The emerging research issues in evidence-based healthcare decision-making and explosion of comparative effectiveness research (CER) are evident proof of the effort to thoroughly incorporate the rich data currently available within the system. The flexibility of Bayesian data mining techniques lends its strength to handle the challenging issues in the biomedical and health care domains. My research focuses primarily on Bayesian data mining techniques for non-traditional data in this domain, which includes, 1. Missing data: Matched-pair studies with fixed marginal totals with application to meta-analysis of dental sealants effectiveness. 2. Data with unusual distribution: Modeling spatial repeated measures with excess zeros and no covariates to estimate U.S. county level natural fluoride concentration. 3. Highly irregular data: Assess overall image regularity in complex wavelet domain to classify mammography image. The goal of my research is to strengthen the link from data to decisions. By using Bayesian data mining techniques including signal and image processing (wavelet analysis), hierarchical Bayesian modeling, clinical trials meta-analyses and spatial statistics, this thesis resolves challenging issues of how to incorporate data to improve the systems of health care and bio fields and ultimately benefit public health.