Modem healthcare organizations generate huge amount of electronic data stored in heterogeneous databases. Even though these data are a valuable resource for mining useful knowledge to support scientific decision making, a representational heterogeneity of such databases requires a customized and flexible method to mine actionable knowledge. In this paper we propose a medical decision support system (MDSS) based on a logical framework. The proposed MDSS includes a knowledge induction component to induce knowledge from clinical data repositories and the induced knowledge can be deployed to pre-treatment data from other sources. The implementation of knowledge induction engine has been presented to express the power of higher-order programming of logic-based language. The flexibility of our mining engine is obtained through the pattern matching and meta-programming facilities provided by logic-based language.