Modelos de sobrevivência aplicados a um plano de saúde do Brasil

Abstract

With a database of a Brazilian health plan, survival analysis was used to determine the length of stay of the user in this plan. This bank contains 110,052 beneficiaries consisting of 60,156 women and 49,896 men with customer records between 1942 and 2018. The study was finished in December 2018 with censored data. As a descriptive analysis was applied to the database resulting in the average length of stay in the plan which was 160 months, with an average age of 38 years, the type of participation that most uses the plan is “D” and the most acquired accommodation is from the ward. In addition, by analyzing the insured's survival with the Kaplan-Meier estimator who obtained the probability of remaining on the 20% plan in recent months for ward accommodation, “T” participation and the male gender. The Cox regression model was inadequate due to the violation of the basic assumption of proportional risks by the covariate ownership. Therefore, the choice of model was made by the Weibull parametric regression method with accommodation / ward, age, title / D, title / E and title / T as significant variables.