Time series model to predict the demand for care of patients with chronic kidney disease, 2022

Abstract

The main objective of this study is to forecast the demand for patients with chronic kidney disease in state-run healthcare facilities in Peru in 2022 using time series models, and to conduct a descriptive analysis of this demand. This study is justified as there are no similar studies in Peru, despite the known deficiencies in equipment and supplies for the treatment of kidney diseases through medical procedures such as dialysis. This is a descriptive and exploratory study; the design is non-experimental, cross-sectional and descriptive. The population consists of 1,064,744 patient records with various information, such as the care period, identification code, name of the healthcare facility, among others, taken from the Open Data Platform of Peru. No sampling was carried out because time series models were built at daily intervals. Statistical techniques such as simple and stacked bar graphs, pie charts and frequency tables were used; A recursive autoregressive forecasting time series model was built using Python through Jupyter Notebook for processing. The most important results show that the highest demand is concentrated in Lima, with a balanced distribution between men and women, and a higher incidence in people aged 50 to 70, especially among those with free insurance. Analyzing the components of the time series and using the Dicky-Fuller test, it was decided to use a recursive autoregressive forecasting model, obtaining an R2 of 96.62%. In addition, after performing a hyperparameter adjustment, an R2 of 94.61% was obtained for the same model, which was less over-adjusted and met most of the time series assumptions [3]. Therefore, we can conclude that the model obtained is good for predicting the demand for care of patients with chronic kidney disease since it has an optimal performance and meets all the assumptions except for autocorrelation.