BN Doebbeling, Program in Health Sciences Research, Iowa City VAMC and University of Iowa Colleges of Medicine and Public Health, Departments of Internal Medicine and Epidemiology; TE Vaughn, University of Iowa College of Public Health, Department of Health Management and Policy; MM Ward, University of Iowa College of Public Health, Department of Health Management and Policy; WR Clarke, University of Iowa College of Public Health, Department of Biostatistics; RF Woolson, University of Iowa College of Public Health, Department of Biostatistics; E Letuchy, University of Iowa College of Medicine, Department of Internal Medicine; BJ BootsMiller, Program in Health Services Research, Iowa City VAMC; T Tripp-Reimer, University of Iowa College of Nursing; BA Sorofman, University of Iowa College of Pharmacy; JB Perlin, The Office of Quality and Performance, VA Central Office, Washington, DC

Objectives: VA uses clinical practice guidelines (CPG) to improve care. Adherence is assessed through annually selected performance measures administered through the External Peer Review Program (EPRP). The purpose of this study was to examine consistency of ranking of VAMC facilities for 11 chronic disease quality-of-care indicators (QI) system-wide.

Methods: EPRP data are a stratified national review of ambulatory care patients' charts, most with >=1 chronic conditions, seen >=3 times annually. We analyzed 4th quarter FY98 data for all VA facilities (N=146) to assess facility-level performance across 11 QI's for five conditions (hypertension, diabetes mellitus [DM], major depressive disorder [MDD], alcohol abuse [AA], tobacco use [TU]). Eleven outcome indicators were calculated as the proportion of eligible patients (N= 28,039) receiving the QI service within the recommended interval, each standardized across facilities. Facilities were ranked two ways: 1) a mean Z score was calculated for 11 standardized outcomes per facility, then combined into five condition-specific Z scores. "Best practices" performance, was defined as a mean Z score in the upper tertile of means and the lowest tertile standard deviation (SD); 2) a percentile rank, based on different a priori cutoff levels, was used to rank order each VAMC nationally on condition-specific outcomes for the same five major conditions. Facilities with the most percentile ranks >80% and the fewest <50% were defined as the "best practices" group. A "quality of fit " test assessed whether performance based on national percentile rank at various cutoffs was above that expected by chance. EPRP and American Hospital Association (AHA) data were linked and regression models estimated to identify macro-organizational characteristics as predictors of the mean facility-level Z score, consistent with "best practices."

Results: Subjects (85% men, mean age = 65) were selected for having hypertension (55%), COPD (45%), DM (30%), ischemic heart disease (4%) or none of the above (6%). Mean performance across facilities on QI's were hypertension screening (91%); DM annual HgBA1c (89%), foot sensation check (76%), dilated exam (72%), annual urine protein (67%), hypertension screen (84%), annual lipid profile (50%); MDD screened (17%), AA screened (74%);TU screened (89%), TU referred (21%). Test results for different percentile cutoffs (50%, 60%, 70%, 80%, and 90%), demonstrated that distribution of the number of ranks above a given cutoff was significantly higher than expected (p<0.01). Twenty-four VAMCs (16%) had "consistently good" performance; predictors included: staffing hours/patient day (B= 0.0074), FTE physicians/1000 outpatient visits (B=-1.027), teaching hospital (B=-0.237), non-special services (B=0.0064) and VISN (B ranges =-0.54-0.74).

Conclusions: Comparing performance in the delivery of multiple indicators of quality of care for chronic disease patients in the VA is possible using currently available data. Adherence to chronic disease QI's varies across facilities nationally, with certain facilities and VISNs performing consistently well, suggesting that implementation, documentation, or contextual (i.e.,VISN) differences may be responsible.

Impact: Lessons from VAMCs and VISNs with "consistently good" performance may be applied elsewhere. Further study is needed to identify factors important in consistent performance in the delivery of quality care and in facilitating effective CPG implementation.