Summary Measures of Quality for Diabetes Care
Leonard M. Pogach MD MBA
East Orange Campus of the VA New Jersey Health Care System, East Orange, NJ
East Orange, NJ
Funding Period: October 2007 - September 2010
To enhance assessments of healthcare quality, the Institute of Medicine (IOM), private sector organizations, the Agency for Healthcare Quality and Research (AHRQ), and health services researchers have proposed the use of quality measures that summarize care for a condition in multiple dimensions, using approaches that account for the relative contribution of each component to mortality and morbidity.
Aim 1: Compare rankings of between- and within-facility performance on diabetes quality of care, determined by a threshold-based approach that uses two different data sources -- medical abstraction from VA External Peer Review Program databases (EPRP), and the administrative Diabetes Epidemiologic Cohort database (DEpiC).
Aim 2: Compare rankings of between- and within-facility performance on diabetes quality of care determined by a threshold- and continuum-based approach, using data from DEpiC only.
Aim 3: Compare rankings of between-facility performance on diabetes quality of care determined by a simulation- and continuum-based approach, using data from DEpiC only.
Aim 4: Compare rankings of between-facility performance on projected outcomes and quality-adjusted life years (QALYs) determined using the CDC Diabetes Model in addition to the United Kingdom Prospective Diabetes Study (UKPDS) model.
Aim 5: Compare projected costs a) using CDC model base case cost assumptions, and b) using CDC model incorporating HERC (Health Economics Resource Center) costs.
This study compared administrative data from 2003-2004 and medical record abstracted data from EPRP databases from the Office of Quality and Performance (OQP). We created measures to assess facility-level performance using three specific approaches (i.e., threshold-, continuum-, and simulation-based).
1. We used the UKPDS model to calculate new measures reflecting 1) the proportion of veterans' QALY "captured" (observed control of these three values/values resulting from guideline-recommended control), and 2) the QALYs "missed" (i.e., the sum of differences in QALY [guideline-recommended vs. observed control] in a10% random sample of VA facilities. Rankings changed by >= 2 in either direction at 4/12 (33%) facilities when based on the current approach vs. one combining information from all three approaches. Although only about one in five subjects met a composite threshold at the facility level, this translated into capturing 95% of QALYs. These results suggest a relatively narrow distribution of the outcome (simulated QALYs), despite a broad distribution of the intermediate outcomes used as inputs for the UKPDS model.
2. Findings show that the assessment of the quality of good glycemic control among VA facilities differs using the National Committee for Quality Assurance (NCQA) - Healthcare Effectiveness Data and Information Set (HEDIS) measure for the overall study population compared to a subset of patients receiving complex glycemic regimens (largely insulin). The use of a continuous measure provided comparable overall rankings (Spearman Rank Correlation test) to the dichotomous measure, but about 15% of the best performing facilities (statistically different) changed to average (not statistically different).
3. We proposed that assessment of efficiency in the treatment of glucose, blood pressure, and cholesterol in persons with diabetes should incorporate evaluation of the future health care benefit that is "purchased" by direct pharmaceutical costs. This will require paradigm shifts in conceptualizing quality measures as being continuous rather than dichotomous, and evaluating benefit in multiple populations that may differ by age and co-morbid conditions. Using a simulation model, we found that using newer classes of medication, even accounting for QALYs related to putative benefits such as less weight gain, come at considerable costs.
We have concluded that despite its conceptual attractiveness, the use of QALYs for evaluating "rankings" probably is of marginal utility to decision makers. Based upon our findings, we have proposed the combined use of continuous rather than threshold measures using risk stratified samples as a more epidemiological (e.g., evidence-based) approach for next generation measures.
In addition to our published impacts, we have also disseminated key findings to VHA and other federal agencies via internal and external operations meetings. Our research on risk stratification by age and chronic complex illness was presented on national Primary Care calls, and our work on glycemic management has led to the development of technical specifications for possible next generation glycemic measures in the Office of Quality and Performance.
DRA: Health Systems
Keywords: Research measure, Diabetes, Quality assessment
MeSH Terms: none