomes.[31, 32]

Despite these more favorable results for RRSs, we still see a large discrepancy between the magnitude of benefit for CA and mortality. This may partly be because the exposure groups are different; most studies examined non‐ICU CA, yet studies reporting mortality used total hospital mortality (ICU and non‐ICU). Additionally, although RRSs may effectively prevent CA, this intervention may have a more limited effect in preventing the patient's ultimate demise (particularly in the ICU).

We also still see that effectiveness reports for RRSs continue to be of low to moderate quality. Many reports give no statistics or denominator data or have missing data. Few control for secular trends in providers, outcomes, and confounders. Outcome measures vary widely, and none conducted blinded outcome assessments. Most studies use a pre‐post design without concurrent controls, substantially increasing the risk of bias. The better‐designed studies that use concurrent controls or cluster randomization (Priestley,[33] Bristow,[34] and the MERIT trial[35]) tend to show lower treatment effects, although interestingly in the MERIT trial, while the cluster‐randomized data showed no benefit, the pre‐post data showed significant improvement in the RRS intervention hospitals. These results have been attributed to the control hospitals using their code teams for RRS activities,[36] negating a comparative improvement in the intervention hospitals.

Can we improve RRS research? Likely, yes. We can begin by being more careful about defining the exposure group. Ideally, studies should not include data from the ICU or the emergency department because these patient populations are not part of the exposure group. Although most studies removed ICU and emergency department data for CA, they did not do so for hospital mortality. ICU mortality is likely biased, because only a small proportion of ICU patients have been exposed to an RRS. Definitions also need to be stringent and uniform. For example, CA may be defined in a variety of ways such as calling the code team versus documented cardiopulmonary resuscitation. Unexpected hospital mortality is often defined as excluding patients with do not resuscitate (DNR) orders, but this may or may not accurately exclude expected deaths. We also need to better attempt to control for confounders and secular trends. Outcomes such as CA and mortality are strongly influenced by changes in patient case‐mix over time, the frequency of care limitation/DNR orders, or by poor triage decisions.[37] Outcomes such as unanticipated ICU admission are indirect and may be heavily influenced by local cultural factors. Finally, authors need to provide robust statistical data and clear numerators and denominators to support their conclusions.

Although we need to do our best to improve the quality of the RRS literature, the near ubiquitous presence of this patient‐safety intervention in North American hospitals raises a crucial question, Do we even need more effectiveness studies and if so what kind? Randomized controlled trials are not likely. It is hard to argue that we still sit at a position of equipoise, and randomizing patients who are deteriorating to standard care versus an RRS is neither practical nor ethical. Finding appropriate concurrent control hospitals that have not implemented some type of RRS would also be very difficult.

We should, however, continue to test the effectiveness of RRSs but in a more diverse manner. RRSs should be more directly compared to other interventions that can improve the problem of failure to rescue such as increased nurse staffing[38, 39, 40] and hospitalist staffing.[41] The low sensitivity and accuracy of monitoring vital signs on general wards by staff is also an area strongly deserving of investigation, as it is likely central to the problem. Researchers have sought to use various combinations of vital signs, including aggregated or weighted scoring systems, and recent data suggest some approaches may be superior to others.[42] Many have advocated for continuous monitoring of a limited set of vital signs similar to the ICU, and there are some recent data indicating that this might be effective.[43, 44] This work is in the early stages, and we do not yet know whether this strategy will affect outcomes. It is conceivable that if the false alarm rate can be kept very low and we can minimize the failure to recognize deteriorating patients (good sensitivity, specificity, and positive predictive value), the need for the RRS response team may be reduced or even eliminated. Additionally, as electronic medical records (EMRs) have expanded, there has been growing interest in leveraging these systems to improve the effectiveness of RRSs.[45] There is a tremendous amount of information within the EMRs that can be used to complement vital‐sign monitoring (manual or continuous), because baseline medical problems, laboratory values, and recent history may have a strong impact on the predictive value of changes in vital signs.

Research should also focus on the possible unintended consequences, costs, and the cost‐effectiveness of RRSs compared with other interventions that can or may reduce the rate of failure to rescue. Certainly, establishing RRSs has costs including staff time and the need to pull staff from other clinical duties to respond. Unintended harm, such as diversion of ICU staff from their usual care, are often mentioned but never rigorously evaluated. Increasing nurse staffing has very substantial costs, but how these costs compare to the costs of the RRS are unclear, although likely the comparison would be very favorable to the RRS, because staffing typically relies on existing employees with expertise in caring for the critically ill as opposed to workforce expansion. Given the current healthcare economic climate, any model that relies on additional employees is not likely to gain support. Establishing continuous monitoring systems have up‐front capital costs, although they may reduce other costs in the long run (eg, staff, medical liability). They also have intangible costs for provider workload if the false alarm rates are too high. Again, this strategy is too new to know the answers to these concerns. As we move forward, such evaluations are needed to guide policy decisions.

We also need more evaluation of RRS implementation science. The optimal way to organize, train, and staff RRSs is unknown. Most programs use physician‐led teams, although some use nurse‐led teams. Few studies have compared the various models, although 1 study that compared a resident‐led to an attending‐led team found no difference.[17] Education is ubiquitous, although actual staff training (simulation for example) is not commonly described. In addition, there is wide variation in the frequency of RRS activation. We know nurses and residents often feel pressured not to activate RRSs, and much of the success of the RRS relies on nurses identifying deteriorating patients and calling the response team. The use of continuous monitoring combined with automatic notification of staff may reduce the barriers to activating RRSs, increasing activation rates, but until then we need more understanding of how to break down these barriers. Family/patient access to activation has also gained ground (1 program demonstrated outcome improvement only after this was established[13]), but is not yet widespread.

The role of the RRS in improving processes of care, such as the appropriate institution of DNR orders, end of life/palliative care discussions, and early goal‐directed therapy for sepsis, have been presented in several studies[46, 47] but remain inadequately evaluated. Here too, there is much to learn about how we might realize the full effectiveness of this patient‐safety strategy beyond outcomes such as CA and hospital mortality. Ideally, if all appropriate patients had DNR orders and we stopped failing to recognize and respond to deteriorating ward patients, CAs on general hospital wards could be nearly eliminated.

RRSs have been described as a band‐aid for a failed model of general ward care.[37] What is clear is that many patients suffer preventable harm from unrecognized deterioration. This needs to be challenged, but are RRSs the best intervention? Despite the Joint Commission's Patient Safety Goal 16, should we still question their implementation? Should we (and the Joint Commission) reconsider our approach and prioritize our efforts elsewhere or should we feel comfortable with the investment that we have made in these systems? Even though there are many unknowns, and the quality of RRS studies needs improvement, the literature is accumulating that RRSs do reduce non‐ICU CA and improve hospital mortality. Without direct comparison studies demonstrating superiority of other expensive strategies, there is little reason to reconsider the RRS concept or question their implementation and our investment. We should instead invest further in this foundational patient‐safety strategy to make it as effective as it can be.

Disclosures: Dr. Pronovost reports the following potential conflicts of interest: grant or contract support from the Agency for Healthcare Research and Quality, and the Gordon and Betty Moore Foundation (research related to patient safety and quality of care), and the National Institutes of Health (acute lung injury research); consulting fees from the Association of Professionals in Infection Control and Epidemiology, Inc.; honoraria from various hospitals, health systems, and the Leigh Bureau to speak on quality and patient safety; book royalties from the Penguin Group; and board membership for the Cantel Medical Group. Dr. Winters reports the following potential conflicts of interest: contract or grant support from Masimo Corporation, honoraria from 3M Corporation and various hospitals and health systems, royalties from Lippincott Williams &Wilkins (UptoDate), and consulting fees from several legal firms for medical legal consulting.

Hospital communication failures are a leading cause of serious errors and adverse events in the United States.[1, 2, 3, 4] With the implementation of duty‐hour restrictions for resident physicians,[5] there has been particular focus on the transfer of information during handoffs at change of shift.[6, 7] Many residency programs have sought to improve the processes of written and verbal handoffs through various initiatives, including: (1) automated linkage of handoff forms to electronic medical records (EMRs)[8, 9, 10]; (2) introduction of oral communication curricula, handoff simulation, or mnemonics[11, 12, 13]; and (3) faculty oversight of housestaff handoffs.[14, 15] Underlying each initiative has been the assumption that improving written and verbal handoff processes will ensure the availability of optimal patient information for on‐call housestaff. There has been little investigation, however, into what clinical questions are actually being asked of on‐call trainees, as well as what sources of information they are using to provide answers.

The aim of our study was to examine the extent to which written and verbal handoffs are utilized by pediatric trainees to derive answers to questions posed during overnight shifts. We also sought to describe both the frequency and types of on‐call questions being asked of trainees. Our primary outcome was trainee use of written handoffs to answer on‐call questions. Secondary outcomes included trainee use of verbal handoffs, as well as their use of alternative information resources to answer on‐call questions, including other clinical staff (ie, attending physicians, senior residents, nursing staff), patients and their families, the medical record, or the Internet. We then examined a variety of trainee, patient, and question characteristics to assess potential predictors of written and verbal handoff use.

METHODS

Institutional approval was granted to prospectively observe pediatric interns at the start of their overnight on‐call shifts on 2 inpatient wards at Boston Children's Hospital during 3 winter months (November through January). Our study was conducted during the postintervention period of a larger study that was designed to examine the effectiveness of a new resident handoff bundle on resident workflow and patient safety.[13] Interns rotating on study ward 1 used a structured, nonautomated tool (Microsoft Word version 2003; Microsoft Corp., Redmond, WA). Interns on study ward 2 used a handoff tool that was developed at the study hospital for use with the hospital's EMR, Cerner PowerChart version 2007.17 (Cerner Corp., Kansas City, MO). Interns on both wards received training on specific communication strategies, including verbal and written handoff processes.[13]

For our study, we recorded all questions being asked of on‐call interns by patients, parents, or other family members, as well as nurses or other clinical providers after completion of their evening handoff. We then directly observed all information resources used to derive answers to any questions asked pertaining to patients discussed in the evening handoff. We excluded any questions about new patient admissions or transfers, as well as nonpatient‐related questions.

Both study wards were staffed by separate day and night housestaff teams, who worked shifts of 12 to 14 hours in duration and had similar nursing schedules. The day team consisted of 3 interns and 1 senior resident per ward. The night team consisted of 1 intern on each ward, supervised by a senior resident covering both wards. Each day intern rotated for 1 week (Sunday through Thursday) during their month‐long ward rotation as part of the night team. We considered any intern on either of the 2 study wards to be eligible for enrollment in this study. Written consent was obtained from all participants.

The night intern received a verbal and written handoff at the shift change (usually performed between 5 and 7_pm) from 1 of the departing day interns prior to the start of the observation period. This handoff was conducted face‐to‐face in a ward conference room typically with the on‐call night intern and supervising resident receiving the handoff together from the departing day intern/senior resident.

RESULTS

Twenty‐eight observation nights (equivalent to 77 hours and 6 minutes of total direct observation time), consisting of 13 sessions on study ward 1 and 15 sessions on study ward 2, were completed. A total of 15 first‐year pediatric interns (5 male, 33%; 10 female, 66.7%), with a median age of 27.5 years (interquartile range [IQR]: 2629 years) participated. Interns on the 2 study wards were comparable with regard to trainee week of service (P=0.43) and consecutive night of call at the time of observation (P=0.45). Each intern was observed for a mean of 2 sessions (range, 13 sessions), with a mean observation time per session of approximately 2 hours and 45 minutes ( 23 minutes).

Answers

Of the 260 recorded questions, 90% (233) had documented answers. For the 10% (27) of questions with undocumented answers, 21 were observed to be verbally deferred by the intern to the day team or another care provider (ie, other physician or nurse), and almost half (42.9% [9]) involved general care‐plan questions; the remainder involved medication (4), diet (2), diagnostic testing (5), or vital sign (1) questions. An additional 6 questions went unanswered during the observation period, and it is unknown if or when they were answered.

Of the answered questions, 90% (209) of answers were provided by trainees within 5 minutes and 9% (21) within 1.5 hours. In all, interns reported using 1 information resource to provide answers for 61% (142) of questions, at least 2 resources for 33% (76) questions, and 3 resources for 6% (15) questions.

Across both study wards, interns reported using information provided in written or verbal handoffs to answer 32.6% of questions. Interns reported using the written handoff, either alone or in combination with other information resources, to provide answers for 7.3% (17) of questions; verbal handoff, either alone or in combination with another resource (excluding written handoff), was reported as a resource for 25.3% (59) of questions. Of note, interns were directly observed to look at the written handoff when answering 21% (49) of questions.

A variety of other resources, including general medical/clinical knowledge, the EMR, and parents or other resources, were used to answer the remaining 67.4% (157) of questions. Intern general medical knowledge (ie, reports of simply knowing the answer to the question in their head[s]) was used to provide answers for 53.2% (124) of questions asked.

Unadjusted univariate regression analyses assessing predictors of written and verbal handoff use are shown in Figure 1. Multivariate logistic regression analyses showed that both dietary questions (odds ratio [OR]: 3.64, 95% confidence interval [CI]: 1.518.76; P=0.004) and interns' consecutive call night (OR: 0.29, 95% CI: 0.090.93; P=0.04) remained significant predictors of written handoff use. After adjusting for risk factors identified above, no differences in written handoff use were seen between the 2 wards.

Figure 1

Multivariate logistic regression for predictors of the verbal handoff use showed that questions regarding patients with longer lengths of stay (OR: 1.97, 95% CI: 1.023.8; P=0.04), those regarding general care plans (OR: 2.07, 95% CI: 1.133.78; P=0.02), as well as those asked by clinical staff (OR: 1.95, 95 CI: 1.043.66; P=0.04), remained significant predictors of reported verbal handoff use.

DISCUSSION

In light of the recent changes in duty hours implemented in July 2011, many pediatric training programs are having trainees work in day and night shifts.[18] Pediatric resident physicians frequently answer questions that pertain to patients handed off between day and night shifts. We found that on average, information provided in the verbal and written handoff was used almost once per hour. Housestaff in our study generally based their answers on information found in 1 or 2 resources, with almost one‐third of all questions involving some use of the written or verbal handoff. Prior research has documented widespread problems with resident handoff practices across programs and a high rate of medical errors due to miscommunications.[3, 4, 19, 20] Given how often information contained within the handoff was used as interns went about their nightly tasks, it is not difficult to understand how errors or omissions in the handoff process may potentially translate into frequent problems in direct patient care.

Trainees reported using written handoff tools to provide answers for 7.3% of questions. As we had suspected, they relied less frequently on their written handoffs as they completed more consecutive call nights. Interestingly, however, even when housestaff did not report using the written handoff, they were observed quite often to look at it before providing an answer. One explanation for this discrepancy between trainee reports and our observations is that the written handoff may serve as a memory tool, even if housestaff do not directly attribute their answers to its content. Our study also found that answers to questions concerning patients' diet and fluids were more likely to be ascribed to information contained in the written handoff. This finding supports the potential value of automated written handoff tools that are linked to the EMR, which can best ensure accuracy of this type of information.

Housestaff in our study also reported using information received during the verbal handoff to answer 1 out of every 4 on‐call questions. Although we did not specifically rate or monitor the quality of verbal handoffs, prior research has demonstrated that resident verbal handoff is often plagued with incomplete and inaccurate data.[3, 4, 19, 21] One investigation found that pediatric interns were prone to overestimating the effectiveness of their verbal handoffs, even as they failed to convey urgent information to their peers.[19] In light of such prior work, our finding that interns frequently rely on the verbal transfer of information supports specific residency training program handoff initiatives that target verbal exchanges.[11, 22, 23]

Although information obtained in the handoff was frequently required by on‐call housestaff, our study found that two‐thirds of all questions were answered using other resources, most often general medical or clinical knowledge. Clearly, background knowledge and experience is fundamental to trainees' ability to perform their jobs. Such reliance on general knowledge for problem solving may not be unique to interns. One recent observational study of senior pediatric cardiac subspecialists reported a high frequency of reliance on their own clinical experience, instinct, or prior training in making clinical decisions.[24] Further investigation may be useful to parse out the exact types of clinical knowledge being used, and may have important implications for how training programs plan for overnight supervision.[25, 26, 27]

Our study has several limitations. First, it was beyond the scope of this study to link housestaff answers to patient outcomes or medical errors. Given the frequency with which the handoff, a known source of vulnerability to medical error, was used by on‐call housestaff, our study suggests that future research evaluating the relationship between questions asked of on‐call housestaff, the answers provided, and downstream patient safety incidents may be merited. Second, our study was conducted in a single pediatric residency program with 1 physician observer midway through the first year of training and only in the early evening hours. This limits the generalizability of our findings, as the use of handoffs to answer on‐call questions may be different at other stages of the training process, within other specialties, or even at different times of the day. We also began our observations after the handoff had taken place; future studies may want to assess how variations in written and verbal handoff processes affect their use. As a final limitation, we note that although collecting information in real time using a direct observational method eliminated the problem of recall bias, there may have been attribution bias.

The results of our study demonstrate that on‐call pediatric housestaff are frequently asked a variety of clinical questions posed by hospital staff, patients, and their families. We found that trainees are apt to rely both on handoff information and other resources to provide answers. By better understanding what resources on‐call housestaff are accessing to answer questions overnight, we may be able to better target interventions needed to improve the availability of patient information, as well as the usefulness of written and verbal handoff tools.[11, 22, 23]

Hospital communication failures are a leading cause of serious errors and adverse events in the United States.[1, 2, 3, 4] With the implementation of duty‐hour restrictions for resident physicians,[5] there has been particular focus on the transfer of information during handoffs at change of shift.[6, 7] Many residency programs have sought to improve the processes of written and verbal handoffs through various initiatives, including: (1) automated linkage of handoff forms to electronic medical records (EMRs)[8, 9, 10]; (2) introduction of oral communication curricula, handoff simulation, or mnemonics[11, 12, 13]; and (3) faculty oversight of housestaff handoffs.[14, 15] Underlying each initiative has been the assumption that improving written and verbal handoff processes will ensure the availability of optimal patient information for on‐call housestaff. There has been little investigation, however, into what clinical questions are actually being asked of on‐call trainees, as well as what sources of information they are using to provide answers.

The aim of our study was to examine the extent to which written and verbal handoffs are utilized by pediatric trainees to derive answers to questions posed during overnight shifts. We also sought to describe both the frequency and types of on‐call questions being asked of trainees. Our primary outcome was trainee use of written handoffs to answer on‐call questions. Secondary outcomes included trainee use of verbal handoffs, as well as their use of alternative information resources to answer on‐call questions, including other clinical staff (ie, attending physicians, senior residents, nursing staff), patients and their families, the medical record, or the Internet. We then examined a variety of trainee, patient, and question characteristics to assess potential predictors of written and verbal handoff use.

METHODS

Institutional approval was granted to prospectively observe pediatric interns at the start of their overnight on‐call shifts on 2 inpatient wards at Boston Children's Hospital during 3 winter months (November through January). Our study was conducted during the postintervention period of a larger study that was designed to examine the effectiveness of a new resident handoff bundle on resident workflow and patient safety.[13] Interns rotating on study ward 1 used a structured, nonautomated tool (Microsoft Word version 2003; Microsoft Corp., Redmond, WA). Interns on study ward 2 used a handoff tool that was developed at the study hospital for use with the hospital's EMR, Cerner PowerChart version 2007.17 (Cerner Corp., Kansas City, MO). Interns on both wards received training on specific communication strategies, including verbal and written handoff processes.[13]

For our study, we recorded all questions being asked of on‐call interns by patients, parents, or other family members, as well as nurses or other clinical providers after completion of their evening handoff. We then directly observed all information resources used to derive answers to any questions asked pertaining to patients discussed in the evening handoff. We excluded any questions about new patient admissions or transfers, as well as nonpatient‐related questions.

Both study wards were staffed by separate day and night housestaff teams, who worked shifts of 12 to 14 hours in duration and had similar nursing schedules. The day team consisted of 3 interns and 1 senior resident per ward. The night team consisted of 1 intern on each ward, supervised by a senior resident covering both wards. Each day intern rotated for 1 week (Sunday through Thursday) during their month‐long ward rotation as part of the night team. We considered any intern on either of the 2 study wards to be eligible for enrollment in this study. Written consent was obtained from all participants.

The night intern received a verbal and written handoff at the shift change (usually performed between 5 and 7_pm) from 1 of the departing day interns prior to the start of the observation period. This handoff was conducted face‐to‐face in a ward conference room typically with the on‐call night intern and supervising resident receiving the handoff together from the departing day intern/senior resident.

RESULTS

Twenty‐eight observation nights (equivalent to 77 hours and 6 minutes of total direct observation time), consisting of 13 sessions on study ward 1 and 15 sessions on study ward 2, were completed. A total of 15 first‐year pediatric interns (5 male, 33%; 10 female, 66.7%), with a median age of 27.5 years (interquartile range [IQR]: 2629 years) participated. Interns on the 2 study wards were comparable with regard to trainee week of service (P=0.43) and consecutive night of call at the time of observation (P=0.45). Each intern was observed for a mean of 2 sessions (range, 13 sessions), with a mean observation time per session of approximately 2 hours and 45 minutes ( 23 minutes).

Answers

Of the 260 recorded questions, 90% (233) had documented answers. For the 10% (27) of questions with undocumented answers, 21 were observed to be verbally deferred by the intern to the day team or another care provider (ie, other physician or nurse), and almost half (42.9% [9]) involved general care‐plan questions; the remainder involved medication (4), diet (2), diagnostic testing (5), or vital sign (1) questions. An additional 6 questions went unanswered during the observation period, and it is unknown if or when they were answered.

Of the answered questions, 90% (209) of answers were provided by trainees within 5 minutes and 9% (21) within 1.5 hours. In all, interns reported using 1 information resource to provide answers for 61% (142) of questions, at least 2 resources for 33% (76) questions, and 3 resources for 6% (15) questions.

Across both study wards, interns reported using information provided in written or verbal handoffs to answer 32.6% of questions. Interns reported using the written handoff, either alone or in combination with other information resources, to provide answers for 7.3% (17) of questions; verbal handoff, either alone or in combination with another resource (excluding written handoff), was reported as a resource for 25.3% (59) of questions. Of note, interns were directly observed to look at the written handoff when answering 21% (49) of questions.

A variety of other resources, including general medical/clinical knowledge, the EMR, and parents or other resources, were used to answer the remaining 67.4% (157) of questions. Intern general medical knowledge (ie, reports of simply knowing the answer to the question in their head[s]) was used to provide answers for 53.2% (124) of questions asked.

Unadjusted univariate regression analyses assessing predictors of written and verbal handoff use are shown in Figure 1. Multivariate logistic regression analyses showed that both dietary questions (odds ratio [OR]: 3.64, 95% confidence interval [CI]: 1.518.76; P=0.004) and interns' consecutive call night (OR: 0.29, 95% CI: 0.090.93; P=0.04) remained significant predictors of written handoff use. After adjusting for risk factors identified above, no differences in written handoff use were seen between the 2 wards.

Figure 1

Multivariate logistic regression for predictors of the verbal handoff use showed that questions regarding patients with longer lengths of stay (OR: 1.97, 95% CI: 1.023.8; P=0.04), those regarding general care plans (OR: 2.07, 95% CI: 1.133.78; P=0.02), as well as those asked by clinical staff (OR: 1.95, 95 CI: 1.043.66; P=0.04), remained significant predictors of reported verbal handoff use.

DISCUSSION

In light of the recent changes in duty hours implemented in July 2011, many pediatric training programs are having trainees work in day and night shifts.[18] Pediatric resident physicians frequently answer questions that pertain to patients handed off between day and night shifts. We found that on average, information provided in the verbal and written handoff was used almost once per hour. Housestaff in our study generally based their answers on information found in 1 or 2 resources, with almost one‐third of all questions involving some use of the written or verbal handoff. Prior research has documented widespread problems with resident handoff practices across programs and a high rate of medical errors due to miscommunications.[3, 4, 19, 20] Given how often information contained within the handoff was used as interns went about their nightly tasks, it is not difficult to understand how errors or omissions in the handoff process may potentially translate into frequent problems in direct patient care.

Trainees reported using written handoff tools to provide answers for 7.3% of questions. As we had suspected, they relied less frequently on their written handoffs as they completed more consecutive call nights. Interestingly, however, even when housestaff did not report using the written handoff, they were observed quite often to look at it before providing an answer. One explanation for this discrepancy between trainee reports and our observations is that the written handoff may serve as a memory tool, even if housestaff do not directly attribute their answers to its content. Our study also found that answers to questions concerning patients' diet and fluids were more likely to be ascribed to information contained in the written handoff. This finding supports the potential value of automated written handoff tools that are linked to the EMR, which can best ensure accuracy of this type of information.

Housestaff in our study also reported using information received during the verbal handoff to answer 1 out of every 4 on‐call questions. Although we did not specifically rate or monitor the quality of verbal handoffs, prior research has demonstrated that resident verbal handoff is often plagued with incomplete and inaccurate data.[3, 4, 19, 21] One investigation found that pediatric interns were prone to overestimating the effectiveness of their verbal handoffs, even as they failed to convey urgent information to their peers.[19] In light of such prior work, our finding that interns frequently rely on the verbal transfer of information supports specific residency training program handoff initiatives that target verbal exchanges.[11, 22, 23]

Although information obtained in the handoff was frequently required by on‐call housestaff, our study found that two‐thirds of all questions were answered using other resources, most often general medical or clinical knowledge. Clearly, background knowledge and experience is fundamental to trainees' ability to perform their jobs. Such reliance on general knowledge for problem solving may not be unique to interns. One recent observational study of senior pediatric cardiac subspecialists reported a high frequency of reliance on their own clinical experience, instinct, or prior training in making clinical decisions.[24] Further investigation may be useful to parse out the exact types of clinical knowledge being used, and may have important implications for how training programs plan for overnight supervision.[25, 26, 27]

Our study has several limitations. First, it was beyond the scope of this study to link housestaff answers to patient outcomes or medical errors. Given the frequency with which the handoff, a known source of vulnerability to medical error, was used by on‐call housestaff, our study suggests that future research evaluating the relationship between questions asked of on‐call housestaff, the answers provided, and downstream patient safety incidents may be merited. Second, our study was conducted in a single pediatric residency program with 1 physician observer midway through the first year of training and only in the early evening hours. This limits the generalizability of our findings, as the use of handoffs to answer on‐call questions may be different at other stages of the training process, within other specialties, or even at different times of the day. We also began our observations after the handoff had taken place; future studies may want to assess how variations in written and verbal handoff processes affect their use. As a final limitation, we note that although collecting information in real time using a direct observational method eliminated the problem of recall bias, there may have been attribution bias.

The results of our study demonstrate that on‐call pediatric housestaff are frequently asked a variety of clinical questions posed by hospital staff, patients, and their families. We found that trainees are apt to rely both on handoff information and other resources to provide answers. By better understanding what resources on‐call housestaff are accessing to answer questions overnight, we may be able to better target interventions needed to improve the availability of patient information, as well as the usefulness of written and verbal handoff tools.[11, 22, 23]

Hospital communication failures are a leading cause of serious errors and adverse events in the United States.[1, 2, 3, 4] With the implementation of duty‐hour restrictions for resident physicians,[5] there has been particular focus on the transfer of information during handoffs at change of shift.[6, 7] Many residency programs have sought to improve the processes of written and verbal handoffs through various initiatives, including: (1) automated linkage of handoff forms to electronic medical records (EMRs)[8, 9, 10]; (2) introduction of oral communication curricula, handoff simulation, or mnemonics[11, 12, 13]; and (3) faculty oversight of housestaff handoffs.[14, 15] Underlying each initiative has been the assumption that improving written and verbal handoff processes will ensure the availability of optimal patient information for on‐call housestaff. There has been little investigation, however, into what clinical questions are actually being asked of on‐call trainees, as well as what sources of information they are using to provide answers.

The aim of our study was to examine the extent to which written and verbal handoffs are utilized by pediatric trainees to derive answers to questions posed during overnight shifts. We also sought to describe both the frequency and types of on‐call questions being asked of trainees. Our primary outcome was trainee use of written handoffs to answer on‐call questions. Secondary outcomes included trainee use of verbal handoffs, as well as their use of alternative information resources to answer on‐call questions, including other clinical staff (ie, attending physicians, senior residents, nursing staff), patients and their families, the medical record, or the Internet. We then examined a variety of trainee, patient, and question characteristics to assess potential predictors of written and verbal handoff use.

METHODS

Institutional approval was granted to prospectively observe pediatric interns at the start of their overnight on‐call shifts on 2 inpatient wards at Boston Children's Hospital during 3 winter months (November through January). Our study was conducted during the postintervention period of a larger study that was designed to examine the effectiveness of a new resident handoff bundle on resident workflow and patient safety.[13] Interns rotating on study ward 1 used a structured, nonautomated tool (Microsoft Word version 2003; Microsoft Corp., Redmond, WA). Interns on study ward 2 used a handoff tool that was developed at the study hospital for use with the hospital's EMR, Cerner PowerChart version 2007.17 (Cerner Corp., Kansas City, MO). Interns on both wards received training on specific communication strategies, including verbal and written handoff processes.[13]

For our study, we recorded all questions being asked of on‐call interns by patients, parents, or other family members, as well as nurses or other clinical providers after completion of their evening handoff. We then directly observed all information resources used to derive answers to any questions asked pertaining to patients discussed in the evening handoff. We excluded any questions about new patient admissions or transfers, as well as nonpatient‐related questions.

Both study wards were staffed by separate day and night housestaff teams, who worked shifts of 12 to 14 hours in duration and had similar nursing schedules. The day team consisted of 3 interns and 1 senior resident per ward. The night team consisted of 1 intern on each ward, supervised by a senior resident covering both wards. Each day intern rotated for 1 week (Sunday through Thursday) during their month‐long ward rotation as part of the night team. We considered any intern on either of the 2 study wards to be eligible for enrollment in this study. Written consent was obtained from all participants.

The night intern received a verbal and written handoff at the shift change (usually performed between 5 and 7_pm) from 1 of the departing day interns prior to the start of the observation period. This handoff was conducted face‐to‐face in a ward conference room typically with the on‐call night intern and supervising resident receiving the handoff together from the departing day intern/senior resident.

RESULTS

Twenty‐eight observation nights (equivalent to 77 hours and 6 minutes of total direct observation time), consisting of 13 sessions on study ward 1 and 15 sessions on study ward 2, were completed. A total of 15 first‐year pediatric interns (5 male, 33%; 10 female, 66.7%), with a median age of 27.5 years (interquartile range [IQR]: 2629 years) participated. Interns on the 2 study wards were comparable with regard to trainee week of service (P=0.43) and consecutive night of call at the time of observation (P=0.45). Each intern was observed for a mean of 2 sessions (range, 13 sessions), with a mean observation time per session of approximately 2 hours and 45 minutes ( 23 minutes).

Answers

Of the 260 recorded questions, 90% (233) had documented answers. For the 10% (27) of questions with undocumented answers, 21 were observed to be verbally deferred by the intern to the day team or another care provider (ie, other physician or nurse), and almost half (42.9% [9]) involved general care‐plan questions; the remainder involved medication (4), diet (2), diagnostic testing (5), or vital sign (1) questions. An additional 6 questions went unanswered during the observation period, and it is unknown if or when they were answered.

Of the answered questions, 90% (209) of answers were provided by trainees within 5 minutes and 9% (21) within 1.5 hours. In all, interns reported using 1 information resource to provide answers for 61% (142) of questions, at least 2 resources for 33% (76) questions, and 3 resources for 6% (15) questions.

Across both study wards, interns reported using information provided in written or verbal handoffs to answer 32.6% of questions. Interns reported using the written handoff, either alone or in combination with other information resources, to provide answers for 7.3% (17) of questions; verbal handoff, either alone or in combination with another resource (excluding written handoff), was reported as a resource for 25.3% (59) of questions. Of note, interns were directly observed to look at the written handoff when answering 21% (49) of questions.

A variety of other resources, including general medical/clinical knowledge, the EMR, and parents or other resources, were used to answer the remaining 67.4% (157) of questions. Intern general medical knowledge (ie, reports of simply knowing the answer to the question in their head[s]) was used to provide answers for 53.2% (124) of questions asked.

Unadjusted univariate regression analyses assessing predictors of written and verbal handoff use are shown in Figure 1. Multivariate logistic regression analyses showed that both dietary questions (odds ratio [OR]: 3.64, 95% confidence interval [CI]: 1.518.76; P=0.004) and interns' consecutive call night (OR: 0.29, 95% CI: 0.090.93; P=0.04) remained significant predictors of written handoff use. After adjusting for risk factors identified above, no differences in written handoff use were seen between the 2 wards.

Figure 1

Multivariate logistic regression for predictors of the verbal handoff use showed that questions regarding patients with longer lengths of stay (OR: 1.97, 95% CI: 1.023.8; P=0.04), those regarding general care plans (OR: 2.07, 95% CI: 1.133.78; P=0.02), as well as those asked by clinical staff (OR: 1.95, 95 CI: 1.043.66; P=0.04), remained significant predictors of reported verbal handoff use.

DISCUSSION

In light of the recent changes in duty hours implemented in July 2011, many pediatric training programs are having trainees work in day and night shifts.[18] Pediatric resident physicians frequently answer questions that pertain to patients handed off between day and night shifts. We found that on average, information provided in the verbal and written handoff was used almost once per hour. Housestaff in our study generally based their answers on information found in 1 or 2 resources, with almost one‐third of all questions involving some use of the written or verbal handoff. Prior research has documented widespread problems with resident handoff practices across programs and a high rate of medical errors due to miscommunications.[3, 4, 19, 20] Given how often information contained within the handoff was used as interns went about their nightly tasks, it is not difficult to understand how errors or omissions in the handoff process may potentially translate into frequent problems in direct patient care.

Trainees reported using written handoff tools to provide answers for 7.3% of questions. As we had suspected, they relied less frequently on their written handoffs as they completed more consecutive call nights. Interestingly, however, even when housestaff did not report using the written handoff, they were observed quite often to look at it before providing an answer. One explanation for this discrepancy between trainee reports and our observations is that the written handoff may serve as a memory tool, even if housestaff do not directly attribute their answers to its content. Our study also found that answers to questions concerning patients' diet and fluids were more likely to be ascribed to information contained in the written handoff. This finding supports the potential value of automated written handoff tools that are linked to the EMR, which can best ensure accuracy of this type of information.

Housestaff in our study also reported using information received during the verbal handoff to answer 1 out of every 4 on‐call questions. Although we did not specifically rate or monitor the quality of verbal handoffs, prior research has demonstrated that resident verbal handoff is often plagued with incomplete and inaccurate data.[3, 4, 19, 21] One investigation found that pediatric interns were prone to overestimating the effectiveness of their verbal handoffs, even as they failed to convey urgent information to their peers.[19] In light of such prior work, our finding that interns frequently rely on the verbal transfer of information supports specific residency training program handoff initiatives that target verbal exchanges.[11, 22, 23]

Although information obtained in the handoff was frequently required by on‐call housestaff, our study found that two‐thirds of all questions were answered using other resources, most often general medical or clinical knowledge. Clearly, background knowledge and experience is fundamental to trainees' ability to perform their jobs. Such reliance on general knowledge for problem solving may not be unique to interns. One recent observational study of senior pediatric cardiac subspecialists reported a high frequency of reliance on their own clinical experience, instinct, or prior training in making clinical decisions.[24] Further investigation may be useful to parse out the exact types of clinical knowledge being used, and may have important implications for how training programs plan for overnight supervision.[25, 26, 27]

Our study has several limitations. First, it was beyond the scope of this study to link housestaff answers to patient outcomes or medical errors. Given the frequency with which the handoff, a known source of vulnerability to medical error, was used by on‐call housestaff, our study suggests that future research evaluating the relationship between questions asked of on‐call housestaff, the answers provided, and downstream patient safety incidents may be merited. Second, our study was conducted in a single pediatric residency program with 1 physician observer midway through the first year of training and only in the early evening hours. This limits the generalizability of our findings, as the use of handoffs to answer on‐call questions may be different at other stages of the training process, within other specialties, or even at different times of the day. We also began our observations after the handoff had taken place; future studies may want to assess how variations in written and verbal handoff processes affect their use. As a final limitation, we note that although collecting information in real time using a direct observational method eliminated the problem of recall bias, there may have been attribution bias.

The results of our study demonstrate that on‐call pediatric housestaff are frequently asked a variety of clinical questions posed by hospital staff, patients, and their families. We found that trainees are apt to rely both on handoff information and other resources to provide answers. By better understanding what resources on‐call housestaff are accessing to answer questions overnight, we may be able to better target interventions needed to improve the availability of patient information, as well as the usefulness of written and verbal handoff tools.[11, 22, 23]

The Centers for Medicaid and Medicare Services' (CMS) Hospital Inpatient Value‐Based Purchasing (VBP) Program, which was signed into law as part of the Patient Protection and Affordable Care Act of 2010, aims to incentivize inpatient providers to deliver high‐value, as opposed to high‐volume, healthcare.[1] Beginning on October 1, 2012, the start of the 2013 fiscal year (FY), hospitals participating in the VBP program became eligible for a variety of performance‐based incentive payments from CMS. These payments are based on an acute care hospital's ability to meet performance measurements in 6 care domains: (1) patient safety, (2) care coordination, (3) clinical processes and outcomes, (4) population or community health, (5) efficiency and cost reduction, and (6) patient‐ and caregiver‐centered experience.[2] The VBP program's ultimate purpose is to enable CMS to improve the health of Medicare beneficiaries by purchasing better care for them at a lower cost. These 3 characteristics of careimproved health, improved care, and lower costsare the foundation of CMS' conception of value.[1, 2] They are closely related to an economic conception of value, which is the difference between an intervention's benefit and its cost.

Although in principle not a new idea, the formal mandate of hospitals to provide high‐value healthcare through financial incentives marks an important change in Medicare and Medicaid policy. In this opportune review of VBP, we first discuss the relevant historical changes in the reimbursement environment of US hospitals that have set the stage for VBP. We then describe the structure of CMS' VBP program, with a focus on which facilities are eligible to participate in the program, the specific outcomes measured and incentivized, how rewards and penalties are allocated, and how the program will be funded. In an effort to anticipate some of the issues that lie ahead, we then highlight a number of potential challenges to the success of VBP, and discuss how VBP will impact the delivery and reimbursement of inpatient care services. We conclude by examining how the VBP program is likely to evolve over time.

HISTORICAL CONTEXT FOR VBP

Over the last decade, CMS has embarked on a number of initiatives to incentivize the provision of higher‐quality and more cost‐effective care. For example, in 2003, CMS implemented a national pay‐for‐performance (P4P) pilot project called the Premier Hospital Quality Incentive Demonstration (HQID).[3, 4] HQID, which ran for 6 years, tracked and rewarded the performance of 216 hospitals in 6 healthcare service domains: (1) acute myocardial infarction (AMI), (2) congestive heart failure (CHF), (3) pneumonia, (4) coronary artery bypass graft surgery, (5) hip and knee replacement surgery, and (6) perioperative management of surgical patients (including prevention of surgical site infections).[4] CMS then introduced its Hospital Compare Web site in 2005 to facilitate public reporting of hospital‐level quality outcomes.[3, 5] This Web site provides the public with access to data on hospital performance across a wide array of measures of process quality, clinical outcomes, spending, and resource utilization.[5] Next, in October 2008, CMS stopped reimbursing hospitals for a number of costly and common hospital‐acquired complications, including hospital‐acquired bloodstream infections and urinary tract infections, patient falls, and pressure ulcers.[3, 6] VBP is the latest and most comprehensive step that CMS has taken in its decade‐long effort to shift from volume to value‐based compensation for inpatient care.

Although CMS appears fully invested in using performance incentives to increase healthcare value, existing evidence of the effects of P4P on patient outcomes remains quite mixed.[7] On one hand, an analysis of an inpatient P4P program sponsored by the United Kingdom's National Health Service's (NHS) suggests that P4P may improve quality and save lives; indeed, hospitals that participated in the NHS P4P program significantly reduced inpatient mortality from pneumonia, saving an estimated 890 lives.[8] Additional empirical work suggests that the HQID was also associated with early improvements in healthcare quality.[9] However, a subsequent long‐term analysis found that participation in HQID had no discernible effect on 30‐day mortality rates.[10] Moreover, a meta‐analysis of P4P incentives for individual practitioners found few methodologically robust studies of P4P for clinicians and concluded that P4P's effects on individual practice patterns and outcomes remain largely uncertain.[11]

VBP: STRUCTURE AND DESIGN

This section reviews the structure of the VBP program. We describe current VBP eligibility criteria and sources of funding for the program, how hospitals participating in VBP are evaluated, and how VBP incentives for FY 2013 have been calculated.

CHALLENGES OF VBP

As the Medicare VBP program evolves, and hospitals confront ever‐larger financial incentives to deliver high‐value as opposed to high‐volume care, it will be important to recognize limitations of the VBP program as they arise. Here we briefly discuss several conceptual and implementation challenges that physicians and policymakers should consider when assessing the merits of VBP in promoting high‐quality healthcare.

High‐Value Care Is Not Always Low‐Cost Care

Not surprisingly, the clinical process measures included in CMS' hospital VBP program evaluate a select and relatively small group of high‐value and low‐cost interventions (eg, appropriate administration of antibiotics and tight control of serum glucose in surgical patients). However, an important body of work has demonstrated that high‐cost care (eg, intensive inpatient hospital care for common acute medical conditions) may also be highly valuable in terms of improving survival.[20, 27, 28, 29, 30] As the hospital VBP program evolves, its overseers will need to consider whether to include additional incentives for high‐value high‐cost healthcare services. Such considerations will likely become increasingly salient as healthcare delivery organizations move toward capitated delivery models. In particular, the VBP program's Medicare Spending Per Beneficiary measure, which quantifies inpatient and subsequent outpatient spending per beneficiary after a given hospitalization episode, will need to distinguish between higher‐spending hospitals that provide highly effective care (eg, care that reduces mortality and readmissions) and facilities that provide less‐effective care.

FUTURE OF VBP

Although the future of VBP is unknown, CMS is likely to modify the program in a number of ways over the next 3 to 5 years. First, CMS will likely expand the breadth and focus of incentivized measures in the VBP program. In FY 2014, for example, CMS is adding a set of 3, 30‐day mortality outcome measures to VBP: 30‐day risk‐adjusted mortality for AMI, CHF, and pneumonia.[1] A hospital's performance with respect to these outcomes will represent 25% of its total performance score in 2014, whereas the clinical process of care and patient experience of care domains will account for 45% and 30% of this score, respectively. In 2015, patient experience and outcome measures will account for 30% each in a hospital's performance score, whereas process and efficiency measures will each account for 20% of this score, respectively. The composition of this performance score evidences a shift away from rewarding process‐based measures and toward incentivizing measures of clinical outcomes and patient satisfaction, the latter of which may be highly subjective and more representative of a hospital's catchment population than of a hospital's care itself.[31] Additional measures in the domains of patient safety, care coordination, population and community health, emergency room wait times, and cost control may also be added to the VBP program in FY 2015 to FY 2017. Furthermore, CMS will continue to reevaluate the appropriateness of measures that are already included in VBP and will stop incentivizing measures that have become topped out, or are no longer supported by the National Quality Forum.[1, 13]

Second, CMS has established an annual gradual increase of 0.25% in the percentage of each hospital's inpatient DRG‐based payment that is at stake under VBP. In FY 2014, for example, participating hospitals will be required to contribute 1.25% of inpatient DRG payments to the VBP program. This percentage is likely to increase to 2% or more by 2017.[1, 32]

Third, expansions of the VBP program complement a number of other quality improvement efforts overseen by CMS, including the Hospital Readmissions Reduction Program. Effective for discharges beginning on October 1, 2012, hospitals with excess readmissions for AMI, CHF, and pneumonia are at risk for reimbursement reductions for all Medicare admissions in proportion to the rate of excess rehospitalizations. Some of the same concerns about the hospital VBP program outlined above have also been raised for this program, namely, whether readmission penalties will be large enough to impact hospital behavior, whether readmissions are even preventable,[33, 34] and whether adjustments in hospital‐level policies will reduce admissions that are known to be heavily influenced by patient economic and social factors that are outside of a hospital's control.[35, 36] Despite the limitations of VBP and the challenges that lie ahead, there is optimism that rewarding hospitals that provide high‐value rather than high‐volume care will not only improve outcomes of hospitalized patients in the United States, but will potentially be able to do so at a lower cost. Encouraging hospitals to improve their quality of care may also have important spillover effects on other healthcare domains. For example, hospitals that adopt systems to ensure prompt delivery of antibiotics to patients with pneumonia may also observe positive spillover effects with the prompt antibiotic management of other acute infectious illnesses that are not covered by VBP. VBP may have spillover effects on medical malpractice liability and defensive medicine as well. Indeed, financial incentives to practice higher‐quality evidenced‐based care may reduce medical malpractice liability and defensive medicine.

The government's ultimate goal in implementing VBP is to identify a broad and clinically relevant set of outcome measures that can be used to incentivize hospitals to deliver high‐quality as opposed to high‐volume healthcare. The first wave of outcome measures has already been instituted. It remains to be seen whether the incentive rewards of Medicare's hospital VBP program will be large enough that hospitals feel compelled to improve and compete for them.

The Centers for Medicaid and Medicare Services' (CMS) Hospital Inpatient Value‐Based Purchasing (VBP) Program, which was signed into law as part of the Patient Protection and Affordable Care Act of 2010, aims to incentivize inpatient providers to deliver high‐value, as opposed to high‐volume, healthcare.[1] Beginning on October 1, 2012, the start of the 2013 fiscal year (FY), hospitals participating in the VBP program became eligible for a variety of performance‐based incentive payments from CMS. These payments are based on an acute care hospital's ability to meet performance measurements in 6 care domains: (1) patient safety, (2) care coordination, (3) clinical processes and outcomes, (4) population or community health, (5) efficiency and cost reduction, and (6) patient‐ and caregiver‐centered experience.[2] The VBP program's ultimate purpose is to enable CMS to improve the health of Medicare beneficiaries by purchasing better care for them at a lower cost. These 3 characteristics of careimproved health, improved care, and lower costsare the foundation of CMS' conception of value.[1, 2] They are closely related to an economic conception of value, which is the difference between an intervention's benefit and its cost.

Although in principle not a new idea, the formal mandate of hospitals to provide high‐value healthcare through financial incentives marks an important change in Medicare and Medicaid policy. In this opportune review of VBP, we first discuss the relevant historical changes in the reimbursement environment of US hospitals that have set the stage for VBP. We then describe the structure of CMS' VBP program, with a focus on which facilities are eligible to participate in the program, the specific outcomes measured and incentivized, how rewards and penalties are allocated, and how the program will be funded. In an effort to anticipate some of the issues that lie ahead, we then highlight a number of potential challenges to the success of VBP, and discuss how VBP will impact the delivery and reimbursement of inpatient care services. We conclude by examining how the VBP program is likely to evolve over time.

HISTORICAL CONTEXT FOR VBP

Over the last decade, CMS has embarked on a number of initiatives to incentivize the provision of higher‐quality and more cost‐effective care. For example, in 2003, CMS implemented a national pay‐for‐performance (P4P) pilot project called the Premier Hospital Quality Incentive Demonstration (HQID).[3, 4] HQID, which ran for 6 years, tracked and rewarded the performance of 216 hospitals in 6 healthcare service domains: (1) acute myocardial infarction (AMI), (2) congestive heart failure (CHF), (3) pneumonia, (4) coronary artery bypass graft surgery, (5) hip and knee replacement surgery, and (6) perioperative management of surgical patients (including prevention of surgical site infections).[4] CMS then introduced its Hospital Compare Web site in 2005 to facilitate public reporting of hospital‐level quality outcomes.[3, 5] This Web site provides the public with access to data on hospital performance across a wide array of measures of process quality, clinical outcomes, spending, and resource utilization.[5] Next, in October 2008, CMS stopped reimbursing hospitals for a number of costly and common hospital‐acquired complications, including hospital‐acquired bloodstream infections and urinary tract infections, patient falls, and pressure ulcers.[3, 6] VBP is the latest and most comprehensive step that CMS has taken in its decade‐long effort to shift from volume to value‐based compensation for inpatient care.

Although CMS appears fully invested in using performance incentives to increase healthcare value, existing evidence of the effects of P4P on patient outcomes remains quite mixed.[7] On one hand, an analysis of an inpatient P4P program sponsored by the United Kingdom's National Health Service's (NHS) suggests that P4P may improve quality and save lives; indeed, hospitals that participated in the NHS P4P program significantly reduced inpatient mortality from pneumonia, saving an estimated 890 lives.[8] Additional empirical work suggests that the HQID was also associated with early improvements in healthcare quality.[9] However, a subsequent long‐term analysis found that participation in HQID had no discernible effect on 30‐day mortality rates.[10] Moreover, a meta‐analysis of P4P incentives for individual practitioners found few methodologically robust studies of P4P for clinicians and concluded that P4P's effects on individual practice patterns and outcomes remain largely uncertain.[11]

VBP: STRUCTURE AND DESIGN

This section reviews the structure of the VBP program. We describe current VBP eligibility criteria and sources of funding for the program, how hospitals participating in VBP are evaluated, and how VBP incentives for FY 2013 have been calculated.