The Learning Curve for AIDS-Related Pneumocystis carinii Pneumonia

Abstract

Previous studies have found lower mortality rates for AIDS-related Pneumocystis carinii pneumonia (PCP) in hospitals with higher levels of experience with PCP. It is not known if patients are selectively referred to better hospitals or if there is a learning curve whereby outcomes improve as physicians gain experience in treating PCP. We assessed cases of PCP at 140 Veterans Administration (VA) Medical Centers in the United States. During 1987-1991, 3,981 patients were hospitalized with first-episode AIDS-related PCP. Mortality at 30 days after admission. For these 3,981 hospitalizations at the 140 study hospitals, the 30-day mortality was 19%. Logistic regression models indicate that older age, race, geographic area, earlier year of treatment, hospitalization in the previous 12 months, and lower levels of hospital experience with AIDS were significant predictors of mortality at 30 days after admission. Compared with hospitals that had treated three cases or fewer of first-episode PCP, the odds of mortality at 30 days at hospitals that treated > 50 cases of first-episode PCP were 0.73 (95% confidence interval 0.58-0.91), after controlling for differences in characteristics of the patients, year, and region. Mortality of patients with AIDS-related PCP decreases as VA hospitals gain experience. Longitudinal analyses over a 5-year period suggest that a learning curve best explains this finding.