3008 — Using Clinical Groupers to Study Episodes of Care
Borzecki AM (Bedford COE), Rosen AK
Episode grouper software programs (EGs) were designed to measure the entire episode of care related to a particular condition. Given the increasing prominence of EGs in quality assessment, understanding how they construct episodes is important, since the algorithm used may impact such assessments. We examined the feasibility of applying two commercial EGs, Medstat’s Medical Episode Grouper (MEG) and Ingenix-Symmetry’s Episode Treatment Groups (ETGs) to VA data, and compared their grouping of claims data representing an episode of care for four chronic conditions: diabetes, depression, congestive heart failure (CHF), and chronic obstructive lung disease (COLD).
We used the VA National Patient Care Database to identify veterans with diabetes, depression, CHF or COLD with > = 2 outpatient visits per year in FY04 and FY05. For disease cohort inclusion, patients had to have > = 2 visits or > = 1 hospitalization with the relevant diagnosis in FY04. To understand how to apply the software to VA data, we tested it on 2 diabetes patients. We then applied the groupers to each cohort using FY05 data and compared agreement on claims grouping, using cross-tabulations and kappa statistics.
Initial testing revealed several basic differences in how EGs construct episodes. For example, ETGs don’t account for inpatient procedure codes; MEG do. They also prioritize diagnoses differently (e.g., MEG prioritize a diagnosis of hypertension ahead of CHF, even if CHF is the primary diagnosis; ETGs assign this to a CHF episode). Overall, the groupers differed significantly in which records were grouped to a given episode; the average overall kappa was 0.49. By condition, the best agreement was seen for depression (kappa 0.69), the worst for CHF (kappa 0.28).
While we demonstrated significant EG differences in individual claims grouping, determining whether these differences affect the “perceived” quality of care delivered around a specific disease requires further work. Additionally, whether one grouper better reflects the actual course of care experienced by a group of patients is unclear.
Although EGs have the potential to broaden the way in which quality is assessed, their use in the VA seems premature. This work is an important initial step in assessing the validity of episode measurement tools in the VA.