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

Health Services Research & Development

Go to the ORD website
Go to the QUERI website

HSR&D Citation Abstract

Search | Search by Center | Search by Source | Keywords in Title

Analyzing the Relationship Between Cost and Quality Using a Bayesian Shrinkage Composite Measure of Quality

Burgess JF, Shwartz M, Stolzman KL. Analyzing the Relationship Between Cost and Quality Using a Bayesian Shrinkage Composite Measure of Quality. Paper presented at: Decision Sciences Institute Annual Meeting; 2011 Nov 19; Boston, MA.


Research Objective Profiling nursing homes on dimensions of cost and quality separately have been widespread in recent years. Intense interest in understanding cost and quality tradeoffs is growing as part of developing profiling value equations. To determine the relationship between cost and quality, it is necessary to combine individual quality measures into a composite measure that can then be compared to costs. We describe the composite measure of quality and then examine its relationship to costs. Study Design We compare cost per day and total cost to a composite measure of quality conceptualized as a formative construct. We control for patient severity using RUGs score. The composite measure is constructed from 28 MDS (Minimum Data Set)-based quality indicators. There is relatively low correlation among most of the 28 indicators, suggesting there is not an underlying latent construct, but rather the indicators capture different dimensions of quality. A multivariate normal model is used to shrink individual indicators in a way that adjusts for the reliability of different indicators in different facilities and reflects any existing correlation among the indicators. The shrunken estimates of the individual indicators are combined into a composite measure using opportunity-based weights (the approach used by CMS in its pay-for-performance programs). We have shown that using shrunken estimates in the composite rather than observed rates leads to better predictions of next year's composite score. We then employ Generalized Estimating Equation methods to adjust for clustering in regressing risk adjusted (by average RUGS score) both average facility cost per day and facility total costs on the Bayesian composite quality measure, using time and facility fixed effects to adjust for secular trends in costs. The cost per day estimates are weighted by length of stay while the total cost estimates are adjusted by a quadratic function of length of stay. Population Studied In this analysis, we consider 112 facilities that in fiscal year 2007 had a denominator for at least one of the quality indicators of at least 10 residents and in which at least a third of all residents were long-stay (based on average daily census). We obtain cost and quality data on these Department of Veterans Affairs Community Living Centers (nursing homes) for Fiscal Years 2005-2007, for a total sample of 336 observations. Principal Findings There is weak, though not statistically significant, support for higher rates of quality complications being related to higher costs using either the cost per day or total cost approaches. The p value associated with the composite measure of quality coefficient in the length of stay weighted cost per day regression is 0.065, while the p value in the total cost regression is 0.112. Conclusions While the cost-quality relationship presented here is not statistically significant, the practical significance of the composite quality indicator impact on costs is high and consistent whether using cost per day or total cost measures of cost. Implications for Policy, Delivery or Practice Though the results are only weakly significant, they do suggest that for VA community living centers (nursing homes) better quality as a composite measure of many indicators is associated with lower costs

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.