The affordability of prescription medications continues to be one of the most pressing public health issues in the United States. Many patients reduce their prescribed doses to make medications last longer or do not fill prescriptions because of cost.1 Cost‐related medication underuse affects patients with and without drug insurance coverage,2 and is likely to become even more problematic as employers scale back on drug benefits3 and drug prices continue to increase.4 The landmark Patient Protection and Affordability Act passed in March 2010 does little to address this issue.5

Existing estimates of cost‐related medication underuse come largely from surveys of ambulatory patients. For example, using data from the Medicare Current Beneficiary Survey, Maden et al. estimated that 11% to 15% of patients reduced medication use in the past year because of cost.6 Tseng and colleagues found very similar rates of cost‐related underuse in managed care beneficiaries with diabetes.7

Hospitalized patients, who have a high burden of disease and tend to use more medications than their ambulatory counterparts, may be particularly vulnerable to cost‐related underuse but, thus far, have been subject to little investigation. New medications, which are frequently prescribed at the time of discharge, may exacerbate these issues further and contribute to preventable readmissions. Accordingly, we surveyed a cohort of medical inpatients at a large academic medical center to estimate the prevalence and predictors of cost‐related medication underuse for hospitalized managed care patients, and to identify strategies that patients perceive as helpful to make medications more affordable.

METHODS

RESULTS

During the study period, 483 potentially‐eligible patients were admitted to the general medicine, cardiology, and oncology services. We excluded 167 because they had been discharged prior to being identified, had been surveyed or already declined participation on a prior admission, or were not managed care enrollees (see Appendix A). Of the remaining 316 subjects, 130 participated in the inpatient survey (response rate = 41%); 93 (75%) of these patients were reached by telephone after hospital discharge and completed the postdischarge survey. The baseline characteristics of our respondents are presented in Table 1. Patients had a mean age of 52 years, were 50% male and two‐thirds of white race, represented a range of household incomes, and almost all had employer‐sponsored prescription coverage. Prior to admission, patients took an average of 5 prescription medications and paid an average copayment of $10.80 and $21.60 for each generic and brand‐name prescription, respectively.

Cost‐Related Medication Underuse

Thirty (23%) of the survey respondents reported at least 1 cost‐related medication underuse strategy in the year prior to their hospital admission (Figure 1), most commonly not filling a prescription at all because of cost (n = 26; 20%). Rates of cost‐related underuse were highest for patients of black race, low income, and women (Figure 2).

Figure 1

Figure 2

In unadjusted analyses, black respondents had 4.60 (95% confidence interval [CI], 1.63 to 13.0) times the odds of reporting cost‐related underuse than non‐Hispanic white respondents (Table 2). The association of black race and cost‐related underuse appears to be confounded, in part, by income (adjusted odds ratio for black race was 4.16; 95% CI, 1.34 to 12.86) and the number of medications patients used on a regular basis (adjusted odds ratio for black race was 4.14; 95% CI, 1.44 to 11.96). After controlling for these variables, as well as age and gender, the relationship between race and cost‐related underuse remained statistically significant (adjusted odds ratio 3.39; 95% CI, 1.05 to 11.02) (Table 2).

Table 2.Predictors of Cost‐Related Underuse

Predictor	Unadjusted Odds Ratio (95% CI)	Adjusted Odds Ratio (95% CI)
Abbreviations: CI = confidence interval.
Age (per additional year)	0.98 (0.941.02)	0.97 (0.931.01)
Male (vs female)	0.84 (0.371.90)	1.03 (0.432.48)
Race (vs white race)
Black	4.60 (1.6313.0)	3.39 (1.0511.02)
Other	1.10 (0.363.37)	0.77 (0.202.99)
No. of medications (per additional medication)	1.10 (1.001.20)	1.10 (1.001.22)
High income (vs low income)	0.62 (0.271.42)	0.71 (0.242.07)

Strategies to Help Make Medications More Affordable

Virtually all respondents (n = 123; 95%) endorsed at least one of the proposed strategies to make medications more affordable (Figure 3). A majority felt that lowering cost sharing (94%), or receiving information about lower‐cost medication options (83%) or programs to subsidize medication costs (83%) would be helpful. Approximately 70% of patients stated that speaking to their outpatient physicians might be helpful, although only 14% reported actually speaking with their primary care provider about medication costs in the past year. Results were mixed for other strategies, including speaking with their inpatient physicians.

Figure 3

DISCUSSION

Almost a quarter of the medical inpatients we surveyed had not filled a medication because of cost, or had skipped doses, reduced dosages, or split pills to make their medicines last longer in the prior year. This amount is larger than that found in many prior studies, conducted in outpatient settings, in which 11% to 19% of patients report cost‐related underuse.68, 10, 11 Our results are particularly striking considering that our study cohort consisted exclusively of patients with commercial health insurance, the vast majority of whom also had employer‐sponsored drug coverage. Cost‐related medication underuse may be even more prevalent among hospitalized patients with less generous benefits, including the uninsured and perhaps even beneficiaries of Medicare Part D.

Reductions in medication use because of cost were particularly high among black patients, whose odds of reporting cost‐related underuse were more than 3 times higher than that of patients of non‐Hispanic white race. Race‐related differences in cost‐related underuse have been observed in outpatient studies,68, 12 and may be an important contributor to racial disparities in evidence‐based medication use.1315 These differences may, in part, reflect racial variations in socioeconomic status; lower income patients, who are more likely to be from a racial or ethnic minority, are more sensitive to cost sharing than higher income individuals.16 Consistent with this, the relationship between race and cost‐related underuse in our study was smaller but still highly significant in multivariable models that adjusted for income.

Not surprisingly, the underuse of effective prescription medications is associated with adverse clinical and economic consequences.17 Heisler et al. found that patients who had restricted medications because of cost were 76% more likely to report a decline in their health status than those who had not.18 The health effects of cost‐related underuse are likely to be particularly significant for hospitalized patients, given their high burden of disease and the frequency with which they are prescribed medications at discharge to treat the condition that led to their initial hospitalization. Thus, targeting efforts to address cost‐related underuse patients who are hospitalized may be an efficient method of improving patient health and reducing preventable readmissions. This is consistent with efforts that address, in the inpatient setting, other health issues that are commonly encountered in the ambulatory arena, such as immunizations and smoking cessation.19

Our survey respondents endorsed numerous strategies as being potentially helpful. Predictably, support for lowering copayments was extremely high. While this may not be practical or even desirable for some medications, lowering copayments for highly effective medications, such as statins and antihypertensives, in the context of value‐based insurance design, is an increasingly adopted strategy that has the potential to simultaneously improve clinical outcomes and reduce overall health spending.20, 21

While the majority of patients felt that talking to their outpatient physicians or pharmacists about medication costs might be helpful, the effectiveness of this strategy is unclear. Consistent with prior results,22, 23 the vast majority of the patients we surveyed had not discussed medication costs prior to their admission or after filling newly prescribed medications. Further, although physicians could help reduce drug expenditures in a variety of ways, including the increased ordering of generic drugs,24 many physicians are uncomfortable talking to their patients about costs,25 have limited knowledge about their patients' out‐of‐pocket expenditures, feel that addressing this issue is not their responsibility,26 or do not have resources, such as electronic formulary information, that could facilitate these discussions in an efficient manner.

An alternative strategy may be to provide patients with better education about medication costs. Virtually none of the patients we surveyed knew how much they would pay for their new prescriptions before visiting the pharmacy. These findings are similar to those observed in the outpatient setting,27 and suggest an opportunity to provide patients with information about the cost of their newly and previously prescribed drugs, and to facilitate discussions between patients and inpatient providers about predischarge prescribing decisions, in the same spirit as other predischarge patient education.28 Of course, issues related to transitions of care between the hospital and community setting, and coordination between inpatient and outpatient providers, must be adequately addressed for this strategy to be effective.

Our study has several notable limitations. It had a relatively small sample size and low response rate. Respondents may have differed systematically from non‐respondents, and we were unable to compare the characteristics of both populations. Further, we studied commercially insured inpatients on internal medicine services at an academic medical center, and thus our results may not be generalizable to patients hospitalized in other settings, or with different types of insurance coverage, including the uninsured. The primary outcome of our study was to determine self‐reported cost‐related underuse. While we used validated measures,8 it is possible that patients who reported reducing their medication use in response to cost may not have actually done so. We did not collect information on education or health literacy, nor did we have access to detailed information about our respondents' pharmacy benefit design structures; these important factors may have confounded our analyses, and/or may have been mediators of our observed results, and should be evaluated further in future studies. We did not have adequate statistical power to evaluate whether patients using specific classes of medications were particularly prone to cost‐related underuse.

Despite these limitations, our study is the first, to our knowledge, to evaluate the impact of medication costs on use in a cohort of hospitalized individuals. The high levels of cost‐related underuse that we observed is concerning. Our results support calls for the further development of interventions to address high medication costs and for the consideration of novel approaches to assist patients around the time of hospital discharge.

The affordability of prescription medications continues to be one of the most pressing public health issues in the United States. Many patients reduce their prescribed doses to make medications last longer or do not fill prescriptions because of cost.1 Cost‐related medication underuse affects patients with and without drug insurance coverage,2 and is likely to become even more problematic as employers scale back on drug benefits3 and drug prices continue to increase.4 The landmark Patient Protection and Affordability Act passed in March 2010 does little to address this issue.5

Existing estimates of cost‐related medication underuse come largely from surveys of ambulatory patients. For example, using data from the Medicare Current Beneficiary Survey, Maden et al. estimated that 11% to 15% of patients reduced medication use in the past year because of cost.6 Tseng and colleagues found very similar rates of cost‐related underuse in managed care beneficiaries with diabetes.7

Hospitalized patients, who have a high burden of disease and tend to use more medications than their ambulatory counterparts, may be particularly vulnerable to cost‐related underuse but, thus far, have been subject to little investigation. New medications, which are frequently prescribed at the time of discharge, may exacerbate these issues further and contribute to preventable readmissions. Accordingly, we surveyed a cohort of medical inpatients at a large academic medical center to estimate the prevalence and predictors of cost‐related medication underuse for hospitalized managed care patients, and to identify strategies that patients perceive as helpful to make medications more affordable.

METHODS

RESULTS

During the study period, 483 potentially‐eligible patients were admitted to the general medicine, cardiology, and oncology services. We excluded 167 because they had been discharged prior to being identified, had been surveyed or already declined participation on a prior admission, or were not managed care enrollees (see Appendix A). Of the remaining 316 subjects, 130 participated in the inpatient survey (response rate = 41%); 93 (75%) of these patients were reached by telephone after hospital discharge and completed the postdischarge survey. The baseline characteristics of our respondents are presented in Table 1. Patients had a mean age of 52 years, were 50% male and two‐thirds of white race, represented a range of household incomes, and almost all had employer‐sponsored prescription coverage. Prior to admission, patients took an average of 5 prescription medications and paid an average copayment of $10.80 and $21.60 for each generic and brand‐name prescription, respectively.

Cost‐Related Medication Underuse

Thirty (23%) of the survey respondents reported at least 1 cost‐related medication underuse strategy in the year prior to their hospital admission (Figure 1), most commonly not filling a prescription at all because of cost (n = 26; 20%). Rates of cost‐related underuse were highest for patients of black race, low income, and women (Figure 2).

Figure 1

Figure 2

In unadjusted analyses, black respondents had 4.60 (95% confidence interval [CI], 1.63 to 13.0) times the odds of reporting cost‐related underuse than non‐Hispanic white respondents (Table 2). The association of black race and cost‐related underuse appears to be confounded, in part, by income (adjusted odds ratio for black race was 4.16; 95% CI, 1.34 to 12.86) and the number of medications patients used on a regular basis (adjusted odds ratio for black race was 4.14; 95% CI, 1.44 to 11.96). After controlling for these variables, as well as age and gender, the relationship between race and cost‐related underuse remained statistically significant (adjusted odds ratio 3.39; 95% CI, 1.05 to 11.02) (Table 2).

Table 2.Predictors of Cost‐Related Underuse

Predictor	Unadjusted Odds Ratio (95% CI)	Adjusted Odds Ratio (95% CI)
Abbreviations: CI = confidence interval.
Age (per additional year)	0.98 (0.941.02)	0.97 (0.931.01)
Male (vs female)	0.84 (0.371.90)	1.03 (0.432.48)
Race (vs white race)
Black	4.60 (1.6313.0)	3.39 (1.0511.02)
Other	1.10 (0.363.37)	0.77 (0.202.99)
No. of medications (per additional medication)	1.10 (1.001.20)	1.10 (1.001.22)
High income (vs low income)	0.62 (0.271.42)	0.71 (0.242.07)

Strategies to Help Make Medications More Affordable

Virtually all respondents (n = 123; 95%) endorsed at least one of the proposed strategies to make medications more affordable (Figure 3). A majority felt that lowering cost sharing (94%), or receiving information about lower‐cost medication options (83%) or programs to subsidize medication costs (83%) would be helpful. Approximately 70% of patients stated that speaking to their outpatient physicians might be helpful, although only 14% reported actually speaking with their primary care provider about medication costs in the past year. Results were mixed for other strategies, including speaking with their inpatient physicians.

Figure 3

DISCUSSION

Almost a quarter of the medical inpatients we surveyed had not filled a medication because of cost, or had skipped doses, reduced dosages, or split pills to make their medicines last longer in the prior year. This amount is larger than that found in many prior studies, conducted in outpatient settings, in which 11% to 19% of patients report cost‐related underuse.68, 10, 11 Our results are particularly striking considering that our study cohort consisted exclusively of patients with commercial health insurance, the vast majority of whom also had employer‐sponsored drug coverage. Cost‐related medication underuse may be even more prevalent among hospitalized patients with less generous benefits, including the uninsured and perhaps even beneficiaries of Medicare Part D.

Reductions in medication use because of cost were particularly high among black patients, whose odds of reporting cost‐related underuse were more than 3 times higher than that of patients of non‐Hispanic white race. Race‐related differences in cost‐related underuse have been observed in outpatient studies,68, 12 and may be an important contributor to racial disparities in evidence‐based medication use.1315 These differences may, in part, reflect racial variations in socioeconomic status; lower income patients, who are more likely to be from a racial or ethnic minority, are more sensitive to cost sharing than higher income individuals.16 Consistent with this, the relationship between race and cost‐related underuse in our study was smaller but still highly significant in multivariable models that adjusted for income.

Not surprisingly, the underuse of effective prescription medications is associated with adverse clinical and economic consequences.17 Heisler et al. found that patients who had restricted medications because of cost were 76% more likely to report a decline in their health status than those who had not.18 The health effects of cost‐related underuse are likely to be particularly significant for hospitalized patients, given their high burden of disease and the frequency with which they are prescribed medications at discharge to treat the condition that led to their initial hospitalization. Thus, targeting efforts to address cost‐related underuse patients who are hospitalized may be an efficient method of improving patient health and reducing preventable readmissions. This is consistent with efforts that address, in the inpatient setting, other health issues that are commonly encountered in the ambulatory arena, such as immunizations and smoking cessation.19

Our survey respondents endorsed numerous strategies as being potentially helpful. Predictably, support for lowering copayments was extremely high. While this may not be practical or even desirable for some medications, lowering copayments for highly effective medications, such as statins and antihypertensives, in the context of value‐based insurance design, is an increasingly adopted strategy that has the potential to simultaneously improve clinical outcomes and reduce overall health spending.20, 21

While the majority of patients felt that talking to their outpatient physicians or pharmacists about medication costs might be helpful, the effectiveness of this strategy is unclear. Consistent with prior results,22, 23 the vast majority of the patients we surveyed had not discussed medication costs prior to their admission or after filling newly prescribed medications. Further, although physicians could help reduce drug expenditures in a variety of ways, including the increased ordering of generic drugs,24 many physicians are uncomfortable talking to their patients about costs,25 have limited knowledge about their patients' out‐of‐pocket expenditures, feel that addressing this issue is not their responsibility,26 or do not have resources, such as electronic formulary information, that could facilitate these discussions in an efficient manner.

An alternative strategy may be to provide patients with better education about medication costs. Virtually none of the patients we surveyed knew how much they would pay for their new prescriptions before visiting the pharmacy. These findings are similar to those observed in the outpatient setting,27 and suggest an opportunity to provide patients with information about the cost of their newly and previously prescribed drugs, and to facilitate discussions between patients and inpatient providers about predischarge prescribing decisions, in the same spirit as other predischarge patient education.28 Of course, issues related to transitions of care between the hospital and community setting, and coordination between inpatient and outpatient providers, must be adequately addressed for this strategy to be effective.

Our study has several notable limitations. It had a relatively small sample size and low response rate. Respondents may have differed systematically from non‐respondents, and we were unable to compare the characteristics of both populations. Further, we studied commercially insured inpatients on internal medicine services at an academic medical center, and thus our results may not be generalizable to patients hospitalized in other settings, or with different types of insurance coverage, including the uninsured. The primary outcome of our study was to determine self‐reported cost‐related underuse. While we used validated measures,8 it is possible that patients who reported reducing their medication use in response to cost may not have actually done so. We did not collect information on education or health literacy, nor did we have access to detailed information about our respondents' pharmacy benefit design structures; these important factors may have confounded our analyses, and/or may have been mediators of our observed results, and should be evaluated further in future studies. We did not have adequate statistical power to evaluate whether patients using specific classes of medications were particularly prone to cost‐related underuse.

Despite these limitations, our study is the first, to our knowledge, to evaluate the impact of medication costs on use in a cohort of hospitalized individuals. The high levels of cost‐related underuse that we observed is concerning. Our results support calls for the further development of interventions to address high medication costs and for the consideration of novel approaches to assist patients around the time of hospital discharge.

The affordability of prescription medications continues to be one of the most pressing public health issues in the United States. Many patients reduce their prescribed doses to make medications last longer or do not fill prescriptions because of cost.1 Cost‐related medication underuse affects patients with and without drug insurance coverage,2 and is likely to become even more problematic as employers scale back on drug benefits3 and drug prices continue to increase.4 The landmark Patient Protection and Affordability Act passed in March 2010 does little to address this issue.5

Existing estimates of cost‐related medication underuse come largely from surveys of ambulatory patients. For example, using data from the Medicare Current Beneficiary Survey, Maden et al. estimated that 11% to 15% of patients reduced medication use in the past year because of cost.6 Tseng and colleagues found very similar rates of cost‐related underuse in managed care beneficiaries with diabetes.7

Hospitalized patients, who have a high burden of disease and tend to use more medications than their ambulatory counterparts, may be particularly vulnerable to cost‐related underuse but, thus far, have been subject to little investigation. New medications, which are frequently prescribed at the time of discharge, may exacerbate these issues further and contribute to preventable readmissions. Accordingly, we surveyed a cohort of medical inpatients at a large academic medical center to estimate the prevalence and predictors of cost‐related medication underuse for hospitalized managed care patients, and to identify strategies that patients perceive as helpful to make medications more affordable.

METHODS

RESULTS

During the study period, 483 potentially‐eligible patients were admitted to the general medicine, cardiology, and oncology services. We excluded 167 because they had been discharged prior to being identified, had been surveyed or already declined participation on a prior admission, or were not managed care enrollees (see Appendix A). Of the remaining 316 subjects, 130 participated in the inpatient survey (response rate = 41%); 93 (75%) of these patients were reached by telephone after hospital discharge and completed the postdischarge survey. The baseline characteristics of our respondents are presented in Table 1. Patients had a mean age of 52 years, were 50% male and two‐thirds of white race, represented a range of household incomes, and almost all had employer‐sponsored prescription coverage. Prior to admission, patients took an average of 5 prescription medications and paid an average copayment of $10.80 and $21.60 for each generic and brand‐name prescription, respectively.

Cost‐Related Medication Underuse

Thirty (23%) of the survey respondents reported at least 1 cost‐related medication underuse strategy in the year prior to their hospital admission (Figure 1), most commonly not filling a prescription at all because of cost (n = 26; 20%). Rates of cost‐related underuse were highest for patients of black race, low income, and women (Figure 2).

Figure 1

Figure 2

In unadjusted analyses, black respondents had 4.60 (95% confidence interval [CI], 1.63 to 13.0) times the odds of reporting cost‐related underuse than non‐Hispanic white respondents (Table 2). The association of black race and cost‐related underuse appears to be confounded, in part, by income (adjusted odds ratio for black race was 4.16; 95% CI, 1.34 to 12.86) and the number of medications patients used on a regular basis (adjusted odds ratio for black race was 4.14; 95% CI, 1.44 to 11.96). After controlling for these variables, as well as age and gender, the relationship between race and cost‐related underuse remained statistically significant (adjusted odds ratio 3.39; 95% CI, 1.05 to 11.02) (Table 2).

Table 2.Predictors of Cost‐Related Underuse

Predictor	Unadjusted Odds Ratio (95% CI)	Adjusted Odds Ratio (95% CI)
Abbreviations: CI = confidence interval.
Age (per additional year)	0.98 (0.941.02)	0.97 (0.931.01)
Male (vs female)	0.84 (0.371.90)	1.03 (0.432.48)
Race (vs white race)
Black	4.60 (1.6313.0)	3.39 (1.0511.02)
Other	1.10 (0.363.37)	0.77 (0.202.99)
No. of medications (per additional medication)	1.10 (1.001.20)	1.10 (1.001.22)
High income (vs low income)	0.62 (0.271.42)	0.71 (0.242.07)

Strategies to Help Make Medications More Affordable

Virtually all respondents (n = 123; 95%) endorsed at least one of the proposed strategies to make medications more affordable (Figure 3). A majority felt that lowering cost sharing (94%), or receiving information about lower‐cost medication options (83%) or programs to subsidize medication costs (83%) would be helpful. Approximately 70% of patients stated that speaking to their outpatient physicians might be helpful, although only 14% reported actually speaking with their primary care provider about medication costs in the past year. Results were mixed for other strategies, including speaking with their inpatient physicians.

Figure 3

DISCUSSION

Almost a quarter of the medical inpatients we surveyed had not filled a medication because of cost, or had skipped doses, reduced dosages, or split pills to make their medicines last longer in the prior year. This amount is larger than that found in many prior studies, conducted in outpatient settings, in which 11% to 19% of patients report cost‐related underuse.68, 10, 11 Our results are particularly striking considering that our study cohort consisted exclusively of patients with commercial health insurance, the vast majority of whom also had employer‐sponsored drug coverage. Cost‐related medication underuse may be even more prevalent among hospitalized patients with less generous benefits, including the uninsured and perhaps even beneficiaries of Medicare Part D.

Reductions in medication use because of cost were particularly high among black patients, whose odds of reporting cost‐related underuse were more than 3 times higher than that of patients of non‐Hispanic white race. Race‐related differences in cost‐related underuse have been observed in outpatient studies,68, 12 and may be an important contributor to racial disparities in evidence‐based medication use.1315 These differences may, in part, reflect racial variations in socioeconomic status; lower income patients, who are more likely to be from a racial or ethnic minority, are more sensitive to cost sharing than higher income individuals.16 Consistent with this, the relationship between race and cost‐related underuse in our study was smaller but still highly significant in multivariable models that adjusted for income.

Not surprisingly, the underuse of effective prescription medications is associated with adverse clinical and economic consequences.17 Heisler et al. found that patients who had restricted medications because of cost were 76% more likely to report a decline in their health status than those who had not.18 The health effects of cost‐related underuse are likely to be particularly significant for hospitalized patients, given their high burden of disease and the frequency with which they are prescribed medications at discharge to treat the condition that led to their initial hospitalization. Thus, targeting efforts to address cost‐related underuse patients who are hospitalized may be an efficient method of improving patient health and reducing preventable readmissions. This is consistent with efforts that address, in the inpatient setting, other health issues that are commonly encountered in the ambulatory arena, such as immunizations and smoking cessation.19

Our survey respondents endorsed numerous strategies as being potentially helpful. Predictably, support for lowering copayments was extremely high. While this may not be practical or even desirable for some medications, lowering copayments for highly effective medications, such as statins and antihypertensives, in the context of value‐based insurance design, is an increasingly adopted strategy that has the potential to simultaneously improve clinical outcomes and reduce overall health spending.20, 21

While the majority of patients felt that talking to their outpatient physicians or pharmacists about medication costs might be helpful, the effectiveness of this strategy is unclear. Consistent with prior results,22, 23 the vast majority of the patients we surveyed had not discussed medication costs prior to their admission or after filling newly prescribed medications. Further, although physicians could help reduce drug expenditures in a variety of ways, including the increased ordering of generic drugs,24 many physicians are uncomfortable talking to their patients about costs,25 have limited knowledge about their patients' out‐of‐pocket expenditures, feel that addressing this issue is not their responsibility,26 or do not have resources, such as electronic formulary information, that could facilitate these discussions in an efficient manner.

An alternative strategy may be to provide patients with better education about medication costs. Virtually none of the patients we surveyed knew how much they would pay for their new prescriptions before visiting the pharmacy. These findings are similar to those observed in the outpatient setting,27 and suggest an opportunity to provide patients with information about the cost of their newly and previously prescribed drugs, and to facilitate discussions between patients and inpatient providers about predischarge prescribing decisions, in the same spirit as other predischarge patient education.28 Of course, issues related to transitions of care between the hospital and community setting, and coordination between inpatient and outpatient providers, must be adequately addressed for this strategy to be effective.

Our study has several notable limitations. It had a relatively small sample size and low response rate. Respondents may have differed systematically from non‐respondents, and we were unable to compare the characteristics of both populations. Further, we studied commercially insured inpatients on internal medicine services at an academic medical center, and thus our results may not be generalizable to patients hospitalized in other settings, or with different types of insurance coverage, including the uninsured. The primary outcome of our study was to determine self‐reported cost‐related underuse. While we used validated measures,8 it is possible that patients who reported reducing their medication use in response to cost may not have actually done so. We did not collect information on education or health literacy, nor did we have access to detailed information about our respondents' pharmacy benefit design structures; these important factors may have confounded our analyses, and/or may have been mediators of our observed results, and should be evaluated further in future studies. We did not have adequate statistical power to evaluate whether patients using specific classes of medications were particularly prone to cost‐related underuse.

Despite these limitations, our study is the first, to our knowledge, to evaluate the impact of medication costs on use in a cohort of hospitalized individuals. The high levels of cost‐related underuse that we observed is concerning. Our results support calls for the further development of interventions to address high medication costs and for the consideration of novel approaches to assist patients around the time of hospital discharge.

In the United States, general medical inpatient care is increasingly provided by hospital‐based physicians, also called hospitalists.1 The field of pediatrics is no exception, and by 2005 there were an estimated 1000 pediatric hospitalists in the workforce.2 Current numbers are likely to be greater than 2500, as the need for pediatric hospitalists has grown considerably.

At the same time, the quality of care delivered by the United States health system has come under increased scrutiny. In 2001, the Institute of Medicine, in its report on the quality of healthcare in America, concluded that between the care we have and what we could have lies not just a gap but a chasm.3 Meanwhile, the cost of healthcare delivery continues to increase. The pressure to deliver cost‐effective, high quality care is among the more important forces driving the proliferation of hospitalists.4

Over the last decade, data supporting the role of hospitalists in improving quality of care for adult patients has continued to accumulate.58 A 2007 retrospective cohort study by Lindenaur et al.7 included nearly 77,000 adult patients and found small reductions in length of stay without adverse effects on mortality or readmission rates, and a 2009 systematic review by Peterson6 included 33 studies and concluded that in general inpatient care of general medical patients by hospitalist physicians leads to decreased hospital cost and length of stay. A 2002 study by Meltzer et al.8 is also interesting, suggesting that improvements in costs and short‐term mortality are related to the disease‐specific experience of hospitalists.

Similar data for pediatric hospitalists has been slower to emerge. A systematic review of the literature by Landrigan et al., which included studies through 2004, concluded that [R]esearch suggests that pediatric hospitalists decrease costs and length of stay . The quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking.9 Since the publication of that review, there have been multiple studies which have sought to evaluate the quality of pediatric hospitalist systems. This review was undertaken to synthesize this new information, and to determine the effect of pediatric hospitalist systems on quality of care.

METHODS

A review of the available English language literature on the Medline database was undertaken in November of 2010 to answer the question, What are the differences in quality of care and outcomes of inpatient medical care provided by hospitalists versus non‐hospitalists in the pediatric population? Care metrics of interest were categorized according to the Society of Hospital Medicine's recommendations for measuring hospital performance.10

Search terms used (with additional medical subject headings [MeSH] terms in parenthesis) were hospital medicine (hospitalist), pediatrics (child health, child welfare), cost (cost and cost analysis), quality (quality indicators, healthcare), outcomes (outcome assessment, healthcare; outcomes and process assessment, healthcare); volume, patient satisfaction, length of stay, productivity (efficiency), provider satisfaction (attitude of health personnel, job satisfaction), mortality, and readmission rate (patient readmission). The citing articles search tool was used to identify other articles that potentially could meet criteria. Finally, references cited in the selected articles, as well as in excluded literature reviews, were searched for additional articles.

Articles were deemed eligible if they were published in a peer‐reviewed journal, if they had a comparative experimental design for hospitalists versus non‐hospitalists, and if they dealt exclusively with pediatric hospitalists. Noncomparative studies were excluded, as were studies that pertained to settings besides that of an inpatient pediatrics ward, such as pediatric intensive care units or emergency rooms. The search algorithm is diagrammed in Figure 1.

Figure 1

The selected articles were reviewed for the relevant outcome measures. The quality of each article was assessed using the Oxford Centre for Evidence‐Based Medicine levels of evidence,11 a widely accepted standard for critical analysis of studies. Levels of evidence are assigned to studies, from 1a (systematic reviews of randomized controlled trials) to 5 (expert opinion only). Well‐conducted prospective cohort studies receive a rating of 2c; those with wide confidence intervals due to small sample size receive a minus () modifier. This system does not specifically address survey studies, which were therefore not assigned a level of evidence.

RESULTS

The screening process yielded 92 possible relevant articles, which were then reviewed individually (by G.M.M.) by title and abstract. A total of 81 articles were excluded, including 48 studies that were either noncomparative or descriptive in nature. Ten of the identified articles were reviews and did not contain primary data. Nine studies were not restricted to the pediatric population. Also excluded were 7 studies that did not have outcomes related to quality (eg, billing performance), and 7 studies of hospitalists in settings besides general pediatric wards (eg, pediatric intensive care units). Ten studies were thus identified. The cited reference tool was used to identify an additional article which met criteria, yielding 11 total articles that were included in the review.

Five of the identified studies published prior to 2005 were previously reviewed by Landrigan et al.9 Since then, 6 additional studies of similar nature have been published and were included here. Articles that met criteria but appeared in an earlier review are included in Table 1; new articles appear in Table 2. The results of all 11 articles were included for this discussion.

DISCUSSION

Of the 11 studies selected for this review, 10 measured length of stay as an outcome, with the majority favoring hospitalists but with mixed results. Three of these studies, those by Dwight et al.,12 Bellet and Whitaker,13 and Landrigan et al.,14 demonstrated 11% to 14% improvement for hospitalist services compared to community pediatricians. Boyd et al.,21 however, found exactly the opposite result, and 2 studies by Conway and Keren20 and Ogershok et al.16 found no difference in length of stay. Two more studies found benefits restricted to certain conditions: Wells et al.15 found 32% shorter lengths of stay for asthma, but not for other conditions; Srivastava et al.17 found a 13% reduction in length of stay for asthma and 11% for dehydration, but none for viral illnesses or when all conditions were combined. Bekmezian et al.18 found shorter lengths of stay on a hospitalist service for hematology and gastroenterology patients, and Simon et al.19 attribute a general trend of decreasing lengths of stay on a surgical service to the implementation of hospital comanagement for a small percentage of patients.

The most common quality measures studied were patient satisfaction, readmission rates, and mortality. Only 1 study by Ogershok et al.16 reported on patient satisfaction and found few differences between hospitalists and community pediatricians. Readmission rate were reported by 6 studies. Bellet and Whitaker13 found a higher readmission rate for pediatric hospitalists, Bekmezian et al.18 found a lower rate but on a subspecialty service. The study with the greatest power for this analysis, by Landrigan et al.14 with nearly 18,000 patients, found no difference, and neither did another 3 studies. Unsurprisingly, no study detected differences in mortality; it would be extremely difficult to adequately power a study to do so in the general pediatric setting, where mortality is rare.

The effect of relative experience of hospitalist physicians is uncertain. Boyd et al.21 speculated that 1 possible cause for the decreased lengths of stay and costs associated with their faculty group compared to hospitalists may have been due to the increased experience of the faculty group. Unfortunately, they were unable to generate statistical significance due to the small numbers of physicians in the study. In contrast, the hospitalists in the report by Dwight et al.12 had decreased lengths of stay but were less experienced. In the adult literature, the study by Meltzer et al.8 suggests that improved outcomes from hospitalist systems may not become apparent for 1 or more years after implementation, but none of the pediatric studies included in our review specifically address this issue. This leaves the possibility open that the hospitalist systems evaluated in some studies had insufficient time in which to develop increased efficiencies.

There were several limitations to our studies. First, due to the heterogeneity and methodological variations among the included studies, we were unable to perform a meta‐analysis. Second, the overall quality of evidence is limited due to the lack of randomized control trials. Third, a lack of agreement on appropriate quality markers has limited the study of quality of care. Published reports continue to focus on financial measures, such as length of stay, despite the recommendation in the previous review by Landrigan et al.9 that such studies would be of limited value. Finally, the current variability of hospitalist models and lack of study of factors that might influence outcomes makes comparisons difficult.

Despite these limitations, several interesting trends emerge from these studies. One such trend is that the more recent studies highlight that simple classification of hospitalist system versus traditional system fails to measure the complexity and nuance of care delivery. The 2006 study by Boyd et al.21 is especially notable because it showed the opposite effect of previous studies, namely, an increase in length of stay and costs for hospitalists at St Joseph's Medical Center in Phoenix, Arizona. In this study, the traditional faculty group was employed by the hospital, and the hospitalist group was a private practice model. The authors suggest that their faculty physicians were therefore operating like hospitalists in that almost all of their time was focused on inpatient care while they were on service. They also had a limited number of general pediatricians, who attended in the inpatient setting, who were more experienced than the private practice groups. Also, the authors theorize that their faculty may have had a closer working relationship with their residents due to additional service responsibilities and locations of the faculty group onsite. Further study of the care models utilized by faculty and hospitalist practices at St Joseph's and other hospitals may reveal important insights about improving the quality and efficiency of inpatient pediatric care in general.

Though there is a clear trend in the adult literature indicating that the use of hospitalists results in superior quality of care, there is less evidence for pediatric systems. The aforementioned previous review by Landrigan et al.9, in 2006 concluded that emerging research suggests that pediatric hospitalist systems decrease cost and length of stay, but also the quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking. Data from the 6 additional studies presented here lend limited support to the first hypothesis, and the presence of only 1 negative study is not sufficient to undermine it.

While data on quality markers such as readmission rate or mortality remain elusive, the 2 studies by Conway et al.20, 22 attempt to evaluate quality by comparing the use of evidence‐based therapies by hospitalists and community pediatricians. Though the use of objective PHIS data for UTI in 2009 did not confirm the conclusion suggested by the 2006 provider survey study, the attempt to find measurable outcomes such as the use of evidence‐based therapies is a start but we need more metrics, including rigorous patient outcome metrics, to define the quality of our care systems. Before the effect of hospitalist systems on quality is fully understood, more work will need to be done defining metrics for comparison.

Unfortunately, over 5 years since the previous review by Landrigan et al.9 called for increased focus on inpatient quality and understanding how to improve, the sophistication of our measurement of pediatric inpatient quality and understanding of the mechanisms underlying improvement is still in its infancy. We propose a solution at multiple levels.

First, the investment in research comparing system‐level interventions (eg, discharge process A vs discharge process B) must be increased. This investment increased significantly due to the over $1 billion in Recovery Act funding for comparative effectiveness research.23 However, the future investment in comparative effectiveness research, often called patient‐centered outcomes research, and proportion of investment focused on delivery system interventions is unclear. We propose that the investment in comparing delivery system interventions is essential to improving not only hospital medicine systems, but, more importantly, the healthcare system broadly. In addition, research investment needs to focus on reliably implementing proven interventions in systems of care, and evaluating both the effects on patient outcomes and cost, and the contextual factors associated with successful implementation.24 A hospital medicine example would be the comparison of the implementation of a guideline for a common disease across a set of hospitals. One could perform a prospective observational design, in which one compares high intensity versus low intensity intervention and assesses the baseline characteristics of the hospital systems, to understand their association with successful implementation and, ultimately, patient outcomes. One could also perform a clustered randomized design.

Second, the development and implementation of pediatric quality of care measures, including in the inpatient setting, needs to increase rapidly. The Children's Health Insurance Program (CHIP) and its focus on an initial core set of quality measures that expands over time, through an investment in measure development and validation, is an opportunity for pediatric hospital medicine. Inpatient measures should be a focus of measure development and implementation. We must move beyond a limited set of inpatient measures to a broader set focused on issues such as patient safety, hospital‐acquired infections, outcomes for common illnesses, and transitions of care. We also need better measures for important pediatric populations, such as children with complex medical conditions.25

Third, our understanding of the mechanisms leading to improvement in hospital medicine systems needs to be developed. Studies of hospital medicine systems should move past simple binary comparisons of hospitalist systems versus traditional systems to understand the effect on patient outcomes and cost of factors such as years of experience, volume of patients seen overall and with a specific condition, staffing model, training, quality improvement knowledge and application, and health information systems. These factors may be additive or multiplicative to the performance of inpatient systems once put into place, but these hypotheses need to be tested.

Fourth, individual hospitalists and their groups must focus on quality measurement and improvement in quality and value delivered. At Cincinnati, we have a portfolio of quality and value projects derived from our strategic objectives, illustrated in Figure 2. The projects have leaders and teams to drive improvement and measure results. Increasingly, we are able to publish these results in peer‐reviewed journals. On a quarterly basis, we review the portfolio via a dashboard and/or run and control charts. We establish new projects and set new goals on at least an annual basis. It is important to note that at the beginning of the 2010‐2011 fiscal year, almost all initiatives identified as priorities were yellow or red. Our group is now planning new initiatives and goals for next year. This is one method applicable to our setting, but a focus on quality and value and measuring results needs to be part of every hospital medicine program. As payer focus on value increases, this will be essential to demonstrate how a hospitalist group improves outcomes and adds value.

Figure 2

CONCLUSION

This review suggests that the use of hospitalists can improve the quality of inpatient care in the pediatric population, but this is not a universal finding and, most importantly, the mechanisms of improvement are poorly understood. We propose 4 components to address these issues so that a systematic review 5 years from now would be much more robust. These are: 1) increased investment in research comparing system‐level interventions and reliable implementation; 2) further development and implementation of pediatric quality of care measures in the inpatient setting; 3) understanding the mechanisms and factors leading to improvement in hospital medicine systems; and 4) an increased focus on quality measurement, and improvement in quality and value delivered by all individual hospitalists and their groups.

In the United States, general medical inpatient care is increasingly provided by hospital‐based physicians, also called hospitalists.1 The field of pediatrics is no exception, and by 2005 there were an estimated 1000 pediatric hospitalists in the workforce.2 Current numbers are likely to be greater than 2500, as the need for pediatric hospitalists has grown considerably.

At the same time, the quality of care delivered by the United States health system has come under increased scrutiny. In 2001, the Institute of Medicine, in its report on the quality of healthcare in America, concluded that between the care we have and what we could have lies not just a gap but a chasm.3 Meanwhile, the cost of healthcare delivery continues to increase. The pressure to deliver cost‐effective, high quality care is among the more important forces driving the proliferation of hospitalists.4

Over the last decade, data supporting the role of hospitalists in improving quality of care for adult patients has continued to accumulate.58 A 2007 retrospective cohort study by Lindenaur et al.7 included nearly 77,000 adult patients and found small reductions in length of stay without adverse effects on mortality or readmission rates, and a 2009 systematic review by Peterson6 included 33 studies and concluded that in general inpatient care of general medical patients by hospitalist physicians leads to decreased hospital cost and length of stay. A 2002 study by Meltzer et al.8 is also interesting, suggesting that improvements in costs and short‐term mortality are related to the disease‐specific experience of hospitalists.

Similar data for pediatric hospitalists has been slower to emerge. A systematic review of the literature by Landrigan et al., which included studies through 2004, concluded that [R]esearch suggests that pediatric hospitalists decrease costs and length of stay . The quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking.9 Since the publication of that review, there have been multiple studies which have sought to evaluate the quality of pediatric hospitalist systems. This review was undertaken to synthesize this new information, and to determine the effect of pediatric hospitalist systems on quality of care.

METHODS

A review of the available English language literature on the Medline database was undertaken in November of 2010 to answer the question, What are the differences in quality of care and outcomes of inpatient medical care provided by hospitalists versus non‐hospitalists in the pediatric population? Care metrics of interest were categorized according to the Society of Hospital Medicine's recommendations for measuring hospital performance.10

Search terms used (with additional medical subject headings [MeSH] terms in parenthesis) were hospital medicine (hospitalist), pediatrics (child health, child welfare), cost (cost and cost analysis), quality (quality indicators, healthcare), outcomes (outcome assessment, healthcare; outcomes and process assessment, healthcare); volume, patient satisfaction, length of stay, productivity (efficiency), provider satisfaction (attitude of health personnel, job satisfaction), mortality, and readmission rate (patient readmission). The citing articles search tool was used to identify other articles that potentially could meet criteria. Finally, references cited in the selected articles, as well as in excluded literature reviews, were searched for additional articles.

Articles were deemed eligible if they were published in a peer‐reviewed journal, if they had a comparative experimental design for hospitalists versus non‐hospitalists, and if they dealt exclusively with pediatric hospitalists. Noncomparative studies were excluded, as were studies that pertained to settings besides that of an inpatient pediatrics ward, such as pediatric intensive care units or emergency rooms. The search algorithm is diagrammed in Figure 1.

Figure 1

The selected articles were reviewed for the relevant outcome measures. The quality of each article was assessed using the Oxford Centre for Evidence‐Based Medicine levels of evidence,11 a widely accepted standard for critical analysis of studies. Levels of evidence are assigned to studies, from 1a (systematic reviews of randomized controlled trials) to 5 (expert opinion only). Well‐conducted prospective cohort studies receive a rating of 2c; those with wide confidence intervals due to small sample size receive a minus () modifier. This system does not specifically address survey studies, which were therefore not assigned a level of evidence.

RESULTS

The screening process yielded 92 possible relevant articles, which were then reviewed individually (by G.M.M.) by title and abstract. A total of 81 articles were excluded, including 48 studies that were either noncomparative or descriptive in nature. Ten of the identified articles were reviews and did not contain primary data. Nine studies were not restricted to the pediatric population. Also excluded were 7 studies that did not have outcomes related to quality (eg, billing performance), and 7 studies of hospitalists in settings besides general pediatric wards (eg, pediatric intensive care units). Ten studies were thus identified. The cited reference tool was used to identify an additional article which met criteria, yielding 11 total articles that were included in the review.

Five of the identified studies published prior to 2005 were previously reviewed by Landrigan et al.9 Since then, 6 additional studies of similar nature have been published and were included here. Articles that met criteria but appeared in an earlier review are included in Table 1; new articles appear in Table 2. The results of all 11 articles were included for this discussion.

DISCUSSION

Of the 11 studies selected for this review, 10 measured length of stay as an outcome, with the majority favoring hospitalists but with mixed results. Three of these studies, those by Dwight et al.,12 Bellet and Whitaker,13 and Landrigan et al.,14 demonstrated 11% to 14% improvement for hospitalist services compared to community pediatricians. Boyd et al.,21 however, found exactly the opposite result, and 2 studies by Conway and Keren20 and Ogershok et al.16 found no difference in length of stay. Two more studies found benefits restricted to certain conditions: Wells et al.15 found 32% shorter lengths of stay for asthma, but not for other conditions; Srivastava et al.17 found a 13% reduction in length of stay for asthma and 11% for dehydration, but none for viral illnesses or when all conditions were combined. Bekmezian et al.18 found shorter lengths of stay on a hospitalist service for hematology and gastroenterology patients, and Simon et al.19 attribute a general trend of decreasing lengths of stay on a surgical service to the implementation of hospital comanagement for a small percentage of patients.

The most common quality measures studied were patient satisfaction, readmission rates, and mortality. Only 1 study by Ogershok et al.16 reported on patient satisfaction and found few differences between hospitalists and community pediatricians. Readmission rate were reported by 6 studies. Bellet and Whitaker13 found a higher readmission rate for pediatric hospitalists, Bekmezian et al.18 found a lower rate but on a subspecialty service. The study with the greatest power for this analysis, by Landrigan et al.14 with nearly 18,000 patients, found no difference, and neither did another 3 studies. Unsurprisingly, no study detected differences in mortality; it would be extremely difficult to adequately power a study to do so in the general pediatric setting, where mortality is rare.

The effect of relative experience of hospitalist physicians is uncertain. Boyd et al.21 speculated that 1 possible cause for the decreased lengths of stay and costs associated with their faculty group compared to hospitalists may have been due to the increased experience of the faculty group. Unfortunately, they were unable to generate statistical significance due to the small numbers of physicians in the study. In contrast, the hospitalists in the report by Dwight et al.12 had decreased lengths of stay but were less experienced. In the adult literature, the study by Meltzer et al.8 suggests that improved outcomes from hospitalist systems may not become apparent for 1 or more years after implementation, but none of the pediatric studies included in our review specifically address this issue. This leaves the possibility open that the hospitalist systems evaluated in some studies had insufficient time in which to develop increased efficiencies.

There were several limitations to our studies. First, due to the heterogeneity and methodological variations among the included studies, we were unable to perform a meta‐analysis. Second, the overall quality of evidence is limited due to the lack of randomized control trials. Third, a lack of agreement on appropriate quality markers has limited the study of quality of care. Published reports continue to focus on financial measures, such as length of stay, despite the recommendation in the previous review by Landrigan et al.9 that such studies would be of limited value. Finally, the current variability of hospitalist models and lack of study of factors that might influence outcomes makes comparisons difficult.

Despite these limitations, several interesting trends emerge from these studies. One such trend is that the more recent studies highlight that simple classification of hospitalist system versus traditional system fails to measure the complexity and nuance of care delivery. The 2006 study by Boyd et al.21 is especially notable because it showed the opposite effect of previous studies, namely, an increase in length of stay and costs for hospitalists at St Joseph's Medical Center in Phoenix, Arizona. In this study, the traditional faculty group was employed by the hospital, and the hospitalist group was a private practice model. The authors suggest that their faculty physicians were therefore operating like hospitalists in that almost all of their time was focused on inpatient care while they were on service. They also had a limited number of general pediatricians, who attended in the inpatient setting, who were more experienced than the private practice groups. Also, the authors theorize that their faculty may have had a closer working relationship with their residents due to additional service responsibilities and locations of the faculty group onsite. Further study of the care models utilized by faculty and hospitalist practices at St Joseph's and other hospitals may reveal important insights about improving the quality and efficiency of inpatient pediatric care in general.

Though there is a clear trend in the adult literature indicating that the use of hospitalists results in superior quality of care, there is less evidence for pediatric systems. The aforementioned previous review by Landrigan et al.9, in 2006 concluded that emerging research suggests that pediatric hospitalist systems decrease cost and length of stay, but also the quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking. Data from the 6 additional studies presented here lend limited support to the first hypothesis, and the presence of only 1 negative study is not sufficient to undermine it.

While data on quality markers such as readmission rate or mortality remain elusive, the 2 studies by Conway et al.20, 22 attempt to evaluate quality by comparing the use of evidence‐based therapies by hospitalists and community pediatricians. Though the use of objective PHIS data for UTI in 2009 did not confirm the conclusion suggested by the 2006 provider survey study, the attempt to find measurable outcomes such as the use of evidence‐based therapies is a start but we need more metrics, including rigorous patient outcome metrics, to define the quality of our care systems. Before the effect of hospitalist systems on quality is fully understood, more work will need to be done defining metrics for comparison.

Unfortunately, over 5 years since the previous review by Landrigan et al.9 called for increased focus on inpatient quality and understanding how to improve, the sophistication of our measurement of pediatric inpatient quality and understanding of the mechanisms underlying improvement is still in its infancy. We propose a solution at multiple levels.

First, the investment in research comparing system‐level interventions (eg, discharge process A vs discharge process B) must be increased. This investment increased significantly due to the over $1 billion in Recovery Act funding for comparative effectiveness research.23 However, the future investment in comparative effectiveness research, often called patient‐centered outcomes research, and proportion of investment focused on delivery system interventions is unclear. We propose that the investment in comparing delivery system interventions is essential to improving not only hospital medicine systems, but, more importantly, the healthcare system broadly. In addition, research investment needs to focus on reliably implementing proven interventions in systems of care, and evaluating both the effects on patient outcomes and cost, and the contextual factors associated with successful implementation.24 A hospital medicine example would be the comparison of the implementation of a guideline for a common disease across a set of hospitals. One could perform a prospective observational design, in which one compares high intensity versus low intensity intervention and assesses the baseline characteristics of the hospital systems, to understand their association with successful implementation and, ultimately, patient outcomes. One could also perform a clustered randomized design.

Second, the development and implementation of pediatric quality of care measures, including in the inpatient setting, needs to increase rapidly. The Children's Health Insurance Program (CHIP) and its focus on an initial core set of quality measures that expands over time, through an investment in measure development and validation, is an opportunity for pediatric hospital medicine. Inpatient measures should be a focus of measure development and implementation. We must move beyond a limited set of inpatient measures to a broader set focused on issues such as patient safety, hospital‐acquired infections, outcomes for common illnesses, and transitions of care. We also need better measures for important pediatric populations, such as children with complex medical conditions.25

Third, our understanding of the mechanisms leading to improvement in hospital medicine systems needs to be developed. Studies of hospital medicine systems should move past simple binary comparisons of hospitalist systems versus traditional systems to understand the effect on patient outcomes and cost of factors such as years of experience, volume of patients seen overall and with a specific condition, staffing model, training, quality improvement knowledge and application, and health information systems. These factors may be additive or multiplicative to the performance of inpatient systems once put into place, but these hypotheses need to be tested.

Fourth, individual hospitalists and their groups must focus on quality measurement and improvement in quality and value delivered. At Cincinnati, we have a portfolio of quality and value projects derived from our strategic objectives, illustrated in Figure 2. The projects have leaders and teams to drive improvement and measure results. Increasingly, we are able to publish these results in peer‐reviewed journals. On a quarterly basis, we review the portfolio via a dashboard and/or run and control charts. We establish new projects and set new goals on at least an annual basis. It is important to note that at the beginning of the 2010‐2011 fiscal year, almost all initiatives identified as priorities were yellow or red. Our group is now planning new initiatives and goals for next year. This is one method applicable to our setting, but a focus on quality and value and measuring results needs to be part of every hospital medicine program. As payer focus on value increases, this will be essential to demonstrate how a hospitalist group improves outcomes and adds value.

Figure 2

CONCLUSION

This review suggests that the use of hospitalists can improve the quality of inpatient care in the pediatric population, but this is not a universal finding and, most importantly, the mechanisms of improvement are poorly understood. We propose 4 components to address these issues so that a systematic review 5 years from now would be much more robust. These are: 1) increased investment in research comparing system‐level interventions and reliable implementation; 2) further development and implementation of pediatric quality of care measures in the inpatient setting; 3) understanding the mechanisms and factors leading to improvement in hospital medicine systems; and 4) an increased focus on quality measurement, and improvement in quality and value delivered by all individual hospitalists and their groups.

In the United States, general medical inpatient care is increasingly provided by hospital‐based physicians, also called hospitalists.1 The field of pediatrics is no exception, and by 2005 there were an estimated 1000 pediatric hospitalists in the workforce.2 Current numbers are likely to be greater than 2500, as the need for pediatric hospitalists has grown considerably.

At the same time, the quality of care delivered by the United States health system has come under increased scrutiny. In 2001, the Institute of Medicine, in its report on the quality of healthcare in America, concluded that between the care we have and what we could have lies not just a gap but a chasm.3 Meanwhile, the cost of healthcare delivery continues to increase. The pressure to deliver cost‐effective, high quality care is among the more important forces driving the proliferation of hospitalists.4

Over the last decade, data supporting the role of hospitalists in improving quality of care for adult patients has continued to accumulate.58 A 2007 retrospective cohort study by Lindenaur et al.7 included nearly 77,000 adult patients and found small reductions in length of stay without adverse effects on mortality or readmission rates, and a 2009 systematic review by Peterson6 included 33 studies and concluded that in general inpatient care of general medical patients by hospitalist physicians leads to decreased hospital cost and length of stay. A 2002 study by Meltzer et al.8 is also interesting, suggesting that improvements in costs and short‐term mortality are related to the disease‐specific experience of hospitalists.

Similar data for pediatric hospitalists has been slower to emerge. A systematic review of the literature by Landrigan et al., which included studies through 2004, concluded that [R]esearch suggests that pediatric hospitalists decrease costs and length of stay . The quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking.9 Since the publication of that review, there have been multiple studies which have sought to evaluate the quality of pediatric hospitalist systems. This review was undertaken to synthesize this new information, and to determine the effect of pediatric hospitalist systems on quality of care.

METHODS

A review of the available English language literature on the Medline database was undertaken in November of 2010 to answer the question, What are the differences in quality of care and outcomes of inpatient medical care provided by hospitalists versus non‐hospitalists in the pediatric population? Care metrics of interest were categorized according to the Society of Hospital Medicine's recommendations for measuring hospital performance.10

Search terms used (with additional medical subject headings [MeSH] terms in parenthesis) were hospital medicine (hospitalist), pediatrics (child health, child welfare), cost (cost and cost analysis), quality (quality indicators, healthcare), outcomes (outcome assessment, healthcare; outcomes and process assessment, healthcare); volume, patient satisfaction, length of stay, productivity (efficiency), provider satisfaction (attitude of health personnel, job satisfaction), mortality, and readmission rate (patient readmission). The citing articles search tool was used to identify other articles that potentially could meet criteria. Finally, references cited in the selected articles, as well as in excluded literature reviews, were searched for additional articles.

Articles were deemed eligible if they were published in a peer‐reviewed journal, if they had a comparative experimental design for hospitalists versus non‐hospitalists, and if they dealt exclusively with pediatric hospitalists. Noncomparative studies were excluded, as were studies that pertained to settings besides that of an inpatient pediatrics ward, such as pediatric intensive care units or emergency rooms. The search algorithm is diagrammed in Figure 1.

Figure 1

The selected articles were reviewed for the relevant outcome measures. The quality of each article was assessed using the Oxford Centre for Evidence‐Based Medicine levels of evidence,11 a widely accepted standard for critical analysis of studies. Levels of evidence are assigned to studies, from 1a (systematic reviews of randomized controlled trials) to 5 (expert opinion only). Well‐conducted prospective cohort studies receive a rating of 2c; those with wide confidence intervals due to small sample size receive a minus () modifier. This system does not specifically address survey studies, which were therefore not assigned a level of evidence.

RESULTS

The screening process yielded 92 possible relevant articles, which were then reviewed individually (by G.M.M.) by title and abstract. A total of 81 articles were excluded, including 48 studies that were either noncomparative or descriptive in nature. Ten of the identified articles were reviews and did not contain primary data. Nine studies were not restricted to the pediatric population. Also excluded were 7 studies that did not have outcomes related to quality (eg, billing performance), and 7 studies of hospitalists in settings besides general pediatric wards (eg, pediatric intensive care units). Ten studies were thus identified. The cited reference tool was used to identify an additional article which met criteria, yielding 11 total articles that were included in the review.

Five of the identified studies published prior to 2005 were previously reviewed by Landrigan et al.9 Since then, 6 additional studies of similar nature have been published and were included here. Articles that met criteria but appeared in an earlier review are included in Table 1; new articles appear in Table 2. The results of all 11 articles were included for this discussion.

DISCUSSION

Of the 11 studies selected for this review, 10 measured length of stay as an outcome, with the majority favoring hospitalists but with mixed results. Three of these studies, those by Dwight et al.,12 Bellet and Whitaker,13 and Landrigan et al.,14 demonstrated 11% to 14% improvement for hospitalist services compared to community pediatricians. Boyd et al.,21 however, found exactly the opposite result, and 2 studies by Conway and Keren20 and Ogershok et al.16 found no difference in length of stay. Two more studies found benefits restricted to certain conditions: Wells et al.15 found 32% shorter lengths of stay for asthma, but not for other conditions; Srivastava et al.17 found a 13% reduction in length of stay for asthma and 11% for dehydration, but none for viral illnesses or when all conditions were combined. Bekmezian et al.18 found shorter lengths of stay on a hospitalist service for hematology and gastroenterology patients, and Simon et al.19 attribute a general trend of decreasing lengths of stay on a surgical service to the implementation of hospital comanagement for a small percentage of patients.

The most common quality measures studied were patient satisfaction, readmission rates, and mortality. Only 1 study by Ogershok et al.16 reported on patient satisfaction and found few differences between hospitalists and community pediatricians. Readmission rate were reported by 6 studies. Bellet and Whitaker13 found a higher readmission rate for pediatric hospitalists, Bekmezian et al.18 found a lower rate but on a subspecialty service. The study with the greatest power for this analysis, by Landrigan et al.14 with nearly 18,000 patients, found no difference, and neither did another 3 studies. Unsurprisingly, no study detected differences in mortality; it would be extremely difficult to adequately power a study to do so in the general pediatric setting, where mortality is rare.

The effect of relative experience of hospitalist physicians is uncertain. Boyd et al.21 speculated that 1 possible cause for the decreased lengths of stay and costs associated with their faculty group compared to hospitalists may have been due to the increased experience of the faculty group. Unfortunately, they were unable to generate statistical significance due to the small numbers of physicians in the study. In contrast, the hospitalists in the report by Dwight et al.12 had decreased lengths of stay but were less experienced. In the adult literature, the study by Meltzer et al.8 suggests that improved outcomes from hospitalist systems may not become apparent for 1 or more years after implementation, but none of the pediatric studies included in our review specifically address this issue. This leaves the possibility open that the hospitalist systems evaluated in some studies had insufficient time in which to develop increased efficiencies.

There were several limitations to our studies. First, due to the heterogeneity and methodological variations among the included studies, we were unable to perform a meta‐analysis. Second, the overall quality of evidence is limited due to the lack of randomized control trials. Third, a lack of agreement on appropriate quality markers has limited the study of quality of care. Published reports continue to focus on financial measures, such as length of stay, despite the recommendation in the previous review by Landrigan et al.9 that such studies would be of limited value. Finally, the current variability of hospitalist models and lack of study of factors that might influence outcomes makes comparisons difficult.

Despite these limitations, several interesting trends emerge from these studies. One such trend is that the more recent studies highlight that simple classification of hospitalist system versus traditional system fails to measure the complexity and nuance of care delivery. The 2006 study by Boyd et al.21 is especially notable because it showed the opposite effect of previous studies, namely, an increase in length of stay and costs for hospitalists at St Joseph's Medical Center in Phoenix, Arizona. In this study, the traditional faculty group was employed by the hospital, and the hospitalist group was a private practice model. The authors suggest that their faculty physicians were therefore operating like hospitalists in that almost all of their time was focused on inpatient care while they were on service. They also had a limited number of general pediatricians, who attended in the inpatient setting, who were more experienced than the private practice groups. Also, the authors theorize that their faculty may have had a closer working relationship with their residents due to additional service responsibilities and locations of the faculty group onsite. Further study of the care models utilized by faculty and hospitalist practices at St Joseph's and other hospitals may reveal important insights about improving the quality and efficiency of inpatient pediatric care in general.

Though there is a clear trend in the adult literature indicating that the use of hospitalists results in superior quality of care, there is less evidence for pediatric systems. The aforementioned previous review by Landrigan et al.9, in 2006 concluded that emerging research suggests that pediatric hospitalist systems decrease cost and length of stay, but also the quality of care in pediatric hospitalist systems is unclear, because rigorous metrics to evaluate quality are lacking. Data from the 6 additional studies presented here lend limited support to the first hypothesis, and the presence of only 1 negative study is not sufficient to undermine it.

While data on quality markers such as readmission rate or mortality remain elusive, the 2 studies by Conway et al.20, 22 attempt to evaluate quality by comparing the use of evidence‐based therapies by hospitalists and community pediatricians. Though the use of objective PHIS data for UTI in 2009 did not confirm the conclusion suggested by the 2006 provider survey study, the attempt to find measurable outcomes such as the use of evidence‐based therapies is a start but we need more metrics, including rigorous patient outcome metrics, to define the quality of our care systems. Before the effect of hospitalist systems on quality is fully understood, more work will need to be done defining metrics for comparison.

Unfortunately, over 5 years since the previous review by Landrigan et al.9 called for increased focus on inpatient quality and understanding how to improve, the sophistication of our measurement of pediatric inpatient quality and understanding of the mechanisms underlying improvement is still in its infancy. We propose a solution at multiple levels.

First, the investment in research comparing system‐level interventions (eg, discharge process A vs discharge process B) must be increased. This investment increased significantly due to the over $1 billion in Recovery Act funding for comparative effectiveness research.23 However, the future investment in comparative effectiveness research, often called patient‐centered outcomes research, and proportion of investment focused on delivery system interventions is unclear. We propose that the investment in comparing delivery system interventions is essential to improving not only hospital medicine systems, but, more importantly, the healthcare system broadly. In addition, research investment needs to focus on reliably implementing proven interventions in systems of care, and evaluating both the effects on patient outcomes and cost, and the contextual factors associated with successful implementation.24 A hospital medicine example would be the comparison of the implementation of a guideline for a common disease across a set of hospitals. One could perform a prospective observational design, in which one compares high intensity versus low intensity intervention and assesses the baseline characteristics of the hospital systems, to understand their association with successful implementation and, ultimately, patient outcomes. One could also perform a clustered randomized design.

Second, the development and implementation of pediatric quality of care measures, including in the inpatient setting, needs to increase rapidly. The Children's Health Insurance Program (CHIP) and its focus on an initial core set of quality measures that expands over time, through an investment in measure development and validation, is an opportunity for pediatric hospital medicine. Inpatient measures should be a focus of measure development and implementation. We must move beyond a limited set of inpatient measures to a broader set focused on issues such as patient safety, hospital‐acquired infections, outcomes for common illnesses, and transitions of care. We also need better measures for important pediatric populations, such as children with complex medical conditions.25

Third, our understanding of the mechanisms leading to improvement in hospital medicine systems needs to be developed. Studies of hospital medicine systems should move past simple binary comparisons of hospitalist systems versus traditional systems to understand the effect on patient outcomes and cost of factors such as years of experience, volume of patients seen overall and with a specific condition, staffing model, training, quality improvement knowledge and application, and health information systems. These factors may be additive or multiplicative to the performance of inpatient systems once put into place, but these hypotheses need to be tested.

Fourth, individual hospitalists and their groups must focus on quality measurement and improvement in quality and value delivered. At Cincinnati, we have a portfolio of quality and value projects derived from our strategic objectives, illustrated in Figure 2. The projects have leaders and teams to drive improvement and measure results. Increasingly, we are able to publish these results in peer‐reviewed journals. On a quarterly basis, we review the portfolio via a dashboard and/or run and control charts. We establish new projects and set new goals on at least an annual basis. It is important to note that at the beginning of the 2010‐2011 fiscal year, almost all initiatives identified as priorities were yellow or red. Our group is now planning new initiatives and goals for next year. This is one method applicable to our setting, but a focus on quality and value and measuring results needs to be part of every hospital medicine program. As payer focus on value increases, this will be essential to demonstrate how a hospitalist group improves outcomes and adds value.

Figure 2

CONCLUSION

This review suggests that the use of hospitalists can improve the quality of inpatient care in the pediatric population, but this is not a universal finding and, most importantly, the mechanisms of improvement are poorly understood. We propose 4 components to address these issues so that a systematic review 5 years from now would be much more robust. These are: 1) increased investment in research comparing system‐level interventions and reliable implementation; 2) further development and implementation of pediatric quality of care measures in the inpatient setting; 3) understanding the mechanisms and factors leading to improvement in hospital medicine systems; and 4) an increased focus on quality measurement, and improvement in quality and value delivered by all individual hospitalists and their groups.