Published Online: June 11, 2013
Rachel Mosher Henke, PhD; Jared Lane Maeda, PhD; William D. Marder, PhD; Barry S. Friedman, PhD; and Herbert S. Wong, PhD
Objectives: To examine the influence of hospital competition on small-area inpatient resource use by payer.
Methods: We measured hospital competition and inpatient resource use using data from the 2008 Healthcare Cost and Utilization Project State Inpatient Databases. Generalized linear models adjusted for patient, population, and market characteristics were used to assess the relationship between inpatient resource use and hospital competition.
Results: Hospital competition had a similar influence on inpatient resource intensity for Medicare and privately insured patients. Hospitals in more competitive markets had significantly lower costs per discharge for both Medicare and privately insured patients. Hospital competition was not significantly associated with length of stay per discharge for either payer.
Conclusion: Findings suggest that policies or incentives that promote or encourage competition in less competitive markets may reduce variation in resource use for both Medicare and private payers.
Am J Manag Care. 2013;19(6):e238-e248
Small areas with more competitive hospital markets tend to have lower facility costs per discharge for both Medicare and privately insured patients.
By using comprehensive data that capture all patients and payers in 43 states, our results add to the evidence that market factors influence small-area healthcare resource use.
Results do not support the hypothesis that payer-specific patterns of healthcare resource variation are a product of differences in how payers determine prices
Small-area variation in healthcare utilization and intensity has been well documented.1-3 Determining the factors associated with this variation is critical to identifying whether the variation in healthcare utilization is appropriate (eg, because of differences in population needs) or whether it represents excessiveness due to exogenous factors. Identifying the sources of the variation can help identify potential policy levers to reduce costs and improve care.
Medicare fee-for-service data are often used to study small-area variations because of their availability and comprehensiveness. The few studies examining data from multiple payers have found that the pattern of variation differs by payer after controlling for population differences.4-6 Market characteristics are a possible source of the observed payer-specific variation in small areas as market factors may impact patient care differently for each payer.
In this study, we examined whether the intensity of competition among hospitals in an area had a differential effect on inpatient resource use for Medicare and privately insured patients. We hypothesized that public payers are less vulnerable to provider market power because they use administrative pricing strategies that are not directly influenced by provider negotiation. Thus, we expected that measures of hospital competition would have more influence on resource use for private payers than would be true for Medicare. This study contributes to the literature by testing this hypothesis using comprehensive inpatient data that capture all discharges at community hospitals in the majority of states in the United States.
We analyzed data from the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID)7 supplemented with other national data to determine the influence of hospital competition on inpatient resource intensity by payer, controlling for patient, population, and market factors. We estimated models for 2 outcomes: length of stay per discharge and cost per discharge.
Data and Time Frame
We created an analytic file that included all inpatient stays at community, nonrehabilitation hospitals from the 42 states that contributed data to the HCUP SID in 2008. These states were Arizona, Arkansas, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Nebraska, Nevada, New Hampshire, New Jersey, New York, North Carolina, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Vermont, Virginia, Washington, West Virginia, Wisconsin, and Wyoming.
Small areas were defined by Core Based Statistical Areas (CBSA) boundaries in this study. We used patient zip code to identify the CBSA associated with each discharge. The CBSAs represent the universe of metropolitan and micropolitan areas in the United States.8 Each CBSA includes a core area with a substantial population that together with the adjacent communities contain a high degree of economic and social integration. Core Based Statistical Areas have been used in previous studies to examine geographic variation2,6,9,10 and are desirable because population and market characteristics can be readily measured at the CBSA level using national data sets such as the Area Resource File.11 We included only those CBSAs for which 99% or more of the population reside in 1 of the states that contributed HCUP data.
We used the HCUP SID data to measure patient characteristics and used the 2 measures of inpatient resource use as outcomes in this study. We linked HCUP SID data with datafrom the Area Resource File, the American Community Survey, the American Hospital Association Annual Survey, the Centers for Medicare & Medicaid Services Hospital Compare, and the Medicaid Statistical Information System at the CBSA level to measure population and market characteristics. We used the HCUP Hospital Market Structure file7 to measure the degree of hospital competition in each area.
The Medicare sample consisted of inpatient stays among patients more than 65 years old, with Medicare assigned as the primary expected payer. This group included patients enrolled in Medicare fee-for-service and Medicare Advantage. Ideally, we would have examined these 2 groups separately, but the data did not allow us to distinguish between them. The private insurance sample consisted of inpatient stays among patients between 40 and 65 years old, with private insurance as the primary expected payer. A limitation of the analysis is that, by definition, the Medicare sample was older than the private insurance sample; thus, age could confound the results. To make the private insurance sample more clinically comparable to the Medicare sample, we excluded patients less than 40 years old and maternity stays.
We examined 2 dimensions of inpatient resource intensity as outcomes. First, we measured length of stay (LOS) per discharge. LOS was determined for each discharge by subtracting the admission date from the discharge date. Second, we measured cost per discharge. Cost was derived from total charges using the Cost-to-Charge Ratios (CCRs) developed for HCUP12 adjusted for area wage index. It is important to underscore that this study examines cost to the facility (resource use) rather than cost to the payer (price).
We measured hospital competition using the Herfindahl-Hirschman Index (HHI). The Herfindahl-Hirschman Index is the sum of the squares of market shares for all of the hospitals in the CBSA. A hospital’s market share is calculated as the number of discharges from that hospital divided by the total number of discharges from all hospitals in the market. More competitive markets have lower HHI values; less competitive markets have higher HHI values. The HHI is the standard measure of hospital competition used by the Department of Justice for antitrust enforcement13 analysis and has been validated through comparison of other competition indicators.14
We measured several patient characteristics to control for the influence of patient mix on resource intensity. Characteristics included patient age, sex, race/ethnicity, and an individual indicator for each of 18 comorbid conditions: addictive disorders, diabetes, general blood disorders, cancer, arthritis, congestive heart failure, chronic obstructive pulmonary disease, hypertension, hypothyroidism, liver disease, fluid and electrolyte disorders, other neurologic disorders, obesity, paralysis, renal failure, solid tumor without metastasis, valvular disease, and weight loss.
We measured patient severity using the All-Payer Severity-adjusted diagnostic related group (APS-DRG) system, which is appropriate for both private payer and Medicare populations and all medical/surgical conditions.15 The APS-DRG system includes LOS and charge weights that measure the predicted LOS and predicted cost of each discharge using diagnoses, procedures, discharge status, age, and LOS as inputs.
We measured population characteristics that may influence resource intensity. These measures included the percentage of adults with a bachelor’s degree or more education, the percentage of households with income below the federal poverty level, average income, income disparity (Gini coefficient), the percentage of persons 16 years or older who were unemployed, the total population of the CBSA, population density (population per square mile), and race/ethnicity (percent white, percent black, percent Hispanic, percent other).
We measured CBSA market characteristics in addition to HHI, which also may influence the area’s inpatient resource intensity. These measures included the number of acute care beds per capita, the number of long-term care beds per capita, the number of rehabilitation beds per capita, the number of primary care physicians per capita, the number of emergency department visits per capita (a proxy for primary care access), the number of physician assistants per capita, the number of specialists per capita, the percentage of acute care beds in high-technology hospitals, and the percentage of discharges from outside the hospital’s CBSA. Following previous work,16-18 we defined high-technology hospitals as those that reported having at least 6 of 8 high-technology services as reported in the American Hospital Association Survey. Finally, we included the state average Medicaid payment per beneficiary as a proxy for Medicaid payment generosity. We obtained this data from the Medicaid Statistical Information System state summary data set.
Quality of care may have a positive or negative association with inpatient resource intensity. We included 3 measures of inpatient quality from the 2008 Centers for Medicare & Medicaid Services Hospital Compare data set aggregated to the CBSA level: heart attack composite, heart failure composite, and pneumonia composite. All quality-of-care, population, and market characteristics were associated with discharges based on patient zip code.
We examined the means and distributions of each patient, population, and market characteristic at the CBSA level. We also examined the variation in inpatient resource intensity (cost per discharge and LOS per discharge) and patient characteristics by payer.
PDF is available on the last page.