Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

Health Services Research & Development

Veterans Crisis Line Badge
Go to the ORD website
Go to the QUERI website

IIR 06-253 – HSR&D Study

New | Current | Completed | DRA | DRE | Portfolios/Projects | Centers | Career Development Projects

IIR 06-253
Improving Quality Measurement Using Quality Adjusted Life Years
Sandeep Vijan MD MS
VA Ann Arbor Healthcare System, Ann Arbor, MI
Ann Arbor, MI
Funding Period: April 2008 - March 2011

BACKGROUND/RATIONALE:
The VHA has been a leader in using quality measures to provide incentives and improve quality of care, and is expanding this with "pay-for-performance" programs. Current measures of quality include measures of processes of care or dichotomized intermediate clinical outcomes. However, there is no metric for comparing the impact of performance across measures as there is no consideration of the magnitude of the link between measures and long-term outcomes. Further, the dichotomization of these measures provides an incentive to make small improvements in those close to goals rather than focusing on those that are farthest from goal and most likely to benefit. There is thus a significant need for a composite metric of quality that can appropriately weight individual measures and can be compared across and within measures and disease processes.

OBJECTIVE(S):
We have conducted a study to develop a measure to examine the use of quality-adjusted life years (QALYs) as a quality measure, using diabetes as a template for quality measurement in chronic disease. The specific aims were as follows:
1. to examine the distributions of the intermediate measures hemoglobin A1c (a1c), blood pressure (BP), and LDL cholesterol, along with age and comorbidities in the VISN 11 diabetes population;
2. to use these values as individual-level inputs into a diabetes simulation model that generates predicted individual-level QALYs and to examine the individual-level potential QALYs gained (QALY-G) by optimal management (defined as achieving guideline goals for each of the intermediate outcome measures);
3. to examine, using regression modeling with QALY-G as the dependent variable, the weights (between and within measures) of the intermediate measures in predicting QALY-G; and
4. to examine, using regression modeling, the profiling characteristics of QALY-G as a tool to profile facilities, teams, and physicians.

METHODS:
We used VHA administrative data to examine quality of care for several key intermediate diabetes outcomes: Hemoglobin A1c level; blood pressure level; and LDL cholesterol level. Values for each condition were used as inputs into simulation models to generate individual-level QALYs. We then ran the same simulations using idealized goals to develop optimal QALYs. The difference between observed and ideal QALYs (QALY-G) was the primary outcome measure and represented "lost" QALYs due to less than ideal care. We used the lost QALYs as dependent variables in a series of multilevel regression analyses. First, we examined, using linear regression, the within-disease predictors of QALYs; for example, in diabetes, we compared the effects of blood pressure vs. glycemic control; additionally, we compared the marginal benefits of gradually increasing levels of control. We developed risk-adjustment models using age and comorbidity to control for effects outside of provider control. Second, we examined levels of clustering and proportion of explained variance in QALY-G at the site, team, and provider level to examine panel size requirements and reliability of profiling using QALY-G.

FINDINGS/RESULTS:
We generated the necessary datasets and completed basic descriptive analyses of the metrics within VISN 11. We found over 50,000 unique veterans who meet the criteria for having diabetes, of which 39,264 are VA users (e.g., more than 1 primary care clinic visit in the prior year). We also collected lab and clinical data describing their diabetes status (mean a1c = 7.6; mean LDL = 102; mean BP = 138/72) and comorbidity levels (Romano adaptation of the Charlson comorbidity index; mean = 2.8). We also collected a variety of key health status indicators (e.g., albuminuria status) to populate the key variables simulation model. We have published 2 papers from the simulation modeling. The first paper uses an extension of the simulation model to examine the achievability of common treatment targets in the broader US population, the likelihood of polypharmacy and side effects. We find that a substantial proportion of the population is unlikely to ever achieve idealized goals, even with optimistic assumptions, which has a direct relevance to quality measurement. In the 2nd paper, we find that QALY-G (quality adjusted life-years gained by optimal management) varies dramatically by underlying risk of developing cardiovascular disease, so that those at lower risk actually experience net harm in quality of life through overly aggressive management. This has clear clinical and quality measurement importance, as it suggests that typical approaches to defining standards of care and quality targets actually may harm some individuals.

Our regression models use QALY-G as the dependent variable and risk factors as independent predictors. This allows us to estimate the relative weights of the risk factors in determining overall gains in QALYs, allowing for the potential to weight aggregate performance measures. We find that contrary to current practice, which weights these equally, there is substantial variation in the effectiveness of risk factor management on outcomes. Of the potential QALY-G, 65.9% were due to blood pressure, 23.4% for LDL cholesterol, and 10.7% for A1c. Thus, a weighting system for an aggregate measure should strongly favor BP management over other risk factor management. Further, age and comoribidity are very large factors in determining degree of benefit, particularly for A1c. We also find that while a composite QALY-G measure is more reliable than the three risk factor levels as a measure of quality, it is not likely to be reliable as a measure for profiling individual physicians, but is reasonable at the facility level.

IMPACT:
VHA is evaluating the effects of providing incentives to improve quality of care. However, current quality measures have significant limitations in that they are all weighted equally and do not provide information on the likely benefit of improving quality. The results of this study will provide a refined set of quality measures. These can be used to provide incentives to improve measures that have the most impact on patient quality of life and longevity, improving quality of care in an effective and cost-efficient manner. Dr. Vijan is working with a national VA committee (formerly based in OQP and now in OABI) to revise performance measures based in part on our findings.

PUBLICATIONS:

Journal Articles

  1. Vijan S, Sussman JB, Yudkin JS, Hayward RA. Effect of patients' risks and preferences on health gains with plasma glucose level lowering in type 2 diabetes mellitus. JAMA internal medicine. 2014 Aug 1; 174(8):1227-34.
  2. Timbie JW, Hayward RA, Vijan S. Variation in the net benefit of aggressive cardiovascular risk factor control across the US population of patients with diabetes mellitus. Archives of internal medicine. 2010 Jun 28; 170(12):1037-44.
  3. Timbie JW, Hayward RA, Vijan S. Diminishing efficacy of combination therapy, response-heterogeneity, and treatment intolerance limit the attainability of tight risk factor control in patients with diabetes. Health services research. 2010 Apr 1; 45(2):437-56.
Conference Presentations

  1. Timbie J, Hayward RA, Vijan S. Implications of Treatment-Response Heterogeneity for the Use of Intermediate Outcomes in Diabetes Quality Measurement. Paper presented at: VA HSR&D National Meeting; 2009 Feb 13; Baltimore, MD.


DRA: Health Systems, Diabetes and Related Disorders
DRE: Prognosis
Keywords: Clinical Performance Measures, Predictive Modeling, Provider Performance Measures, Quality assessment, Quality Indicators, Quality of life, Research measure, Statistical Methods
MeSH Terms: none

Questions about the HSR&D website? Email the Web Team.

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.