Análise comparativa de escores prognósticos para pacientes críticos com HIV/AIDS em Unidade de Terapia Intensiva: coorte retrospectiva

Abstract

Despite access to highly active antiretroviral therapy (HAART), many people living with Human Immunodeficiency Virus and Acquired Immunodeficiency Syndrome (PLWHA) require care in Intensive Care Units (ICU). Thus, it is important to establish the prognosis of these individuals at admission to direct care. There are several mortality predictor scores, however, there is a lack of data on which of these is most suitable for PLWHA in the ICU. Objective: To compare the performance of prognostic scores in PLWHA admitted to the ICU in the first 24 hours and 72 hours of hospitalization. Method: Retrospective cohort study. Clinical and laboratory data were collected from 103 adult PLWHA, in the first 24 and 72 hours of ICU admission, in order to extract data for 10 prognostic scores. To compare the performance of the scores, logistic regression was used; discrimination was determined by comparing the areas under the curves (AUC) Receiver Operating Characteristic (ROC), using the method proposed by DeLong et al. (1988). The calibration of the models was evaluated using the Hosmer-Lemeshow test. Statistical analysis was performed using R-Statistics, version 4.0.5. Results: During the 24 hours after ICU admission, seven prognostic scores were evaluated, of which five had modest and similar discriminatory power, with the 24h MPM II having the highest discrimination (AUC: 0.665), followed by ODIN (AUC: 0.649), MPM II Admission (AUC: 0.641), APACHE II (AUC: 0.619) and SAPS III (AUC: 0.617), with no statistically significant difference between them (p-value > 0.050). All models had adequate calibration (p-value > 0.050). In the 72 hours, three prognostic scores were analyzed. MPM II 72h showed good discriminatory power (AUC: 0.802), while SOFA and MODS presented acceptable discriminatory power (AUC: 0.712; AUC: 0.780, respectively). There was a significant difference (p-value < 0.050) in the AUC of the SOFA score for the others, but there was no difference between the AUC of the MPM II 72h and the AUC of the MODS (p-value > 0.050). All models calibrated properly (p-value > 0.050). Conclusion: After comparing the discriminatory performance of seven prognostic scores within 24 hours of admission, there was no difference between them. However, the 24h MPM II showed greater sensitivity and accuracy. At 72 hours after admission, three prognostic scores were compared. MODS and MPM II 72h showed better discriminatory power than SOFA. In addition, the MPM II 72h demonstrated greater sensitivity and accuracy in PLWHA admitted to the ICU.