Objective: To define our ICU experience with AIDS patients, Pneumocystis carinii pneumonia, and respiratory failure, and to delineate factors predictive of hospital survival.
Design: A retrospective study in which logistic regression analysis was applied to data obtained during the first 144 hrs of ICU admission.
Setting: A university hospital medical ICU associated with a national AIDS treatment center.
Patients: Twenty-seven male patients with AIDS, P. carinii pneumonia, and respiratory failure who desired full supportive and resuscitative care.
Measurements and main results: Of 27 patients who met study criteria, 19 (70%) were nonsurvivors and eight (30%) were survivors. The relative risk of death was 2.2 times greater in patients who exhibited the combination of pH less than 7.35 and a base deficit greater than 4 mEq/L, at any time in their ICU course, than in patients who did not (95% confidence interval = 1.01, 4.81). Furthermore, the relative risk of death was 3.7 times greater in patients who required positive end-expiratory pressure greater than 10 cm H2O after 96 hrs of ICU care than in those patients who did not (95% confidence interval = 1.09, 12.33). Indices of oxygen transfer, severity of chest radiograph abnormalities, concurrent lung infections, and most laboratory studies on hospital admission were not different between the two groups nor predictive of hospital survival.
Conclusions: When dealing with AIDS/P. carinii pneumonia/ICU patients, it is not possible to distinguish who will survive to hospital discharge based on information routinely available before ICU admission. Those patients with the greatest chance of survival demonstrate a significant decrease in the required level of respiratory support within the first 4 days of ICU care. The presence of a metabolic acidemia (pH less than 7.35 and base deficit greater than 4 mEq/L), at any time during the ICU course, is a poor prognostic sign. We suggest that such objective variables should be included in the development of any new outcome predictor model for this group of ICU patients.