Developing a hybrid decision support model to discover evidence based knowledge of the elderly with depression

Abstract

Data mining is the process to extract hidden patterns from enormous amount of data that is commonly used in a range of areas including marketing, fraud detection, scientific discovery as well as health care. The study was conducted to ensure high accuracy in assessing of elderly depression and to build useful decision rules by developing a very reliable evidence based decision support model with the combination of statistical analysis and decision tree algorithms. A large data set of 2008 Korean Elderly Survey (KES) was used consisted of 14,970 elderly data. Having depression as target variable, input variables were demographic, health related and socioeconomic characteristics of the Korean elderly population. Statistical analysis was conducted as a feature selection procession that includes the Chi-square, Fisher's exact test, the Mann-Whitney U-test and Wald logistic regression Using the C5.0 decision tree algorithm of Clementine 12.0, the final decision support models were built and C5.0 tree showed a high accuracy level of 81.6%. The decision model developed in this study can improve healthcare providers' ability in making decisions, increasing vigilance with suspected depression in elderly population.