Aditi P. Sen, PhD

 Self-insured employers cover more people than Medicare, Medicaid, or direct purchasers of private insurance.This study examined the ability of self-insured employers to negotiate hospital prices and the relationship between hospital prices and employer market power in the United States.

 Repeated cross-section analysis of commercial claims.

 We used the US Census Bureau County Business Patterns data to estimate employer market power at the metropolitan statistical area (MSA)–year level and used the Truven Health MarketScan commercial claims to estimate mean hospital prices and price ratios at the MSA-year level (2010-2016). We calculated descriptive statistics for employer market power, mean hospitalization prices, and a case mix–adjusted price ratio measure during the study period and analyzed the 10 most concentrated labor markets. We estimated MSA-year–level ordinary least squares regressions of hospitalization price and the price ratio measure on employer market power.

 Large self-insured employers had concentrated market power in very few MSAs in 2016. The mean value of our employer market power measure was 62 for 2016, compared with the mean value of 5410 for hospital market power in the United States. Regression analyses find a slight relationship: A 1-point increase in employer market power was associated with a $6.61 decrease in the hospitalization price (mean = $20,813), but this result becomes statistically insignificant once the models control for hospital wages.

 Employer market power is low in most MSAs. Self-insured employers may consider building purchase alliances with state and local government employee groups to enhance their market power and lower negotiated prices for hospital services.

In most areas of the United States, self-insured employers lack market power to negotiate hospital prices.

There is no evidence that greater employer market power is associated with lower hospital prices in the employer-sponsored insurance markets.

Employer coalitions would need to enroll a considerable number of employers to have a sufficient market power to negotiate with hospitals.

Self-insured employers may consider building purchase alliances with state and local government employee groups to enhance their market power and lower negotiated prices for hospital services.

fty-five percent of Americans younger than 65 years receive their health insurance from their employers.

 Recently, hospital prices paid by employers have increased much faster than inflation and outpaced what public payers pay.

 The price differential paid by employers for hospital services increased from 10% above the Medicare rate to more than 100% above the Medicare rate in many metropolitan statistical areas (MSAs) between 2005 and 2019.

 If the differential between public- and private-sector prices continues to widen, it could continue to cause slower or negative effective wage growth for US workers, whose total compensation increases have been absorbed by more expensive health benefits.

 meaning that employers bear financial risk for their health plans. Previous research in this area has focused on how insurer concentration could lead to lower private-sector hospital prices.

 However, most insurers are paid through administrative services only (ASO) contracts, do not assume risk, and may not have a direct incentive for lower prices.

 Insurers may use their entire book of business and compete for potential clients (ie, self-insured employers) by offering them low-premium plans by obtaining low health care prices.

 However, insurers can also compete for employers based on administrative fees instead of prices, as prices are often not shared with employers during the bidding process. Furthermore, ASO contracts are sometimes paid as a percentage of overall claims,

 which would actually create a negative incentive to negotiate lower prices, or on a per-member per-month basis, which would attenuate the incentive.

Empirically, this has been borne out in the data. A recent National Bureau of Economic Research working paper by Craig et al has examined within-payer price variation at hospitals in Massachusetts,

 finding that those enrolled in ASO plans pay higher prices for the same services at the same hospital than those enrolled in fully insured plans administered by the same payer. This empirical evidence suggests that insurers, because they lack the incentive, may not be negotiating lower prices for their ASO enrollees. Although insurers benefit from having self-insured employers in their book of business,

 they may not want to alienate hospitals who are among the largest employers in the community and important potential customers.

The growing differential between Medicare and private-sector prices shows that the private health care payment system is not keeping pace with consolidation in the hospital sector.

 In response to this growing differential, some large self-insured employers—for example, Berkshire Hathaway, JPMorgan Chase, and Amazon—have created an entity to address the cost of their health care benefits, and some large self-insured employers have begun to directly negotiate prices with hospitals.

 However, the success of these efforts is likely to be a function of employers’ negotiating power in health care markets. On one hand, employers may solicit bids from a variety of insurers and work with consultants to choose insurers that pay lower prices. If this is the case, insurers should work to achieve low prices and employer negotiating power should not matter. On the other hand, large employers with significant market power may have some leverage against hospital systems. If ASO insurers are not negotiating lower prices, then an employer’s market power could function as a counterforce to provider and insurer market power (eg, the dominant hospital system may want to serve the dominant employer). These large employers may be able to use their market power in negotiations by engaging in direct contracting with providers, demanding that insurers negotiate lower prices, or to remove gag clauses that prevent the employers from seeing the prices up front. There is some evidence of this occurring both in the private sector

 The extent to which this is happening across the country in areas where employers do have market power is unknown, a priori.

In this study, we examined the extent of employer market power in the context of hospital price negotiation and explored whether employers can use their market power to obtain lower prices for hospitalizations. We used Truven MarketScan data to examine inpatient hospital prices across MSAs and constructed a measure of employer market power adapted from the labor economics literature.

 We examined the level of employer market power across MSAs in the United States, identifying markets with particularly high and low levels of employer market power, and examined the association between employer market power and hospital prices.

We used the US Census Bureau County Business Patterns (CBP) data to estimate employer market power at the MSA level for each year between 2010 and 2016.

 The CBP data provide annual statistics for all businesses with paid employees within the United States. At a county level, the CBP data provide estimates of the number of firms with the following ranges of employees: 1 to 4, 5 to 9, 10 to 19, 20 to 49, 50 to 99, 100 to 249, 250 to 499, 500 to 999, 1000 to 1499, 1500 to 2499, 2500 to 4999, and 5000 or more.

We used the Truven Health MarketScan commercial claims and annual enrollment data files, sourced from large employers, for the years 2010-2016 to estimate MSA-year–level prices.

 The Truven MarketScan data include information from 350 insurers offering private-sector employer-sponsored coverage to approximately 25% of all the commercially insured individuals and their families in the United States. This database has been used to estimate local-level hospital prices in prior research.

Additional MSA-year–level characteristics were obtained from the Current Population Survey (CPS), the American Hospital Association (AHA) Annual Survey, and the American Medical Association (AMA) Competition in Health Insurance annual reports. Demographic data and unemployment rates were obtained from CPS at the county level and were aggregated to the MSA-year level. Hospital concentration measures were obtained from AHA at the county level and aggregated to the MSA-year level. Insurer concentration data were available from the AMA reports at an MSA-year level. To adjust hospital prices for local wages in the hospital setting, we used the CMS wage index, a standardized measure of local hospital wages relative to national hospital wages.

 The CMS wage index is used to adjust prospective payments to hospitals based on local hospital wages and defines labor market areas by Core-Based Statistical Areas, which align with our MSA-year–level analyses.

Our key measure was employer market power. The employer market power measure was defined as the sum of the squared market shares in the labor market for each employer (similar to the Herfindahl-Hirschman Index [HHI]),

 following the literature in labor economics.

 In the labor economics literature, each industry is evaluated separately; however, we aggregated industries for our analysis because the industry is not relevant with respect to hospital prices and negotiations. To later match to the hospital price data, we constructed the employer market power measure at the MSA level. Because labor markets are generally smaller than an MSA, we constructed the employer market power measure at the county level and then took a weighted mean up to the MSA level. Our process followed 3 steps.

First, we estimated the number of employees within a county by assuming that each employer within a range (eg, 5-9) employs the midpoint of that range (eg, 7.5) and then summed the total employees across firms within a county. For employers reporting 5000 or more employees, we estimated the number of employees by subtracting the estimated number of employees in all the smaller categories from the total number of employees in the county, and then divided by the number of employers with 5000 or more employees.

Second, we took the estimates derived above and calculated a county-level measure of employer market power. A value close to 0 indicates a highly competitive labor market and a value of 10,000 indicates a perfectly monopsonistic employer.

Third, we took the weighted mean of these county-level measures at the MSA level, weighted by the employers’ market shares in that MSA, to calculate each MSA’s employer market concentration.

Our dependent variable was hospitalization price. We constructed 2 hospitalization price measures using the Truven Health MarketScan commercial claims data for inpatient admissions. First, we computed the mean inpatient hospitalization price for each MSA-year. To account for differences in case mix, we also computed a price ratio measure that controls for differences across markets. Specifically, the expected price in each MSA-year was calculated by multiplying the national prices for each diagnosis-related group (DRG) code by the number of cases for each DRG in that MSA. This price ratio measure accounts for differences in case mix that may otherwise drive differences in prices across markets by adjusting for the local frequency of DRG codes.

We calculated descriptive statistics, mapped the distribution of employer market power across the country, and analyzed the 10 most concentrated labor markets, focusing on the largest employer in each market. To analyze the descriptive relationship between inpatient prices and employer market power, we estimated MSA-year–level ordinary least squares regressions of hospital prices on employer market power. We estimated these models with year fixed effects to account for time trends, then added market-level controls, and finally added the CMS wage index to the regressions to examine the relative contribution of wages. We estimated models separately with a linear specification of employer market power and with the linear specification plus quartiles of employer market power. The nonlinear specifications are incorporated to test the hypothesis that employer market power is strongest at the highest levels of the distribution, as has been found in prior literature on employer monopsony power and wages.

 We also estimated log-transformed models as a sensitivity check.

We performed 3 key robustness checks. First, to account for the potential endogeneity of employer market power, we estimated an instrumental variable (IV) model where we instrumented MSA-level employer market power with the employer market power of the nearest MSA, identifying the nearest MSA based on the geographic distance between the centroids of MSAs.

 Second, because employers may have more bargaining power in more competitive hospital markets, we estimated a series of models where we interacted employer market power with hospital HHI. Third, markets with significant employer market power may also have a health care organization as the largest employer, potentially diminishing the association between employer market power and hospital prices. To examine this, we estimated models excluding all 819 (33.0%) MSA-years in which a health care organization was the largest employer.

, we show descriptive statistics related to the key measures of employer market power and prices over time. In 2016, the average employer market was very unconcentrated, with a mean (SD) of 61.59 (70.88) and median of 46.27. Overall, labor markets were becoming less concentrated over time, with a mean (SD) of 63.70 (63.37) in 2010 and 61.59 (70.88) in 2016.

The MSA-level mean (SD) hospitalization price in 2016 was $23,911 ($6789), with a median (interquartile range) of $22,377 ($19,777-$26,323), consistent with previous MSA-level estimates.33 The mean hospitalization price increased from $17,898 in 2010 to $23,911 in 2016. The price ratio measure, by nature, has a mean close to 1 (mean [SD], 0.98 [0.28] in 2016). In 2016, the price ratio measure ranged from 0.57 to 2.56 (ie, hospitalization prices 43% less than the national mean to 156% higher than the national mean, controlling for case mix).

, we show the geographic distribution of employer market power across the United States in 2016. Larger, more populous MSAs tended to have less concentrated markets, and smaller, less populous MSAs tended to have more concentrated markets.

 displays the MSAs with the 10 highest values of employer market power in 2016, along with select characteristics. Markets with relatively high levels of employer market power were generally smaller cities. The largest employers in the more concentrated MSAs spanned a range of industries but generally fell into 3 categories: large universities (Ithaca, New York; Blacksburg-Christiansburg-Radford, Virginia; Morgantown, West Virginia; and Lafayette-West Lafayette, Indiana); health systems (Rochester, Minnesota; Madera, California; Weirton-Steubenville, West Virginia-Ohio; and Winchester, Virginia-West Virginia); and manufacturing (Sioux City, Iowa-Nebraska-South Dakota; and Kokomo, Indiana). These concentrated MSAs generally had hospital prices lower than the mean but had a wide range: from $15,319 in Weirton-Steubenville, West Virginia-Ohio, to $30,140 in Morgantown, West Virginia. Adjusting the prices for the case mix in each MSA, the price ratio measure followed a similar trend. Detailed sources and methods underlying this table can be found in 

, the relationship between hospital prices and employer market power is examined. Panel A of Table 3 presents the relationship with a linear specification of employer market power. Columns 1 through 3 model the mean hospitalization price and employer market power, whereas columns 4 through 6 model the price ratio measure. With and without market-level controls, there was a statistically significant relationship between employer market power and hospitalization prices. Adjusting for market controls, including sex, race/ethnicity, education level, unemployment, insurer HHI, and hospital HHI, a 1-SD increase in employer market power (ie, a 65-point increase) was associated with a $430 lower mean hospitalization price. Accounting for differences in differences in case mix, a 1-SD increase in employer market power was associated with 2.1% lower prices. This relationship was substantially weaker and no longer statistically significant after controlling for the CMS hospital wage index, which has an extremely strong relationship with prices. Panel B of Table 3 shows the relationship between hospital prices and employer market power, with a nonlinear specification of employer market power that includes the continuous measure along with a categorical variable for the quartile of employer market power. We find qualitatively similar results. As a sensitivity check, we reestimated our models with log-transformed outcomes (

). The results did not qualitatively change.

, we present results from the IV model. In the first stage, we found a strong 

 statistic of 12.34. The second-stage results were not qualitatively different from the main results presented in Table 3.

 displays results interacting employer market power and hospital HHI. We found no statistically significant interaction between these 2 measures, with point estimates around 0.

, we examined how our results may change when excluding MSA-years in which a health care organization was the largest employer. Although the overall results were unchanged, the association between employer market power and hospital prices was slightly larger, yet still not statistically different from 0 once we controlled for hospital wages.

Our results showed that large, self-insured employers did not have concentrated market power in most MSAs. The mean value of employer market power was 62 for 2016, compared with the mean value of 5410 for hospital market power in the United States.

 Our regression analyses find a slight (but economically insignificant) negative relationship between employer market power and hospital prices when we estimate models without hospital wage controls, and this result becomes statistically insignificant once the models control for hospital wages. Given these 2 alternate specifications, we do not find evidence that employers are using bargaining power to negotiate lower hospital prices.

It has been suggested that employers should take an active role in negotiating prices with hospitals, rather than relying on ASO insurers to do so.

 Our study suggests that almost all employers, operating alone, simply do not have the market power to impose a threat of effective negotiation. Given that health insurers have not been effective in negotiating hospital prices, alternative approaches may be needed to constrain hospital prices for self-insured employers.

Instead of directly negotiating prices, some large self-insured employers have adopted other approaches to contain their health care spending, such as directing patients toward centers of excellence,

 improving incentives for in-network care, offering high-deductible health plans in combination with personal medical accounts,

 and changing drug formularies to encourage generic drug usage.

 However, the effectiveness of these tactics can be limited by local market hospital concentration, which forces employers to accept the prices that hospitals charge.

Creating purchasing alliances is another strategy for employers to increase their purchasing power.

 To be an effective negotiator, such purchasing alliances need to have substantial market power in their geographic area. In most MSAs, achieving substantial market power would require that many large employers would all agree on the same approach. Achieving this level of collective action may be a challenge; in many cities, the combined market share of major employers is relatively small. One approach that may hold promise is creating purchase alliances of private and public employee groups,

 which can take a more active role in negotiating with hospitals using their combined large market share.

This study had several important limitations. First, our analysis was not causal. Instead, we focused on the descriptive nature of employer market power and the descriptive relationship between employer market power and hospital prices. Second, we estimated our analysis at the MSA level, as the MarketScan data do not have finer geography. Although some prior studies have measured employer market power at the commuting zone–industry level,

 we believe that an MSA-level analysis is appropriate as it has been used in prior studies of hospital prices

 and because the employer’s total market share, not their market share within a geographic market-industry combination, should be key in price negotiations. Third, we were not able to observe which employers were specifically included in the MarketScan data. Although the MarketScan data are sourced from large employers, there could be markets in which the largest employer is not included. This may bias the descriptive relationship between employer market power and inpatient prices toward a null result of no relationship. However, in 7 of the 10 most concentrated markets, we found that MarketScan enrollment substantially exceeded the estimated number of employees at the largest employer, enhancing the credibility of our estimates.

This study examined inpatient hospital prices and employer market power across the United States between 2010 and 2016. Employer market power was low in most MSAs, a potential explanation for why employers historically have not taken a more active role in price negotiation. Employer coalitions would need to enroll a considerable number of employers to have a sufficient market power to negotiate with hospitals. Self-insured employers may consider building purchase alliances with state and local government employee groups to enhance their market power and negotiate lower prices for hospital services.

 Department of Health Policy and Management, Bloomberg School of Public Health (MDE, MKM, GB, APS, GA), and Carey Business School (GB), Johns Hopkins University, Baltimore, MD.

 This project was supported by Arnold Ventures and grant number T32HS000029 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (MDE, MKM, GB, APS, GA); acquisition of data (MDE, MKM, GA); analysis and interpretation of data (MDE, MKM, APS); drafting of the manuscript (MDE, MKM, GA); critical revision of the manuscript for important intellectual content (MDE, MKM, GB, APS); statistical analysis (MKM); obtaining funding (GA); administrative, technical, or logistic support (GB); and supervision (GA).

 Matthew D. Eisenberg, PhD, Bloomberg School of Public Health, Johns Hopkins University, 624 N Broadway, Rm 406, Baltimore, MD 21205. Email: eisenberg@jhu.edu.

1. Berchick ER, Barnett JC, Upton RD. Health insurance coverage in the United States: 2018. United States Census Bureau. November 2019. Accessed January 10, 2020. https://www.census.gov/content/dam/Census/library/publications/2019/demo/p60-267.pdf

2. NHE fact sheet. CMS. Updated December 16, 2020. Accessed December 20, 2020. https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/nationalhealthexpenddata/nhe-fact-sheet.html

3. Whaley CM, Briscombe B, Kerber R, O’Neill B, Kofner A. Nationwide evaluation of health care prices paid by private health plans: findings from round 3 of an employer-led transparency initiative. RAND Corp. 2020. Accessed September 22, 2020. https://www.rand.org/pubs/research_reports/RR4394.html

4. Anderson GF, Hussey P, Petrosyan V. It’s still the prices, stupid: why the US spends so much on health care, and a tribute to Uwe Reinhardt. 

. 2019;38(1):87-95. doi:10.1377/hlthaff.2018.05144

5. Bai G, Anderson GF. Market power: price variation among commercial insurers for hospital services. 

. 2018;37(10):1615-1622. doi:10.1377/hlthaff.2018.0567

6. 2017 Employer Health Benefits Survey: section 10: plan funding. Kaiser Family Foundation. September 19, 2017. Accessed August 22, 2019. https://www.kff.org/report-section/ehbs-2017-section-10-plan-funding/

7. Trish EE, Herring BJ. How do health insurer market concentration and bargaining power with hospitals affect health insurance premiums? 

. 2015;42:104-114. doi:10.1016/j.jhealeco.2015.03.009

8. Barrette E, Gowrisankaran G, Town R. Countervailing market power and hospital competition. National Bureau of Economic Research working paper No. 27005. April 2020. Updated September 2020. Accessed November 1, 2020. http://www.nber.org/papers/w27005

9. Sachdev G, White C, Bai G. Self-insured employers are using price transparency to improve contracting with health care providers: the Indiana experience. 

. October 7, 2019. Accessed November 1, 2020. https://www.healthaffairs.org/do/10.1377/hblog20191003.778513/full/

10. Jeng HW. Third party administrator (TPA) service pricing and incentive contracts. 

. 1996;4:209-227. http://digitalcommons.unl.edu/joap/112

11. Craig SV, Ericson KM, Starc A. How important is price variation between health insurers? National Bureau of Economic Research working paper No. 25190. October 2018. Accessed May 27, 2020. http://www.nber.org/papers/w25190

12. Moriya AS, Vogt WB, Gaynor M. Hospital prices and market structure in the hospital and insurance industries. 

. 2010;5(4):459-479. doi:10.1017/S1744133110000083

13. Ho K, Lee RS. Insurer competition in health care markets. 

. 2017;85(2):379-417. doi:10.3982/ECTA13570

14. McKellar MR, Naimer S, Landrum MB, Gibson TB, Chandra A, Chernew M. Insurer market structure and variation in commercial health care spending. 

. 2014;49(3):878-892. doi:10.1111/1475-6773.12131

15. Dauda S. Hospital and health insurance markets concentration and inpatient hospital transaction prices in the U.S. health care market. 

. 2018;53(2):1203-1226. doi:10.1111/1475-6773.12706

16. Koller CF, Khullar D. The commercial differential for hospital prices: responses from states and employers. 

. 2019;322(8):723-724. doi:10.1001/jama.2019.9275

17. White C, Whaley CM. Prices paid to hospitals by private health plans are high relative to Medicare and vary widely. RAND Corp. 2019. Accessed August 26, 2019. https://www.rand.org/pubs/research_reports/RR3033.html

18. Henry Ford Health System launches “direct to employer” healthcare contract with General Motors. News release. Henry Ford Health System. August 6, 2018. Accessed January 10, 2020. https://www.henryford.com/news/2018/08/direct-to-employer-announcement

19. Navarro NV. “It was a crisis for our community”: this nonprofit is working to lower the sky-high health care costs in Summit County. CPR News. March 9, 2020. Accessed November 1, 2020. https://www.cpr.org/2020/03/09/it-was-a-crisis-for-our-community-this-nonprofit-is-working-to-lower-the-sky-high-health-care-costs-in-summit-county/

20. Autor D, Dorn D, Katz LF, Patterson C, Van Reenen J. The fall of the labor share and the rise of superstar firms. 

. 2020;135(2):645-709. doi:10.1093/qje/qjaa004

21. Azar J, Marinescu I, Steinbaum MI. Labor market concentration. National Bureau of Economic Research working paper No. 24147. December 2017. Updated February 2019. Accessed August 22, 2019. http://www.nber.org/papers/w24147

22. Azar JA, Marinescu I, Steinbaum MI, Taska B. Concentration in US labor markets: evidence from online vacancy data. 

. 2020;66. doi:10.1016/j.labeco.2020.101886

23. Benmelech E, Bergman N, Kim H. Strong employers and weak employees: how does employer concentration affect wages? National Bureau of Economic Research working paper No. 24307. February 2018. Accessed August 22, 2019. http://www.nber.org/papers/w24307

24. CBP datasets. United States Census Bureau. Accessed August 22, 2019. https://www.census.gov/programs-surveys/cbp/data/datasets.html

25. IBM MarketScan research databases. Accessed June 2, 2021. https://www.ibm.com/products/marketscan-research-databases?cm_sp=Scheduler-_-CopyChng2-_-C

26. Neprash HT, Chernew ME, Hicks AL, Gibson T, McWilliams JM. Association of financial integration between physicians and hospitals with commercial health care prices. 

. 2015;175(12):1932-1939. doi:10.1001/jamainternmed.2015.4610

27. Neprash HT, Wallace J, Chernew ME, McWilliams JM. Measuring prices in health care markets using commercial claims data. 

. 2015;50(6):2037-2047. doi:10.1111/1475-6773.12304

28. Chen JL, Hicks AL, Chernew ME. Prices for physician services in Medicare Advantage versus traditional Medicare. 

29. Wage index files. CMS. Accessed May 27, 2020. https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Wage-Index-Files

30. Herfindahl-Hirschman Index. US Department of Justice. Updated July 21, 2018. Accessed December 4, 2019. https://www.justice.gov/atr/herfindahl-hirschman-index

31. Nevo A. Measuring market power in the ready-to-eat cereal industry. 

32. Hausman J. Valuation of new goods under perfect and imperfect competition. In: Bresnahan T, Gordon R. 

National Bureau of Economic Research; 1996:207-248. Accessed November 20, 2020. https://econpapers.repec.org/bookchap/nbrnberch/6068.htm

33. Maeda JL, Nelson L. An analysis of private-sector prices for hospital admissions: working paper 2017-02. Congressional Budget Office. April 4, 2017. Accessed January 10, 2020. https://www.cbo.gov/publication/52567

34. Cooper Z, Craig SV, Gaynor M, Van Reenen J. The price ain’t right? hospital prices and health spending on the privately insured. National Bureau of Economic Research working paper No. 21815. December 2015. Updated May 2018. Accessed September 3, 2019. http://www.nber.org/papers/w21815

35. Robinson JC, MacPherson K. Payers test reference pricing and centers of excellence to steer patients to low-price and high-quality providers. 

. 2012;31(9):2028-2036. doi:10.1377/hlthaff.2011.1313

36. Haviland AM, Eisenberg MD, Mehrotra A, Huckfeldt PJ, Sood N. Do “consumer-directed” health plans bend the cost curve over time? 

. 2016;46:33-51. doi:10.1016/j.jhealeco.2016.01.001

37. Goldberg SE, Fragala MS, Wohlgemuth JG. Self-insured employer health benefits strategy established a negative cost trend while improving performance. 

. 2019;22(6):547-554. doi:10.1089/pop.2018.0184

38. Blumenthal D, Gustafsson L, Bishop S. To control health care costs, U.S. employers should form purchasing alliances. 

. November 2, 2018. Accessed August 22, 2019. https://hbr.org/2018/11/to-control-health-care-costs-u-s-employers-should-form-purchasing-alliances

39. Hansard S. Insurance payments tied to Medicare intrigue states, employers. 

. August 22, 2019. Accessed August 26, 2019. https://news.bloomberglaw.com/health-law-and-business/insurance-payments-tied-to-medicare-intrigue-states-employers

40. Newman D, Parente ST, Barrette E, Kennedy K. Prices for common medical services vary substantially among the commercially insured. 

. 2016;35(5):923-927. doi:10.1377/hlthaff.2015.1379

41. Fulton BD. Health care market concentration trends in the United States: evidence and policy responses. 

. 2017;36(9):1530-1538. doi:10.1377/hlthaff.2017.0556

42. Scheffler RM, Arnold DR. Insurer market power lowers prices in numerous concentrated provider markets. 

. 2017;36(9):1539-1546. doi:10.1377/hlthaff.2017.0552

Articles

This study examines the ability of self-insured employers to negotiate hospital prices and the relationship between hospital prices and employer market power in the United States.

Large Self-insured Employers Lack Power to Effectively Negotiate Hospital Prices

 Medicare Part B payment methods incentivize the use of more expensive injectable and infused drugs. We examined prescribing patterns in the context of intravenous (IV) iron, for which multiple similarly safe and efficacious formulations exist, with wide variations in price.

 We conducted a retrospective cohort analysis of IV iron utilization and payment in the Medicare population between 2015 and 2017.

 This analysis used a national, random 20% sample of Medicare fee-for-service beneficiaries with Part B claims for IV iron between January 2015 and December 2017—a period before, during, and after a national shortage of iron dextran. This sample included 66,710 Medicare fee-for-service beneficiaries with at least 1 Part B claim for IV iron.

 The greatest increase in utilization occurred in the most expensive iron formulation, ferric carboxymaltose; its market share rose from 27.4% of use in 2015 to 47.7% in 2017. The use of a less expensive formulation, iron dextran, decreased from 26.7% to 18.7% over the same period. An alternative payment model in Maryland hospitals was associated with markedly less utilization of ferric carboxymaltose, accounting for 4.7% of IV iron utilization in Maryland hospitals.

 There was an increase in the dispensing of a higher-priced IV iron formulation associated with a shortage of a less expensive drug that persisted once the shortage ended. These findings in IV iron have broader implications for Part B drug payment policy because the price of the drug determines the physician and health system payment.

When there was flux in the intravenous iron market, higher-cost formulations were adopted, potentially due to financial incentives in Medicare Part B to use higher-priced drugs.

The external shock of a shortage in iron dextran led to shifts in market share, which did not return to baseline after resolution of the shortage.

The market share of ferric carboxymaltose increased markedly, despite being one of the most expensive formulations.

A novel hospital payment system in Maryland avoided the increase in utilization of higher-cost formulations of intravenous iron.

Current Part B payment policies promote the use of more expensive drugs.

For drugs covered under Medicare Part B (physician-administered injectable and infused drugs), payment is based on the average sales price (ASP) for the prescription drug plus a percentage add-on payment.

 Since the Medicare Modernization Act of 2003, the physician or hospital-owned infusion clinic administering the drug is paid the ASP plus 6%, which was reduced to 4.3% under the Budget Control Act of 2011. Although it was intended to cover variability in the purchase price of physician-administered drugs and costs to obtain and store the drugs, this additional payment has been criticized for incentivizing the prescription of more expensive formulations.

There is little research on the impact of payment incentives on clinical practices or the associated cost implications, particularly when multiple alternative drugs are available. Understanding the implications of financial incentives under Part B should inform ongoing policy debates around Part B drug payment policy. During periods of shifting market share due to an external shock (eg, drug shortages, new entrants into the market, expanded label indications), when clinical practice patterns change, payment incentives may be particularly influential.

Drug shortages have occurred with increasing frequency since 2001 and are related to multiple weather, industrial, and corporate factors.

 Drug shortages cause providers to seek alternative medications, resulting in new prescribing patterns. However, it is unclear if drug shortages influence physicians to reevaluate financial incentives favoring higher-cost drugs.

Iron deficiency anemia is the most common cause of anemia worldwide. Patients with poor tolerance or poor absorption of oral iron supplementation require intravenous (IV) iron.

 Five types of IV iron are available in the United States: ferric carboxymaltose, low-molecular-weight iron dextran, ferric gluconate, iron sucrose, and ferumoxytol.

 These formulations are similarly safe and efficacious.

 High-molecular-weight iron dextran was associated with potentially serious infusion toxicity and was removed from the market in the United States in 2009 (“iron dextran” throughout this article refers to low-molecular-weight iron dextran).

Based on Medicare Part B Drug and Biological ASP data, these formulations have markedly different prices (

 Iron dextran is among the lower-cost alternatives, with a January 2016 ASP for iron dextran of $243.80 per 1000 mg compared with $1057.00 per 1000 mg for ferric carboxymaltose.

 A 4.3% add-on payment in January 2016 would have been $10.48 per 1000 mg of iron dextran and $45.45 for 1000 mg of ferric carboxymaltose. Ferric carboxymaltose comes in 750-mg vials, with a cost of $792.75 based upon January 2016 ASP.

 Due to those cost differences and variations in dosage, median payment for the infusion of iron dextran was $73 (including ASP-based payment plus the add-on payment) compared with $782 for ferric carboxymaltose (

]). Practices and hospital outpatient departments (HOPDs) also receive a separate payment to administer IV iron based on the duration of infusion.

A national shortage of iron dextran was identified by the American Society of Health-System Pharmacists (ASHP) from January to May 2016.

 Both the FDA and the ASHP maintain records regarding drug shortages. Shortages recognized by the ASHP represent a spectrum of scarcity, from absolute lack of supply to restricted ability for individual purchasers to consistently acquire new orders.

 The iron dextran shortage from January to May 2016 falls into this latter category of shortage; drug utilization declined with an identified shortage, but supplies did not run out entirely.

We used the case of IV iron to investigate the broader implications of payment policy on drug utilization, particularly when alternatives with different financial incentives are available.

We sought to answer 4 questions using IV iron as a case study: First, what IV iron market share changes occurred between 2015 and 2017? Second, what were the spending implications of any changes in short- and long-term prescribing patterns by physicians? Third, was product selection occurring at the practice group/HOPD level or at the beneficiary level? Finally, to increase our confidence in assessing the role of financial incentives (under the ASP-based payment system) in physician prescribing, we compared utilization trends in Maryland, which implemented an all-payer global budget program for acute care hospitals in 2014, with national trends.

 Under the Maryland global budget, hospitals do not face the same financial incentives embedded in the ASP-based payment system; instead, they are incentivized to choose the least expensive of clinically effective alternatives.

Our primary data source for this study was a national, random 20% sample of Medicare fee-for-service (FFS) beneficiaries. We examined their claims data from the period of January 1, 2015, through December 31, 2017. We included claims from the Outpatient and Carrier files with Healthcare Common Procedure Coding System (HCPCS) codes for ferric carboxymaltose (J1439), iron dextran (J1750), iron sucrose (J1756), ferric gluconate (J2916), and ferumoxytol (Q0138 and Q0139). This resulted in a total of 1,713,068 Medicare claims. Consistent with the Medicare drug spending dashboard, we excluded claims where Medicare was not the primary payer (3754 claims) and claims associated with a critical access hospital (11,133 claims). In our national analysis, we also excluded claims where total spending was $0 (11,366 claims), claims where the IV iron infusion HCPCS code(s) had $0 in payment (1,432,483 claims), and those billed in Maryland (used as a comparison group to examine differences in payment policy in the state) (9420 claims). Claims where IV iron HCPCS codes had $0 spending were related to bundled payments, such as for dialysis services, and were excluded given differences in payment structure and incentives. Our primary analysis included 244,912 claims, covering 66,710 unique Medicare beneficiaries and 16,572 unique providers (based on unique National Provider Identifier numbers).

We first examined the characteristics of beneficiaries receiving different iron formulations during the study period. We then analyzed utilization (measured in grams) and spending by HCPCS code for each iron formulation by month. We scaled estimates from the 20% sample by a factor of 5 to approximate total utilization and spending for all Medicare FFS beneficiaries. Two formulations, iron dextran and ferric carboxymaltose, were included in a detailed analysis because they accounted for most changes in market share during the study period. Changes in prescribing patterns for iron dextran and ferric carboxymaltose were tested for statistical significance using 2-sided 

 tests to compare average claim spending on the formulations in each of the 3 years before, during, and after a shortage of iron dextran (2015, 2016, and 2017).

We estimated spending implications for the Medicare program due to changing prescribing patterns. We calculated the estimated spending based on total units of all IV iron formulations administered from 2015 to 2017 if market share across formulations had remained constant. Using mean ASP plus 4.3% for each quarter, we calculated expected spending and compared this with actual spending to identify excess spending within the Medicare program attributable to changes in formulation market share.

We then examined the percentage of practice groups or HOPDs where a single IV iron formulation was used. We restricted this analysis to practice groups and HOPDs that had at least 5 claims for IV iron in the 20% sample of FFS claims during the beginning and ending periods of the study (January-June 2015 and July-December 2017).

Finally, utilization from the state of Maryland was analyzed separately given the unique payment system for hospital-based services.

 We compared utilization by IV iron formulation in HOPDs vs physician offices because regulated spaces such as HOPDs in Maryland were subject to the global budget whereas physician offices were not.

This study used a retrospective, deidentified, national data set and was exempt from institutional review board review.

In the 20% sample, 66,710 Medicare beneficiaries, after taking into account the exclusion criteria, had at least 1 paid claim for IV iron during the study period. Beneficiaries who received IV iron were more likely to be women than men and had a mean age of 74 years, similar to the overall Medicare Part B population (

). There were no clinically significant differences in mean Charlson Comorbidity Index score between beneficiaries prescribed iron dextran and ferric carboxymaltose (Table 2).

 shows utilization of the various IV iron formulations over the study period, January 2015 to December 2017. In 2015, an average of 14,358 g per month was reported for all formulations of IV iron. In 2016, overall utilization of IV iron increased to 16,772 g per month. In 2017, utilization of IV iron continued to increase to an average of 18,600 g per month. Corresponding with the increase in units administered during the study period, there was an increase in the number of beneficiaries receiving IV iron (

The market share of different formulations of IV iron shifted during the study period. Ferric carboxymaltose, iron dextran, and ferumoxytol had the greatest use during the study period, accounting for 235,512 g (39.5%), 127,276 g (21.3%), and 168,468 g (28.2%), respectively. Iron sucrose represented 55,427 g (9.3%) of claims, whereas ferric gluconate claims totaled 10,071 g (1.7%).

Despite being one of the less costly formulations, iron dextran utilization fell both absolutely and as a percentage of claims. In 2015, an average of 3839 g of iron dextran was administered each month (26.7% of utilization). This utilization dropped to an average of 3293 g per month (19.6% of utilization) in 2016. Finally, absolute usage rose slightly in 2017 to an average of 3474 g per month of iron dextran, but usage as a percentage of total IV iron continued to decline (18.7% of utilization).

Ferric carboxymaltose experienced the greatest increase in utilization during the study period. Use of ferric carboxymaltose increased from an average of 3938 g per month (27.4% of utilization) in 2015, to 6807 g per month (40.6% of utilization) in 2016, to 8880 g per month (47.7% of utilization) in 2017. Use of the other IV iron formulations stayed relatively consistent over the study period, with the exception of ferumoxytol, which had stable monthly utilization but a decline in market share (Figure 1).

Spending on IV iron over this period rose with the overall increase in utilization and the shift toward more expensive formulations (Figure 1). Total observed IV iron spending was $432.7 million from 2015 to 2017. Had IV iron formulation market share remained stable from the beginning of the study period, the total cost for the units of iron used in this sample would have been $393.3 million. Thus, we estimate that shifting prescribing patterns during the study period led to an additional $39.4 million in Medicare Part B spending from 2015 to 2017.

 tests confirm that the increase in ferric carboxymaltose spending over time is statistically significant. When comparing all IV iron claims in 2016 vs 2015, there was an average per-claim increase in spending on ferric carboxymaltose of $90 (95% CI, $85-$94; 

 tests also confirm that there was a statistically significant drop in iron dextran spending in 2016 relative to 2015. The average per-claim spending on iron dextran decreased by $7 (95% CI, $6-$8; 

 < .001). In 2017, spending on iron dextran increased but did not return to earlier levels despite resolution of the shortage. Spending on ferric carboxymaltose continued to rise in 2017 but at a slower rate (

). These increases in per-claim spending for ferric carboxymaltose reflect the increased utilization during this period.

Single-Product Utilization by Practice Groups/Hospitals

To understand the decision-making approach that providers used to select IV iron formulations, we examined the percentage of practice groups and hospitals utilizing only 1 formulation of IV iron (

). From January to June 2015, 78.1% of hospitals and 66.7% of practice groups administered only 1 formulation, decreasing to 52.7% of hospitals and 56.1% of practice groups by July to December 2017. Decreasing use of iron dextran and ferumoxytol accounted for most of the decrease in single-formulation utilization. Single-formulation utilization for the 2 most expensive agents, ferric carboxymaltose and ferumoxytol, were more common in hospital-based systems than independent practice groups (with 45.0% of hospitals and 21.9% of groups exclusively administering 1 of the 2 most expensive agents from July-December 2017).

Effect of Removing Cost-Based Incentives on IV Iron Utilization in a Novel Payment System in Maryland

Because Maryland hospitals are not subject to the same cost-based financial incentives in Part B, we compared utilization of IV iron formulations in Maryland HOPDs with utilization in the rest of the country. In HOPDs, which were subject to a global budget in Maryland, utilization trends in Maryland were distinct from the non-Maryland Medicare Part B population. Compared with other states over the entire period, HOPDs in Maryland were much more likely to use the less-costly iron dextran than the more-costly ferric carboxymaltose (

). Specifically, ferric carboxymaltose accounted for 4.7% of all iron units in Maryland HOPDs compared with 42.7% of the non-Maryland HOPD sample. In physician offices, where payment incentives were the same in Maryland as in other states, utilization trends were similar between the Maryland and non-Maryland providers (Figure 3).

This study has 4 main findings. First, we found that iron dextran’s market share fell during the study period while a more expensive formulation, ferric carboxymaltose, gained in market share. The external shock of the iron dextran shortage catalyzed rapid shifts in market share; however, prescribing did not return to baseline after resolution of the shortage. Second, we found that the shift in prescribing patterns led to an increase in Medicare spending on IV iron by $39.4 million. We also found that more than half of hospitals and practice groups administered only 1 formulation of IV iron and that hospital systems were more likely than practice groups to utilize the most expensive agents.

Finally, we found lower utilization of higher-cost ferric carboxymaltose in HOPDs in Maryland compared with the rest of the country, consistent with the financial incentives that Maryland hospitals face under global budgets. This finding supports the hypothesis that the financial incentives in ASP-based payment promote the use of higher-cost Part B drugs.

This study has several limitations. First, this analysis is limited to the Medicare FFS population, and prescribing patterns in this population may not be generalizable. Second, IV iron is often included in bundled payments, as with end-stage renal disease, and these outpatient infusions are not reflected in the current analysis.

During a period of rapid fluctuation in the intravenous iron market, there was increased use of a more expensive drug that afforded providers additional revenue.

Aditi P. Sen, PhD

Articles

Large Self-insured Employers Lack Power to Effectively Negotiate Hospital Prices

Payment Incentives and the Use of Higher-Cost Drugs: A Retrospective Cohort Analysis of Intravenous Iron in the Medicare Population

Comparing Patient OOP Spending for Specialty Drugs in Medicare Part D and Employer-Sponsored Insurance

Innovative Payment to Scale Up Access to Medications for Opioid Use Disorder