Hospitalists and the Quality of Care in Hospitals

Background  Little is known about the link between hospitalists and performance on hospital-level quality indicators.

Methods  From October 1, 2005, through September 31, 2006, we linked the Hospital Quality Alliance (HQA) data to the American Hospital Association data on the presence of hospitalists. Main outcome measures included composite measurements of hospital-level quality of care for 3 conditions (acute myocardial infarction [AMI], congestive heart failure [CHF], and pneumonia) and 2 dimensions of care (treatment and diagnosis, as well as counseling and prevention). We fitted a series of logistic regression models to examine the relationship between hospitalists and overall quality of care for each condition, controlling for all other hospital characteristics.

Results  Of 3619 hospitals reporting HQA data, 1461 (40.4%) had hospitalists. Hospitals with hospitalists tended to be large, private, not-for-profit, teaching institutions located in the southern United States. The mean unadjusted composite scores were higher for hospitals with hospitalists vs those with no hospitalists for all 3 conditions (93% vs 86% for AMI, 82% vs 72% for CHF, and 75% vs 71% for pneumonia) and both dimensions of care (87% vs 77% for treatment and diagnosis and 75% vs 66% for counseling and prevention) (P < .001 for all comparisons). After multivariable adjustment, hospitals with hospitalists continued to perform significantly better than those without hospitalists across all composite scores except for CHF.

Conclusion  Hospitals with hospitalists were associated with better performance on HQA indicators for AMI, pneumonia, and the domains of overall disease treatment and diagnosis, as well as counseling and prevention.

Hospitalists are physicians whose primary professional focus is the general medical care of hospitalized patients and whose activities include patient care, teaching, research, and leadership-related inpatient care. As of 2007, there were an estimated 20 000 hospitalists in the United States.¹ Recent survey results have shown that 29% of hospitals overall and 55% of hospitals with more than 200 beds have hospitalists.² Studies presented in the literature have demonstrated that the use of hospitalists is associated with lower inpatient costs and shorter lengths of stay compared with the use of general internists and family physicians,^3-5 and such savings did not have a detrimental effect on the rates of death or readmission.⁶

Little is known about the association between the use of hospitalists and quality of care as measured by existing national standards. The Hospital Quality Alliance (HQA) is a collaboration between the Centers for Medicaid and Medicare Services and provider organizations that publicly report benchmark quality data through the Hospital Compare Web site.⁷ Previous analyses have shown significant variability across indicators in the quality of hospital care provided based on hospital characteristics. Jha et al⁸ found slightly but significantly better performance at not-for-profit hospitals, academic hospitals, and those located in the Northeast or Midwest, whereas Landon et al⁹ found that hospitals with high registered nurse to patient staffing ratios and those with more technology availability were associated with higher-quality care. However, these previous analyses lacked information about the presence of hospitalists.

Hospitalists solely care for inpatients, and this specialization may increase their expertise in managing disease conditions, which in turn may favorably affect the quality of care provided.^3,10,11 Prior research has shown that the use of hospitalists leads to reduced length of stay,⁶ closer adherence to treatment guidelines,^12-15 and better postdischarge follow-up.¹⁶ Given the large number of hospitalists and their integral role in the inpatient clinical delivery system, it is important to understand how this specialty affects the quality of care provided.^17,18 The aim of this study was to examine the link between hospitalists and performance as measured by the HQA benchmark quality measures for acute myocardial infarction (AMI), congestive heart failure (CHF), and pneumonia.

The HQA provides performance information that is updated quarterly. We used these data to measure hospital-level process indicators of care for AMI, CHF, and pneumonia from October 1, 2005, through September 31, 2006. For each of the 3 conditions, we examined the quality of care at 3619 hospitals reporting data to the HQA. Two of us (L.L. and L.S.H.), who are hospitalists, examined the list of indicators and chose only those related to hospitalist care (Table 1). We excluded 8 measures that are typically provided in the emergency department, such as the administration of aspirin on arrival for the treatment of AMI.

To these data, we linked data on hospital characteristics, including whether a hospitalist program was present, from the 2005 American Hospital Association National Survey of Hospitals.¹⁹ We restricted our analyses to general medical and surgical hospitals. Data obtained from the National Survey of Hospitals also included geographic region, type of setting (urban vs rural), presence or absence of an intensive care unit, number of beds, ratio of registered full-time nursing to patient staffing per 1000 patient-days, teaching status (defined as membership in the Council of Teaching Hospitals), ownership (for-profit, not-for-profit, government, or military), and the number of Medicare and Medicaid discharges. Registered nurse staffing levels and the proportions of admissions covered by Medicare and Medicaid were divided into quartiles.

Composite scores, representing the overall quality of care delivered for each disease, were created by dividing the number of times a hospital performed the appropriate care across all indicators for that condition by the number of opportunities the hospital had to provide appropriate care for that condition. This method is described by the Joint Commission on the Accreditation of Healthcare Organizations.²⁰ To create stable and reliable composite measures, only hospitals that reported data from at least 25 patients were used. Based on the previously published results of factor analysis, we also created 2 additional functional composite measures representing the underlying domains of (1) treatment and diagnosis and (2) counseling and prevention across each disease category.⁹ Quality indicators selected for each condition, stratified by domain, are presented in Table 1.

We used χ² tests to compare characteristics of hospitals with and without hospitalists. For each process indicator and composite score, we calculated the mean performance and the standard deviation and compared the mean performances between hospitals with and without hospitalists using 2-tailed t tests. Next, we examined the relationship between the presence of hospitalists and overall quality of care for each condition, controlling for all other hospital characteristics.

We fitted a series of binomial logistic regression models for the grouped hospital data that included all hospital characteristics listed in the “Data Sources” subsection as covariates. These variables were included because they are essential characteristics of hospitals and they have been included in most studies using the American Hospital Association and HQA data sets.^8,9,13,21 Therefore, all covariates were included in the model regardless of bivariate association with the outcome. The models estimated the number of times an indicator was successfully met at each hospital per the total number of opportunities the hospital had to provide care for the indicator. Logistic regression models were estimated separately for each disease and functional composite score. These models are marginal models in which the main purpose was to estimate covariate effects while accounting for the clustering of scores at the hospital level. Hence, generalized estimating equations (the GENMOD procedure in SAS statistical software, version 9.1; SAS Institute Inc, Cary, North Carolina) were used because our focus is on estimating the average response for the entire group of analyzed hospitals (“population-averaged” effects) rather than the specific regression parameters that would enable prediction of the effect for a given hospital.^22,23 To address concerns of potential omitted variable bias, we conducted secondary analyses assessing the association between the presence of hospitalists and the 8 excluded nonhospitalist clinical care indicators to confirm our a priori determination that the presence of hospitalists would not be associated with differences in quality for those indicators.

Of 3619 hospitals, 1461 (40.4%) had hospitalists. Compared with hospitals without hospitalists, those with hospitalists were significantly more likely to be located in the southern United States and in large metropolitan statistical areas (Table 2). These hospitals also were more likely to be large, private, not-for-profit, teaching centers with more than 200 beds and an intensive care unit. Compared with hospitals without hospitalists, those with hospitalists cared for a lower proportion of Medicare patients, had a higher proportion of Medicaid patients, and had a significantly higher proportion of nurses to patients (P < .001 for all comparisons).

In our bivariable analyses, the mean scores for each disease indicator were significantly higher for hospitals with hospitalists compared with those without hospitalists (P < .001 for all) (Table 3). In addition, hospitals with hospitalists had higher composite scores for each disease and for the dimensions of overall treatment and diagnosis and for counseling and prevention (P < .001 for all).

We found no significant association between the excluded nonhospitalist care indicators and the presence of hospitalists. Compared with hospitals without hospitalists, those with hospitalists continued to have significantly higher odds of meeting applicable indicators for AMI, pneumonia, treatment and diagnosis, and counseling and prevention (Table 4). In addition, hospitals with a high ratio of nursing staff per 1000 patient days were associated with higher scores across all quality indicators. Lower scores across all measures were consistently found for smaller and public and/or municipal hospitals. Hospitals with high proportions of Medicaid patients had a nonsignificant trend toward providing lower quality care.

In this study of 3619 hospitals, we found that hospitals with hospitalists were associated with better performance on quality indicators for AMI and pneumonia and in the composite domains of disease treatment and diagnosis and counseling and prevention, controlling for hospital characteristics such as size, location, ownership type, and staffing availability. Our findings expand upon prior research on the hospitalist model that has focused on measuring quality of care by cost reduction and length of stay.⁵ The results of early single-institution studies showed large reductions in these measures, yet a recent large retrospective analysis found only a moderately sized reduction in length of stay and costs.⁶ There is limited evidence concerning the effect of hospitalists on the quality of care using national standards.^12-15 Our study contributes to the growing hospitalist literature by focusing on the quality of care provided as measured by process indicators in a national sample of hospitals.

Previous research using the HQA measures have found substantial variability and room for improvement in the care of hospitalized patients across medical conditions.^8,9 For example, half the hospitals examined by Jha et al⁸ had mean performance scores that were at least 90% for 5 of 10 quality indicators examined, especially those related to AMI. However, the scores were much lower for measures related to CHF and pneumonia. Hospitalists may provide an opportunity for improving inpatient care for these conditions.

Our data demonstrated that the quality of care for the conditions examined was better in hospitals with hospitalists. Hospitalists solely care for inpatients, and this specialization increases their expertise in managing inpatient disease conditions and their adherence to treatment guidelines.^3,10,11 Findings from previous single-institution studies using disease-specific measures of clinical processes have shown that hospitalists have specific practice patterns that result in improved efficiency and better outcomes. For example, Lindenauer et al¹³ demonstrated that patients with CHF treated by hospitalists were more likely to have had their left ventricular ejection fraction assessed. Roytman et al¹⁴ have shown that hospitalists, compared with nonhospitalists, use angiotensin-converting enzyme inhibitors and angiotensin receptor blockers earlier, have an aggressive approach to diuresis, and have greater use of social work services and decreased use of serial chest radiography, medical consultants, and serial brain natriuretic peptide measurements. In addition, 30-day follow-up was more common for patients with CHF treated by academic hospitalists compared with academic nonhospitalists.¹⁶ In patients with pneumonia, Rifkin and colleagues¹⁵ have demonstrated that hospitalists converted intravenous to oral antibiotics quicker than nonhospitalists. Our study's national scope, adjustment for variability in hospital characteristics, and use of evidence-based HQA quality measures distinguishes it from previous research on hospital care.

In our study, we found that few hospital characteristics consistently predicted score performance across all 3 disease conditions. However, hospitals with higher registered nurse to patient staffing ratios were consistently associated with better care. Previous research has demonstrated that higher nurse to patient ratios are associated with decreased length of stay and lower rates of medical complications and lower mortality.^24-27 For example, after adjusting for patient and institutional characteristics, Needleman and colleagues²⁵ demonstrated that a higher proportion of hours of nursing care provided by registered nurses is associated with fewer medical complications and lower rates of death among patients who develop serious complications. Nursing staff provide significant skilled care and clinical assessments and are able to spend more time with patients than physicians. These aspects of hospital nursing care may explain the association between nurse staffing and quality of care demonstrated in multiple studies.

We also found that public and/or municipal hospitals and smaller hospitals were consistently associated with lower quality of care. These findings highlight the importance of adequate financing for providing higher quality care. For example, Landon et al⁹ demonstrated that hospitals that invest in technology score higher on HQA measures. Our study underscores the need to adequately fund chronically underfunded public and other safety-net hospitals in order to ensure the delivery of high-quality care.²⁸

Our study has several limitations. First, our data measure hospital-level performance. We were unable to match individual patient care to whether only a hospitalist provided medical care. We were also unable to show whether there were important differences in the care individuals receive within hospitals. Second, although our study focused on 3 diseases, hospitalists provide care for a broader range of medical conditions. Thus, it is not known the extent to which these data are indicative of quality of care for other conditions. In addition, although HQA data are widely used measures of quality, performance has been shown to have a modest relationship with clinical outcomes.^21,29-33 However, modest mortality differences have been estimated to translate into a clinically significant number of saved lives.^21,30-33 The modest differences we found in performance may not represent differences in clinical outcomes between hospitals with and without hospitalists. Because a limited number of focused process indicators are measured, they likely omit important contributory clinical care processes. These unmeasured clinical processes, such as staffing patterns, effective use of information technology, and clinical coordination, may explain why the link between process and outcome measures is modest.³⁴ In addition, process measurement does not capture how well an indicator is provided to the patient, as in the case of smoking cessation counseling. Thus, similar process scores could mask important differences in possible outcomes. Third, we examined quality scores only among those hospitals that reported data with numbers that were adequate to create summary scores and those hospitals that reported having or not having a hospitalist program. Hence, our findings should not be generalized to all US hospitals. Fourth, although the HQA scores have specific inclusion and exclusion criteria, they may not fully account for case mix or the severity of health conditions. Next, because our data on hospitals' characteristics do not include potentially important facets of the delivery of care, there is a potential for omitted variable confounding. However, we found no significant association between the nonhospitalist-related indicators that were excluded and the presence of hospitalists, strengthening our findings. Last, our data are cross-sectional and, thus, associations found are not proof of causality. For example, there may be unmeasured variables that represent inclusion of hospitalists at hospitals with better infrastructure and other processes that lead to better quality of care over and above what we were able to control.

In summary, we found that hospitals with hospitalists are associated with better hospital performance. Complex factors affect the quality of hospital care delivered to patients. Hospitalists are likely to be one of several factors that contribute to high-quality care. With the continued growth of the hospitalist inpatient care model, further research is needed to delineate the specific hospitalist model characteristics associated with improved quality and outcomes of care. The involvement of hospitalists in the acute care of hospitalized patients may contribute to improved quality of care for patients with common inpatient diagnoses.

Correspondence: Lenny López, MD, MPH, Institute for Health Policy, Massachusetts General Hospital, 50 Staniford St, Ninth Floor, Boston, MA 02114 (llopez1@partners.org).

Accepted for Publication: May 7, 2009.

Author Contributions: Drs López and Hicks had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: López, McKean, and Weissman. Acquisition of data: López and Weissman. Analysis and interpretation of data: López, Hicks, Cohen, and Weissman. Drafting of the manuscript: López. Critical revision of the manuscript for important intellectual content: López, Hicks, Cohen, McKean, and Weissman. Statistical analysis: López and Cohen. Administrative, technical, and material support: Weissman. Study supervision: Hicks, McKean, and Weissman.

Financial Disclosure: Dr Hicks consults for WellPoint Inc. The other authors have no disclosures or conflicts of interest to report.

Funding/Support: This study was supported by Institutional National Research Service Award 5 T32 HP11001-19 (Dr López).

Previous Presentations: Results of this study were presented as a poster at the 2008 New England regional (March 14, 2008; Worcester, Massachusetts) and national Society of General Internal Medicine (April 9-12, 2008; Pittsburgh, Pennsylvania) meetings and the 2008 Society of Hospital Medicine meeting (April 3-5, 2008; San Diego, California).