Abstract
Analysts interested in physician market concentration often have access to tax identification numbers (TINs), but not the number of truly independent negotiating units (NUs). Health plans do know the true number of NUs, and, using 2014 claims data for Minnesota physicians from a large midwestern health plan, we compare Herfindahl-Hirschman Index (HHI) measures of physician market concentration using TINs versus NUs at the county and metropolitan statistical area (MSA) levels for thirteen specialties. We found that HHIs computed using TINs versus NUs were similar across Minnesota. Two MSAs in Minnesota met the Department of Justice’s definition of highly concentrated markets. There is reason to believe that the discrepancy between TIN and NU HHIs may vary by insurance product and region of the country, and so we encourage other researchers to work with health plans to replicate our study.
Keywords
I. Introduction
Market concentration and the resulting pricing power of physicians and hospitals are growing concerns. Increased pricing power results not only in higher unit prices 1 but also increases the ability of providers to resist other cost-containment efforts such tiered cost-sharing systems. 2
A central difficulty in studies of physician market concentration and prices is identifying the number of independent “bargaining” or “negotiating” units (NUs). An NU consists of physicians who negotiate with health plans as a group and independently from other physicians in the market. In theory, an NU could consist of a single physician, a large group practice, multiple practices owned by a hospital or other integrated delivery system, or a network of physician practices. The number of NUs for a given specialty in a market area represents the number of different entities providing similar services with which a health plan can negotiate prices.
Physicians generally can be identified in claims data by their National Provider Identifier (NPI) number or their tax identification number (TIN). TINs frequently have been used to group NPIs into bargaining units, 3 but physicians can negotiate prices with health plans in groups larger than their TIN. Some datasets attempt to identify NUs that are larger than TINs. For example, SK&A data 4 link individuals to larger organizations, and the American Medical Association also maintains a group practice dataset, 5 but these groupings are not designed to identify NUs.
Although researchers may not know how many independent NUs exist in a market area, health plans must negotiate prices with the providers in their networks and, thus, are aware of this information. For this study, we worked with a health plan to compare measures of physician market concentration computed using TINs versus independent NUs.
II. Measuring Market Concentration
Market concentration often is measured by the Herfindahl-Hirschman index or HHI. The HHI is computed by summing the squared market shares of all the independent producers in the market area. For example, in a market with three producers with shares equal to 10%, 50%, and 40%, the HHI would be (0.1)2 + (0.5)2 + (0.4)2 = 0.42. HHIs usually are multiplied by 10,000 for presentation, and so the HHI in the previous example would be presented as 4,200.
According to the Department of Justice (DOJ), 6 markets with HHIs between 1,500 and 2,500 are considered moderately concentrated, while HHIs greater than 2,500 indicate a “highly concentrated market.” The DOJ states, “Transactions that increase the HHI by more than 200 points in highly concentrated markets are presumed likely to enhance market power under the Horizontal merger Guidelines issued by the Department of Justice and the Federal Trade Commission.”
III. Data and Setting
We analyzed market concentration among physicians in the State of Minnesota using claims data from a large health plan that offers coverage throughout the Upper Midwest. The health plan is one of three large plans serving the Twin Cities area and the second largest health insurer in the entire state. In 2014, the data represent 144,824 enrollees, 11,079 NPIs in 1,072 practice locations, and 730,266 claims.
In many cases, claims data would not be a good source of information on physician market concentration. The health plan might contract selectively with only a few providers in the market area, or the structure of benefits might favor one set of providers over another, thus skewing the distribution of claims across providers. However, this health plan offers a broad-choice product to its commercially insured members that includes virtually every physician in Minnesota. Moreover, the plan does not use coverage differences to direct patients to one physician versus another. Thus, the data are likely to represent the market shares of providers undistorted by health plan contracting arrangements or benefit design.
Claims were counted at the “header” level, that is, one claim per visit. All inpatient and outpatient professional claims reported on CMS-1500 forms were counted 7 and grouped by specialty, where specialties are identified by the health plan’s credentialing codes. 8
The proper geographic market for health care services can be difficult to define and may vary between urban and rural areas. The smallest geographic unit in HHI assessments typically is a zip code. Zip-code-level HHIs can be weighted up to any market area, such as the county or metropolitan statistical area (MSA), using the proportion of patients from the county or MSA who originate from each zip code as weights. In rare cases, such as tertiary care centers (e.g., the University of Minnesota or the Mayo Clinic), HHIs might be reported for areas larger than an MSA. We conducted our basic analysis at the zip code level, but then aggregated the results to the county and MSA levels.
We followed Dunn and Shapiro’s 9 approach to compute HHIs. Using the most common service codes for each specialty, we computed the distance from the centroid of each zip code to each provider seen by residents of that zip code for that service. The set of clinics that captured 95% of the claims by enrollees living in that zip code became the “choice set” of clinics for residents of that zip code. We limited the analysis to Minnesota patients who were treated by Minnesota physicians.
The probability that the enrollee chose the ith location for a specific visit was modeled as [1 – (Di /Dmax)], where Di is the distance from the zip code centroid to the physician’s location, and Dmax is the maximum distance to a physician location in the 95% choice set for that zip code. If multiple physicians practiced at the same location under the same TIN or NU, we multiplied the probability of visiting that location by the number of doctors in the clinic at that location. Then we scaled the probabilities for the clinics visited by patients from each zip code to sum to 1.0. Table 1 shows a hypothetical example representing data from one zip code.
Computing Market Shares.
If the same TIN or NU operated in multiple locations, we added the market shares of clinics belonging to that TIN or NU. For example, if Clinics 1 and 2 in Table 1 were in the same NU, then the probability of visiting that NU would be .208 + .667 = .875. The HHI for the zip code is the sum of squared market shares of TINs or NUs serving that zip code. If Clinics 1 and 2 were in the same NU, then the NU-level HHI would equal (.875)2 + (.125)2 = .782 or 7,820 because HHIs usually are multiplied by 10,000.
We analyzed the data at two levels: (1) eighty-seven Minnesota counties and (2) Minnesota’s eight MSAs plus four quadrants of the state that comprise the non-MSA areas (northeast, northwest, southeast, and southwestern nonmetropolitan Minnesota). A map of the geographic areas is shown in Figure 1. The La Crosse (WI), Fargo (ND), and Grand Forks (ND) MSAs extend into Minnesota, but most of the population resides in the other states. The Twin Cities and Duluth MSAs extend into Wisconsin, but most of the population resides in Minnesota.
Minnesota MSAs.
IV. Results
A. HHIs by Specialty and Geographic Area
Table 2 shows the 2014 HHIs computed for TINs and NUs by specialty at the county level. 10 Only allergists have average county-level HHIs above 2,500, thus qualifying as a highly concentrated market according to the DOJ criteria. However, dermatologists, endocrinologists, oncologists, pediatricians, pathologists, and rheumatologists all have HHIs greater than 1,500 computed using either TINs or NUs and, thus, are moderately concentrated.
Specialty HHIs Computed Using TIN Versus NUs Averaged Across Counties (2014).
Tables 3 and 4 show the HHIs aggregated to the MSA level. Six MSAs exceed the 1,500 threshold for moderate concentration. Compared to the county-level results, cardiologists replace endocrinologists on the list of moderately concentrated markets. The La Crosse and Rochester MSAs have HHIs well above the “highly concentrated” level of 2,500 averaged across specialties.
Specialty HHIs Computed Using TIN Versus NUs Averaged Across MSAs (2014).
MSA HHIs by Tax Identification Number and Negotiating Unit (NU) Averaged Across Specialties (2014).
B. Differences Between HHIs Based on NUs Versus TINs
HHIs by specialty, averaged across geographic areas, are shown in the fourth columns of Tables 1 and 2. HHIs based on TINs and NUs are very similar and in many cases are identical.
Figure 2 compares the county-level HHIs computed using TINs versus NUs for all 1,025 combinations of counties and specialties. In 2014, there was no difference in the TIN and NU HHIs for 634 (62%) of the county-specialty combinations, and many of the discrepancies were small. The average discrepancy between TIN and NU-based HHIs at the MSA level was 8 including the zeros, with a maximum discrepancy of 1,364.
Discrepancies between MSA and TIN-based HHIs at the county level.
At the MSA level, 52% of the NU and TIN-based HHIs were identical across the 150 MSA-specialty combinations. The average difference was 22, with a maximum of 1,364. Figures 2 and 3 show the discrepancies between HHIs based on NUs versus TINs diagrammatically at the county-specialty and MSA-specialty levels in 2014. Several of the MSA outliers occur in counties that are part of MSAs, with the majority of their counties in states bordering Minnesota.
Discrepancies between MSA and TIN-based HHIs at the MSA level.
V. Discussion and Conclusions
Computation of market concentration measures is central to the study of health care organizations and the subsequent effects of concentration on prices. Researchers typically have incomplete information on the negotiating arrangements between health care providers and health plans.
We analyzed data from a Minnesota health plan to compare measures of market concentration based on TINs, which generally are available to researchers, versus independent NUs, which generally are not available to researchers. We found that the two measures generally yielded very similar results across both specialties and geographic areas in 2014.
We found evidence of “highly concentrated” markets in two MSAs. High levels of market concentration may result in increased market pricing power, allowing physicians to resist health plans’ attempts to negotiate competitive prices and install additional cost-saving measures. 11
High correspondence between TIN and NU-based HHIs could be due to several factors. First, when two provider organizations merge, the acquiring entity could incorporate the acquired entity quickly under the acquiring entity’s TIN. Second, the high correspondence could imply that the health care providers comprising negotiating units in our data are owned or employed by large systems, rather than belonging to networks that negotiate with health plans on behalf of multiple independent TINs. Third, our findings are likely to be affected by the organization of physician practices in the Upper Midwest. Using TIN-level data, Welch et al. 12 found that physicians in New England, the Upper Midwest, and the Northwest were more likely to practice in large groups than physicians in other areas of the country. In that case, the differences we find between TIN and NU HHIs in Minnesota could be lower than in other states. Finally, it could be the case that physicians in an NU have their own TINs but choose to bill under only one member’s TIN.
Researchers should be aware that physician networks and other organizations representing multiple medical practices may negotiate payment rates differently across physicians and insurance products. During the analysis, we found a large physician network that does not negotiate payment rates for some of its member physicians in the broad-choice insurance product on which our results are based. However, it does negotiate payment rates for all its member physicians for a narrow-network insurance product sold by the same health plan. Thus, our results may be conservative in the sense of understating the degree of market concentration for more restrictive health plan products.
We encourage other researchers to work with other local, regional, and national health plans to obtain information regarding providers’ TINs versus NUs. The results could improve the accuracy of research results and provide more reliable information on market concentration to policymakers.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.