Editorial

The Affordable Care Act has made hospital readmissions a major public policy target by tying Medicare hospital payments to readmission rates for certain diseases. Since then, debate has spiked over the factors contributing to hospital readmissions, with particular attention being paid to the impact of socioeconomic status and access to care. The Massachusetts healthcare reform of 2006 is a useful natural experiment to help disentangle some of these effects. Although reform did little to change patients' income, education, health literacy, or other determinants of socioeconomic status, it did dramatically reduce uninsurance rates. State‐wide uninsurance rates dropped from 8.4% prereform to 3.4% postreform.[1] Most important, a gain in insurance appeared to translate to genuine improvements in access to outpatient and preventive care. Massachusetts residents postreform were more likely to report a usual source of care, were more likely to have had outpatient office visits, and less likely to use the emergency department.[1, 2, 3, 4] Thus, the 2006 Massachusetts health reform legislation appears to have genuinely increased access both to insurance and to outpatient care, while reducing the need for preventable hospital‐based care.

Contrary to popular belief, patients without insurance have low unadjusted readmission rates for most conditions, often even lower than rates among those who have private insurance, perhaps because uninsured patients tend to be younger and healthier than the general population, or perhaps because they avoid costly healthcare services such as hospitalizations and rehospitalizations.[5, 6] A priori, it is therefore possible that obtaining insurance would encourage such patients to seek care, increasing readmission rates. On the other hand, access to insurance might increase use of outpatient preventive and follow‐up care and treatments that would reduce readmission risk. A recent study by Lasser et al. found that, on a patient level, healthcare reform was associated with fairly minimal changes in readmission rates in Massachusetts, compared with trends in states not adopting health insurance reform.[7] The authors further found that there was no improvement in readmission rates among Hispanic and black patients in Massachusetts compared with other states, nor was there differential improvement in counties with the highest baseline uninsurance rates compared to other Massachusetts counties.

The question raised by Chen et al. in this issue of the Journal of Hospital Medicine, however, is whether the Massachusetts reform affected hospital‐level aggregate readmission rates, not individual patient‐level risk of readmission.[8] Because public policy regarding readmissions is directed at hospitals, not patients, a hospital‐level examination can shed light on likely implications for hospitals of new insurance gains prompted by the Affordable Care Act. Some commentators have expressed concern that payment penalty programs for excess readmissions may harm safety‐net hospitals.[9] Although uninsured patients may have low readmission rates, hospitals with high proportions of uninsured patients (safety‐net hospitals) tend to have slightly higher readmission rates than other hospitals, probably because they also have higher proportions of highreadmission‐risk Medicaid patients. Reducing the rate of uninsurance at these hospitals could theoretically have a number of different hospital‐level effects. Patients obtaining insurance might elect to seek care elsewhere, changing the distribution of patients among hospitals, and potentially affecting readmission rates. Hospitals might be more prone to readmit insured patients, increasing their readmission rate if more of their patients gain insurance. They might use new revenue from newly insured patients to provide better care‐coordination services, potentially reducing readmission risk, or as happened in Massachusetts, safety‐net hospitals may find themselves unexpectedly losing revenue because of elimination of other subsidies, potentially reducing their ability to provide care transition services.[10]

Examining the effect of health insurance reform on hospital readmission rates empirically, Chen et al. find that readmission rates rose 0.6 percentage points in the group of hospitals with the highest prereform rates of uninsurance, but that after risk adjustment for age, gender, race, and comorbidity, there was no significant change relative to other hospitals. What accounts for these results? One possibility is that some patients gaining health insurance who previously received care at safety‐net hospitals began to seek care at other institutions, but that the redistribution of patients occurred more among healthier newly insured patients than among the more chronically ill newly insured. In such a case, the hospitals with the highest prereform uninsurance rates might be left with a sicker population, increasing their unadjusted readmission rates but leaving adjusted readmission rates unchanged. One study did show a 2.3% decrease in the annual volume at Massachusetts safety‐net hospitals, compared to a 1.9% increase at nonsafety‐net hospitals postreform. The relative difference, however, was not statistically significant.[10] Nonetheless, Chen et al. did find that the comorbidity index increased postreform only for patients in the highest prereform uninsurance quartile, suggesting that their patient population might have become sicker. An alternate hypothesis is that safety‐net hospitals were the most financially adversely affected by the Massachusetts reform and may therefore have cut services that would have reduced readmission risk. However, the fact that risk‐adjusted readmission rates were unchanged makes that hypothesis less probable. Finally, the authors suggest that the population who were newly insured were themselves sicker than the general population, as illustrated by the increase in comorbidity postreform. Given the national profile of uninsured patients as generally healthier than insured patients, however, it is unlikely that Massachusetts' newly insured patients were much less healthy than previously insured patients, especially because Massachusetts enrolled a very large fraction of its previously uninsured patients. It is more probable that either there was a shift of healthier patients out of the safety‐net hospitals, that hospital billing departments began paying more attention to documenting comorbidity in patients for whom hospitalizations would now be reimbursed, or that new access to care enabled preexisting conditions to be diagnosed and documented.

So what does Chen et al.'s study mean for hospitals as they face an influx of newly insured patients through the Affordable Care Act health exchanges or Medicaid expansions? First, it is important to note that although risk adjustment eliminated any change in readmission rates postreform, the risk adjustment model used by these investigators is not the same as that used by Medicare. Medicare does not adjust for race, as the investigators did. Therefore, if some of the change in readmission rate is driven by changes in patient population at safety‐net hospitals, and if such moves occur differentially in different patient populations, increases in readmission rates at safety‐net hospitals might still be present even after Medicare‐type risk adjustment. Even so, the impact on Medicare‐driven readmission penalties for hospitals is likely to be fairly minimal, because Medicare coverage is relatively unaffected by the Affordable Care Act and by the Massachusetts reforms. Medicare enrollment rates are driven by age and disability, neither of which is directly relevant to universal coverage schemes. Studies of Massachusetts reform found very little change in Medicare enrollment postreform despite large declines in uninsurance.[2] Because payment penalties are determined based purely on readmission rates among Medicare patients over 65 years old, changes in overall hospital readmissions have little financial consequences for hospitals at the present moment. In the future, however, if private insurers and state Medicaid programs begin to focus on readmissions as well, then the situation may change.

Overall, Chen et al.'s study lends heartening support to accumulating evidence that the lessons of the RAND Corporation's health insurance experiment may not apply to high‐intensity care such as hospitalizationsat least in Massachusetts.[2, 11] Increasingly, it is becoming clear that gaining insurance does not necessarily mean receipt of more inpatient care. If you build it, they may not come.

Disclosures: Nothing to report.

The Affordable Care Act has made hospital readmissions a major public policy target by tying Medicare hospital payments to readmission rates for certain diseases. Since then, debate has spiked over the factors contributing to hospital readmissions, with particular attention being paid to the impact of socioeconomic status and access to care. The Massachusetts healthcare reform of 2006 is a useful natural experiment to help disentangle some of these effects. Although reform did little to change patients' income, education, health literacy, or other determinants of socioeconomic status, it did dramatically reduce uninsurance rates. State‐wide uninsurance rates dropped from 8.4% prereform to 3.4% postreform.[1] Most important, a gain in insurance appeared to translate to genuine improvements in access to outpatient and preventive care. Massachusetts residents postreform were more likely to report a usual source of care, were more likely to have had outpatient office visits, and less likely to use the emergency department.[1, 2, 3, 4] Thus, the 2006 Massachusetts health reform legislation appears to have genuinely increased access both to insurance and to outpatient care, while reducing the need for preventable hospital‐based care.

Contrary to popular belief, patients without insurance have low unadjusted readmission rates for most conditions, often even lower than rates among those who have private insurance, perhaps because uninsured patients tend to be younger and healthier than the general population, or perhaps because they avoid costly healthcare services such as hospitalizations and rehospitalizations.[5, 6] A priori, it is therefore possible that obtaining insurance would encourage such patients to seek care, increasing readmission rates. On the other hand, access to insurance might increase use of outpatient preventive and follow‐up care and treatments that would reduce readmission risk. A recent study by Lasser et al. found that, on a patient level, healthcare reform was associated with fairly minimal changes in readmission rates in Massachusetts, compared with trends in states not adopting health insurance reform.[7] The authors further found that there was no improvement in readmission rates among Hispanic and black patients in Massachusetts compared with other states, nor was there differential improvement in counties with the highest baseline uninsurance rates compared to other Massachusetts counties.

The question raised by Chen et al. in this issue of the Journal of Hospital Medicine, however, is whether the Massachusetts reform affected hospital‐level aggregate readmission rates, not individual patient‐level risk of readmission.[8] Because public policy regarding readmissions is directed at hospitals, not patients, a hospital‐level examination can shed light on likely implications for hospitals of new insurance gains prompted by the Affordable Care Act. Some commentators have expressed concern that payment penalty programs for excess readmissions may harm safety‐net hospitals.[9] Although uninsured patients may have low readmission rates, hospitals with high proportions of uninsured patients (safety‐net hospitals) tend to have slightly higher readmission rates than other hospitals, probably because they also have higher proportions of highreadmission‐risk Medicaid patients. Reducing the rate of uninsurance at these hospitals could theoretically have a number of different hospital‐level effects. Patients obtaining insurance might elect to seek care elsewhere, changing the distribution of patients among hospitals, and potentially affecting readmission rates. Hospitals might be more prone to readmit insured patients, increasing their readmission rate if more of their patients gain insurance. They might use new revenue from newly insured patients to provide better care‐coordination services, potentially reducing readmission risk, or as happened in Massachusetts, safety‐net hospitals may find themselves unexpectedly losing revenue because of elimination of other subsidies, potentially reducing their ability to provide care transition services.[10]

Examining the effect of health insurance reform on hospital readmission rates empirically, Chen et al. find that readmission rates rose 0.6 percentage points in the group of hospitals with the highest prereform rates of uninsurance, but that after risk adjustment for age, gender, race, and comorbidity, there was no significant change relative to other hospitals. What accounts for these results? One possibility is that some patients gaining health insurance who previously received care at safety‐net hospitals began to seek care at other institutions, but that the redistribution of patients occurred more among healthier newly insured patients than among the more chronically ill newly insured. In such a case, the hospitals with the highest prereform uninsurance rates might be left with a sicker population, increasing their unadjusted readmission rates but leaving adjusted readmission rates unchanged. One study did show a 2.3% decrease in the annual volume at Massachusetts safety‐net hospitals, compared to a 1.9% increase at nonsafety‐net hospitals postreform. The relative difference, however, was not statistically significant.[10] Nonetheless, Chen et al. did find that the comorbidity index increased postreform only for patients in the highest prereform uninsurance quartile, suggesting that their patient population might have become sicker. An alternate hypothesis is that safety‐net hospitals were the most financially adversely affected by the Massachusetts reform and may therefore have cut services that would have reduced readmission risk. However, the fact that risk‐adjusted readmission rates were unchanged makes that hypothesis less probable. Finally, the authors suggest that the population who were newly insured were themselves sicker than the general population, as illustrated by the increase in comorbidity postreform. Given the national profile of uninsured patients as generally healthier than insured patients, however, it is unlikely that Massachusetts' newly insured patients were much less healthy than previously insured patients, especially because Massachusetts enrolled a very large fraction of its previously uninsured patients. It is more probable that either there was a shift of healthier patients out of the safety‐net hospitals, that hospital billing departments began paying more attention to documenting comorbidity in patients for whom hospitalizations would now be reimbursed, or that new access to care enabled preexisting conditions to be diagnosed and documented.

So what does Chen et al.'s study mean for hospitals as they face an influx of newly insured patients through the Affordable Care Act health exchanges or Medicaid expansions? First, it is important to note that although risk adjustment eliminated any change in readmission rates postreform, the risk adjustment model used by these investigators is not the same as that used by Medicare. Medicare does not adjust for race, as the investigators did. Therefore, if some of the change in readmission rate is driven by changes in patient population at safety‐net hospitals, and if such moves occur differentially in different patient populations, increases in readmission rates at safety‐net hospitals might still be present even after Medicare‐type risk adjustment. Even so, the impact on Medicare‐driven readmission penalties for hospitals is likely to be fairly minimal, because Medicare coverage is relatively unaffected by the Affordable Care Act and by the Massachusetts reforms. Medicare enrollment rates are driven by age and disability, neither of which is directly relevant to universal coverage schemes. Studies of Massachusetts reform found very little change in Medicare enrollment postreform despite large declines in uninsurance.[2] Because payment penalties are determined based purely on readmission rates among Medicare patients over 65 years old, changes in overall hospital readmissions have little financial consequences for hospitals at the present moment. In the future, however, if private insurers and state Medicaid programs begin to focus on readmissions as well, then the situation may change.

Overall, Chen et al.'s study lends heartening support to accumulating evidence that the lessons of the RAND Corporation's health insurance experiment may not apply to high‐intensity care such as hospitalizationsat least in Massachusetts.[2, 11] Increasingly, it is becoming clear that gaining insurance does not necessarily mean receipt of more inpatient care. If you build it, they may not come.

Disclosures: Nothing to report.

The Affordable Care Act has made hospital readmissions a major public policy target by tying Medicare hospital payments to readmission rates for certain diseases. Since then, debate has spiked over the factors contributing to hospital readmissions, with particular attention being paid to the impact of socioeconomic status and access to care. The Massachusetts healthcare reform of 2006 is a useful natural experiment to help disentangle some of these effects. Although reform did little to change patients' income, education, health literacy, or other determinants of socioeconomic status, it did dramatically reduce uninsurance rates. State‐wide uninsurance rates dropped from 8.4% prereform to 3.4% postreform.[1] Most important, a gain in insurance appeared to translate to genuine improvements in access to outpatient and preventive care. Massachusetts residents postreform were more likely to report a usual source of care, were more likely to have had outpatient office visits, and less likely to use the emergency department.[1, 2, 3, 4] Thus, the 2006 Massachusetts health reform legislation appears to have genuinely increased access both to insurance and to outpatient care, while reducing the need for preventable hospital‐based care.

Contrary to popular belief, patients without insurance have low unadjusted readmission rates for most conditions, often even lower than rates among those who have private insurance, perhaps because uninsured patients tend to be younger and healthier than the general population, or perhaps because they avoid costly healthcare services such as hospitalizations and rehospitalizations.[5, 6] A priori, it is therefore possible that obtaining insurance would encourage such patients to seek care, increasing readmission rates. On the other hand, access to insurance might increase use of outpatient preventive and follow‐up care and treatments that would reduce readmission risk. A recent study by Lasser et al. found that, on a patient level, healthcare reform was associated with fairly minimal changes in readmission rates in Massachusetts, compared with trends in states not adopting health insurance reform.[7] The authors further found that there was no improvement in readmission rates among Hispanic and black patients in Massachusetts compared with other states, nor was there differential improvement in counties with the highest baseline uninsurance rates compared to other Massachusetts counties.

The question raised by Chen et al. in this issue of the Journal of Hospital Medicine, however, is whether the Massachusetts reform affected hospital‐level aggregate readmission rates, not individual patient‐level risk of readmission.[8] Because public policy regarding readmissions is directed at hospitals, not patients, a hospital‐level examination can shed light on likely implications for hospitals of new insurance gains prompted by the Affordable Care Act. Some commentators have expressed concern that payment penalty programs for excess readmissions may harm safety‐net hospitals.[9] Although uninsured patients may have low readmission rates, hospitals with high proportions of uninsured patients (safety‐net hospitals) tend to have slightly higher readmission rates than other hospitals, probably because they also have higher proportions of highreadmission‐risk Medicaid patients. Reducing the rate of uninsurance at these hospitals could theoretically have a number of different hospital‐level effects. Patients obtaining insurance might elect to seek care elsewhere, changing the distribution of patients among hospitals, and potentially affecting readmission rates. Hospitals might be more prone to readmit insured patients, increasing their readmission rate if more of their patients gain insurance. They might use new revenue from newly insured patients to provide better care‐coordination services, potentially reducing readmission risk, or as happened in Massachusetts, safety‐net hospitals may find themselves unexpectedly losing revenue because of elimination of other subsidies, potentially reducing their ability to provide care transition services.[10]

Examining the effect of health insurance reform on hospital readmission rates empirically, Chen et al. find that readmission rates rose 0.6 percentage points in the group of hospitals with the highest prereform rates of uninsurance, but that after risk adjustment for age, gender, race, and comorbidity, there was no significant change relative to other hospitals. What accounts for these results? One possibility is that some patients gaining health insurance who previously received care at safety‐net hospitals began to seek care at other institutions, but that the redistribution of patients occurred more among healthier newly insured patients than among the more chronically ill newly insured. In such a case, the hospitals with the highest prereform uninsurance rates might be left with a sicker population, increasing their unadjusted readmission rates but leaving adjusted readmission rates unchanged. One study did show a 2.3% decrease in the annual volume at Massachusetts safety‐net hospitals, compared to a 1.9% increase at nonsafety‐net hospitals postreform. The relative difference, however, was not statistically significant.[10] Nonetheless, Chen et al. did find that the comorbidity index increased postreform only for patients in the highest prereform uninsurance quartile, suggesting that their patient population might have become sicker. An alternate hypothesis is that safety‐net hospitals were the most financially adversely affected by the Massachusetts reform and may therefore have cut services that would have reduced readmission risk. However, the fact that risk‐adjusted readmission rates were unchanged makes that hypothesis less probable. Finally, the authors suggest that the population who were newly insured were themselves sicker than the general population, as illustrated by the increase in comorbidity postreform. Given the national profile of uninsured patients as generally healthier than insured patients, however, it is unlikely that Massachusetts' newly insured patients were much less healthy than previously insured patients, especially because Massachusetts enrolled a very large fraction of its previously uninsured patients. It is more probable that either there was a shift of healthier patients out of the safety‐net hospitals, that hospital billing departments began paying more attention to documenting comorbidity in patients for whom hospitalizations would now be reimbursed, or that new access to care enabled preexisting conditions to be diagnosed and documented.

So what does Chen et al.'s study mean for hospitals as they face an influx of newly insured patients through the Affordable Care Act health exchanges or Medicaid expansions? First, it is important to note that although risk adjustment eliminated any change in readmission rates postreform, the risk adjustment model used by these investigators is not the same as that used by Medicare. Medicare does not adjust for race, as the investigators did. Therefore, if some of the change in readmission rate is driven by changes in patient population at safety‐net hospitals, and if such moves occur differentially in different patient populations, increases in readmission rates at safety‐net hospitals might still be present even after Medicare‐type risk adjustment. Even so, the impact on Medicare‐driven readmission penalties for hospitals is likely to be fairly minimal, because Medicare coverage is relatively unaffected by the Affordable Care Act and by the Massachusetts reforms. Medicare enrollment rates are driven by age and disability, neither of which is directly relevant to universal coverage schemes. Studies of Massachusetts reform found very little change in Medicare enrollment postreform despite large declines in uninsurance.[2] Because payment penalties are determined based purely on readmission rates among Medicare patients over 65 years old, changes in overall hospital readmissions have little financial consequences for hospitals at the present moment. In the future, however, if private insurers and state Medicaid programs begin to focus on readmissions as well, then the situation may change.

Overall, Chen et al.'s study lends heartening support to accumulating evidence that the lessons of the RAND Corporation's health insurance experiment may not apply to high‐intensity care such as hospitalizationsat least in Massachusetts.[2, 11] Increasingly, it is becoming clear that gaining insurance does not necessarily mean receipt of more inpatient care. If you build it, they may not come.

Disclosures: Nothing to report.

Nurse practitioners (NPs) and physician assistants (PAs) have been caring for patients since the mid‐1960s.[1] Although both roles grew out of a need for more primary care providers, more recently there has been an increase in the utilization of NPs and PAs in acute care roles. This meteoric rise of advanced practice providers in the inpatient setting has been driven by stressors from residency work‐hour reforms and from growing financial pressures in healthcare systems, where NPs and PAs are seen as less expensive alternatives.[2, 3] Inadequate physician supply to meet the needs of growing healthcare service is also a driving factor. Despite increasing numbers of enrollees and increasing numbers of medical schools, many sources estimate a physician shortage of 50,000 providers by year 2025.[4] To address this growing shortage, the number of NP and PA providers in acute care continues to grow as Kartha and colleagues[5] clearly demonstrate in their study, published in this issue of Journal of Hospital Medicine. Their research shows that within hospitals in the Veterans Health Administration (VHA)the largest coordinated healthcare association in the United Statesfully half of all inpatient medical teams are utilizing NPs and PAs in some capacity, most commonly in staffing models working directly with attending physicians or on teams with housestaff.[5]

Many different practice models exist that incorporate NPs and PAs into acute care settings, including models in general medicine and intensive care settings, as well as in specialty care populations such as patients with diabetes or congestive heart failure.[1, 6] Few studies, however, delineate specific roles for NPs or PAs in inpatient acute care or provide outcomes‐based evidence in support of the proposed models. This is in contrast to research available regarding NP and PA staffing models in the outpatient setting.[7, 8] In the current study, Kartha et al.[5] shed light on the use of NPs and PAs in inpatient medical units at the VHA. Their findings show that the majority of NPs and PAs on the inpatient team function mostly autonomously and perform tasks including performing histories and physicals, writing progress notes, placing orders, and communicating with primary care providers and consultants. Almost half also serve on hospital committees and participate in quality improvement activities. Interestingly, although the training and regulation of NPs and PAs differ considerably,[1] Kartha et al. found that the scope of practice of these providers is generally the same. PAs are more likely to perform procedures and teach nonphysician students but otherwise function similarly to NPs. The clinical workload for NPs and PAs also does not differ, with an average of 6.5 patients seen per day. This information is crucial when analyzing the cost‐effectiveness of these providers, especially in light of evidence suggesting that hospitalist physicians typically care for approximately twice as many patients.[9]

Although Kartha et al.[5] focus primarily on describing the scope of NPs and PAs in hospital medicine, they also report on outcomes. Their findings show that presence of NPs and PAs on inpatient teams did not alter patient or nurse satisfaction nor were there any consistent improvements in the perception of care coordination. Of note, assessment of care coordination was based on survey responses from nurse managers and chiefs of medicine, individuals who are not necessarily direct members of the inpatient team, thus questioning the validity of this measure. Other studies on NP/PA models have also focused on patient‐centered outcomes. A study by Roy et al.[10] found that an inpatient PA‐run service supervised by hospitalists was comparable with a traditional resident‐run service, with no significant differences in risk‐adjusted length of stay (LOS), mortality, intensive care unit (ICU) transfers, or hospital readmissions. Although total costs were lower on the PA service, this difference was minimal. Gershengorn et al.[11] examined the impact of nonphysician staffing in an ICU setting and again found equivalent care. In this study, an ICU team staffed by NPs and PAs had similar hospital mortality and LOS as compared with a standard housestaff ICU service. Both these studies have limitations in that they are retrospective analyses rather than randomized controlled trials, and they were conducted at academic medical centers, thus narrowing their generalizability. Moreover, purity of data is difficult to achieve, as few systems exist where NPs and PAs are the sole providers managing patients without interaction or coverage from physician colleagues.

Given the considerable presence of NPs and PAs in acute care hospitals as documented by Kartha et al.,[5] providing appropriate training in hospital medicine to these clinicians is important. A study by Dhuper and Choksi[12] evaluated a 2‐year PA postgraduate training program in hospital medicine. PAs spent 40 hours per week on direct patient care while rotating on general medical floors and ICUs, along with 16 hours per week in didactic instruction. When compared with a traditional 3‐year medical residency at the same institution, the PA training program had similar outcomes on patient care including similar number of adverse events, readmissions, and patient satisfaction scores. A more formal postgraduate training program for PAs has been established at the Mayo Clinic Arizona.[13] This 12‐month program, based on the Society of Hospital Medicine's (SHM) Core Competencies, consists of general medicine and inpatient medical subspecialty rotations, didactic instruction, and self‐directed teaching modules to learn systems‐based practices. The Adult Hospital Medicine Boot Camp, sponsored by the SHM and the American Academy of Physician Assistants, is another training opportunity for both NPs and PAs who currently work in or are planning to practice hospital medicine.[14] Finally, in accordance with the move to provide standardized training for providers who practice in acute care settings, professional nursing organizations have developed the Consensus Model for Advanced Practice Registered Nurse Regulation that contains recommendations ensuring similar education and licensure requirements for those who practice in acute care.[15]

Although the optimal utilization of NPs and PAs in hospital medicine is still unknown, the reality is that the number of NPs and PAs actually working in this capacity is significant, as Kartha and his colleagues report.[5] A study of academic medical centers also found that among the institutions that responded to a survey, 31% and 42% used PAs and NPs, respectively, in hospitalist roles.[16] Current evidence suggests that NP‐ and PA‐based care with physician collaboration in an inpatient setting can result in comparable outcomes with physician‐only care models. However, much of this evidence is of poor quality or cannot be generalized to all settings. Kartha et al.[5] have provided a good first step in describing the role of NPs and PAs within hospital medicine. Though their education and training backgrounds are different, the ultimate scope of practice for these 2 groups of providers is very similar. Future research should focus on defining the best practice model for utilization of NPs and PAs in hospital medicine with emphasis on measurable goals. These can include standard outcomes such as LOS but also specific measures of quality and safety such as days of urinary catheter use or percentage of patients receiving venous thromboprophylaxis.[17] By understanding the scope of NP and PA practice, collecting more robust data regarding outcomes, and emphasizing training for NPs and PAs within hospital medicine, there is opportunity to impact the quality and efficiency of care of hospitalized patients.

Nurse practitioners (NPs) and physician assistants (PAs) have been caring for patients since the mid‐1960s.[1] Although both roles grew out of a need for more primary care providers, more recently there has been an increase in the utilization of NPs and PAs in acute care roles. This meteoric rise of advanced practice providers in the inpatient setting has been driven by stressors from residency work‐hour reforms and from growing financial pressures in healthcare systems, where NPs and PAs are seen as less expensive alternatives.[2, 3] Inadequate physician supply to meet the needs of growing healthcare service is also a driving factor. Despite increasing numbers of enrollees and increasing numbers of medical schools, many sources estimate a physician shortage of 50,000 providers by year 2025.[4] To address this growing shortage, the number of NP and PA providers in acute care continues to grow as Kartha and colleagues[5] clearly demonstrate in their study, published in this issue of Journal of Hospital Medicine. Their research shows that within hospitals in the Veterans Health Administration (VHA)the largest coordinated healthcare association in the United Statesfully half of all inpatient medical teams are utilizing NPs and PAs in some capacity, most commonly in staffing models working directly with attending physicians or on teams with housestaff.[5]

Many different practice models exist that incorporate NPs and PAs into acute care settings, including models in general medicine and intensive care settings, as well as in specialty care populations such as patients with diabetes or congestive heart failure.[1, 6] Few studies, however, delineate specific roles for NPs or PAs in inpatient acute care or provide outcomes‐based evidence in support of the proposed models. This is in contrast to research available regarding NP and PA staffing models in the outpatient setting.[7, 8] In the current study, Kartha et al.[5] shed light on the use of NPs and PAs in inpatient medical units at the VHA. Their findings show that the majority of NPs and PAs on the inpatient team function mostly autonomously and perform tasks including performing histories and physicals, writing progress notes, placing orders, and communicating with primary care providers and consultants. Almost half also serve on hospital committees and participate in quality improvement activities. Interestingly, although the training and regulation of NPs and PAs differ considerably,[1] Kartha et al. found that the scope of practice of these providers is generally the same. PAs are more likely to perform procedures and teach nonphysician students but otherwise function similarly to NPs. The clinical workload for NPs and PAs also does not differ, with an average of 6.5 patients seen per day. This information is crucial when analyzing the cost‐effectiveness of these providers, especially in light of evidence suggesting that hospitalist physicians typically care for approximately twice as many patients.[9]

Although Kartha et al.[5] focus primarily on describing the scope of NPs and PAs in hospital medicine, they also report on outcomes. Their findings show that presence of NPs and PAs on inpatient teams did not alter patient or nurse satisfaction nor were there any consistent improvements in the perception of care coordination. Of note, assessment of care coordination was based on survey responses from nurse managers and chiefs of medicine, individuals who are not necessarily direct members of the inpatient team, thus questioning the validity of this measure. Other studies on NP/PA models have also focused on patient‐centered outcomes. A study by Roy et al.[10] found that an inpatient PA‐run service supervised by hospitalists was comparable with a traditional resident‐run service, with no significant differences in risk‐adjusted length of stay (LOS), mortality, intensive care unit (ICU) transfers, or hospital readmissions. Although total costs were lower on the PA service, this difference was minimal. Gershengorn et al.[11] examined the impact of nonphysician staffing in an ICU setting and again found equivalent care. In this study, an ICU team staffed by NPs and PAs had similar hospital mortality and LOS as compared with a standard housestaff ICU service. Both these studies have limitations in that they are retrospective analyses rather than randomized controlled trials, and they were conducted at academic medical centers, thus narrowing their generalizability. Moreover, purity of data is difficult to achieve, as few systems exist where NPs and PAs are the sole providers managing patients without interaction or coverage from physician colleagues.

Given the considerable presence of NPs and PAs in acute care hospitals as documented by Kartha et al.,[5] providing appropriate training in hospital medicine to these clinicians is important. A study by Dhuper and Choksi[12] evaluated a 2‐year PA postgraduate training program in hospital medicine. PAs spent 40 hours per week on direct patient care while rotating on general medical floors and ICUs, along with 16 hours per week in didactic instruction. When compared with a traditional 3‐year medical residency at the same institution, the PA training program had similar outcomes on patient care including similar number of adverse events, readmissions, and patient satisfaction scores. A more formal postgraduate training program for PAs has been established at the Mayo Clinic Arizona.[13] This 12‐month program, based on the Society of Hospital Medicine's (SHM) Core Competencies, consists of general medicine and inpatient medical subspecialty rotations, didactic instruction, and self‐directed teaching modules to learn systems‐based practices. The Adult Hospital Medicine Boot Camp, sponsored by the SHM and the American Academy of Physician Assistants, is another training opportunity for both NPs and PAs who currently work in or are planning to practice hospital medicine.[14] Finally, in accordance with the move to provide standardized training for providers who practice in acute care settings, professional nursing organizations have developed the Consensus Model for Advanced Practice Registered Nurse Regulation that contains recommendations ensuring similar education and licensure requirements for those who practice in acute care.[15]

Although the optimal utilization of NPs and PAs in hospital medicine is still unknown, the reality is that the number of NPs and PAs actually working in this capacity is significant, as Kartha and his colleagues report.[5] A study of academic medical centers also found that among the institutions that responded to a survey, 31% and 42% used PAs and NPs, respectively, in hospitalist roles.[16] Current evidence suggests that NP‐ and PA‐based care with physician collaboration in an inpatient setting can result in comparable outcomes with physician‐only care models. However, much of this evidence is of poor quality or cannot be generalized to all settings. Kartha et al.[5] have provided a good first step in describing the role of NPs and PAs within hospital medicine. Though their education and training backgrounds are different, the ultimate scope of practice for these 2 groups of providers is very similar. Future research should focus on defining the best practice model for utilization of NPs and PAs in hospital medicine with emphasis on measurable goals. These can include standard outcomes such as LOS but also specific measures of quality and safety such as days of urinary catheter use or percentage of patients receiving venous thromboprophylaxis.[17] By understanding the scope of NP and PA practice, collecting more robust data regarding outcomes, and emphasizing training for NPs and PAs within hospital medicine, there is opportunity to impact the quality and efficiency of care of hospitalized patients.

Nurse practitioners (NPs) and physician assistants (PAs) have been caring for patients since the mid‐1960s.[1] Although both roles grew out of a need for more primary care providers, more recently there has been an increase in the utilization of NPs and PAs in acute care roles. This meteoric rise of advanced practice providers in the inpatient setting has been driven by stressors from residency work‐hour reforms and from growing financial pressures in healthcare systems, where NPs and PAs are seen as less expensive alternatives.[2, 3] Inadequate physician supply to meet the needs of growing healthcare service is also a driving factor. Despite increasing numbers of enrollees and increasing numbers of medical schools, many sources estimate a physician shortage of 50,000 providers by year 2025.[4] To address this growing shortage, the number of NP and PA providers in acute care continues to grow as Kartha and colleagues[5] clearly demonstrate in their study, published in this issue of Journal of Hospital Medicine. Their research shows that within hospitals in the Veterans Health Administration (VHA)the largest coordinated healthcare association in the United Statesfully half of all inpatient medical teams are utilizing NPs and PAs in some capacity, most commonly in staffing models working directly with attending physicians or on teams with housestaff.[5]

Many different practice models exist that incorporate NPs and PAs into acute care settings, including models in general medicine and intensive care settings, as well as in specialty care populations such as patients with diabetes or congestive heart failure.[1, 6] Few studies, however, delineate specific roles for NPs or PAs in inpatient acute care or provide outcomes‐based evidence in support of the proposed models. This is in contrast to research available regarding NP and PA staffing models in the outpatient setting.[7, 8] In the current study, Kartha et al.[5] shed light on the use of NPs and PAs in inpatient medical units at the VHA. Their findings show that the majority of NPs and PAs on the inpatient team function mostly autonomously and perform tasks including performing histories and physicals, writing progress notes, placing orders, and communicating with primary care providers and consultants. Almost half also serve on hospital committees and participate in quality improvement activities. Interestingly, although the training and regulation of NPs and PAs differ considerably,[1] Kartha et al. found that the scope of practice of these providers is generally the same. PAs are more likely to perform procedures and teach nonphysician students but otherwise function similarly to NPs. The clinical workload for NPs and PAs also does not differ, with an average of 6.5 patients seen per day. This information is crucial when analyzing the cost‐effectiveness of these providers, especially in light of evidence suggesting that hospitalist physicians typically care for approximately twice as many patients.[9]

Although Kartha et al.[5] focus primarily on describing the scope of NPs and PAs in hospital medicine, they also report on outcomes. Their findings show that presence of NPs and PAs on inpatient teams did not alter patient or nurse satisfaction nor were there any consistent improvements in the perception of care coordination. Of note, assessment of care coordination was based on survey responses from nurse managers and chiefs of medicine, individuals who are not necessarily direct members of the inpatient team, thus questioning the validity of this measure. Other studies on NP/PA models have also focused on patient‐centered outcomes. A study by Roy et al.[10] found that an inpatient PA‐run service supervised by hospitalists was comparable with a traditional resident‐run service, with no significant differences in risk‐adjusted length of stay (LOS), mortality, intensive care unit (ICU) transfers, or hospital readmissions. Although total costs were lower on the PA service, this difference was minimal. Gershengorn et al.[11] examined the impact of nonphysician staffing in an ICU setting and again found equivalent care. In this study, an ICU team staffed by NPs and PAs had similar hospital mortality and LOS as compared with a standard housestaff ICU service. Both these studies have limitations in that they are retrospective analyses rather than randomized controlled trials, and they were conducted at academic medical centers, thus narrowing their generalizability. Moreover, purity of data is difficult to achieve, as few systems exist where NPs and PAs are the sole providers managing patients without interaction or coverage from physician colleagues.

Given the considerable presence of NPs and PAs in acute care hospitals as documented by Kartha et al.,[5] providing appropriate training in hospital medicine to these clinicians is important. A study by Dhuper and Choksi[12] evaluated a 2‐year PA postgraduate training program in hospital medicine. PAs spent 40 hours per week on direct patient care while rotating on general medical floors and ICUs, along with 16 hours per week in didactic instruction. When compared with a traditional 3‐year medical residency at the same institution, the PA training program had similar outcomes on patient care including similar number of adverse events, readmissions, and patient satisfaction scores. A more formal postgraduate training program for PAs has been established at the Mayo Clinic Arizona.[13] This 12‐month program, based on the Society of Hospital Medicine's (SHM) Core Competencies, consists of general medicine and inpatient medical subspecialty rotations, didactic instruction, and self‐directed teaching modules to learn systems‐based practices. The Adult Hospital Medicine Boot Camp, sponsored by the SHM and the American Academy of Physician Assistants, is another training opportunity for both NPs and PAs who currently work in or are planning to practice hospital medicine.[14] Finally, in accordance with the move to provide standardized training for providers who practice in acute care settings, professional nursing organizations have developed the Consensus Model for Advanced Practice Registered Nurse Regulation that contains recommendations ensuring similar education and licensure requirements for those who practice in acute care.[15]

Although the optimal utilization of NPs and PAs in hospital medicine is still unknown, the reality is that the number of NPs and PAs actually working in this capacity is significant, as Kartha and his colleagues report.[5] A study of academic medical centers also found that among the institutions that responded to a survey, 31% and 42% used PAs and NPs, respectively, in hospitalist roles.[16] Current evidence suggests that NP‐ and PA‐based care with physician collaboration in an inpatient setting can result in comparable outcomes with physician‐only care models. However, much of this evidence is of poor quality or cannot be generalized to all settings. Kartha et al.[5] have provided a good first step in describing the role of NPs and PAs within hospital medicine. Though their education and training backgrounds are different, the ultimate scope of practice for these 2 groups of providers is very similar. Future research should focus on defining the best practice model for utilization of NPs and PAs in hospital medicine with emphasis on measurable goals. These can include standard outcomes such as LOS but also specific measures of quality and safety such as days of urinary catheter use or percentage of patients receiving venous thromboprophylaxis.[17] By understanding the scope of NP and PA practice, collecting more robust data regarding outcomes, and emphasizing training for NPs and PAs within hospital medicine, there is opportunity to impact the quality and efficiency of care of hospitalized patients.

The management of perioperative medications is a tale of progressive scientific enquiry. Although long‐term use of agents such as aspirin or statins improves clinical outcomes, use during surgery ranges from problematic to protective. The delicate balance between proven long‐term benefits in the nonoperative setting versus short‐term uncertainty in the perioperative setting must be assessed using a lens that incorporates patient risk, surgical process, and pharmacodynamic principles. Advances in our understanding of perioperative physiology, coupled with robust clinical and outcome data, have led to new knowledge and insights regarding how best to manage these medications. As well illustrated by the near 180 change in the use of perioperative ‐blockers to prevent adverse cardiovascular outcomes, these new data have led to substantial progress in surgical safety and patient outcomes.

In this issue of the Journal of Hospital Medicine, 2 retrospective cohort studies add to this growing body of evidence by examining risks associated with use of angiotensin‐converting enzyme (ACE) inhibitors in the perioperative setting. In the first study, conducted at a single academic medical center, Nielson and colleagues evaluate the association of preoperative ACE‐inhibitor use with hypotension and acute kidney injury in patients undergoing major elective orthopedic surgery.[1] The authors report that patients receiving ACE‐inhibitors were not only more likely to experience hypotension after induction of anesthesia (12.2% vs 6.7%, P = 0.005), but also were more likely to develop postoperative acute kidney injury (odds ratio [OR]: 2.68, 95% confidence interval [CI]: 1.25‐1.99). In the second study which focused solely on patients who were receiving preoperative ACE‐inhibitor therapy, Mudumbai and colleagues used a national Veterans Affairs database to examine the association between failure to resume ACE‐inhibitor treatment after surgery and outcomes at 30 days.[2] The authors found that failure to resume treatment 14 days after surgery was not only common (affecting 1 in 4 patients), but was also associated with increased 30‐day mortality (hazard ratio: 3.44, 95% CI: 3.30‐3.60).[2] Taken together, these 2 studies shed new light on clinical practice and policy implications for the use of these agents in the surgical period. Both sets of authors should be congratulated on moving the needle forward in this enquiry.

ACE‐inhibitors physiologically mediate their effects by preventing the formation of the potent vasoconstrictor angiotensin‐II from its precursor angiotensin‐I. In doing so, they decrease arterial resistance, increase venous capacitance, decrease glomerular filtration pressure, and promote natriuresis. These vascular effects have key benefits for the management of a number of chronic diseases including hypertension, congestive heart failure, and diabetic nephropathy. However, these clinical alterations are often problematic in the perioperative setting. For example, ACE‐inhibitors may cause vasoplegia during anesthetic administration, commonly manifested as hypotension during induction.[3] This hemodynamic alteration has been viewed as being so precarious, that some authors recommend withholding ACE‐inhibitors prior to major cardiovascular procedures such as coronary artery bypass grafting.[4] Although hypotension leads to management challenges (often increasing vasoconstrictor requirements), several studies report that ACE‐inhibitor use during surgery may also be associated with increased risks of acute kidney injury and mortality.[5, 6] However, despite these data, existing literature has not shown a consistent association between ACE‐inhibitor use and adverse postoperative outcomes. For example, a propensity‐matched cohort study of 79,228 patients at the Cleveland Clinic found no difference in hemodynamic characteristics, vasopressor requirements, or cardiorespiratory complications among patients who were or were not using ACE‐inhibitors during noncardiac surgery.[7] Furthermore, some research has also found that withdrawal of ACE‐inhibitors may itself lead to harm. For instance, a contemporary study reported that withdrawal of ACE‐inhibitor therapy in patients undergoing coronary artery bypass was associated with increased in‐hospital cardiovascular events.[8] Given this uncertainty, it is not surprising that management of ACE‐inhibitors in the perioperative period remains a subject of ongoing controversy with clinical reviews often recommending consideration of the risks and benefits associated with use of these agents.[9]

It is important to note that driver of this ambiguity is the very design of relevant studies. For instance, studies that focus on patients undergoing coronary artery bypass grafting surgery have limited external validity, as these patients are very different from those who undergo elective hip or knee replacement (such as those included by Nielsen and colleagues). Additionally, many studies suffer from methodological constraints or biases. For instance, retrospective observational studies often suffer from selection bias and residual confounding; that is, the individuals chosen for inclusion in the study and the variables available for analysis are often limited by available data, curtailing the conclusions that can be generated. Although the use of multivariable regression or propensity score techniques helps address these limitations, residual confounding by unmeasured variables always remains a threat to statistical inference. The potential influence for this bias is particularly relevant when examining the results of the study by Nielsen and colleagues. Another important limitation is survivor bias, a problem inherent in the study by Mudumbai and colleagues. Put simply, for a patient to resume ACE‐inhibitors after surgery, this same patient, must also survive for at least 15 to 30 days after surgery. Thus, for some patients, failure to resume this treatment may simply be a marker of early mortality rather than failure to resume the ACE itself. The potential influence of this bias is supported by an included sensitivity analysis, where a large change in the adjusted OR was observed when patients suffering early mortality were excluded. This swing in effect size suggests that biases related to comparing patients who survived to those who did not with respect to ACE use may, in part, account for the results of the study.

These limitations aside, the studies brought forth by both authors help inform practice with respect to the use of these agents during surgery. In this context, 3 paradigms are relevant for practicing hospitalists.

First, if a patient is maintained on an ACE‐inhibitor before surgery, should the medication be temporarily held before surgery to minimize hypotension during anesthesia? The study by Nielsen and colleagues (comparing those on ACE‐inhibitor treatment to those without), in addition to the evidence generated from other studies in this area[10, 11, 12] suggest that this is a rational decision. Although the existence of a withdrawal state from abrupt cessation of ACE‐inhibitor use is theoretically plausible, this has yet to be reliably reported in the literature. Given the short half‐life of most ACE‐inhibitors, cessation 24 hours before surgery appears to be the most pragmatic clinical approach.

Second, if a patient is on an ACE‐inhibitor before surgery, when should the medication be resumed after surgery? The findings from the study by Mudumbai and colleagues, in addition to contemporary evidence,[7, 8, 13] support the resumption of these agents as soon as possible following operative intervention. Once hemodynamic stability and volume status have been assured, risks associated with postoperative ACE‐inhibitor use appear to be outweighed by benefits, though specific care is likely necessary in those with preexisting renal dysfunction.[14] A program that ensures reconciliation of medications in the postoperative setting may be valuable in ensuring that such treatment is restarted.

Third, if a patient is not on ACE‐inhibitor therapy, should this be started before surgery for perioperative or long‐term benefit? Although neither study examines this issue, the potential for significant risk make this an unattractive option. Future interventional studies with thoughtfully weighed safety parameters may be necessary to assess whether such a paradigm may be valuable.

The studies included in this issue of the Journal of Hospital Medicine suggest that the use of ACE‐inhibitors during the perioperative period may be considered a function of time and place. Resuming ACE‐inhibitors and cessation of treatment at specific intervals in relation to surgery can help ensure positive outcomes. Hospitalists have an important role in this regard, as they are ideally situated to manage these agents in the many patients undergoing surgery across the United States.

Disclosures: Dr. Wijeysundera is supported by a Clinician‐Scientist Award from the Canadian Institutes of Health Research, and a Merit Award from the Department of Anesthesia at the University of Toronto. Dr. Chopra is supported by a Career‐Development Award (1K08HS022835‐01) from the Agency of Healthcare Research and Quality.

The management of perioperative medications is a tale of progressive scientific enquiry. Although long‐term use of agents such as aspirin or statins improves clinical outcomes, use during surgery ranges from problematic to protective. The delicate balance between proven long‐term benefits in the nonoperative setting versus short‐term uncertainty in the perioperative setting must be assessed using a lens that incorporates patient risk, surgical process, and pharmacodynamic principles. Advances in our understanding of perioperative physiology, coupled with robust clinical and outcome data, have led to new knowledge and insights regarding how best to manage these medications. As well illustrated by the near 180 change in the use of perioperative ‐blockers to prevent adverse cardiovascular outcomes, these new data have led to substantial progress in surgical safety and patient outcomes.

In this issue of the Journal of Hospital Medicine, 2 retrospective cohort studies add to this growing body of evidence by examining risks associated with use of angiotensin‐converting enzyme (ACE) inhibitors in the perioperative setting. In the first study, conducted at a single academic medical center, Nielson and colleagues evaluate the association of preoperative ACE‐inhibitor use with hypotension and acute kidney injury in patients undergoing major elective orthopedic surgery.[1] The authors report that patients receiving ACE‐inhibitors were not only more likely to experience hypotension after induction of anesthesia (12.2% vs 6.7%, P = 0.005), but also were more likely to develop postoperative acute kidney injury (odds ratio [OR]: 2.68, 95% confidence interval [CI]: 1.25‐1.99). In the second study which focused solely on patients who were receiving preoperative ACE‐inhibitor therapy, Mudumbai and colleagues used a national Veterans Affairs database to examine the association between failure to resume ACE‐inhibitor treatment after surgery and outcomes at 30 days.[2] The authors found that failure to resume treatment 14 days after surgery was not only common (affecting 1 in 4 patients), but was also associated with increased 30‐day mortality (hazard ratio: 3.44, 95% CI: 3.30‐3.60).[2] Taken together, these 2 studies shed new light on clinical practice and policy implications for the use of these agents in the surgical period. Both sets of authors should be congratulated on moving the needle forward in this enquiry.

ACE‐inhibitors physiologically mediate their effects by preventing the formation of the potent vasoconstrictor angiotensin‐II from its precursor angiotensin‐I. In doing so, they decrease arterial resistance, increase venous capacitance, decrease glomerular filtration pressure, and promote natriuresis. These vascular effects have key benefits for the management of a number of chronic diseases including hypertension, congestive heart failure, and diabetic nephropathy. However, these clinical alterations are often problematic in the perioperative setting. For example, ACE‐inhibitors may cause vasoplegia during anesthetic administration, commonly manifested as hypotension during induction.[3] This hemodynamic alteration has been viewed as being so precarious, that some authors recommend withholding ACE‐inhibitors prior to major cardiovascular procedures such as coronary artery bypass grafting.[4] Although hypotension leads to management challenges (often increasing vasoconstrictor requirements), several studies report that ACE‐inhibitor use during surgery may also be associated with increased risks of acute kidney injury and mortality.[5, 6] However, despite these data, existing literature has not shown a consistent association between ACE‐inhibitor use and adverse postoperative outcomes. For example, a propensity‐matched cohort study of 79,228 patients at the Cleveland Clinic found no difference in hemodynamic characteristics, vasopressor requirements, or cardiorespiratory complications among patients who were or were not using ACE‐inhibitors during noncardiac surgery.[7] Furthermore, some research has also found that withdrawal of ACE‐inhibitors may itself lead to harm. For instance, a contemporary study reported that withdrawal of ACE‐inhibitor therapy in patients undergoing coronary artery bypass was associated with increased in‐hospital cardiovascular events.[8] Given this uncertainty, it is not surprising that management of ACE‐inhibitors in the perioperative period remains a subject of ongoing controversy with clinical reviews often recommending consideration of the risks and benefits associated with use of these agents.[9]

It is important to note that driver of this ambiguity is the very design of relevant studies. For instance, studies that focus on patients undergoing coronary artery bypass grafting surgery have limited external validity, as these patients are very different from those who undergo elective hip or knee replacement (such as those included by Nielsen and colleagues). Additionally, many studies suffer from methodological constraints or biases. For instance, retrospective observational studies often suffer from selection bias and residual confounding; that is, the individuals chosen for inclusion in the study and the variables available for analysis are often limited by available data, curtailing the conclusions that can be generated. Although the use of multivariable regression or propensity score techniques helps address these limitations, residual confounding by unmeasured variables always remains a threat to statistical inference. The potential influence for this bias is particularly relevant when examining the results of the study by Nielsen and colleagues. Another important limitation is survivor bias, a problem inherent in the study by Mudumbai and colleagues. Put simply, for a patient to resume ACE‐inhibitors after surgery, this same patient, must also survive for at least 15 to 30 days after surgery. Thus, for some patients, failure to resume this treatment may simply be a marker of early mortality rather than failure to resume the ACE itself. The potential influence of this bias is supported by an included sensitivity analysis, where a large change in the adjusted OR was observed when patients suffering early mortality were excluded. This swing in effect size suggests that biases related to comparing patients who survived to those who did not with respect to ACE use may, in part, account for the results of the study.

These limitations aside, the studies brought forth by both authors help inform practice with respect to the use of these agents during surgery. In this context, 3 paradigms are relevant for practicing hospitalists.

First, if a patient is maintained on an ACE‐inhibitor before surgery, should the medication be temporarily held before surgery to minimize hypotension during anesthesia? The study by Nielsen and colleagues (comparing those on ACE‐inhibitor treatment to those without), in addition to the evidence generated from other studies in this area[10, 11, 12] suggest that this is a rational decision. Although the existence of a withdrawal state from abrupt cessation of ACE‐inhibitor use is theoretically plausible, this has yet to be reliably reported in the literature. Given the short half‐life of most ACE‐inhibitors, cessation 24 hours before surgery appears to be the most pragmatic clinical approach.

Second, if a patient is on an ACE‐inhibitor before surgery, when should the medication be resumed after surgery? The findings from the study by Mudumbai and colleagues, in addition to contemporary evidence,[7, 8, 13] support the resumption of these agents as soon as possible following operative intervention. Once hemodynamic stability and volume status have been assured, risks associated with postoperative ACE‐inhibitor use appear to be outweighed by benefits, though specific care is likely necessary in those with preexisting renal dysfunction.[14] A program that ensures reconciliation of medications in the postoperative setting may be valuable in ensuring that such treatment is restarted.

Third, if a patient is not on ACE‐inhibitor therapy, should this be started before surgery for perioperative or long‐term benefit? Although neither study examines this issue, the potential for significant risk make this an unattractive option. Future interventional studies with thoughtfully weighed safety parameters may be necessary to assess whether such a paradigm may be valuable.

The studies included in this issue of the Journal of Hospital Medicine suggest that the use of ACE‐inhibitors during the perioperative period may be considered a function of time and place. Resuming ACE‐inhibitors and cessation of treatment at specific intervals in relation to surgery can help ensure positive outcomes. Hospitalists have an important role in this regard, as they are ideally situated to manage these agents in the many patients undergoing surgery across the United States.

Disclosures: Dr. Wijeysundera is supported by a Clinician‐Scientist Award from the Canadian Institutes of Health Research, and a Merit Award from the Department of Anesthesia at the University of Toronto. Dr. Chopra is supported by a Career‐Development Award (1K08HS022835‐01) from the Agency of Healthcare Research and Quality.

The management of perioperative medications is a tale of progressive scientific enquiry. Although long‐term use of agents such as aspirin or statins improves clinical outcomes, use during surgery ranges from problematic to protective. The delicate balance between proven long‐term benefits in the nonoperative setting versus short‐term uncertainty in the perioperative setting must be assessed using a lens that incorporates patient risk, surgical process, and pharmacodynamic principles. Advances in our understanding of perioperative physiology, coupled with robust clinical and outcome data, have led to new knowledge and insights regarding how best to manage these medications. As well illustrated by the near 180 change in the use of perioperative ‐blockers to prevent adverse cardiovascular outcomes, these new data have led to substantial progress in surgical safety and patient outcomes.

In this issue of the Journal of Hospital Medicine, 2 retrospective cohort studies add to this growing body of evidence by examining risks associated with use of angiotensin‐converting enzyme (ACE) inhibitors in the perioperative setting. In the first study, conducted at a single academic medical center, Nielson and colleagues evaluate the association of preoperative ACE‐inhibitor use with hypotension and acute kidney injury in patients undergoing major elective orthopedic surgery.[1] The authors report that patients receiving ACE‐inhibitors were not only more likely to experience hypotension after induction of anesthesia (12.2% vs 6.7%, P = 0.005), but also were more likely to develop postoperative acute kidney injury (odds ratio [OR]: 2.68, 95% confidence interval [CI]: 1.25‐1.99). In the second study which focused solely on patients who were receiving preoperative ACE‐inhibitor therapy, Mudumbai and colleagues used a national Veterans Affairs database to examine the association between failure to resume ACE‐inhibitor treatment after surgery and outcomes at 30 days.[2] The authors found that failure to resume treatment 14 days after surgery was not only common (affecting 1 in 4 patients), but was also associated with increased 30‐day mortality (hazard ratio: 3.44, 95% CI: 3.30‐3.60).[2] Taken together, these 2 studies shed new light on clinical practice and policy implications for the use of these agents in the surgical period. Both sets of authors should be congratulated on moving the needle forward in this enquiry.

ACE‐inhibitors physiologically mediate their effects by preventing the formation of the potent vasoconstrictor angiotensin‐II from its precursor angiotensin‐I. In doing so, they decrease arterial resistance, increase venous capacitance, decrease glomerular filtration pressure, and promote natriuresis. These vascular effects have key benefits for the management of a number of chronic diseases including hypertension, congestive heart failure, and diabetic nephropathy. However, these clinical alterations are often problematic in the perioperative setting. For example, ACE‐inhibitors may cause vasoplegia during anesthetic administration, commonly manifested as hypotension during induction.[3] This hemodynamic alteration has been viewed as being so precarious, that some authors recommend withholding ACE‐inhibitors prior to major cardiovascular procedures such as coronary artery bypass grafting.[4] Although hypotension leads to management challenges (often increasing vasoconstrictor requirements), several studies report that ACE‐inhibitor use during surgery may also be associated with increased risks of acute kidney injury and mortality.[5, 6] However, despite these data, existing literature has not shown a consistent association between ACE‐inhibitor use and adverse postoperative outcomes. For example, a propensity‐matched cohort study of 79,228 patients at the Cleveland Clinic found no difference in hemodynamic characteristics, vasopressor requirements, or cardiorespiratory complications among patients who were or were not using ACE‐inhibitors during noncardiac surgery.[7] Furthermore, some research has also found that withdrawal of ACE‐inhibitors may itself lead to harm. For instance, a contemporary study reported that withdrawal of ACE‐inhibitor therapy in patients undergoing coronary artery bypass was associated with increased in‐hospital cardiovascular events.[8] Given this uncertainty, it is not surprising that management of ACE‐inhibitors in the perioperative period remains a subject of ongoing controversy with clinical reviews often recommending consideration of the risks and benefits associated with use of these agents.[9]

It is important to note that driver of this ambiguity is the very design of relevant studies. For instance, studies that focus on patients undergoing coronary artery bypass grafting surgery have limited external validity, as these patients are very different from those who undergo elective hip or knee replacement (such as those included by Nielsen and colleagues). Additionally, many studies suffer from methodological constraints or biases. For instance, retrospective observational studies often suffer from selection bias and residual confounding; that is, the individuals chosen for inclusion in the study and the variables available for analysis are often limited by available data, curtailing the conclusions that can be generated. Although the use of multivariable regression or propensity score techniques helps address these limitations, residual confounding by unmeasured variables always remains a threat to statistical inference. The potential influence for this bias is particularly relevant when examining the results of the study by Nielsen and colleagues. Another important limitation is survivor bias, a problem inherent in the study by Mudumbai and colleagues. Put simply, for a patient to resume ACE‐inhibitors after surgery, this same patient, must also survive for at least 15 to 30 days after surgery. Thus, for some patients, failure to resume this treatment may simply be a marker of early mortality rather than failure to resume the ACE itself. The potential influence of this bias is supported by an included sensitivity analysis, where a large change in the adjusted OR was observed when patients suffering early mortality were excluded. This swing in effect size suggests that biases related to comparing patients who survived to those who did not with respect to ACE use may, in part, account for the results of the study.

These limitations aside, the studies brought forth by both authors help inform practice with respect to the use of these agents during surgery. In this context, 3 paradigms are relevant for practicing hospitalists.

First, if a patient is maintained on an ACE‐inhibitor before surgery, should the medication be temporarily held before surgery to minimize hypotension during anesthesia? The study by Nielsen and colleagues (comparing those on ACE‐inhibitor treatment to those without), in addition to the evidence generated from other studies in this area[10, 11, 12] suggest that this is a rational decision. Although the existence of a withdrawal state from abrupt cessation of ACE‐inhibitor use is theoretically plausible, this has yet to be reliably reported in the literature. Given the short half‐life of most ACE‐inhibitors, cessation 24 hours before surgery appears to be the most pragmatic clinical approach.

Second, if a patient is on an ACE‐inhibitor before surgery, when should the medication be resumed after surgery? The findings from the study by Mudumbai and colleagues, in addition to contemporary evidence,[7, 8, 13] support the resumption of these agents as soon as possible following operative intervention. Once hemodynamic stability and volume status have been assured, risks associated with postoperative ACE‐inhibitor use appear to be outweighed by benefits, though specific care is likely necessary in those with preexisting renal dysfunction.[14] A program that ensures reconciliation of medications in the postoperative setting may be valuable in ensuring that such treatment is restarted.

Third, if a patient is not on ACE‐inhibitor therapy, should this be started before surgery for perioperative or long‐term benefit? Although neither study examines this issue, the potential for significant risk make this an unattractive option. Future interventional studies with thoughtfully weighed safety parameters may be necessary to assess whether such a paradigm may be valuable.

The studies included in this issue of the Journal of Hospital Medicine suggest that the use of ACE‐inhibitors during the perioperative period may be considered a function of time and place. Resuming ACE‐inhibitors and cessation of treatment at specific intervals in relation to surgery can help ensure positive outcomes. Hospitalists have an important role in this regard, as they are ideally situated to manage these agents in the many patients undergoing surgery across the United States.

Disclosures: Dr. Wijeysundera is supported by a Clinician‐Scientist Award from the Canadian Institutes of Health Research, and a Merit Award from the Department of Anesthesia at the University of Toronto. Dr. Chopra is supported by a Career‐Development Award (1K08HS022835‐01) from the Agency of Healthcare Research and Quality.

Only 20 years ago, science from peer‐reviewed journals was still distributed and consumed in the same fashion that evolved from the earliest days of medical science: in print at monthly or weekly intervals. The Internet radically accelerated this paradigm but left the essential processes intact; journals could publish the information and readers could read it more easily, but the basic forums for interaction and discussion over the content remained the same. Enter Web 2.0 and the era of social media. Authors, editors, and readers can now interact easily with each other over the content in real time and across great distances.

Social media may not have changed the way science is produced and reviewed, but it is certainly changing how people consume and use the science. Some have suggested that social media activity around particular articles or journals may be a more important measure of impact than traditional measures of citation,[1] and others have suggested that Twitter activity in particular has changed both the speed and quality of discussion about new studies within the scientific community.[2] In the face of these trends, the Journal of Hospital Medicine (JHM) has decided to develop a bold strategy for leadership in this emerging area, with an initial focus on increasing JHM's activity and visibility on Twitter.

As part of this initial focus, JHM has successfully developed and implemented a protocol for use by authors to compose 2 Tweets describing their publications: the first announces the article's publication (e.g., New evidence on white coats and risk for hospital‐acquired infections), and the second promotes a key point from the article (e.g., Does the doctor's white coat spread hospital infection?). These Tweets are encouraged (but not required) from the corresponding author for every article in every edition, and JHM's editorial staff works with individual authors to refine their message and maximize their impact. To help authors, we have developed several tips for effective tweeting (Table 1).

Even after just 1 year of this Twitter‐focused strategy, we are already seeing noteworthy impact and have learned several lessons.

AUTHORS CAN AND WILL GENERATE TWEETS FOR THEIR ARTICLES

When we started asking authors to generate tweets for their articles, Twitter was relatively new, and we were unsure if authors would be willing and able to participate. Since we started, we have noticed a steady increase in the number of author‐generated tweets. Today, more than three‐quarters of tweets per issue are author generated. Anecdotal feedback has been very positive, and authors have expressed interest in the plan for tweeting as well as feedback on how well their tweets were written. If authors or institutions are on Twitter, we also encourage using the Twitter name or handle in the tweet, which serves as a way for others on Twitter to identify directly with the author or institution and often results in greater interest in a particular tweet. Of note, authors have no obligation to become regular users of Twitter or engage with followers of JHM's Twitter feed, but many find themselves following the journal's feed more closely (and responding to posts by other authors) once they have joined Twitter and tweeted about their own work via JHM.

#HASHTAGS MAKE IT HAPPEN

Because Twitter users are a very large crowd of people with diverse interests, it is important to target tweets to the groups that would be most interested in studies. The use of hashtags makes it easy to index tweets. One of the major edits of author‐generated tweets that we provide is to index the articles to the most popular hashtags. For example, medical education studies can be indexed under #meded, which is a popular hashtag for clinician educators. Other important hashtags for hospitalists include #ptsafety, #readmissions, #healthpolicy, #healthcosts, or #infectiousdisease. To select hashtags, we have found the healthcare hashtag directory maintained by Symplur (Upland, CA; http://www.symplur.com/healthcare‐hashtags) to be a helpful resource for figuring out what the most popular ways to index tweets are and also for identifying hashtags for areas that are less well known to hospitalists, such as #histmedicine, which is for the history of medicine.

HIGH IMPACT STUDIES MAKE A BIGGER IMPACT ON TWITTER

We observed a high number of retweets and comments about articles that were the most viewed studies on JHM online, referring to Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) and the Society of Hospital Medicine's Choosing Wisely campaign. This is not surprising given the national focus on readmissions as well as cost‐conscious care. Moreover, our experience is in line with observations that Twitter provides an additional source of page views and article downloads for medical journals[3] and research, which demonstrates that studies that are tweeted will eventually be cited more.[4, 5]

TECHNOLOGY STUDIES ARE ADORED BY TWITTER

Studies and research examining the use of smartphones, apps, or social media in healthcare draw a lot of attention on Twitter, particularly from other technophiles in healthcare who often use the #hscm healthcare social media hashtag. Such studies often resonate with Twitter users, who tend to be engaged in technology at a high level and are interested in how to advance the use of technology in the healthcare workplace.

JHM's social media strategy has already been very successful in its early implementation; the JHM twitter feed has >600 followers. Although most authors submit their own tweets (71/117 or 61% of articles over the last year), JHM has also created social media roles for editors to fill in tweets when missing and ensure timely and consistent output from the JHM feed. We have also started a Facebook page, with a rapidly growing number of followers, and we continue to see our social media influence scores rise. In the next year we hope to develop a JHM blog, with invited commentary as well as a process for unsolicited submissions from our readership.

Increasingly, a journal's impact (small i) is measured not only in the traditional metric of impact factor (a representation of the number of papers cited in a given journal publication year), but also by the journal's ability to disseminate knowledge and awareness of issues key to the field. Social media is a major element of the next phase of evidence dissemination, and JHM is pleased to be developing and growing its footprint in the digital world.

Only 20 years ago, science from peer‐reviewed journals was still distributed and consumed in the same fashion that evolved from the earliest days of medical science: in print at monthly or weekly intervals. The Internet radically accelerated this paradigm but left the essential processes intact; journals could publish the information and readers could read it more easily, but the basic forums for interaction and discussion over the content remained the same. Enter Web 2.0 and the era of social media. Authors, editors, and readers can now interact easily with each other over the content in real time and across great distances.

Social media may not have changed the way science is produced and reviewed, but it is certainly changing how people consume and use the science. Some have suggested that social media activity around particular articles or journals may be a more important measure of impact than traditional measures of citation,[1] and others have suggested that Twitter activity in particular has changed both the speed and quality of discussion about new studies within the scientific community.[2] In the face of these trends, the Journal of Hospital Medicine (JHM) has decided to develop a bold strategy for leadership in this emerging area, with an initial focus on increasing JHM's activity and visibility on Twitter.

As part of this initial focus, JHM has successfully developed and implemented a protocol for use by authors to compose 2 Tweets describing their publications: the first announces the article's publication (e.g., New evidence on white coats and risk for hospital‐acquired infections), and the second promotes a key point from the article (e.g., Does the doctor's white coat spread hospital infection?). These Tweets are encouraged (but not required) from the corresponding author for every article in every edition, and JHM's editorial staff works with individual authors to refine their message and maximize their impact. To help authors, we have developed several tips for effective tweeting (Table 1).

Even after just 1 year of this Twitter‐focused strategy, we are already seeing noteworthy impact and have learned several lessons.

AUTHORS CAN AND WILL GENERATE TWEETS FOR THEIR ARTICLES

When we started asking authors to generate tweets for their articles, Twitter was relatively new, and we were unsure if authors would be willing and able to participate. Since we started, we have noticed a steady increase in the number of author‐generated tweets. Today, more than three‐quarters of tweets per issue are author generated. Anecdotal feedback has been very positive, and authors have expressed interest in the plan for tweeting as well as feedback on how well their tweets were written. If authors or institutions are on Twitter, we also encourage using the Twitter name or handle in the tweet, which serves as a way for others on Twitter to identify directly with the author or institution and often results in greater interest in a particular tweet. Of note, authors have no obligation to become regular users of Twitter or engage with followers of JHM's Twitter feed, but many find themselves following the journal's feed more closely (and responding to posts by other authors) once they have joined Twitter and tweeted about their own work via JHM.

#HASHTAGS MAKE IT HAPPEN

Because Twitter users are a very large crowd of people with diverse interests, it is important to target tweets to the groups that would be most interested in studies. The use of hashtags makes it easy to index tweets. One of the major edits of author‐generated tweets that we provide is to index the articles to the most popular hashtags. For example, medical education studies can be indexed under #meded, which is a popular hashtag for clinician educators. Other important hashtags for hospitalists include #ptsafety, #readmissions, #healthpolicy, #healthcosts, or #infectiousdisease. To select hashtags, we have found the healthcare hashtag directory maintained by Symplur (Upland, CA; http://www.symplur.com/healthcare‐hashtags) to be a helpful resource for figuring out what the most popular ways to index tweets are and also for identifying hashtags for areas that are less well known to hospitalists, such as #histmedicine, which is for the history of medicine.

HIGH IMPACT STUDIES MAKE A BIGGER IMPACT ON TWITTER

We observed a high number of retweets and comments about articles that were the most viewed studies on JHM online, referring to Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) and the Society of Hospital Medicine's Choosing Wisely campaign. This is not surprising given the national focus on readmissions as well as cost‐conscious care. Moreover, our experience is in line with observations that Twitter provides an additional source of page views and article downloads for medical journals[3] and research, which demonstrates that studies that are tweeted will eventually be cited more.[4, 5]

TECHNOLOGY STUDIES ARE ADORED BY TWITTER

Studies and research examining the use of smartphones, apps, or social media in healthcare draw a lot of attention on Twitter, particularly from other technophiles in healthcare who often use the #hscm healthcare social media hashtag. Such studies often resonate with Twitter users, who tend to be engaged in technology at a high level and are interested in how to advance the use of technology in the healthcare workplace.

JHM's social media strategy has already been very successful in its early implementation; the JHM twitter feed has >600 followers. Although most authors submit their own tweets (71/117 or 61% of articles over the last year), JHM has also created social media roles for editors to fill in tweets when missing and ensure timely and consistent output from the JHM feed. We have also started a Facebook page, with a rapidly growing number of followers, and we continue to see our social media influence scores rise. In the next year we hope to develop a JHM blog, with invited commentary as well as a process for unsolicited submissions from our readership.

Increasingly, a journal's impact (small i) is measured not only in the traditional metric of impact factor (a representation of the number of papers cited in a given journal publication year), but also by the journal's ability to disseminate knowledge and awareness of issues key to the field. Social media is a major element of the next phase of evidence dissemination, and JHM is pleased to be developing and growing its footprint in the digital world.

Only 20 years ago, science from peer‐reviewed journals was still distributed and consumed in the same fashion that evolved from the earliest days of medical science: in print at monthly or weekly intervals. The Internet radically accelerated this paradigm but left the essential processes intact; journals could publish the information and readers could read it more easily, but the basic forums for interaction and discussion over the content remained the same. Enter Web 2.0 and the era of social media. Authors, editors, and readers can now interact easily with each other over the content in real time and across great distances.

Social media may not have changed the way science is produced and reviewed, but it is certainly changing how people consume and use the science. Some have suggested that social media activity around particular articles or journals may be a more important measure of impact than traditional measures of citation,[1] and others have suggested that Twitter activity in particular has changed both the speed and quality of discussion about new studies within the scientific community.[2] In the face of these trends, the Journal of Hospital Medicine (JHM) has decided to develop a bold strategy for leadership in this emerging area, with an initial focus on increasing JHM's activity and visibility on Twitter.

As part of this initial focus, JHM has successfully developed and implemented a protocol for use by authors to compose 2 Tweets describing their publications: the first announces the article's publication (e.g., New evidence on white coats and risk for hospital‐acquired infections), and the second promotes a key point from the article (e.g., Does the doctor's white coat spread hospital infection?). These Tweets are encouraged (but not required) from the corresponding author for every article in every edition, and JHM's editorial staff works with individual authors to refine their message and maximize their impact. To help authors, we have developed several tips for effective tweeting (Table 1).

Even after just 1 year of this Twitter‐focused strategy, we are already seeing noteworthy impact and have learned several lessons.

AUTHORS CAN AND WILL GENERATE TWEETS FOR THEIR ARTICLES

When we started asking authors to generate tweets for their articles, Twitter was relatively new, and we were unsure if authors would be willing and able to participate. Since we started, we have noticed a steady increase in the number of author‐generated tweets. Today, more than three‐quarters of tweets per issue are author generated. Anecdotal feedback has been very positive, and authors have expressed interest in the plan for tweeting as well as feedback on how well their tweets were written. If authors or institutions are on Twitter, we also encourage using the Twitter name or handle in the tweet, which serves as a way for others on Twitter to identify directly with the author or institution and often results in greater interest in a particular tweet. Of note, authors have no obligation to become regular users of Twitter or engage with followers of JHM's Twitter feed, but many find themselves following the journal's feed more closely (and responding to posts by other authors) once they have joined Twitter and tweeted about their own work via JHM.

#HASHTAGS MAKE IT HAPPEN

Because Twitter users are a very large crowd of people with diverse interests, it is important to target tweets to the groups that would be most interested in studies. The use of hashtags makes it easy to index tweets. One of the major edits of author‐generated tweets that we provide is to index the articles to the most popular hashtags. For example, medical education studies can be indexed under #meded, which is a popular hashtag for clinician educators. Other important hashtags for hospitalists include #ptsafety, #readmissions, #healthpolicy, #healthcosts, or #infectiousdisease. To select hashtags, we have found the healthcare hashtag directory maintained by Symplur (Upland, CA; http://www.symplur.com/healthcare‐hashtags) to be a helpful resource for figuring out what the most popular ways to index tweets are and also for identifying hashtags for areas that are less well known to hospitalists, such as #histmedicine, which is for the history of medicine.

HIGH IMPACT STUDIES MAKE A BIGGER IMPACT ON TWITTER

We observed a high number of retweets and comments about articles that were the most viewed studies on JHM online, referring to Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) and the Society of Hospital Medicine's Choosing Wisely campaign. This is not surprising given the national focus on readmissions as well as cost‐conscious care. Moreover, our experience is in line with observations that Twitter provides an additional source of page views and article downloads for medical journals[3] and research, which demonstrates that studies that are tweeted will eventually be cited more.[4, 5]

TECHNOLOGY STUDIES ARE ADORED BY TWITTER

Studies and research examining the use of smartphones, apps, or social media in healthcare draw a lot of attention on Twitter, particularly from other technophiles in healthcare who often use the #hscm healthcare social media hashtag. Such studies often resonate with Twitter users, who tend to be engaged in technology at a high level and are interested in how to advance the use of technology in the healthcare workplace.

JHM's social media strategy has already been very successful in its early implementation; the JHM twitter feed has >600 followers. Although most authors submit their own tweets (71/117 or 61% of articles over the last year), JHM has also created social media roles for editors to fill in tweets when missing and ensure timely and consistent output from the JHM feed. We have also started a Facebook page, with a rapidly growing number of followers, and we continue to see our social media influence scores rise. In the next year we hope to develop a JHM blog, with invited commentary as well as a process for unsolicited submissions from our readership.

Increasingly, a journal's impact (small i) is measured not only in the traditional metric of impact factor (a representation of the number of papers cited in a given journal publication year), but also by the journal's ability to disseminate knowledge and awareness of issues key to the field. Social media is a major element of the next phase of evidence dissemination, and JHM is pleased to be developing and growing its footprint in the digital world.

Hospital readmission is not a new problem, but ever since the Centers for Medicaid and Medicare Services (CMS) announced that hospital reimbursement would be linked to readmission rates, the quest to understand drivers of this outcome has taken on new and remarkable vigor. Despite the avalanche of new studies on readmission factors[1] and transition interventions,[2, 3] surprisingly few have focused on conditions more prevalent in the aging Medicare population such as functional limitations. This trend in the literature reflects what is perhaps the greatest irony of the CMS readmission policy itself: while focused on improving care for a predominantly over 65‐year‐old population, it is agnostic to core geriatric vulnerabilities like function and cognition.

In this issue of the Journal of Hospital Medicine, Hoyer and colleagues take an important first step toward exploring such vulnerabilities.[4] Although it may not surprise many hospitalists that these play a role in complex outcomes such as readmission, the effects reported here are striking. The odds for readmission were 300% higher for patients with the lowest functional scores compared to those with highest scores after adjusting for other known factors such as comorbidities, age, and severity of illness. In terms of readmission rates, 29% of functionally impaired medical patients were readmitted compared to 11% of those with high function. Similar but less profound trends were seen in patients discharged from neurology and orthopedic services.

Although this was a single‐site study, and functional assessments were made on admission to an acute rehabilitation facility after hospital discharge, these findings are compelling and suggest many important areas for future research. First, these results suggest a need for replication in nationally representative data to better understand their scope and generalizability. Certainly, the number of participants (9405 patients) gives this study plenty of power; however, the sample is limited in that presumably all patients had some level of functional decline, but enough potential for functional recovery to warrant discharge to acute rehabilitation. We do not know what effects functional limitations might have on patients discharged to other settings (eg, community with home rehabilitation or skilled nursing facility with rehabilitation). Thus, future research should examine whether the impact of functional limitations described in this sample applies to the larger universe of hospital discharges.

We also do not know anything about the functional status of these patients at admission or their functional trajectory prior to hospitalization, which limits conclusions about whether the disabilities observed were hospital acquired. Functional ability, like vital signs, can be quite variable during the course of acute illness and should be interpreted in the context of each patient's baseline. The functional trajectory for a patient who was impaired at the time of hospital discharge, but independent before hospitalization, is likely very different than one who was chronically impaired at baseline. Thus, postdischarge is only half the story at best, and future research should explore the functional status and trajectory of patients before admission too.

Finally, to assess functional status, the authors of this study used the Functional Independence Measure (FIM) score, a well‐validated instrument used in rehabilitation facilities. One advantage of using this measure to predict readmission is that in addition to 12 items that assess physical domains overlapping with the Activities of Daily Living (ADL) measures commonly used in hospitals, it also includes 5 items about cognition and thus gives an overall view of both physical and mental status in context of functional ability. On the down side, the FIM score is less well known in the acute care setting and does not include instrumental ADLs, such as shopping, housekeeping, food preparation/cleanup, telephone, transportation, and technology like computers, that are often important for patients returning home. Given the interesting findings by Hoyer et al., future research should explore possible associations with these activities in patients discharged to community as well.

The results by Hoyer et al. also have important implications for policy and practice. At the level of national policy and ongoing healthcare reform, Medicare should consider ways to incentivize hospitals to collect data on functional status of patients more consistently. Currently, there is no International Classification of Diseases, 9th Revision code to capture functional limitation during hospitalization as a diagnosis or comorbidity (whether hospital acquired or not), which precludes any discussion about including functional status as an adjustor in the current CMS model for expected readmission rates for hospitals. Regardless of CMS policy and performance incentives or penalties, a lot more could be done at the level of hospital policy and practice to improve screening for functional vulnerabilities on admission and prior to discharge. Although this may require greater investment in standardizing physical therapy evaluation for most patients (especially those over 65 years old), the increased readmission rates found by Hoyer et al. in functionally impaired patients suggest it would be penny wise but pound foolish not to do so. In other words, if hospitals want to reduce their readmission rates by identifying and intervening on high‐risk patients, identifying functionally impaired patients seems to be the low‐hanging fruit.

In summary, Hoyer and colleagues have made an important contribution to the ever‐expanding literature on readmission risk factors, but they have likely just identified the tip of the iceberg. As Medicare enrollment continues to climb with the growth of baby boomers over 65 years old, the demand for acute care in older adults will continue to grow.[5] Moreover, as pressure mounts to improve the quality and reduce the costs of hospital care, greater understanding of geriatric vulnerabilities in this population will be increasingly important.

Hospital readmission is not a new problem, but ever since the Centers for Medicaid and Medicare Services (CMS) announced that hospital reimbursement would be linked to readmission rates, the quest to understand drivers of this outcome has taken on new and remarkable vigor. Despite the avalanche of new studies on readmission factors[1] and transition interventions,[2, 3] surprisingly few have focused on conditions more prevalent in the aging Medicare population such as functional limitations. This trend in the literature reflects what is perhaps the greatest irony of the CMS readmission policy itself: while focused on improving care for a predominantly over 65‐year‐old population, it is agnostic to core geriatric vulnerabilities like function and cognition.

In this issue of the Journal of Hospital Medicine, Hoyer and colleagues take an important first step toward exploring such vulnerabilities.[4] Although it may not surprise many hospitalists that these play a role in complex outcomes such as readmission, the effects reported here are striking. The odds for readmission were 300% higher for patients with the lowest functional scores compared to those with highest scores after adjusting for other known factors such as comorbidities, age, and severity of illness. In terms of readmission rates, 29% of functionally impaired medical patients were readmitted compared to 11% of those with high function. Similar but less profound trends were seen in patients discharged from neurology and orthopedic services.

Although this was a single‐site study, and functional assessments were made on admission to an acute rehabilitation facility after hospital discharge, these findings are compelling and suggest many important areas for future research. First, these results suggest a need for replication in nationally representative data to better understand their scope and generalizability. Certainly, the number of participants (9405 patients) gives this study plenty of power; however, the sample is limited in that presumably all patients had some level of functional decline, but enough potential for functional recovery to warrant discharge to acute rehabilitation. We do not know what effects functional limitations might have on patients discharged to other settings (eg, community with home rehabilitation or skilled nursing facility with rehabilitation). Thus, future research should examine whether the impact of functional limitations described in this sample applies to the larger universe of hospital discharges.

We also do not know anything about the functional status of these patients at admission or their functional trajectory prior to hospitalization, which limits conclusions about whether the disabilities observed were hospital acquired. Functional ability, like vital signs, can be quite variable during the course of acute illness and should be interpreted in the context of each patient's baseline. The functional trajectory for a patient who was impaired at the time of hospital discharge, but independent before hospitalization, is likely very different than one who was chronically impaired at baseline. Thus, postdischarge is only half the story at best, and future research should explore the functional status and trajectory of patients before admission too.

Finally, to assess functional status, the authors of this study used the Functional Independence Measure (FIM) score, a well‐validated instrument used in rehabilitation facilities. One advantage of using this measure to predict readmission is that in addition to 12 items that assess physical domains overlapping with the Activities of Daily Living (ADL) measures commonly used in hospitals, it also includes 5 items about cognition and thus gives an overall view of both physical and mental status in context of functional ability. On the down side, the FIM score is less well known in the acute care setting and does not include instrumental ADLs, such as shopping, housekeeping, food preparation/cleanup, telephone, transportation, and technology like computers, that are often important for patients returning home. Given the interesting findings by Hoyer et al., future research should explore possible associations with these activities in patients discharged to community as well.

The results by Hoyer et al. also have important implications for policy and practice. At the level of national policy and ongoing healthcare reform, Medicare should consider ways to incentivize hospitals to collect data on functional status of patients more consistently. Currently, there is no International Classification of Diseases, 9th Revision code to capture functional limitation during hospitalization as a diagnosis or comorbidity (whether hospital acquired or not), which precludes any discussion about including functional status as an adjustor in the current CMS model for expected readmission rates for hospitals. Regardless of CMS policy and performance incentives or penalties, a lot more could be done at the level of hospital policy and practice to improve screening for functional vulnerabilities on admission and prior to discharge. Although this may require greater investment in standardizing physical therapy evaluation for most patients (especially those over 65 years old), the increased readmission rates found by Hoyer et al. in functionally impaired patients suggest it would be penny wise but pound foolish not to do so. In other words, if hospitals want to reduce their readmission rates by identifying and intervening on high‐risk patients, identifying functionally impaired patients seems to be the low‐hanging fruit.

In summary, Hoyer and colleagues have made an important contribution to the ever‐expanding literature on readmission risk factors, but they have likely just identified the tip of the iceberg. As Medicare enrollment continues to climb with the growth of baby boomers over 65 years old, the demand for acute care in older adults will continue to grow.[5] Moreover, as pressure mounts to improve the quality and reduce the costs of hospital care, greater understanding of geriatric vulnerabilities in this population will be increasingly important.

Hospital readmission is not a new problem, but ever since the Centers for Medicaid and Medicare Services (CMS) announced that hospital reimbursement would be linked to readmission rates, the quest to understand drivers of this outcome has taken on new and remarkable vigor. Despite the avalanche of new studies on readmission factors[1] and transition interventions,[2, 3] surprisingly few have focused on conditions more prevalent in the aging Medicare population such as functional limitations. This trend in the literature reflects what is perhaps the greatest irony of the CMS readmission policy itself: while focused on improving care for a predominantly over 65‐year‐old population, it is agnostic to core geriatric vulnerabilities like function and cognition.

In this issue of the Journal of Hospital Medicine, Hoyer and colleagues take an important first step toward exploring such vulnerabilities.[4] Although it may not surprise many hospitalists that these play a role in complex outcomes such as readmission, the effects reported here are striking. The odds for readmission were 300% higher for patients with the lowest functional scores compared to those with highest scores after adjusting for other known factors such as comorbidities, age, and severity of illness. In terms of readmission rates, 29% of functionally impaired medical patients were readmitted compared to 11% of those with high function. Similar but less profound trends were seen in patients discharged from neurology and orthopedic services.

Although this was a single‐site study, and functional assessments were made on admission to an acute rehabilitation facility after hospital discharge, these findings are compelling and suggest many important areas for future research. First, these results suggest a need for replication in nationally representative data to better understand their scope and generalizability. Certainly, the number of participants (9405 patients) gives this study plenty of power; however, the sample is limited in that presumably all patients had some level of functional decline, but enough potential for functional recovery to warrant discharge to acute rehabilitation. We do not know what effects functional limitations might have on patients discharged to other settings (eg, community with home rehabilitation or skilled nursing facility with rehabilitation). Thus, future research should examine whether the impact of functional limitations described in this sample applies to the larger universe of hospital discharges.

We also do not know anything about the functional status of these patients at admission or their functional trajectory prior to hospitalization, which limits conclusions about whether the disabilities observed were hospital acquired. Functional ability, like vital signs, can be quite variable during the course of acute illness and should be interpreted in the context of each patient's baseline. The functional trajectory for a patient who was impaired at the time of hospital discharge, but independent before hospitalization, is likely very different than one who was chronically impaired at baseline. Thus, postdischarge is only half the story at best, and future research should explore the functional status and trajectory of patients before admission too.

Finally, to assess functional status, the authors of this study used the Functional Independence Measure (FIM) score, a well‐validated instrument used in rehabilitation facilities. One advantage of using this measure to predict readmission is that in addition to 12 items that assess physical domains overlapping with the Activities of Daily Living (ADL) measures commonly used in hospitals, it also includes 5 items about cognition and thus gives an overall view of both physical and mental status in context of functional ability. On the down side, the FIM score is less well known in the acute care setting and does not include instrumental ADLs, such as shopping, housekeeping, food preparation/cleanup, telephone, transportation, and technology like computers, that are often important for patients returning home. Given the interesting findings by Hoyer et al., future research should explore possible associations with these activities in patients discharged to community as well.

The results by Hoyer et al. also have important implications for policy and practice. At the level of national policy and ongoing healthcare reform, Medicare should consider ways to incentivize hospitals to collect data on functional status of patients more consistently. Currently, there is no International Classification of Diseases, 9th Revision code to capture functional limitation during hospitalization as a diagnosis or comorbidity (whether hospital acquired or not), which precludes any discussion about including functional status as an adjustor in the current CMS model for expected readmission rates for hospitals. Regardless of CMS policy and performance incentives or penalties, a lot more could be done at the level of hospital policy and practice to improve screening for functional vulnerabilities on admission and prior to discharge. Although this may require greater investment in standardizing physical therapy evaluation for most patients (especially those over 65 years old), the increased readmission rates found by Hoyer et al. in functionally impaired patients suggest it would be penny wise but pound foolish not to do so. In other words, if hospitals want to reduce their readmission rates by identifying and intervening on high‐risk patients, identifying functionally impaired patients seems to be the low‐hanging fruit.

In summary, Hoyer and colleagues have made an important contribution to the ever‐expanding literature on readmission risk factors, but they have likely just identified the tip of the iceberg. As Medicare enrollment continues to climb with the growth of baby boomers over 65 years old, the demand for acute care in older adults will continue to grow.[5] Moreover, as pressure mounts to improve the quality and reduce the costs of hospital care, greater understanding of geriatric vulnerabilities in this population will be increasingly important.

Hospital readmissions, which account for a substantial proportion of healthcare expenditures, have increasingly become a focus for hospitals and health systems. Hospitals now assume greater responsibility for population health, and face financial penalties by federal and state agencies that consider readmissions a key measure of the quality of care provided during hospitalization. Consequently, there is broad interest in identifying approaches to reduce hospital reutilization, including emergency department (ED) revisits and hospital readmissions. In this issue of the Journal of Hospital Medicine, Auger et al.[1] report the results of a systematic review, which evaluates the effect of discharge interventions on hospital reutilization among children.

As Auger et al. note, the transition from hospital to home is a vulnerable time for children and their families, with 1 in 5 parents reporting major challenges with such transitions.[2] Auger and colleagues identified 14 studies spanning 3 pediatric disease processes that addressed this issue. The authors concluded that several interventions were potentially effective, but individual studies frequently used multifactorial interventions, precluding determination of discrete elements essential to success. The larger body of care transitions literature in adult populations provides insights for interventions that may benefit pediatric patients, as well as informs future research and quality improvement priorities.

The authors identified some distinct interventions that may successfully decrease hospital reutilization, which share common themes from the adult literature. The first is the use of a dedicated transition coordinator (eg, nurse) or coordinating center to assist with the patient's transition home after discharge. In adult studies, this bridging strategy[3, 4] (ie, use of a dedicated transition coordinator or provider) is initiated during the hospitalization and continues postdischarge in the form of phone calls or home visits. The second theme illustrated in both this pediatric review[1] and adult reviews[3, 4, 5] focuses on enhanced or individualized patient education. Most studies have used a combination of these strategies. For example, the Care Transitions Intervention (one of the best validated adult discharge approaches) uses a transition coach to aid the patient in medication self‐management, creation of a patient‐centered record, scheduling follow‐up appointments, and understanding signs and symptoms of a worsening condition.[6] In a randomized study, this intervention demonstrated a reduction in readmissions within 90 days to 16.7% in the intervention group, compared with 22.5% in the control group.[6] One of the pediatric studies highlighted in the review by Auger et al. achieved a decrease in 14‐day ED revisits from 8% prior to implementation of the program to 2.7% following implementation of the program.[7] This program was for patients discharged from the neonatal intensive care unit and involved a nurse coordinator (similar to a transition coach) who worked closely with families and ensured adequate resources prior to discharge as well as a home visitation program.[7]

Although Auger et al. identify some effective approaches to reducing hospital reutilization after discharge in children, their review and the complementary adult literature bring to light 4 main unresolved questions for hospitalists seeking to improve care transitions: (1) how to dissect diverse and heterogeneous interventions to determine the key driver of success, (2) how to interpret and generally apply interventions from single centers where they may have been tailored to a specific healthcare environment, (3) how to generalize the findings of many disease‐specific interventions to other populations, and (4) how to evaluate the cost and assess the costbenefit of implementing many of the more resource intensive interventions. An example of a heterogeneous intervention addressed in this pediatric systematic review was described by Ng et al.,[8] in which the intervention group received a combination of an enhanced discharge education session, disease‐specific nurse evaluation, an animated education booklet, and postdischarge telephone follow‐up, whereas the control group received a shorter discharge education session, a disease‐specific nurse evaluation only if referred by a physician, a written education booklet, and no telephone follow‐up. Investigators found that intervention patients were less likely to be readmitted or revisit the ED as compared with controls. A similarly multifaceted intervention introduced by Taggart et al.[9] was unable to detect a difference in readmissions or ED revisits. It is unclear whether or not the differences in outcomes were related to differences in the intervention bundle itself or institutional or local contextual factors, thus limiting application to other hospitals. Generalizability of interventions is similarly complicated in adults.

The studies presented in this pediatric review article are specific to 3 disease processes: cancer, asthma, and neonatal intensive care (ie, premature) populations. Beyond these populations, there were no other pediatric conditions that met inclusion criteria, thus limiting the generalizability of the findings. As described by Rennke et al.,[3] adult systematic reviews that have focused only on disease‐specific interventions to reduce hospital reutilization are also difficult to generalize to broader populations. Two of the 3 recent adult transition intervention systematic reviews excluded disease‐specific interventions in an attempt to find more broadly applicable interventions but struggled with the same heterogeneity discussed in this review by Auger et al.[3, 4] Although disease‐specific interventions were included in the third adult systematic review and the evaluation was restricted to randomized controlled trials, the authors still grappled with finding 1 or 2 common, successful intervention components.[5] The fourth unresolved question involves understanding the financial burden of implementing more resource‐intensive interventions such as postdischarge home nurse visits. For example, it may be difficult to justify the business case for hiring a transition coach or initiating home nurse visits when the cost and financial implications are unclear. Neither the pediatric nor adult literature describes this well.

Some of the challenges in identifying effective interventions differ between adult and pediatric populations. Adults tend to have multiple comorbid conditions, making them more medically complex and at greater risk for adverse outcomes, medication errors, and hospital utilization.[10] Although a small subset of the pediatric population with complex chronic medical conditions accounts for a majority of hospital reutilization and cost,[11] most hospitalized pediatric patients are otherwise healthy with acute illnesses.[12] Additionally, pediatric patients have lower overall hospital reutilization rates when compared with adults. Adult 30‐day readmission rates are approximately 20%[13] compared with pediatric patients whose mean 30‐day readmission rate is 6.5%.[14] With readmission being an outcome upon which studies are basing intervention success or failure, the relatively low readmission rates in the pediatric population make shifting that outcome more challenging.

There is also controversy about whether policymakers should be focusing on decreasing 30‐day readmission rates as a measure of success. We believe that efforts should focus on identifying more meaningful outcomes, especially outcomes important to patients and their families. No single metric is likely to be an adequate measure of the quality of care transitions, but a combination of outcome measures could potentially be more informative both for patients and clinicians. Patient satisfaction with the discharge process is measured as part of standard patient experience surveys, and the 3‐question Care Transitions Measure[15] has been validated and endorsed as a measure of patient perception of discharge safety in adult populations. There is a growing consensus that 30‐day readmission rates are lacking as a measure of discharge quality, and therefore, measuring shorter‐term7‐ or 14‐dayreadmission rates along with short‐term ED utilization after discharge would likely be more helpful for identifying care transitions problems. Attention should also be paid to measuring rates of specific adverse events in the postdischarge period, such as adverse drug events or failure to follow up on pending test results, as these failures are often implicated in reutilization.

In reflecting upon the published data on adult and pediatric transitions of care interventions and the lingering unanswered questions, we propose a few considerations for future direction of the field. First, engagement of the primary care provider may be beneficial. In many interventions describing a care transition coordinator, nursing fulfilled this role; however, there are opportunities for the primary care provider to play a greater role in this arena. Second, the use of factorial design in future studies may help elucidate which specific parts of each intervention may be the most crucial.[16] Finally, readmission rates are a controversial quality measure in adults. Pediatric readmissions are relatively uncommon, making it difficult to track measurements and show improvement. Clinicians, patients, and policymakers should prioritize outcome measures that are most meaningful to patients and their families that occur at a much higher rate than that of readmissions.

Hospital readmissions, which account for a substantial proportion of healthcare expenditures, have increasingly become a focus for hospitals and health systems. Hospitals now assume greater responsibility for population health, and face financial penalties by federal and state agencies that consider readmissions a key measure of the quality of care provided during hospitalization. Consequently, there is broad interest in identifying approaches to reduce hospital reutilization, including emergency department (ED) revisits and hospital readmissions. In this issue of the Journal of Hospital Medicine, Auger et al.[1] report the results of a systematic review, which evaluates the effect of discharge interventions on hospital reutilization among children.

As Auger et al. note, the transition from hospital to home is a vulnerable time for children and their families, with 1 in 5 parents reporting major challenges with such transitions.[2] Auger and colleagues identified 14 studies spanning 3 pediatric disease processes that addressed this issue. The authors concluded that several interventions were potentially effective, but individual studies frequently used multifactorial interventions, precluding determination of discrete elements essential to success. The larger body of care transitions literature in adult populations provides insights for interventions that may benefit pediatric patients, as well as informs future research and quality improvement priorities.

The authors identified some distinct interventions that may successfully decrease hospital reutilization, which share common themes from the adult literature. The first is the use of a dedicated transition coordinator (eg, nurse) or coordinating center to assist with the patient's transition home after discharge. In adult studies, this bridging strategy[3, 4] (ie, use of a dedicated transition coordinator or provider) is initiated during the hospitalization and continues postdischarge in the form of phone calls or home visits. The second theme illustrated in both this pediatric review[1] and adult reviews[3, 4, 5] focuses on enhanced or individualized patient education. Most studies have used a combination of these strategies. For example, the Care Transitions Intervention (one of the best validated adult discharge approaches) uses a transition coach to aid the patient in medication self‐management, creation of a patient‐centered record, scheduling follow‐up appointments, and understanding signs and symptoms of a worsening condition.[6] In a randomized study, this intervention demonstrated a reduction in readmissions within 90 days to 16.7% in the intervention group, compared with 22.5% in the control group.[6] One of the pediatric studies highlighted in the review by Auger et al. achieved a decrease in 14‐day ED revisits from 8% prior to implementation of the program to 2.7% following implementation of the program.[7] This program was for patients discharged from the neonatal intensive care unit and involved a nurse coordinator (similar to a transition coach) who worked closely with families and ensured adequate resources prior to discharge as well as a home visitation program.[7]

Although Auger et al. identify some effective approaches to reducing hospital reutilization after discharge in children, their review and the complementary adult literature bring to light 4 main unresolved questions for hospitalists seeking to improve care transitions: (1) how to dissect diverse and heterogeneous interventions to determine the key driver of success, (2) how to interpret and generally apply interventions from single centers where they may have been tailored to a specific healthcare environment, (3) how to generalize the findings of many disease‐specific interventions to other populations, and (4) how to evaluate the cost and assess the costbenefit of implementing many of the more resource intensive interventions. An example of a heterogeneous intervention addressed in this pediatric systematic review was described by Ng et al.,[8] in which the intervention group received a combination of an enhanced discharge education session, disease‐specific nurse evaluation, an animated education booklet, and postdischarge telephone follow‐up, whereas the control group received a shorter discharge education session, a disease‐specific nurse evaluation only if referred by a physician, a written education booklet, and no telephone follow‐up. Investigators found that intervention patients were less likely to be readmitted or revisit the ED as compared with controls. A similarly multifaceted intervention introduced by Taggart et al.[9] was unable to detect a difference in readmissions or ED revisits. It is unclear whether or not the differences in outcomes were related to differences in the intervention bundle itself or institutional or local contextual factors, thus limiting application to other hospitals. Generalizability of interventions is similarly complicated in adults.

The studies presented in this pediatric review article are specific to 3 disease processes: cancer, asthma, and neonatal intensive care (ie, premature) populations. Beyond these populations, there were no other pediatric conditions that met inclusion criteria, thus limiting the generalizability of the findings. As described by Rennke et al.,[3] adult systematic reviews that have focused only on disease‐specific interventions to reduce hospital reutilization are also difficult to generalize to broader populations. Two of the 3 recent adult transition intervention systematic reviews excluded disease‐specific interventions in an attempt to find more broadly applicable interventions but struggled with the same heterogeneity discussed in this review by Auger et al.[3, 4] Although disease‐specific interventions were included in the third adult systematic review and the evaluation was restricted to randomized controlled trials, the authors still grappled with finding 1 or 2 common, successful intervention components.[5] The fourth unresolved question involves understanding the financial burden of implementing more resource‐intensive interventions such as postdischarge home nurse visits. For example, it may be difficult to justify the business case for hiring a transition coach or initiating home nurse visits when the cost and financial implications are unclear. Neither the pediatric nor adult literature describes this well.

Some of the challenges in identifying effective interventions differ between adult and pediatric populations. Adults tend to have multiple comorbid conditions, making them more medically complex and at greater risk for adverse outcomes, medication errors, and hospital utilization.[10] Although a small subset of the pediatric population with complex chronic medical conditions accounts for a majority of hospital reutilization and cost,[11] most hospitalized pediatric patients are otherwise healthy with acute illnesses.[12] Additionally, pediatric patients have lower overall hospital reutilization rates when compared with adults. Adult 30‐day readmission rates are approximately 20%[13] compared with pediatric patients whose mean 30‐day readmission rate is 6.5%.[14] With readmission being an outcome upon which studies are basing intervention success or failure, the relatively low readmission rates in the pediatric population make shifting that outcome more challenging.

There is also controversy about whether policymakers should be focusing on decreasing 30‐day readmission rates as a measure of success. We believe that efforts should focus on identifying more meaningful outcomes, especially outcomes important to patients and their families. No single metric is likely to be an adequate measure of the quality of care transitions, but a combination of outcome measures could potentially be more informative both for patients and clinicians. Patient satisfaction with the discharge process is measured as part of standard patient experience surveys, and the 3‐question Care Transitions Measure[15] has been validated and endorsed as a measure of patient perception of discharge safety in adult populations. There is a growing consensus that 30‐day readmission rates are lacking as a measure of discharge quality, and therefore, measuring shorter‐term7‐ or 14‐dayreadmission rates along with short‐term ED utilization after discharge would likely be more helpful for identifying care transitions problems. Attention should also be paid to measuring rates of specific adverse events in the postdischarge period, such as adverse drug events or failure to follow up on pending test results, as these failures are often implicated in reutilization.

In reflecting upon the published data on adult and pediatric transitions of care interventions and the lingering unanswered questions, we propose a few considerations for future direction of the field. First, engagement of the primary care provider may be beneficial. In many interventions describing a care transition coordinator, nursing fulfilled this role; however, there are opportunities for the primary care provider to play a greater role in this arena. Second, the use of factorial design in future studies may help elucidate which specific parts of each intervention may be the most crucial.[16] Finally, readmission rates are a controversial quality measure in adults. Pediatric readmissions are relatively uncommon, making it difficult to track measurements and show improvement. Clinicians, patients, and policymakers should prioritize outcome measures that are most meaningful to patients and their families that occur at a much higher rate than that of readmissions.

Hospital readmissions, which account for a substantial proportion of healthcare expenditures, have increasingly become a focus for hospitals and health systems. Hospitals now assume greater responsibility for population health, and face financial penalties by federal and state agencies that consider readmissions a key measure of the quality of care provided during hospitalization. Consequently, there is broad interest in identifying approaches to reduce hospital reutilization, including emergency department (ED) revisits and hospital readmissions. In this issue of the Journal of Hospital Medicine, Auger et al.[1] report the results of a systematic review, which evaluates the effect of discharge interventions on hospital reutilization among children.

As Auger et al. note, the transition from hospital to home is a vulnerable time for children and their families, with 1 in 5 parents reporting major challenges with such transitions.[2] Auger and colleagues identified 14 studies spanning 3 pediatric disease processes that addressed this issue. The authors concluded that several interventions were potentially effective, but individual studies frequently used multifactorial interventions, precluding determination of discrete elements essential to success. The larger body of care transitions literature in adult populations provides insights for interventions that may benefit pediatric patients, as well as informs future research and quality improvement priorities.

The authors identified some distinct interventions that may successfully decrease hospital reutilization, which share common themes from the adult literature. The first is the use of a dedicated transition coordinator (eg, nurse) or coordinating center to assist with the patient's transition home after discharge. In adult studies, this bridging strategy[3, 4] (ie, use of a dedicated transition coordinator or provider) is initiated during the hospitalization and continues postdischarge in the form of phone calls or home visits. The second theme illustrated in both this pediatric review[1] and adult reviews[3, 4, 5] focuses on enhanced or individualized patient education. Most studies have used a combination of these strategies. For example, the Care Transitions Intervention (one of the best validated adult discharge approaches) uses a transition coach to aid the patient in medication self‐management, creation of a patient‐centered record, scheduling follow‐up appointments, and understanding signs and symptoms of a worsening condition.[6] In a randomized study, this intervention demonstrated a reduction in readmissions within 90 days to 16.7% in the intervention group, compared with 22.5% in the control group.[6] One of the pediatric studies highlighted in the review by Auger et al. achieved a decrease in 14‐day ED revisits from 8% prior to implementation of the program to 2.7% following implementation of the program.[7] This program was for patients discharged from the neonatal intensive care unit and involved a nurse coordinator (similar to a transition coach) who worked closely with families and ensured adequate resources prior to discharge as well as a home visitation program.[7]

Although Auger et al. identify some effective approaches to reducing hospital reutilization after discharge in children, their review and the complementary adult literature bring to light 4 main unresolved questions for hospitalists seeking to improve care transitions: (1) how to dissect diverse and heterogeneous interventions to determine the key driver of success, (2) how to interpret and generally apply interventions from single centers where they may have been tailored to a specific healthcare environment, (3) how to generalize the findings of many disease‐specific interventions to other populations, and (4) how to evaluate the cost and assess the costbenefit of implementing many of the more resource intensive interventions. An example of a heterogeneous intervention addressed in this pediatric systematic review was described by Ng et al.,[8] in which the intervention group received a combination of an enhanced discharge education session, disease‐specific nurse evaluation, an animated education booklet, and postdischarge telephone follow‐up, whereas the control group received a shorter discharge education session, a disease‐specific nurse evaluation only if referred by a physician, a written education booklet, and no telephone follow‐up. Investigators found that intervention patients were less likely to be readmitted or revisit the ED as compared with controls. A similarly multifaceted intervention introduced by Taggart et al.[9] was unable to detect a difference in readmissions or ED revisits. It is unclear whether or not the differences in outcomes were related to differences in the intervention bundle itself or institutional or local contextual factors, thus limiting application to other hospitals. Generalizability of interventions is similarly complicated in adults.

The studies presented in this pediatric review article are specific to 3 disease processes: cancer, asthma, and neonatal intensive care (ie, premature) populations. Beyond these populations, there were no other pediatric conditions that met inclusion criteria, thus limiting the generalizability of the findings. As described by Rennke et al.,[3] adult systematic reviews that have focused only on disease‐specific interventions to reduce hospital reutilization are also difficult to generalize to broader populations. Two of the 3 recent adult transition intervention systematic reviews excluded disease‐specific interventions in an attempt to find more broadly applicable interventions but struggled with the same heterogeneity discussed in this review by Auger et al.[3, 4] Although disease‐specific interventions were included in the third adult systematic review and the evaluation was restricted to randomized controlled trials, the authors still grappled with finding 1 or 2 common, successful intervention components.[5] The fourth unresolved question involves understanding the financial burden of implementing more resource‐intensive interventions such as postdischarge home nurse visits. For example, it may be difficult to justify the business case for hiring a transition coach or initiating home nurse visits when the cost and financial implications are unclear. Neither the pediatric nor adult literature describes this well.

Some of the challenges in identifying effective interventions differ between adult and pediatric populations. Adults tend to have multiple comorbid conditions, making them more medically complex and at greater risk for adverse outcomes, medication errors, and hospital utilization.[10] Although a small subset of the pediatric population with complex chronic medical conditions accounts for a majority of hospital reutilization and cost,[11] most hospitalized pediatric patients are otherwise healthy with acute illnesses.[12] Additionally, pediatric patients have lower overall hospital reutilization rates when compared with adults. Adult 30‐day readmission rates are approximately 20%[13] compared with pediatric patients whose mean 30‐day readmission rate is 6.5%.[14] With readmission being an outcome upon which studies are basing intervention success or failure, the relatively low readmission rates in the pediatric population make shifting that outcome more challenging.

There is also controversy about whether policymakers should be focusing on decreasing 30‐day readmission rates as a measure of success. We believe that efforts should focus on identifying more meaningful outcomes, especially outcomes important to patients and their families. No single metric is likely to be an adequate measure of the quality of care transitions, but a combination of outcome measures could potentially be more informative both for patients and clinicians. Patient satisfaction with the discharge process is measured as part of standard patient experience surveys, and the 3‐question Care Transitions Measure[15] has been validated and endorsed as a measure of patient perception of discharge safety in adult populations. There is a growing consensus that 30‐day readmission rates are lacking as a measure of discharge quality, and therefore, measuring shorter‐term7‐ or 14‐dayreadmission rates along with short‐term ED utilization after discharge would likely be more helpful for identifying care transitions problems. Attention should also be paid to measuring rates of specific adverse events in the postdischarge period, such as adverse drug events or failure to follow up on pending test results, as these failures are often implicated in reutilization.

In reflecting upon the published data on adult and pediatric transitions of care interventions and the lingering unanswered questions, we propose a few considerations for future direction of the field. First, engagement of the primary care provider may be beneficial. In many interventions describing a care transition coordinator, nursing fulfilled this role; however, there are opportunities for the primary care provider to play a greater role in this arena. Second, the use of factorial design in future studies may help elucidate which specific parts of each intervention may be the most crucial.[16] Finally, readmission rates are a controversial quality measure in adults. Pediatric readmissions are relatively uncommon, making it difficult to track measurements and show improvement. Clinicians, patients, and policymakers should prioritize outcome measures that are most meaningful to patients and their families that occur at a much higher rate than that of readmissions.

There is consensus that the hospital is an appropriate place to start chronic medications for conditions that caused the hospitalization (e.g., aspirin for a patient admitted with acute myocardial infarction). However, little is known about physician attitudes toward starting chronic medications for conditions unrelated to the reason for hospitalization (e.g., aspirin in a patient with a history of myocardial infarction admitted for cellulitis). Although hospitalists can identify and remedy potential gaps in the management of chronic conditions, changes in such medications during the hospital stay can create a number of problems. Contextual factors, such as prior medication trials, patient preferences, and longstanding patterns of disease management, may be unknown to the inpatient clinician, and medication confusion, nonadherence, and adverse effects can result from multiple medication changes.[1, 2] The lack of consensus about changing chronic medications for conditions unrelated to the reason for admission reflects a lack of clarity regarding the risk‐benefit equation in this area.

The study by Breu and colleagues[3] in this issue provides one of the first studies of hospitalist and primary care physician (PCP) attitudes about changing chronic medications during hospitalization for conditions unrelated to the reason for admission. The authors had hospitalists and PCPs consider six cases, half involving a medication change related to the reason for admission and half involving a medication change unrelated to the reason for admission. They found that PCPs were more likely than hospitalists to feel that inpatient interventions were appropriate when unrelated to the reason for admission. However, the majority of both hospitalists and PCPs did not feel interventions in these cases were appropriate.

Although this study provides useful insight into the attitudes of physicians toward these issues, it is likely that even more physicians would be skeptical of initiating chronic medications in the hospital if the scenarios reflected the messy reality that often faces clinicians when patients are hospitalized. The study asked physician respondents to assume full outpatient electronic medical record (EMR) access and communication at discharge. However, in practice, inpatient physicians often do not have full outpatient EMR access. If they do have full access to records, they typically do not have the time to thoroughly review the chart, leading to over half of internal medicine patients having at least one medication discrepancy at admission.[4] In addition, communication between hospitalists and PCPs occurs infrequently, and discharge summaries are often not available by the time of the first postdischarge clinic visit and lack important information, such as diagnostic test results and discharge medications.[2]

We believe that in most clinical settings, the serious problems that accompany changing medications in hospitalized patients argue for a judicious approach to modifying medications for chronic conditions not related to the reason for hospitalization. However, the more important question is how the prescribing process in hospitalized patients can be re‐envisioned in a manner that allows individualization of these decisions to serve both the short‐ and long‐term needs of patients. Because the success and appropriateness of long‐term treatment decisions often depends on contextual factors, PCP follow‐up, and patient medication compliance, in most cases decisions about initiating long‐term therapy for conditions not central to the hospital admission should involve each of these circumstances. Shared decision making models involve clinicians and patients sharing information, expressing treatment preferences, deliberating the options, and coming to an agreement on a treatment plan,[5] and these models have been associated with improved adherence and disease‐specific outcomes.[6] Shared decision making in many cases could be done quickly and efficiently through a quick check‐in with the PCP and a brief discussion with the patient. When consensus cannot be reached with these methods, then raising the issue with the PCP and patient but deferring the final decision until after discharge would be appropriate.

In hospitalized patients, less is often more, and minimizing the number of nonessential medication changes may ultimately yield better outcomes. Although inpatient clinicians can identify important gaps in care, the best solutions come from discussions that can bridge the inpatient‐outpatient divide and ultimately serve the long‐term needs of patients.

There is consensus that the hospital is an appropriate place to start chronic medications for conditions that caused the hospitalization (e.g., aspirin for a patient admitted with acute myocardial infarction). However, little is known about physician attitudes toward starting chronic medications for conditions unrelated to the reason for hospitalization (e.g., aspirin in a patient with a history of myocardial infarction admitted for cellulitis). Although hospitalists can identify and remedy potential gaps in the management of chronic conditions, changes in such medications during the hospital stay can create a number of problems. Contextual factors, such as prior medication trials, patient preferences, and longstanding patterns of disease management, may be unknown to the inpatient clinician, and medication confusion, nonadherence, and adverse effects can result from multiple medication changes.[1, 2] The lack of consensus about changing chronic medications for conditions unrelated to the reason for admission reflects a lack of clarity regarding the risk‐benefit equation in this area.

The study by Breu and colleagues[3] in this issue provides one of the first studies of hospitalist and primary care physician (PCP) attitudes about changing chronic medications during hospitalization for conditions unrelated to the reason for admission. The authors had hospitalists and PCPs consider six cases, half involving a medication change related to the reason for admission and half involving a medication change unrelated to the reason for admission. They found that PCPs were more likely than hospitalists to feel that inpatient interventions were appropriate when unrelated to the reason for admission. However, the majority of both hospitalists and PCPs did not feel interventions in these cases were appropriate.

Although this study provides useful insight into the attitudes of physicians toward these issues, it is likely that even more physicians would be skeptical of initiating chronic medications in the hospital if the scenarios reflected the messy reality that often faces clinicians when patients are hospitalized. The study asked physician respondents to assume full outpatient electronic medical record (EMR) access and communication at discharge. However, in practice, inpatient physicians often do not have full outpatient EMR access. If they do have full access to records, they typically do not have the time to thoroughly review the chart, leading to over half of internal medicine patients having at least one medication discrepancy at admission.[4] In addition, communication between hospitalists and PCPs occurs infrequently, and discharge summaries are often not available by the time of the first postdischarge clinic visit and lack important information, such as diagnostic test results and discharge medications.[2]

We believe that in most clinical settings, the serious problems that accompany changing medications in hospitalized patients argue for a judicious approach to modifying medications for chronic conditions not related to the reason for hospitalization. However, the more important question is how the prescribing process in hospitalized patients can be re‐envisioned in a manner that allows individualization of these decisions to serve both the short‐ and long‐term needs of patients. Because the success and appropriateness of long‐term treatment decisions often depends on contextual factors, PCP follow‐up, and patient medication compliance, in most cases decisions about initiating long‐term therapy for conditions not central to the hospital admission should involve each of these circumstances. Shared decision making models involve clinicians and patients sharing information, expressing treatment preferences, deliberating the options, and coming to an agreement on a treatment plan,[5] and these models have been associated with improved adherence and disease‐specific outcomes.[6] Shared decision making in many cases could be done quickly and efficiently through a quick check‐in with the PCP and a brief discussion with the patient. When consensus cannot be reached with these methods, then raising the issue with the PCP and patient but deferring the final decision until after discharge would be appropriate.

In hospitalized patients, less is often more, and minimizing the number of nonessential medication changes may ultimately yield better outcomes. Although inpatient clinicians can identify important gaps in care, the best solutions come from discussions that can bridge the inpatient‐outpatient divide and ultimately serve the long‐term needs of patients.

There is consensus that the hospital is an appropriate place to start chronic medications for conditions that caused the hospitalization (e.g., aspirin for a patient admitted with acute myocardial infarction). However, little is known about physician attitudes toward starting chronic medications for conditions unrelated to the reason for hospitalization (e.g., aspirin in a patient with a history of myocardial infarction admitted for cellulitis). Although hospitalists can identify and remedy potential gaps in the management of chronic conditions, changes in such medications during the hospital stay can create a number of problems. Contextual factors, such as prior medication trials, patient preferences, and longstanding patterns of disease management, may be unknown to the inpatient clinician, and medication confusion, nonadherence, and adverse effects can result from multiple medication changes.[1, 2] The lack of consensus about changing chronic medications for conditions unrelated to the reason for admission reflects a lack of clarity regarding the risk‐benefit equation in this area.

The study by Breu and colleagues[3] in this issue provides one of the first studies of hospitalist and primary care physician (PCP) attitudes about changing chronic medications during hospitalization for conditions unrelated to the reason for admission. The authors had hospitalists and PCPs consider six cases, half involving a medication change related to the reason for admission and half involving a medication change unrelated to the reason for admission. They found that PCPs were more likely than hospitalists to feel that inpatient interventions were appropriate when unrelated to the reason for admission. However, the majority of both hospitalists and PCPs did not feel interventions in these cases were appropriate.

Although this study provides useful insight into the attitudes of physicians toward these issues, it is likely that even more physicians would be skeptical of initiating chronic medications in the hospital if the scenarios reflected the messy reality that often faces clinicians when patients are hospitalized. The study asked physician respondents to assume full outpatient electronic medical record (EMR) access and communication at discharge. However, in practice, inpatient physicians often do not have full outpatient EMR access. If they do have full access to records, they typically do not have the time to thoroughly review the chart, leading to over half of internal medicine patients having at least one medication discrepancy at admission.[4] In addition, communication between hospitalists and PCPs occurs infrequently, and discharge summaries are often not available by the time of the first postdischarge clinic visit and lack important information, such as diagnostic test results and discharge medications.[2]

We believe that in most clinical settings, the serious problems that accompany changing medications in hospitalized patients argue for a judicious approach to modifying medications for chronic conditions not related to the reason for hospitalization. However, the more important question is how the prescribing process in hospitalized patients can be re‐envisioned in a manner that allows individualization of these decisions to serve both the short‐ and long‐term needs of patients. Because the success and appropriateness of long‐term treatment decisions often depends on contextual factors, PCP follow‐up, and patient medication compliance, in most cases decisions about initiating long‐term therapy for conditions not central to the hospital admission should involve each of these circumstances. Shared decision making models involve clinicians and patients sharing information, expressing treatment preferences, deliberating the options, and coming to an agreement on a treatment plan,[5] and these models have been associated with improved adherence and disease‐specific outcomes.[6] Shared decision making in many cases could be done quickly and efficiently through a quick check‐in with the PCP and a brief discussion with the patient. When consensus cannot be reached with these methods, then raising the issue with the PCP and patient but deferring the final decision until after discharge would be appropriate.

In hospitalized patients, less is often more, and minimizing the number of nonessential medication changes may ultimately yield better outcomes. Although inpatient clinicians can identify important gaps in care, the best solutions come from discussions that can bridge the inpatient‐outpatient divide and ultimately serve the long‐term needs of patients.

The Hospital Readmission Reduction Program (HRRP)[1] contained within the Affordable Care Act focused national and local attention on hospital resources and efforts to reduce hospital readmissions. Driven by the Centers for Medicare and Medicaid Services' (CMS) desire to pay for value instead of volume, the response of hospitals and health systems appears to be yielding change across the United States.[2] A number of recent publications in the Journal of Hospital Medicine (JHM) exemplify the keen interest in reducing readmissions, while providing guidance regarding interventions and where we might target future research. Evidence from an exemplary systematic review of the pediatric literature confirms some experience in adults regarding effective interventionsall studies were multifacetedand highlights the importance of identifying a single healthcare provider or centrally coordinated hub to assume responsibility for extended care transition and follow‐up.[3] Notably, studies of pediatric patients and their families document the effectiveness of enhanced inpatient education and engagement while in the hospital.[3] Unfortunately, a study among adults at a top‐ranked academic institution indicates poor communication among nurses and physicians regarding patient discharge education.[4] Efforts to improve nursephysician communication by redesigning the hospitalist model of care delivery at a Veterans Affairs (VA) institution appeared to enhance perceptions of communication among the care team members and reduced length of stay, but disappointingly there was no reduction in readmission rates.[5] Studies such as this are essential in identifying which specific interventions may actually change outcomes such as readmission rates.

In 1984, a diminutive elderly woman provocatively squawked Where's the beef?, launching a highly successful advertising campaign for Wendy's hamburger chain.[6] This catchphrase may aptly describe Bradley and colleague's survey study of the State Action on Avoidable Rehospitalization (STAAR) and Hospital‐to‐Home (H2H) campaigns.[7] Auerbach and colleagues eloquently stated in a 2007 New England Journal of Medicine perspective[8] how they had witnessed recent initiatives that emphasize dissemination of innovative but unproven strategies, an approach that runs counter to the principle of following the evidence[9] in selecting interventions that meet quality and safety goals.[10] I firmly agree with this assessment, and 6 years later believe we should be more thoughtful about potentially repeating implementation of unproven strategies.

Do we know if the interventions recommended by H2H and STAAR are what hospital care teams should be attempting? Even the authors mention that definitive evidence on their effectiveness is lacking. The H2H and STAAR programs certainly encourage some theoretically laudable activitiesmedication reconciliation by nurses, alerting outpatient physicians within 48 hours of patient discharge, and providing skilled nursing facilities the direct contact number of the inpatient treating physician for patients transferred. However, do these efforts actually improve patient outcomes? Before embarking on state or national campaigns to improve care, we should consider carefully what are the best evidence‐based interventions. Remarkably, some prior evidence indicates that direct communication between the hospital‐based physician and primary care provider (PCP) may not actually impact patient outcomes.[11] Newer research published in JHM confirms my belief that the PCP needs to be engaged by hospitalists during a hospitalization. Lindquist's research group at Northwestern nicely demonstrated how communication between a patient's PCP and the admitting hospitalist, complemented by contact between the PCP and patient within 24 hours postdischarge, reduced the probability of a medication discrepancy by 70%.[12] Although no evaluation of the effect on readmissions was reported, this study may provide information on causality related to the importance of PCP involvement in the care of hospitalized patients.

Numerous publications now document research on successfully implemented programs that lowered hospital readmissions, and are cited by CMS as evidence‐based interventions.[13] Projects Re‐Engineered Discharge (RED)[14] and Better Outcomes by Optimizing Safe Transitions[15] target the hospital discharge process, and both appear to lower hospital readmission rates. The Care Transitions Intervention (CTI),[16] Transitional Care Model (TCM),[17] and the Guided Care model[18] all leverage nurse practitioners or nurses to protect elderly patients during what can be a perilous care transition from hospital to home. CTI and TCM have been further validated in effectiveness studies.[19, 20] Two recent systematic reviews provide further insight into the complexity of efforts to reduce 30‐day rehospitalizations, but unfortunately do not reveal a desired silver bullet. The first focused exclusively on interventions to reduce 30‐day rehospitalization, and concluded that no single intervention was successful alone, but identified interventions bridging the hospital‐to‐home transition (eg, CTI), and a bundle of interventions such as Project RED as showing efficacy.[21] The second review more broadly sought to evaluate the effectiveness of hospital‐initiated strategies to prevent postdischarge adverse events (AEs) such as readmissions and emergency department visits,[22] stating Because of scant evidence, no conclusions could be reached on methods to prevent postdischarge AEs. The researchers' sobering conclusion stated that strategies to improve patient safety at hospital discharge remain unclear.

With rising federal penalties for higher‐than‐expected readmission rates, many hospital leaders eagerly join collaboratives aiming to reduce hospital readmissions. H2H appears to be among the largest, reporting >600 hospital participants, and STAAR has been active since 2009, with a recently published qualitative study identifying gaps in evidence for effective interventions, and deficits in quality improvement capabilities among some organizations as implementation challenges.[23] Notably, the survey by Bradley and colleagues documented that just half of the hospitals had a quality improvement (QI) team focused on reducing readmissions. Although laudable in their goals, H2H and STAAR may represent expensive commitments of staff and time to efforts that may not improve outcomes. Importantly, recently published research evaluating QI studies showed concerning results among patients with chronic obstructive pulmonary disease (COPD). A randomized controlled trial (RCT) conducted at 6 Glasgow hospitals evaluated supported self‐management (home visits by nurses and thorough education) by patients with moderate to severe COPD, but documented no changes in hospitalization or mortality.[24]Another RCT at 20 sites evaluated a comprehensive care management program to prevent hospitalizations among 960 VA patients with COPD.[25] It had to be stopped early due to elevated all‐cause mortality in the intervention group, and there was no difference in hospitalization rates.

Moving forward, QI efforts to reduce hospital readmissions should utilize proven interventions unless they are part of a rigorous trial. The emerging field of implementation science (the scientific study of methods to promote the systematic uptake of research findings and other evidence‐based practices into routine practice, and hence, to improve the quality and effectiveness of health services[26]) needs to be applied to additional research in this area.[27] Another consideration would be for CMS and funders such as the Commonwealth Foundation or The Robert Wood Johnson Foundation to encourage and fund merging of current initiatives to move away from competition and provide clarity to community hospitals. Regardless, such collaboration should still undertake formal evaluation to discern best approaches to implementation. I applaud the authors for recognizing that Input from hospitalists who are often critical links among inpatient and outpatient care and between patients and their families is strongly needed to ensure hospitals focus on what strategies are most effective for successful transitions from hospital to home. Yet, I wonder why neither of the large STAAR and H2H initiatives actively partnered with hospitalists and their specialty society (Society of Hospital Medicine) directly in the leadership of these initiatives? On the other hand, why not ask medical societies engaged in delivery of primary care (eg, American Academy for Family Practice, American College of Physicians, or Society of General Internal Medicine), especially to elderly patients (American Geriatric Society), to contribute directly? Involvement on an advisory board is likely not sufficient. Prior efforts document the willingness of these organizations to collaborate and achieve consensus on principles for transitions of care.[28] As powerfully articulated 6 years ago, [W]e must pursue the solutions to quality and safety problems in a way that does not blind us to harms, squander scarce resources, or delude us about the effectiveness of our efforts.[8]

The Hospital Readmission Reduction Program (HRRP)[1] contained within the Affordable Care Act focused national and local attention on hospital resources and efforts to reduce hospital readmissions. Driven by the Centers for Medicare and Medicaid Services' (CMS) desire to pay for value instead of volume, the response of hospitals and health systems appears to be yielding change across the United States.[2] A number of recent publications in the Journal of Hospital Medicine (JHM) exemplify the keen interest in reducing readmissions, while providing guidance regarding interventions and where we might target future research. Evidence from an exemplary systematic review of the pediatric literature confirms some experience in adults regarding effective interventionsall studies were multifacetedand highlights the importance of identifying a single healthcare provider or centrally coordinated hub to assume responsibility for extended care transition and follow‐up.[3] Notably, studies of pediatric patients and their families document the effectiveness of enhanced inpatient education and engagement while in the hospital.[3] Unfortunately, a study among adults at a top‐ranked academic institution indicates poor communication among nurses and physicians regarding patient discharge education.[4] Efforts to improve nursephysician communication by redesigning the hospitalist model of care delivery at a Veterans Affairs (VA) institution appeared to enhance perceptions of communication among the care team members and reduced length of stay, but disappointingly there was no reduction in readmission rates.[5] Studies such as this are essential in identifying which specific interventions may actually change outcomes such as readmission rates.

In 1984, a diminutive elderly woman provocatively squawked Where's the beef?, launching a highly successful advertising campaign for Wendy's hamburger chain.[6] This catchphrase may aptly describe Bradley and colleague's survey study of the State Action on Avoidable Rehospitalization (STAAR) and Hospital‐to‐Home (H2H) campaigns.[7] Auerbach and colleagues eloquently stated in a 2007 New England Journal of Medicine perspective[8] how they had witnessed recent initiatives that emphasize dissemination of innovative but unproven strategies, an approach that runs counter to the principle of following the evidence[9] in selecting interventions that meet quality and safety goals.[10] I firmly agree with this assessment, and 6 years later believe we should be more thoughtful about potentially repeating implementation of unproven strategies.

Do we know if the interventions recommended by H2H and STAAR are what hospital care teams should be attempting? Even the authors mention that definitive evidence on their effectiveness is lacking. The H2H and STAAR programs certainly encourage some theoretically laudable activitiesmedication reconciliation by nurses, alerting outpatient physicians within 48 hours of patient discharge, and providing skilled nursing facilities the direct contact number of the inpatient treating physician for patients transferred. However, do these efforts actually improve patient outcomes? Before embarking on state or national campaigns to improve care, we should consider carefully what are the best evidence‐based interventions. Remarkably, some prior evidence indicates that direct communication between the hospital‐based physician and primary care provider (PCP) may not actually impact patient outcomes.[11] Newer research published in JHM confirms my belief that the PCP needs to be engaged by hospitalists during a hospitalization. Lindquist's research group at Northwestern nicely demonstrated how communication between a patient's PCP and the admitting hospitalist, complemented by contact between the PCP and patient within 24 hours postdischarge, reduced the probability of a medication discrepancy by 70%.[12] Although no evaluation of the effect on readmissions was reported, this study may provide information on causality related to the importance of PCP involvement in the care of hospitalized patients.

Numerous publications now document research on successfully implemented programs that lowered hospital readmissions, and are cited by CMS as evidence‐based interventions.[13] Projects Re‐Engineered Discharge (RED)[14] and Better Outcomes by Optimizing Safe Transitions[15] target the hospital discharge process, and both appear to lower hospital readmission rates. The Care Transitions Intervention (CTI),[16] Transitional Care Model (TCM),[17] and the Guided Care model[18] all leverage nurse practitioners or nurses to protect elderly patients during what can be a perilous care transition from hospital to home. CTI and TCM have been further validated in effectiveness studies.[19, 20] Two recent systematic reviews provide further insight into the complexity of efforts to reduce 30‐day rehospitalizations, but unfortunately do not reveal a desired silver bullet. The first focused exclusively on interventions to reduce 30‐day rehospitalization, and concluded that no single intervention was successful alone, but identified interventions bridging the hospital‐to‐home transition (eg, CTI), and a bundle of interventions such as Project RED as showing efficacy.[21] The second review more broadly sought to evaluate the effectiveness of hospital‐initiated strategies to prevent postdischarge adverse events (AEs) such as readmissions and emergency department visits,[22] stating Because of scant evidence, no conclusions could be reached on methods to prevent postdischarge AEs. The researchers' sobering conclusion stated that strategies to improve patient safety at hospital discharge remain unclear.

With rising federal penalties for higher‐than‐expected readmission rates, many hospital leaders eagerly join collaboratives aiming to reduce hospital readmissions. H2H appears to be among the largest, reporting >600 hospital participants, and STAAR has been active since 2009, with a recently published qualitative study identifying gaps in evidence for effective interventions, and deficits in quality improvement capabilities among some organizations as implementation challenges.[23] Notably, the survey by Bradley and colleagues documented that just half of the hospitals had a quality improvement (QI) team focused on reducing readmissions. Although laudable in their goals, H2H and STAAR may represent expensive commitments of staff and time to efforts that may not improve outcomes. Importantly, recently published research evaluating QI studies showed concerning results among patients with chronic obstructive pulmonary disease (COPD). A randomized controlled trial (RCT) conducted at 6 Glasgow hospitals evaluated supported self‐management (home visits by nurses and thorough education) by patients with moderate to severe COPD, but documented no changes in hospitalization or mortality.[24]Another RCT at 20 sites evaluated a comprehensive care management program to prevent hospitalizations among 960 VA patients with COPD.[25] It had to be stopped early due to elevated all‐cause mortality in the intervention group, and there was no difference in hospitalization rates.

Moving forward, QI efforts to reduce hospital readmissions should utilize proven interventions unless they are part of a rigorous trial. The emerging field of implementation science (the scientific study of methods to promote the systematic uptake of research findings and other evidence‐based practices into routine practice, and hence, to improve the quality and effectiveness of health services[26]) needs to be applied to additional research in this area.[27] Another consideration would be for CMS and funders such as the Commonwealth Foundation or The Robert Wood Johnson Foundation to encourage and fund merging of current initiatives to move away from competition and provide clarity to community hospitals. Regardless, such collaboration should still undertake formal evaluation to discern best approaches to implementation. I applaud the authors for recognizing that Input from hospitalists who are often critical links among inpatient and outpatient care and between patients and their families is strongly needed to ensure hospitals focus on what strategies are most effective for successful transitions from hospital to home. Yet, I wonder why neither of the large STAAR and H2H initiatives actively partnered with hospitalists and their specialty society (Society of Hospital Medicine) directly in the leadership of these initiatives? On the other hand, why not ask medical societies engaged in delivery of primary care (eg, American Academy for Family Practice, American College of Physicians, or Society of General Internal Medicine), especially to elderly patients (American Geriatric Society), to contribute directly? Involvement on an advisory board is likely not sufficient. Prior efforts document the willingness of these organizations to collaborate and achieve consensus on principles for transitions of care.[28] As powerfully articulated 6 years ago, [W]e must pursue the solutions to quality and safety problems in a way that does not blind us to harms, squander scarce resources, or delude us about the effectiveness of our efforts.[8]

The Hospital Readmission Reduction Program (HRRP)[1] contained within the Affordable Care Act focused national and local attention on hospital resources and efforts to reduce hospital readmissions. Driven by the Centers for Medicare and Medicaid Services' (CMS) desire to pay for value instead of volume, the response of hospitals and health systems appears to be yielding change across the United States.[2] A number of recent publications in the Journal of Hospital Medicine (JHM) exemplify the keen interest in reducing readmissions, while providing guidance regarding interventions and where we might target future research. Evidence from an exemplary systematic review of the pediatric literature confirms some experience in adults regarding effective interventionsall studies were multifacetedand highlights the importance of identifying a single healthcare provider or centrally coordinated hub to assume responsibility for extended care transition and follow‐up.[3] Notably, studies of pediatric patients and their families document the effectiveness of enhanced inpatient education and engagement while in the hospital.[3] Unfortunately, a study among adults at a top‐ranked academic institution indicates poor communication among nurses and physicians regarding patient discharge education.[4] Efforts to improve nursephysician communication by redesigning the hospitalist model of care delivery at a Veterans Affairs (VA) institution appeared to enhance perceptions of communication among the care team members and reduced length of stay, but disappointingly there was no reduction in readmission rates.[5] Studies such as this are essential in identifying which specific interventions may actually change outcomes such as readmission rates.

In 1984, a diminutive elderly woman provocatively squawked Where's the beef?, launching a highly successful advertising campaign for Wendy's hamburger chain.[6] This catchphrase may aptly describe Bradley and colleague's survey study of the State Action on Avoidable Rehospitalization (STAAR) and Hospital‐to‐Home (H2H) campaigns.[7] Auerbach and colleagues eloquently stated in a 2007 New England Journal of Medicine perspective[8] how they had witnessed recent initiatives that emphasize dissemination of innovative but unproven strategies, an approach that runs counter to the principle of following the evidence[9] in selecting interventions that meet quality and safety goals.[10] I firmly agree with this assessment, and 6 years later believe we should be more thoughtful about potentially repeating implementation of unproven strategies.

Do we know if the interventions recommended by H2H and STAAR are what hospital care teams should be attempting? Even the authors mention that definitive evidence on their effectiveness is lacking. The H2H and STAAR programs certainly encourage some theoretically laudable activitiesmedication reconciliation by nurses, alerting outpatient physicians within 48 hours of patient discharge, and providing skilled nursing facilities the direct contact number of the inpatient treating physician for patients transferred. However, do these efforts actually improve patient outcomes? Before embarking on state or national campaigns to improve care, we should consider carefully what are the best evidence‐based interventions. Remarkably, some prior evidence indicates that direct communication between the hospital‐based physician and primary care provider (PCP) may not actually impact patient outcomes.[11] Newer research published in JHM confirms my belief that the PCP needs to be engaged by hospitalists during a hospitalization. Lindquist's research group at Northwestern nicely demonstrated how communication between a patient's PCP and the admitting hospitalist, complemented by contact between the PCP and patient within 24 hours postdischarge, reduced the probability of a medication discrepancy by 70%.[12] Although no evaluation of the effect on readmissions was reported, this study may provide information on causality related to the importance of PCP involvement in the care of hospitalized patients.

Numerous publications now document research on successfully implemented programs that lowered hospital readmissions, and are cited by CMS as evidence‐based interventions.[13] Projects Re‐Engineered Discharge (RED)[14] and Better Outcomes by Optimizing Safe Transitions[15] target the hospital discharge process, and both appear to lower hospital readmission rates. The Care Transitions Intervention (CTI),[16] Transitional Care Model (TCM),[17] and the Guided Care model[18] all leverage nurse practitioners or nurses to protect elderly patients during what can be a perilous care transition from hospital to home. CTI and TCM have been further validated in effectiveness studies.[19, 20] Two recent systematic reviews provide further insight into the complexity of efforts to reduce 30‐day rehospitalizations, but unfortunately do not reveal a desired silver bullet. The first focused exclusively on interventions to reduce 30‐day rehospitalization, and concluded that no single intervention was successful alone, but identified interventions bridging the hospital‐to‐home transition (eg, CTI), and a bundle of interventions such as Project RED as showing efficacy.[21] The second review more broadly sought to evaluate the effectiveness of hospital‐initiated strategies to prevent postdischarge adverse events (AEs) such as readmissions and emergency department visits,[22] stating Because of scant evidence, no conclusions could be reached on methods to prevent postdischarge AEs. The researchers' sobering conclusion stated that strategies to improve patient safety at hospital discharge remain unclear.

With rising federal penalties for higher‐than‐expected readmission rates, many hospital leaders eagerly join collaboratives aiming to reduce hospital readmissions. H2H appears to be among the largest, reporting >600 hospital participants, and STAAR has been active since 2009, with a recently published qualitative study identifying gaps in evidence for effective interventions, and deficits in quality improvement capabilities among some organizations as implementation challenges.[23] Notably, the survey by Bradley and colleagues documented that just half of the hospitals had a quality improvement (QI) team focused on reducing readmissions. Although laudable in their goals, H2H and STAAR may represent expensive commitments of staff and time to efforts that may not improve outcomes. Importantly, recently published research evaluating QI studies showed concerning results among patients with chronic obstructive pulmonary disease (COPD). A randomized controlled trial (RCT) conducted at 6 Glasgow hospitals evaluated supported self‐management (home visits by nurses and thorough education) by patients with moderate to severe COPD, but documented no changes in hospitalization or mortality.[24]Another RCT at 20 sites evaluated a comprehensive care management program to prevent hospitalizations among 960 VA patients with COPD.[25] It had to be stopped early due to elevated all‐cause mortality in the intervention group, and there was no difference in hospitalization rates.

Moving forward, QI efforts to reduce hospital readmissions should utilize proven interventions unless they are part of a rigorous trial. The emerging field of implementation science (the scientific study of methods to promote the systematic uptake of research findings and other evidence‐based practices into routine practice, and hence, to improve the quality and effectiveness of health services[26]) needs to be applied to additional research in this area.[27] Another consideration would be for CMS and funders such as the Commonwealth Foundation or The Robert Wood Johnson Foundation to encourage and fund merging of current initiatives to move away from competition and provide clarity to community hospitals. Regardless, such collaboration should still undertake formal evaluation to discern best approaches to implementation. I applaud the authors for recognizing that Input from hospitalists who are often critical links among inpatient and outpatient care and between patients and their families is strongly needed to ensure hospitals focus on what strategies are most effective for successful transitions from hospital to home. Yet, I wonder why neither of the large STAAR and H2H initiatives actively partnered with hospitalists and their specialty society (Society of Hospital Medicine) directly in the leadership of these initiatives? On the other hand, why not ask medical societies engaged in delivery of primary care (eg, American Academy for Family Practice, American College of Physicians, or Society of General Internal Medicine), especially to elderly patients (American Geriatric Society), to contribute directly? Involvement on an advisory board is likely not sufficient. Prior efforts document the willingness of these organizations to collaborate and achieve consensus on principles for transitions of care.[28] As powerfully articulated 6 years ago, [W]e must pursue the solutions to quality and safety problems in a way that does not blind us to harms, squander scarce resources, or delude us about the effectiveness of our efforts.[8]