The Applicability of Lee-Carter Method to Forecast Health Services Use in Brazil

Cristina Guimarães Rodrigues, Centro de Desenvolvimento e Planejamento Regional (CEDEPLAR), UFMG
Monica V. Andrade, Centro de Desenvolvimento e Planejamento Regional (CEDEPLAR), UFMG
Carla J. Machado, Centro de Desenvolvimento e Planejamento Regional (CEDEPLAR), UFMG

This paper aims to forecast hospital admission rates in Brazil from 2008 to 2020 using the Lee-Carter method. We test the applicability of the Lee-Carter approach, usually applied to forecast life expectancy, comparing the results to other methods traditionally used in the literature. We used data about hospital admissions from public health care in Brazil from 1993 to 2007, the only available period of time with information about age and sex of the patients. We also compare the Lee-Carter model with the traditional models of forecasting health services utilization which use simple trends extrapolation. The results indicate that Lee-Carter method fits well to the data, and better than traditional methods, especially to forecast rates by age. We have shown that useful forecasting of health services utilization can be derived from the Lee-Carter method, even when one does not have a long time-series.