Predictors of short- and long-term survival in HIV-infected patients admitted to the ICU

To evaluate the prognosis of HIV-infected patients admitted to ICUs and to identify factors predictive of short- and long-term survival.

A prospective study from January 1, 1990, to December 31, 1992, including all consecutive HIV-infected patients admitted to our ICU for the first time. ICU survivors were followed up until January 1, 1994.

An 18-bed infectious diseases ICU in a 1,300-bed university hospital in Paris.

Four hundred twenty-one HIV-related admissions were recorded during the study period (33.5% of 1,258 admissions to ICU); 354 HIV-infected patients were first ICU admissions and were analyzed.

Predictive factors on univariate and multivariate analyses (logistic regression and Cox model) for short- and long-term mortality were performed. Respiratory failure was the main cause of admission (49.2%), followed by neurologic disorders (26.8%), sepsis (10.2%), heart failure (4.5%), and miscellaneous disorders (9.3%). For these groups, in-ICU and in-hospital mortality rates were as follows: 16.7% and 33.9%; 23.2% and 41.1%; 38.9% and 58.3%; 25% and 68.8%; and 12.1% and 24.2%, respectively. In-ICU and in-hospital mortality rates were significantly different across the groups (p=0.026 and 0.002, respectively). Multivariate analysis showed that the in-hospital outcome was significantly associated with functional status (p=0.05), time since AIDS diagnosis (p=0.04), HIV disease stage (0.016), simplified acute physiology score (SAPS I) (p=0.06), need for mechanical ventilation (p<0.000001), and its duration (p=0.0001). In the 281 patients who were discharged alive from the ICU, cumulative survival rates were 51%+/-38% at 6 months, 28%+/-38% at 12 months, and 18%+/-30% at 24 months. Median and crude mean+/-SD survival times were 199 days and 316+/-343 days. Multivariate analysis showed that the long-term outcome was significantly associated with functional status (p=0.000001), weight loss (p=0.00001), the CD4 count (p=0.00001), the HIV disease stage (p=0.01), the duration of AIDS (p=0.001), the admission cause group (p=0.03), and the SAPS I at admission (p=0.00001).

The short-term (in-ICU and in-hospital) outcome of HIV-infected patients was mainly related to the severity of the acute illness (SAPS I, cause of admission, need for and duration of mechanical ventilation), and to the preadmission health status, based on functional status and weight loss. Some of these parameters, in particular the SAPS I and preadmission health status, also influenced the long-term outcome. Whereas HIV-related variables had little impact on the in-ICU outcome, they were closely related with the in-hospital outcome and even more strikingly with the long-term outcome. Thus, the life expectancy of HIV-infected patients, which depends primarily on the natural history of the HIV infection, is the most powerful determinant of the long-term prognosis. Our results confirm that ICU support for HIV-infected patients should not be considered futile.