1018 — EpiTools: Medication History Estimator
Sauer BC, Pickard S, He T, Samore M, and Nebeker JR, SLC IDEAS;
Epidemiological and health services research has been criticized as being unreliable. Scientific evidence is strengthened when the study procedures of important findings are transparent, open for review, and easily reproduced by different investigators and in various settings. This study demonstrates the Medication History Estimator (MHE) and how it is used to infer drug exposure from order, dispensing, and administration data.
The MHE was developed in the VINCI environment using pharmacy data from Regions 1 and 4 during 2004-2010. This tool was designed to allow the user to define specific inputs that include criteria for incident courses, gap criteria to end a course, calculation of proportion of days covered (PDC), output data structure for intention-to-treat analysis, and on-protocol analysis with the data structured by unit of time. The program also automatically generates reports suitable for publication. We developed and tested the tool by evaluating exposures to ACE-inhibitors (ACE-I) and Angiotensin Receptor Blockers (ARBs). We report courses based on a 180-day criteria for incident course and 90-day gap criteria to end the course.
There were ~970,000 Veterans with 47% receiving at least one ACE-inhibitor and 9% receiving at least one ARB. There were a total of 498,250 ACE-I courses. The average duration and standard deviation (SD) of the 66,608 incident and 431,642 established ACE-I courses were 151 days (SD 111-days) and 245 (128), respectively. There were a total of 91,861 ARB courses. The average duration and SD of the 10,507 incident and 81,354 established courses were 156 days (115 days), and 249 days (127 days), respectively. Course (~90%) and period (~70%) level PDC were similar between the treatments.
When investigators study medication exposure, they are confronted with a series of decisions concerning how to characterize treatment histories and classify treatment groups. The MHE standardizes these decisions, makes the logic transparent and is designed for rapid response evaluation where parameters can easily be varied to test the robustness of the findings to design specification.
Development of modular programs, such as the MHE, that combine to produce epidemiological pipelines to automate components of the research process will support transparent and rapid response to nationally important clinical questions.