The implementation of clinical practice electronic records will provide access to comprehensive data of outcomes in response to clinical interventions for large populations. Analyzed appropriately, this information compiled into databases will have enormous potential to assess treatment effectiveness, both expanding and complementing the results of Randomized Controlled Trials (RCTs). However, because of pitfalls with observational Studies and validity of data from clinical records, proof of principle is needed prior to capitalizing appropriately on the enormous potential of this data. Our long-term goal is to determine whether and how non-experimental data from an electronic Ambulatory Medical Record Database can be used to reliably inform the practice of medicine. This application will examine the validity of Observational Studies using data from the UK General Practice Research Database (GPRD) to assess treatment effectiveness based on rigorous comparisons with published RCTs of cardiovascular disease. Each GPRD Study will use a modeled prospective cohort design that, to the extent possible, will reconstruct the RCT with the exception of randomization, by employing similar inclusion/exclusion criteria, treatment protocol and outcome measures. Outcomes will be analyzed by adjustment for measured confounders using propensity scores and/or multivariable regression. We also will test whether specific study characteristics predict the likelihood for obtaining a valid result with a GPRD Study. Therefore the aggregated results from 20 RCTGPRD comparative studies will be examined for predictors of concordance. If individual GPRD Studies are concordant, they will be repeated using broader entry criteria in order to examine their utility to expand generalizability. The expanded GPRD study will be compared with the initial restricted Study and with meta-analyses relevant to the comparable RCT to examine their usefulness for this purpose. Thus the Specific Aims of this study will be to assess: 1. Concordance between outcomes to therapy when GPRD Studies are subjected to rigorous comparison with comparable published RCTs of cardiovascular disease, 2. Whether particular characteristics predict concordance (or lack thereof) between results of RCTs and GPRD Studies 3. Whether GPRD Studies can verify as well as expand the generalizability of findings from RCTs.
