Wattanavadee Sriwattanapongse, Metta Kuning,, and Naratip Jansakul

pp. 207 - 214

Abstract

This study is based on the individual hospital case records of malaria routinely reported from 1999 to 2004 in the North-western area of Thailand, which included Mae Hong Son and Tak provinces. The objective of this study was to model the patterns of hospital-diagnosed malaria incidences by month, district and age-group for the two North-western border provinces in Thailand. The model used linear regression, Poisson regression and negative binomial regression to forecast the districts and age groups in which epidemics are likely to occur in the near future in order to prevent the disease by using suitable measures. Among the models fitted, the best were chosen based on the analysis of deviance and the negative binomial generalized linear model was clearly preferable. The model contains additive effects associated with the season of the year,
district, age group and the malaria incidence rates in previous months, and can be used to provide useful short-term forecasts. Having a model that provides such forecasts of disease outbreaks, even if based purely on statistical data analysis, can provide a useful basis for allocation of resources for disease prevention.