BACKGROUND/RATIONALE:
There is increasing recognition of the need to measure care across inpatient and outpatient settings to account for the entire episode of care related to a given condition. Episode groupers (EGs) were designed explicitly for such purposes and are being used by growing numbers of non-VA healthcare organizations. However, their clinical validity has not been determined. Thus, further evaluation of these products is necessary to determine their utility to the VA.

OBJECTIVE(S):
To examine the feasibility of applying two commercially available EGs, Medstat's Medical Episode Grouper (MEG) and Ingenix-Symmetry Episode Treatment Groups (ETGs) to VA data, and to compare them with respect to their ability to group claims data representing an episode of care for each of four chronic medical conditions: diabetes, depression, congestive heart failure (CHF), and chronic obstructive lung disease (COLD).

METHODS:
This was a retrospective observational pilot study. Our sample consisted of veterans with a diagnosis of diabetes, depression, CHF or COLD with at least 2 outpatient visits per year in consecutive years during FY04-FY05. We obtained outpatient and inpatient information from the VA National Patient Care Database. Patients had to have at least 2 outpatient visits or one inpatient stay with the relevant diagnosis in FY04 for inclusion in the disease cohort. To understand how to apply the grouper software to VA data, we first tested it on 2 patients with a diagnosis of diabetes in FY05. We then applied the groupers to patients from each diagnostic cohort using FY05 data and compared MEG and ETG output with respect to several episode criteria by condition. Analyses included cross-tabulations and kappa statistics to compare EGs on specific outputs.

FINDINGS/RESULTS:
From our initial analysis of 2 patients, we found the following basic differences with respect to record grouping: ETGs do not account for inpatient procedures because they only use CPT, not ICD-9 procedure codes; MEGs use both. ETGs attempt to group records that lack diagnostic codes such as laboratory or radiographic tests; MEGs consider records without diagnostic codes as ungroupable. The groupers have a slightly different prioritization with respect to the episode to which a given record is assigned. The groupers differed significantly in terms of which records were grouped to a given episode. Kappas were relatively low; the average overall kappa was 0.49. By condition, the best agreement was seen for depression (mean kappa 0.69), the worst for CHF (mean kappa 0.28).

IMPACT:
Given the limitations of current groupers based on our preliminary analysis, it seems premature at this time to recommend that either current grouper system be adopted by VA researchers or policy makers. This work represents an important step toward further assessment of these episode measurement tools, and in understanding which diagnoses group better than others. Future work will include determining whether such systems perform better in characterizing care for acute conditions or in other chronic conditions, and trying to construct new grouping algorithms for patients with chronic diseases that better reflect their course of care.

---

External Links for this Project

---

PUBLICATIONS:

None at this time.

DRA: Health Systems
DRE: none
Keywords: Research measure, Research method
MeSH Terms: none