Article Title

Expectation vs. Reality: Drug Coverage and Pharmacy Fill Rates at Group Health

Publication Date



data capture, coverage


Background/Aims: It seems elementary that data capture of pharmacy fills at a health maintenance organization (HMO) should be suspect for members without drug coverage. Coverage of prescriptions is a major incentive for filling prescriptions at the HMO pharmacy. Even better –– even if there are scripts filled at outside pharmacies, they should generally result in the filing of a claim, which also generates fill data. Imagine our surprise, then, when it was observed that over a period of some years, Group Health members without drug coverage had significantly higher rates of pharmacy fills than did members who had drug coverage. The proposed presentation describes the investigation into this seeming anomaly.

Methods: After independently verifying that we do indeed observe higher fill rates in uncovered patients over a period of four years, we proceeded from the theory that the result was indication of a virtual data warehouse data problem. Either we incorrectly computed the DRUGCOV field in enrollment data, or we somehow mishandled the fills data. To investigate problems with the fills data, we recalculated rates of 30-day fills using native Group Health pharmacy data from our enterprise data warehouse. For the drug coverage flag, we isolated the set of people whose rates were anomalous, characterized their plans and coverages, and verified their drug coverage status.

Results: To our great surprise, we could readily reproduce the finding in native Group Health pharmacy data. Similarly, we found no defect in the drug coverage information. Notwithstanding this, the investigation did arrive at a plausible (if surprising) explanation for the results.

Discussion: As the saying goes, “the map is not the territory.” While our intuitions and expectations about our data are necessary and powerful assets for focusing our uses of data to pursue research, it is always good to investigate and verify that our assumptions are true, and doubly important to investigate and document findings that defy conventional wisdom.