Professional Fees for U.S. Hospital Care, 2016-2020.

This study used publicly available data. We identified admissions and ED visits among a large all-ages convenience sample of individuals with employer-sponsored insurance or Medicaid reported in the 2016–2020 Merative MarketScan Commercial, Medicare Supplemental, and Medicaid databases. MarketScan reports clinical diagnoses and associated payments (charges submitted by providers are not reported) to health care facilities and providers from a selection of large employers and employer-sponsored health plans (Commercial/Medicare Supplemental) or state Medicaid agencies and Medicaid-contracted health plans (Medicaid). Study outcome measures were associations between PFR and selected patient and service characteristics, adjusted mean PFR for all admissions and ED visits by payer type (Commercial or Medicaid), and adjusted mean PFR by selected diagnostic categories representing all clinical diagnoses: Major Diagnostic Category (MDC; 25 categories), Clinical Classification Software Refined (CCSR; > 500 categories), and Diagnostic Related Group (DRG; > 900 inpatient categories). Dollar values were adjusted to 2020 medical prices. 6 PFR was calculated as total admission or ED visit payment divided by facility-only payment as reported in the data source. This measure can be multiplied by facility-only costs for hospital care to yield a total cost estimate. For example, if the admission facility cost in a hospital discharge data source is $1000 and the corresponding estimated PFR for the admission clinical diagnosis is 1.240, the total estimated direct medical cost of the admission can be calculated as $1240. We combined patients’ inpatient (and preceding ED) and outpatient ED payment records and clinical information for services beginning on the same date. Where > 1 International Classification of Diseases, Tenth Revision, Clinical Modification diagnosis was reported as the admission primary diagnosis (< 0.2% of analyzed admissions), we classified the admission using the first-listed primary diagnosis. We identified the ED visit primary diagnosis based on the first-listed diagnosis to which the facility payment was attributed; ED visits with > 1 set of diagnoses with associated facility payments (< 0.6% of the potential sample) were excluded. We excluded admissions and ED visits with missing or illogical diagnostic information (ie, ICD-9-CM diagnosis codes) or illogical payment values (ie, negative or zero total payments or facility payments or total payments less than facility payments). We excluded outliers with the lowest 1% value of facility payments for the sample per hospitalized day (ie, <$396 per day for Commercial insurance and <$74 per day for Medicaid admissions) or ED visit (ie, <$31 total facility payment for Commercial and <$13 for Medicaid ED visits). Adult (≥ 18 y old) comorbidities among inpatient admissions were identified using HCUP Comorbidity Software 7 and child (< 18 y old) comorbidities were identified using the Child Comorbidity Index. 8 Surgery was identified by DRG (classified as surgical or medical) for inpatient admissions and Current Procedural Terminology codes (10021–69990) for ED visits. SAS 9.4 was used for sample selection and Stata 17 was used for regression modeling. Generalized linear regression models with log links controlling for selected patient and service characteristics were used to calculate adjusted mean PFR per admission or ED visit. Models controlled for patient age, sex, race/ethnicity (Medicaid only), health insurance plan type (eg, health maintenance organization), ED services preceding an inpatient admission, number of patient comorbidities (admissions only), whether the admission or ED visit included surgical procedures, length of inpatient stay (admissions only), discharge status (admissions only), U.S. Census region (Commercial only), and DRG (admissions models) or CCSR (ED visit models). Adjusted mean PFR per year or diagnostic classification was calculated as the mean value of the model-predicted PFR for each admission or visit (Stata “margins” program). PFR for clinical classifications with <100 admissions or ED visits was not calculated. Machine-readable PFR estimates by payer type for all analyzed clinical classifications are reported in Supplemental Digital Content 1 ( http://links.lww.com/MLR/C695 ).

Analysis included 6.7 million Commercial admissions, 4.2 million Medicaid payer admissions, 22.2 million Commercial ED visits, and 17.7 million Medicaid payer ED visits ( Fig. 1 ). Higher patient age, non-White race/ethnicity, and longer inpatient stay were associated with lower PFR, as was female sex—except among Medicaid payer ED visits ( Supplemental Digital Content 2 , http://links.lww.com/MLR/C696 ). Commercial comprehensive health plans (ie, no incentive for patients to use particular providers) were generally associated with lower PFR. Medicaid health maintenance organization and preferred provider organization plans were associated with higher PFR, and Medicaid point of service with capitation plans had a mixed relationship with PFR (ie, associated with lower PFR for admissions and higher ED visits). Admissions with preceding ED care, a higher number of patient comorbidities, and non-home inpatient discharge destination were associated with higher PFR. ED visits with surgical procedures were associated with higher PFR for Commercial visits but lower PFR for Medicaid payer ED visits. Hospitals in the Northeast were associated with higher PFR for Commercial admissions and ED visits compared with hospitals in the West, lower PFR compared with hospitals in the South, and a mixed relationship (lower for admissions, higher for ED visits) compared with hospitals in the North Central region. Adjusted mean PFR for 2016–2020 admissions was 1.224 for Commercial admissions and 1.178 for Medicaid admissions, indicating professional payments on average increased total payments by 22.4% and 17.8%, respectively, above facility-only payments ( Table 1 ). This is a 9% and 3% decline in average PFR, respectively, compared with 2004 estimates (1.342 and 1.211). 5 Adjusted mean PFR for ED visits during 2016–2020 was 1.283 for Commercial and 1.415 for Medicaid visits. This is a 12% and 5% decline in average PFR, respectively, compared with 2004 estimates (1.452 and 1.490). PFR was highest by MDC for admissions with MDC 14 “pregnancy, childbirth, and the puerperium” (Commercial PFR: 1.485; Medicaid: 1.391) and lowest for Commercial admissions with MDC 20 “alcohol or drug use or induced organic mental disorders” (PFR: 1.062) and Medicaid admissions with MDC 17 “myeloproliferative diseases and disorders, poorly differentiated neoplasms” (PFR: 1.090) ( Table 2 ). PFR was highest by MDC for Commercial ED visits with MDC 9 “diseases and disorders of the skin, subcutaneous tissue, and breast” (PFR: 1.367) and Medicaid ED visits with MDC 3 “diseases and disorders of the ear, nose, mouth, and throat” (PFR: 1.496) and lowest for MDC 17 “myeloproliferative diseases and disorders, poorly differentiated neoplasms” (commercial PFR: 1.102; Medicaid PFR: 1.241). PFR was highest by CCSR for Commercial ED visits with CCSR NEO066 “malignant neuroendocrine tumors” (PFR: 1.467) and Medicaid ED visits with CCSR END013 “pituitary disorders” (PFR: 1.789) ( Table 3 ). PFR was highest by DRG for Commercial admissions with DRG 583 “mastectomy for malignancy without complication or comorbidity/major complication or comorbidity (PFR: 1.803) and Medicaid admissions with 785 “cesarean section with sterilization without complication or comorbidity/major complication or comorbidity” (PFR: 1.575) ( Table 4 ).

In this study, we updated estimates of the amount by which facility-only financial data reported in hospital discharge data sources can underestimate the full cost of medical care patients receive during hospital admissions and ED visits by excluding professional fees. Financial information in this study’s analyzed data source facilitated diagnosis-specific PFR estimates, adjusted for multiple patient and service factors, and the PFR estimates reported here are designed to be directly applied to hospital discharge data sources for cost of illness analysis. This study’s results suggest that professional fees comprised a declining proportion of hospital-based care costs over approximately the last 2 decades. These results are consistent with our previous PFR investigation of 2004–2012 data years, 5 a subsequent similar analysis of 2007–2014 data years by other researchers using a different data source, 9 and analyses of specific hospital-based services and diagnoses. 10 , 11 Another report on aggregate health care expenditures using sources such as the National Health Expenditure Accounts has pointed to overall spending increases during the same period for both inpatient and professional services, but these topics were not investigated in the manner presented here; that is, this study examined professional fees specific to hospital-based care. 12 , 13 This study had several limitations. Investigation into why PFRs changed over the study period is beyond the scope of this study. Different hospital prices for similar services, financial incentives to improve physician quality, and efforts to improve hospital price transparency and comparability for consumers and health care payers are the subject of direct investigation in other studies. 14 – 16 MarketScan Commercial data are not nationally representative of the population with employer-sponsored insurance nor Medicare coverage and the MarketScan Medicaid sample included a limited number of states. U.S. Census region is a crude indicator of geographic differences in health care costs; we lacked consistent data to further control for geographic variation, such as urban/rural location. Although our previous 2004–2012 PFR estimates did not include Commercial patients age older than 65 years (ie, those with Medicare supplemental plans), a separate analysis for the present study restricted to age 0–64 patients was not materially different compared with the all-age estimates. This study controlled for observable patient and insurance characteristics, including health plan type, which addressed patients enrolled in managed care plans. However, this study could not control for provider characteristics, such as physician specialty, and hospital facility characteristics, such as ownership, organization, and geographic location, which influence health care costs. 2 , 17 – 20 Hospitals’ costs vary widely by service type; for example, maternity services—a frequent cause for inpatient admission—are known outliers 21 ; therefore, PFR estimates by clinical classification (comprehensively reported in Supplemental Digital Content 1 , http://links.lww.com/MLR/C695 ) may be most relevant for some health services research questions. This study estimated PFR per admission and ED visit based on payments that hospitals and physicians received for medical services, whereas hospital charges typically reported in hospital discharge data sources multiplied by CCR provide an estimate of hospitals’ costs to provide services. Both approaches yield recognized estimates of medical costs, but this means that PFR estimates are not precisely complementary to facility cost estimates from hospital discharge data. This issue might be mitigated given that CCR can be a reasonable proxy for price (or payments)-to-charge ratios, which are more directly analogous to the PFR estimates presented here. Despite what might be modest differences in the nature of financial data underlying our PFR estimates versus that underlying hospital discharge data, we propose that our approach offers a reasonable option for improving cost estimates from hospital discharge data by accounting for professional fees. By comparing hospital-based professional versus facility fees over time, it seems that professional fees comprised a declining proportion of hospital-based care costs during approximately the past 2 decades. Still, adjustments for professional fees remain an important analytic step when hospital facility-only financial data are used to estimate health care costs. The PFR estimates generated in this study offer an opportunity to address the systematic and substantial underestimation of health care service costs using facility-only costs reported in hospital discharge data.