Predicting Price-to-Charge Ratios for Community Hospitals in the United States

Purpose and Background.   Three major elements of hospital costing are the charge to payers, the total payment (or price) received from payers, and the cost to produce the services provided. Reliable price data could enable consumers to choose providers that offer better value than others, eventually leading to market-level gains in quality and efficiency. The objective of this study is to predict price-to-charge ratios for community hospital stays on the basis of charge data and information about individual stays, hospitals, market areas, and states.

Data and Methods. We used Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID) from California, Florida, Massachusetts, Nevada, New Jersey, Virginia, West Virginia and Wisconsin for the year 2006. We merged SID files with hospitals’ financial information to estimate revenue per stay by payer. We predicted PCRs for major payer categories for over 1,000 community hospitals as a function of state, market, hospital, and patient characteristics using exponential conditional mean (ECM) models with a log link and gamma-distributed errors. ECM was chosen on the basis of skewness in residuals of OLS models. The unit of analysis was the hospital. Person-level characteristics were represented by hospital-level means among all stays in the dataset. We used a two-stage iterative estimation in which equations are estimated individually and the errors saved. In the next step, each payer equation is estimated a second time with the first-stage errors of the other payer equations as new independent variables. We assessed goodness of fit through model characteristics and by assessing the match of actual and predicted PCRs. The specific criteria include Efron’s pseudo-R-squared statistic, predicted mean, mean absolute error, and root mean squared error (RMSE).

Principal Findings. Average demographic characteristics among a hospital’s inpatients were significant predictors of PCRs for Medicare and Private insurance, but not for Medicaid or Self-Pay. One or more hospital characteristics were significantly related to every PCR category. Being a critical-access hospital was associated with strikingly higher PCRs for Medicare (.523) and Medicaid (.394), all else equal. Sole community provider status was also associated with higher PCRs for Medicare and Private stays. Conversely, rural referral status was not significantly related to any PCR. Teaching hospital status was associated with significantly greater PCRs for Medicare, Medicaid, and Private insurance holders. The Herfindahl-Hirschman Index based on 15-mile radius, a measure of hospital concentration, had a positive and significant association with Private insurance. Greater numbers of Medicare and Medicaid discharges were associated with significantly lower PCRs. Higher regional wages were associated with lower PCRs, although the relation was significant for only three PCR categories. Higher PCRs were most often associated with worse economic conditions and greater state generosity in Medicaid eligibility and spending.

 Conclusions. Inpatient encounter prices paid by Medicare, Medicaid, and private insurance can be estimated with acceptable accuracy for community hospital stays. Stays funded by other insurance types or by patients were harder to predict and simultaneously have almost no correlation with CCRs.