The relation between hospital experience and mortality for patients with AIDS.

OBJECTIVE

To determine whether there is an association between mortality and hospital acquired immune deficiency syndrome (AIDS) experience for all AIDS-related diagnoses, and to determine whether the association is explained by differences in severity of illness, earlier discharge of terminally ill patients, or differences in resource use.

DESIGN AND SETTING

Population-based statewide cohort study. All hospitalizations identified for a cohort of AIDS patients diagnosed during 1987 in 40 Massachusetts hospitals were included.

PATIENTS

All women and all male intravenous drug users (n = 151), and a random sample of all male non-intravenous drug users diagnosed with AIDS during 1987 in Massachusetts (n = 149).

MAIN OUTCOME MEASURES

Primary end points studied were (1) inpatient mortality and (2) 30-day mortality. Length of stay, cost, and intensive care unit use were also examined.

RESULTS

In 806 hospitalizations at 40 hospitals inpatient mortality was 13.2%. Relative risk of mortality at low AIDS experience hospitals was 2.16 (95% confidence interval, 1.43 to 3.26) compared with high AIDS experience hospitals. When logistic regression was used to control for age, sex, race, human immunodeficiency virus transmission mode, severity, payer, admission type, hospital ownership, and teaching status, low hospital experience with AIDS remained a significant predictor of inpatient mortality (relative risk, 2.92; 95% confidence interval, 1.37 to 6.22). Comparisons of 30-day mortality by hospital AIDS experience yielded similar results. Length of stay and intensive care unit use were also significantly higher at low-experience hospitals after controlling for severity of illness (P < .05).

CONCLUSIONS

We conclude that mortality of AIDS patients is higher at hospitals with less AIDS experience. This finding is not because of greater severity of illness, differences in discharge patterns of the terminally ill, or less intensive use of resources.