Exploring Cohort-design Methods and Enhancing the Existing HSIU Method

A simple cohort approach, HSIU, was developed in the previous Indiana University/FNIH grant to screen associations between medications and non-specified conditions in very large clinical and administrative databases. The research showed that through OMOP’s extensive testing of all methods that analytic results are extremely sensitive to design features of methods. To better control the balance between sensitivity and specificity and to improve the overall accuracy through bias reduction, we anticipate that the implementation of a series of options that allow flexibility in the study design will help to achieve better classification of drug-outcome associations and enhance the computational speed.

In this work, we are exploring cohort-design methods, and will develop a program that will allow to study outcomes with flexible design features. Specifically, we are developing a generic cohort program by enhancing and refining the existing HSIU method by adding or modifying various design features, for example, requirement of certain baseline duration, options for follow-up and options for comparator selection by either an algorithm based on the vocabulary hierarchy, or user input, to better define exposed patient cohort and ‘control’ patient cohort.

Deliverables of this research will be posted below as available.