Avaliação de desempenho de modelos de ajustamento pelo risco

Abstract

RESUMO - Contexto: A mortalidade adequadamente ajustada pelo risco é um indicador de efetividade dos cuidados prestados e indesejado para o doente, prestadores de cuidados de saúde, gestores e a sociedade em geral. Também é utiliza para comparar desempenho entre diversos prestadores de cuidados de saúde. Esta última aptidão é obtida com operacionalização dos sistemas de classificação de doentes que atribuem aos doentes um determinado risco de morte com base nas conceções de produto hospitalar e dimensões do risco. No presente estudo foram utilizados quatro sistemas de classificação de doentes, dois genéricos (APR-DRG e Disease Staging) e dois clínicos (APACHE II e SAPS II). Do estudo conclui-se que os sistemas com significância clínica, como Disease Staging, APACHE II e SAPS II têm maior sensibilidade na previsão da mortalidade e estão relacionados com a procura de cuidados de saúde

ABSTRACT - The study had as general objective to evaluate the adustment of the four risk adjustment methods (APR-DRG, Disease Staging APACHE II and SAPS II) which measure the severity of the disease and the probability of hospital mortality. In the specific objectives we aimed to analyze if the models had similar expected mortality; analyze differences and similarities; to analyze the relationship between disease severity and mortality in the four systems. The population under study: discharge summaries, hospitalization episodes of the Centro Hospitalar Lisboa Central-EPE and internal data from UUM and UCIP-HSJ. The evaluation of the adjustment of the 4 models as well as their variants according to the variables introduced from the literature review, revealed a good calibration and with significant results for all systems except SAPS II. Regarding the relationship between the probability of death estimated for the 4 systems given by the percentiles and verified effective deaths, it should be noted that in all systems there is a dependence relation between the variables (p < 0.01). In all systems, there is a relationship between the two variables, with the highest percentile being the highest percentage of deaths. The systems with the highest percentages of deaths in the last percentile are SAPS II and Disease Staging, curiously in equal percentage. Disease Staging is the only model in which, in the third percentile, there is a higher percentage of deaths compared to patients classified as discharge alive from hospital. The two generic systems have the lowest percentages of deaths in the first and second percentiles. More results can be checked in detail from the charts. From the study, it was concluded that systems with clinical significance, such as Disease Staging, APACHE II and SAPS II, are more sensitive in the prediction of mortality and are related to the demand for health care.