Tempo de permanência no complexo hospitalar da UFC (2017 a 2021) via GAMLSS

• Ano de defesa: 2023
• Autor(a) principal: Vasconcelos, Hemerson Bruno da Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: GAMLSS; permanência hospitalar; Dados assimétricos; Internação; Length of stay; Skewed data; Hospitalization; CNPQ::CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA
• Link de acesso: http://repositorio.ufc.br/handle/riufc/77950

The hospital bed is a strategic and high-cost resource, which must be used whenever there is medical indication and in a rational manner, restoring the health of the population that needs it. In this context, the management of the Hospital Complex of the Federal University of Ceará (CH-UFC) establishes the reduction in the average length of stay as one of its strategic goals for the 2021-2023 cycle. The present work aims to study the characteristics of hospital stays at a local level and from the perspective of statistical modelling. Records of hospitalizations that occurred between 2017 and 2021 were analyzed, based on secondary data available in information systems. It is possible to verify, both in the literature review and by analyzing the data found, that length of stay is a variable with strong asymmetry to the right. The proposed technique is the regression model of the GAMLSS (Generalized Linear Models for Location, Scale and Shape) class, which has, among its advantages, the possibility of adjusting all parameters of the distribution assumed for the response explicitly as a function of the explanatory variables. During the model selection process, priority was given to probability distributions supported by (0, ∞) that have a well-defined position parameter. As the minimum criteria for fit quality, the wormplot with more than 95% of the points in the confidence bands is adopted. Based on hospitalizations of adult patients, segmented by diagnostic group, 55 regression models were selected. Of these, 52 presented a good fit to the data (given the adopted criteria) and 45 were estimated under a distribution that has a mean, median or mode equal to μ. Among the biggest contributions to increased length of stay are the number of exam requests, the maximum exam waiting time and the day of the week on which medical discharge occurred.