The aim of the study was to choose and validate the tool(s) to predict the number of hospitalized patients by testing three predictive algorithms: a linear regression model, Auto-Regressive Moving Average (ARMA) model, and Generalized Auto-Regressive Conditional Heteroskedasticity (GARCH) model. The study used data from the collection of data on inflammatory bowel diseases (IBD) from the public database of the National Health Fund for the years 2009-2017, data recalculation taking into account the population of provinces and the country in particular years, and prediction making for the number of patients who would require hospitalization in 2017. The anticipated numbers were compared with real data and percentage prediction errors were calculated. Results of prediction for 2017 indicated the number of hospitalizations for Crohn's disease (CD) and ulcerative colitis (UC) at 17 and 16 respectively per 100,000 persons and 72 per 100,000 persons for all IBD cases. The actual outcomes were 21 for both CD and UC (81% and 75% accuracy of prediction, respectively), and 99 for all IBD cases (73% accuracy). The prediction results do not differ significantly from the actual outcome, this means that the prediction tool (in the form of a linear regression) actually gives good results. Our study showed that the newly developed tool may be used to predict with good enough accuracy the number of patients hospitalized due to IBD in order to organize appropriate therapeutic resources.
Keywords: inflammatory bowel disease; Crohn’s disease; epidemiology; large intestine; medical statistics; small intestine; ulcerative colitis.