Generic drugs account for approximately 86% of all prescriptions dispensed in the US. Generics are approved for marketing on the basis of demonstrating bioequivalence with their brand-name counterparts and there is no formal requirement for post-marketing surveillance. However, the need for post-marketing surveillance of generics to ensure patient safety is highlighted in recent years in light of reports of suboptimal performance of certain generics. The most efficient approach to monitor the effectiveness and safety of generics in the post- marketing period is to use data collected as a part of patients' routine healthcare to conduct large observational studies comparing outcomes between users of generic and brand-name versions of a drug. However, such studies are prone to confounding bias. Therefore, we propose to conduct research aimed at improving the methodology of post-marketing studies of generics to address confounding and generate valid inferences from these studies. Specifically, our aims will be;
Aim 1 : Evaluate novel data sources to augment comparative studies of generics. In this Aim, we will assess the feasibility of linking novel data sources, including electronic health records from the Partners HealthCare System and data from the US census, to health insurance claims from the US Medicare program in order to identify additional confounders such as lifestyle variables and socioeconomic indicators which are usually not measured in insurance claims.
Aim 2 : Identify a set of generally applicable confounders that could be used in comparative studies of generics. In this Aim, we will identify a general set of confounders for observational studies of generic drugs by considering a broad array of variables across the three data sources linked in Aim 1.
Aim 3 : Compare the performance of various methods for confounding control in comparative studies of generics. In this Aim, we will compare a number of candidate methods for confounding control in observational studies of generics including propensity score (PS)-matching, high dimesional-PS matching, disease risk score-matching; and traditional multivariable regression adjustment.
Generic drugs account for approximately 86% of all prescriptions dispensed in the US, yet there is very little methodological research is available dedicated to post-marketing surveillance studies of generics. Therefore, we propose to conduct research aimed at improving the methodology of post-marketing studies of generics by linking multiple data sources and evaluating cutting-edge analytic techniques to generate valid inferences from these studies.
