BACKGROUND/RATIONALE:
The VA is a national leader in performance measurement. However, methods of VA performance measurement may not be keeping pace with this reputation because they: (1) are based on small samples of patients at the facility level; (2) are primarily cross-sectional "snapshots" of care; and, (3) do not provide information that readily lends itself to quality improvement. The goals of this project are to develop methods to assess facility-level performance that addresses these issues, and then to compare performance among like peer facilities. Further, we examine the impact on facility-level performance of excluding patients in whom aggressive care may be inappropriate, e.g., those with limited life expectancy (LLE).

OBJECTIVE(S):
Our project objectives include (1) to report facility-level data on performance measures for a common chronic disease (diabetes) using VA administrative data; (2) to compare performance using our previously developed VA peer facility groups; and (3) to evaluate the effect of excluding patients in whom aggressive treatment may be inappropriate, e.g., those with LLE.

METHODS:
We used VA administrative data to identify a cohort of patients with diabetes utilizing primary care in fiscal year (FY) 2007. We determined the proportion that had glycemic control at index (defined as hemoglobin A1C < 7%) and those who received appropriate follow-up care (e.g., medication adjustment) to uncontrolled A1C levels within 8 weeks. Overall quality of care was defined as the sum of those with control at index and those who received appropriate follow-up care. We also assessed the proportion of patients with poor control defined as A1C reading > 9%. We conducted a review of relevant literature and convened a panel of expert clinicians to develop condition-specific algorithms to identify those with LLE. We compared facility-level performance across VA peer groupings using logistic regression, adjusting for patient-level risk factors including age, race, and illness burden and assessing the impact of patients with LLE.

FINDINGS/RESULTS:
We identified 830,328 patients with diabetes in FY 2007 at 105 facilities. Of these, 87% had at least 1 A1C level recorded within 12 months of their index date. Only 47% had controlled A1C levels at index, ranging from 34.7% to 67.1% across facilities. Among eligible patients, 36.7% received appropriate follow-up care (range, 23.8% to 50.2%). Two-thirds of the cohort had overall good quality of care (range, 54.0% to 82.4%). Six percent had at least 1 LLE condition. Compared to diabetic patients without LLE, those identified as having LLE were more likely to have A1C levels controlled at index (51.6% vs. 47.0%, p < 0.001) and as likely to receive overall good quality (66.9% vs. 66.6%, p = 0.209). However, LLE patients were less likely to receive treatment intensification in response to elevated A1C levels (31.7% vs. 37.0%, p < 0.001). The adjusted odds of diabetic patients having glycemic control at index at a non-tertiary facility were similar to the group mean of diabetic patients of 12 similar hospitals (OR 0.97, 95% CI, 0.78-1.20). However, at the highest ranked facility, diabetics were 67% more likely to have A1C < 7% at index (OR 1.67, 95% CI, 1.35-2.06). Peer rankings for the performance measures were not impacted when LLE patients were excluded.

IMPACT:
Performance measures that more accurately reflect the quality of diabetes care being provided at the facility-level can be developed using available VA administrative data sources. Such measures provide a cost-effective, efficient, and clinically valid method of quality assessment. Further research is needed to extend these findings to other chronic medical conditions.

---

External Links for this Project

---

PUBLICATIONS:

Journal Other

1. Woodard L, Urech T, Robinson C, Kuebeler M, Petersen LA. Differences in therapy intensification for glycemic and lipid control in diabetic patients with and without limited life expectancy. [Abstract]. Journal of general internal medicine. 2009 Apr 1; 24(S1):S55-S56. [view]

Conference Presentations

1. Pietz K, Woodard L, Petersen LA. Clustering to achieve normality in generalized linear mixed models. Poster session presented at: Joint Statistical Annual Meeting; 2009 Aug 4; Washington, DC. [view]
2. Woodard L, Urech T, Robinson C, Kuebeler M, Petersen LA. Differences in therapy intensification for glycemic and lipid control in diabetic patients with and without limited life expectancy. Paper presented at: Society of General Internal Medicine Annual Meeting; 2009 May 13; Miami, FL. [view]
3. Woodard LD. Treatment intensification in patients with diabetes and limited life expectancy. Paper presented at: VA HSR&D National Meeting; 2009 Feb 12; Baltimore, MD. [view]

DRA: Health Systems Science
DRE: none
Keywords: Diabetes, Quality assessment
MeSH Terms: none