Effectiveness and safety of medical products are scientifically investigated based on either interventional or experimental studies and observational studies. More than often conflicting results may be reached from multiple medical research; that is, results often are not reproduced when similar studies are conducted by another researcher. In randomized clinical trials, reproducibility issues is enhanced by controlling for known and unknown factors with the use of randomization. The reproducibility of results from observational studies has been called into question many times in the recent past, with several analyses showing that well over half of the reported findings are subsequently refuted (Journal of the National Cancer Institute, May 2007). Confounding of variables that occurs as a result of not having randomization is of the several issues that affect the reproducibility in observational studies. Observational data (also called Real World Data (RWD)) are large databases that contain electronic health records, insurance claims or registry data in typical clinical care. Discussions on RWD are included in a recent article (NEJM, 2016) written by FDA officers and in the FDA 2017 guidance document on RWD. A general theme that arises from all these publications is the need for new thinking and innovative design and methodologies including that of Bayesian methods. In this investigation, we will use a Bayesian hierarchical model and Bayes factors to evaluate the reproducibility of conclusions based on a specific endpoint (e.g., health care resource utilization). After investigating reproducibility within a database, we will extend this investigation to another database. Consistency of results will be investigated within a specific database and also across databases using priors developed initially within a database. Specific disease areas and endpoints (e.g., diabetes) will be used as example for this investigation for illustration.