Effects of Hospitalists on Mortality in a Multicenter Study of Academic Hospitalists

Background:

A few studies have found lower mortality for patients treated by hospitalists compared to patients treated by non‐hospitalists, and one study found mortality differences after discharge that increased over time. Other studies have found no differences in mortality between hospitalist and non‐hospitalist patients. However, almost all these studies have had inadequate power to exclude clininically meaningful differences in mortality and/or been limited by non‐random assignment of patients to hospitalists versus non‐hospitalists. We studied mortality of patients admitted to hospitalists and nonhospitalists based on a predetermined call schedule in six academic medical centers.

Method:

From July 1, 2001 to June 30, 2003 we collected administrative data on all general medicine patients assigned to hospitalists or non‐hospitalists based on a predtermined call schedule at six academic medical medical centers. Data collected included age, gender, race, primary and secondary diagnoses using ICD‐9 codes, and social security number. We then linked this data to the National Death Index using social security number and performed multivariate logistic regression analyses and proportional hazard modelling to examine mortality for hospitalist and non‐hospitalist patients after discharge.

Summary of Results:

31,013 patients were admitted to the general medicine services. 26,282 had complete data to allow potential matching to the national death index by social security number with follow‐up for at least six months. Patients cared for by hospitalists and non‐hospitalists were not‐statistically different in age, gender, race, or primary and secondary diagnoses using ICD‐9 codes.There were no statistically significant differences in mortality between patients cared for by hospitalists and non‐hospitalists over any time interval. The odds ratio for the effect of hospitalists relative to non‐hospitalists at 7 days was 0.96 with a 95% CI (0.86 −1.08); at 30 days 1.00 (0.90‐1.11); at 180 days 1.03 (0.95‐1.11).

Statement of Conclusions:

Patients cared for by hospitalist and non‐hospitalist attending physicians did not have statistically different patterns of mortality. Confidence intervals for estimates of the relative risk of mortality were fairly narrow, ruling out large differenes in mortality in this study. Nevertheless, differences in mortality on the order of a 5‐10% difference in relative risk that might be considered clinically significant cannot be excluded. Future studies that aim to rule out smaller effects of hospitalists on mortality will require very large sample sizes. Alternative outcome measures, including process of care, are essential in evaluating the effects of hospitalist programs.

Author Disclosure Block:

D.O. Meltzer, None; J. Zhang, None; J. Siddique, None; A. Auerbach, None; D. Gonzales, None; P. Kaboli, None; J. Schnipper, None; T. Wettterneck, None; V. Arora, None; R. Wachter, None.