Análise de regressão logística para predição de custos assistenciais no setor de saúde suplementar

• Autor(a) principal: Aquino, Luana Ramos de
• Data de Publicação: 2017
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFS
• Texto Completo: https://ri.ufs.br/handle/riufs/6727
• topic: Ciências atuariais; Ensino de estatística; Actuarial science; Teaching statistics; OUTROS::CIÊNCIAS ATUARIAIS

One commonly observed phenomenon in health plans is the distribution of costs in such a way that a small portion (less than 5%) of the population accounts for a large portion (more than two thirds) of expenditure. In this way, the objective of this work is to explain the health sector as a whole, as well as to obtain a logistic model of prediction that can classify the beneficiaries of a supplementary healthcare provider, with a focus on those with higher healthcare costs. This paper addresses the problem of how to optimize the management of health care costs with supplementary health through the identification of beneficiaries who are more likely to belong to this small group of high costs. The logistic regression technique was applied to a sample of 42 thousand beneficiaries of a large health plan operator (150 thousand patients) from Brazil in order to group patients according to their respective expenditures in 2013, from their care costs in the previous year and age. The database used in the study comprises a total of 42 variables, 41 of which are independent based on the information of sex, age and care use in 2012 and 1 dependent variable (cost of care in 2013). The beneficiaries were classified as low, medium and high cost users according to their care costs. The logistic method presented a 96.6% accuracy in the sample considered, but considering the target public (high cost clients) the method presented difficulties in identifying these beneficiaries, correctly predicting 13% of these clients. The result obtained by the logistic model obtained a good accuracy due to the number of correct predictions of the low cost beneficiaries and showed a considerable prediction for the individuals of high cost, since those of high cost are of difficult prediction.