Aims: Since the start of the Cleveland Health Quality Choice (CHQC) Program in 1991, in-hospital mortality for 6 key medical conditions declined 16-45 percent, and length of stay declined 24-37 percent. It is not known whether there have been declines in longer-term mortality or whether the decrease results from declining admission severity of illness. To examine this, our 4 aims are to: 1) examine trends between 1991-96 in condition-specific observed in-hospital, early post-discharge, 30-day, and 180-day mortality for all Medicare beneficiaries admitted to hospitals in Greater Cleveland for the 6 CHQC medical conditions, with further examination of these trends according to patients' DNR status and discharge disposition (e.g.; home without assistance vs. extended post-discharge care); 2) develop condition-specific models of admission severity of illness that are valid for the entire study period; 3) examine trends in admission severity of illness between 1991-96; and 4) analyze trends in hospital performance from 1991-96 based upon risk-adjusted mortality rates. Methods: This 24-month investigation will merge hospital data from the rich CHQC dataset with long-term administrative data (MEDPAR) for all admissions between 7/1/91-12/31/96 of Medicare beneficiaries with one of 6 medical conditions: acute myocardial infarction, congestive heart failure, GI hemorrhage, stroke, pneumonia, and obstructive airway disease (N of CHQC admissions=138,659). Databases will be merged using 6 common elements, and condition-specific databases will be created. Temporal trends in observed mortality for each condition and each time interval (e.g. 30-days post admission) will be analyzed using logistic regression (Aim 1). Condition-specific models of admission severity of illness (i.e. predicted probability of death) will be developed based on 1991-92 data using logistic regression (Aim 2). The models will then be used to calculate the admission severity of illness for all hospitalizations. Time trends in admission severity of illness will be analyzed using linear regression with the admission severity of illness as the dependent variable (Aim 3). Finally, the admission severity of illness from Aim 2 will be used as a covariate to analyze trends in risk-adjusted mortality rates (Aim 4), with further analysis by DNR status, discharge disposition, and individual participating hospitals (N=27). Significance: The study results will be important to policymakers, payors, hospitals, physicians, and patients. If trends for 30-day risk-adjusted mortality show similar declines as for observed in-hospital mortality (during a period when length of stay has also declined dramatically), this would suggest that hospital efficiency has improved. Conversely, if 30-day risk-adjusted mortality rates are not improving, this would suggest that the apparent improvement in hospital mortality is merely a shift of dying and vulnerable patients from the hospital to other sites. This study will also help us understand whether all hospitals are improving or whether hospitals with the worst performance in 1991 have improved more than others.
| Katzan, I L; Dawson, N V; Thomas, C L et al. (2007) The cost of pneumonia after acute stroke. Neurology 68:1938-43 |
