Removing a Constraint on Hospital Utilization: A Natural Experiment in Maryland
Published Online: June 27, 2014
Noah S. Kalman, MD; Bradley G. Hammill, MS; Robert B. Murray, MA, MBA; and Kevin A. Schulman, MD
CMS designed inpatient diagnosis-related groupbased payments and outpatient payments to cover the cost of an average admission. Such costs include the variable or direct patient care costs (nurse staffing, medications, etc) as well as fixed or nonpatient care costs (building operation and maintenance, administrative staff, etc).1 Although direct costs increase for every admission, fixed nonpatient care costs are less sensitive to changes in patient volume. The marginal profitability of every additional patient encounter within this payment structure incentivizes hospitals to increase service volume.2 Studies have examined the effect of overall increases and decreases in payment rates on hospital behavior. However, it is unclear whether directly addressing the marginal profitability of increasing patient volume influences hospital behavior.
Maryland has a unique hospital payment regulation scheme. Through a CMS waiver, the Maryland Health Services Cost Review Commission (HSCRC) sets inpatient and outpatient hospital payment rates paid by Medicare, Medicaid, and all commercial payers in Maryland. The Maryland system embodies the same incentives as Medicare’s inpatient prospective payment system: the HSCRC sets case-rate constraints based on diagnosis-related groups with the goal of incentivizing hospital efficiency and quality.3
Unlike Medicare, the HSCRC attempted to limit growth in hospital costs by making a distinction between payments for baseline patient volume and payments based on changes in patient volume. Under the HSCRC payment adjustment policy in the 1990s, payers reimbursed hospitals 85% of the case rate for additional patient volume above the previous year’s volume to more closely approximate the variable, direct costs of patient care (although it did not precisely calculate fixed or marginal costs of care in the assessment). In the case of a hospital with 110 admissions in a given year compared with 100 admissions in the previous year, the Commission would direct payers to provide 100% of the case rate for the first 100 patients but to provide only 85% of the case rate for the additional 10 admissions. (Conversely, the HSCRC would compensate hospitals with 15% of the payment rate for reductions in patient volume to contribute to fixed costs.) This policy also applied to hospital outpatient services. By reducing the profitability of incremental patient volume, the formula was expected to reduce the incentive to increase utilization at the hospital level.
Beginning in 2001, the HSCRC eliminated this payment adjustment component of the reimbursement scheme. Hospitals subsequently received full case rates for incremental increases in patient volume, identical to the current Medicare system. This policy change should have provided an incentive for hospitals to increase patient volume, because incremental volume would now be reimbursed at 100% of the case rate. In this paper, we explore trends in hospital utilization and finances in Maryland before and after this payment change. We hypothesized that hospital revenues, costs, and profits would increase following the payment adjustment repeal.
We used data from the HSCRC Disclosure of Hospital Financial and Statistical Information, an annual report that compiles mandatory reporting data from Maryland hospitals. We obtained hospital-level financial and statistical information for fiscal years 1991 through 2008.4
The study population consisted of hospitals in Maryland that reported uninterrupted data to the HSCRC between fiscal years 1991 and 2008. One additional hospital, which opened in fiscal year 1994, was included. Hospitals that closed or merged with other hospitals in this time period were excluded from the analysis. Hospitals that joined hospital systems within the state were included if they retained a unique hospital identifier. Four hospitals that had been subject to a form of global capitation in the 1990s, rather than case-based constraints with the 85% payment adjustment policy, were excluded.
The primary outcomes of interest were hospital-level trends in annual inpatient admissions, outpatient equivalent volume, equivalent admissions, operating revenue, operating costs, and operating profit (Table 1). In a subgroup analysis, we divided hospitals into highoccupancy (≥55% occupancy in fiscal year 2000) and lowoccupancy (<55% occupancy in fiscal year 2000) categories at the time of the policy change to examine whether capacity constraints were associated with hospitals’ responses to the new incentive.
Each hospital’s time series was interrupted by the repeal of the payment adjustment policy effective July 1, 2000 (ie, the beginning of fiscal year 2001). We analyzed these interrupted time-series data using a segmented regression analysis which assumed that hospitals adjusted their behavior in the first year following the regulatory change. To estimate the level and slope of the trend line both before and after the policy change, we included fiscal year, an indicator for the postrepeal period, and an interaction between these 2 variables as explanatory variables in each model.
We estimated the statistical models using hierarchical generalized linear model methods. Due to the skewed nature of the time-series values across hospitals, we specified a log link with gamma distributed errors to estimate the model effects on an exponential growth, or relative, scale.5 This approach enabled comparisons among hospitals on the same scale, regardless of time-series quantity. Use of hierarchical methods was necessary for simultaneous analysis of time-series data across multiple hospitals. Specifically, we allowed for random variation by hospital around the intercept and around each explanatory parameter. This approach had the effect of accounting for the autocorrelation between time-series values within each hospital over time. It also resulted in conditional, or within-hospital, estimates of effect. We report the parameter estimates for each explanatory variable for the average hospital. For subgroup analyses, we estimated the regression models after adding a group indicator and fully interacting that indicator with the other variables described above.
PDF is available on the last page.