Dear Mr. Spears:
This letter responds to your request on behalf of your clients, the Pacific Business Group on Health (“PBGH”), the California Public Employees’ Retirement System (“CalPERS”), and the California Health Care Coalition (“CHCC”), for the issuance of a business review letter pursuant to the Department of Justice’s Business Review Procedure, 28 C.F.R. § 50.6. Your request proposes a data exchange program for hospital services called the Hospital Value Initiative (“HVI”). You request a statement of the Department of Justice’s present enforcement intentions regarding the HVI. For the reasons discussed below, the Department has no present intention to challenge the formation or operation of the HVI.
I. HVI Representations
You made the following representations in the letter (and the attachments thereto) that you and Anne B. Clairborne sent to Christine A. Varney, Assistant Attorney General, Antitrust Division, on June 19, 2009 (hereinafter “Letter”) (1):
A. The HVI Organizers and Objective
The HVI is sponsored by three broad-based associations that represent various public and private group purchasers of health care services: PBGH, CalPERS, and CHCC. (Letter at 2) (2). These organizations, located in California, comprise a large number of companies and other entities purchasing health care services that act on behalf of more than seven million people. (Letter at 3-4). You assert that HVI, as a complement to ongoing hospital quality measurement initiatives, seeks to improve transparency in the California health care marketplace, particularly with respect to the cost and quality of hospital services. (Letter at 2).
The HVI will (1) analyze the claims data that major third-party health plans (hereinafter “payors”) receive from hospitals; (2) develop index scores from the data that will allow for comparison of the relative cost and resource utilization efficiency of hospitals in California; and (3) distribute these index scores to hospitals, payors, and group purchasers of health care services. (Letter at 8-9, 12). You assert that distributing the index scores will allow payors and group purchasers of hospital services to make more informed purchasing decisions and spur competition among hospitals to operate more efficiently and to provide better quality services at lower costs. (Letter at 7-8).
B. Operation of the HVI
The HVI will collect and aggregate claims data from five payors that have agreed in principle to participate in the program. (Letter at 4-5). The data will be generated by the approximately 330 hospitals that are expected to participate in the HVI data exchange. (Letter at 5). These five payors, which collectively account for approximately 67% of all HMO and PPO health insurance enrollment in California, are: (1) California Physicians Service, d/b/a Blue Shield of California; (2) Health Net of California, Inc. and Health Net Life Insurance Company; (3) United HealthCare Services, Inc. and its affiliates, PacifiCare Life and Health Insurance Company and PacifiCare of California; (4) Wellpoint/Blue Cross of California, Inc.; and (5) Aetna Health of California, Inc. and Aetna Life Insurance Company. (Letter at 4-5). The approximately 330 participating hospitals represent more than 70% of the hospitals in the state.
Our analysis is predicated on the accuracy of your projections of the number of payor and hospital participants in the HVI. Therefore, our analysis could change if the actual number of participants proved to be materially lower than projected.
After each calendar year, Milliman, Inc., an independent consultant retained by the HVI, will collect from the participating payors all the inpatient and outpatient hospital services claims data for the preceding year for services provided to their insureds by each participating hospital. (Letter at 8, 12). The collected claims data will include a hospital’s “Allowed Amount” for a given service, which is the total amount that a hospital receives from a payor and its insured (through deductibles or coinsurance) for providing a service. (Letter at 8).
You anticipate that the data compiled by Milliman for a given year will contain more than one million claims, covering every hospital service provided by every participating hospital to all the insureds enrolled with the five participating payors. (Letter at 8-9). Milliman will aggregate and analyze the submitted claims data to develop two types of index scores that will allow comparisons of hospitals’ costs and relative efficiency: (1) the “Buyer Cost Index” or “BCI”; and (2) the “Resource-Use Efficiency ” score or “RUE.” (Letter at 6). Due to the time requirements needed to ensure that the entire year’s claims data have been collected and analyzed, all of the data that Milliman will use to compute the BCI and RUE scores will be between ten and twenty-two months old when the HVI distributes these scores. (Letter at 12).
The BCI is intended to allow payors, group purchasers, and hospitals to determine how a given hospital’s charges for providing a specified service compare to the average charges for providing the same service at all of the California hospitals participating in the HVI. (Letter at 9).
Milliman will calculate two core types of BCI scores, an All-Payor BCI and a Payor-Specific BCI, for each participating hospital. (Letter at 9, n. 5). For the All-Payor BCI, the numerator will be the Allowed Amounts that all five participating payors paid to a specific hospital for all discharges for a specific type of service. (Letter, Appendix C at 1; see Letter at 9). The denominator will be the Allowed Amounts that would have been paid to that hospital for those discharges if it were paid the average rate paid by all payors to all participating hospitals. (Letter, Appendix C at 1; see Letter at 9, n.5). For example, if the All-Payor BCI statistic were 1.2 for a hypothetical hospital, Hospital A, for APR-DRG 540 (Cesarean deliveries) (3), it would reflect that over the previous year, Hospital A received a 20% higher reimbursement for APR-DRG 540 than the average amount paid by all five participating payors to all of the participating hospitals in California that provide Cesarean deliveries.
Milliman will calculate the Payor-Specific BCI using only one payor’s claims data. (Letter at 14). The numerator will be the Allowed Amount that a payor paid a hospital for discharges that received a specific type of service; the denominator will be how much that payor would have paid the hospital if it paid the average Allowed Amount that it paid all of the participating hospitals. (Letter, Appendix C at 7; see Letter at 9, n. 5). Significantly, as discussed further below, a payor will receive only its own Payor-Specific BCI score; it will not receive the Payor-Specific score for any other payor. (Letter at 14). Furthermore, each payor already possesses all of the data needed to calculate its own Payor-Specific BCI. (Letter, Appendix C at 7).
Milliman will calculate both the All-Payor and Payor-Specific BCI for inpatient services at three different levels of granularity: (1) each APR-DRG; (4) (2) each Major Diagnostic Category (“MDC”); (5) and (3) an aggregate of all of a hospital’s APR-DRGs. (Letter at 10, Appendix C at Attachment A).
Milliman will use a different classification system to calculate BCI scores for outpatient services than for inpatient services. For outpatient services, Milliman will apply its proprietary Resource Based Relative Value System (“RBRVS”) methodology to establish statewide benchmarks for each of the approximately 14,000 procedures contained in the Healthcare Common Procedure Coding System. (Letter at 10). In contrast to inpatient services, where Milliman will report BCI scores for hundreds of procedures, Milliman will use the RBRVS to report aggregated BCI scores for only five broad categories: Emergency Room, Outpatient Surgery, Radiology, Lab/Pathology, and Other Outpatient Services. (Letter at 10-11).
Milliman will calculate RUE scores for inpatient services. The purpose of these scores is to identify hospitals’ resource utilization levels on a per “bed-day” basis for given types of procedures. (Letter at 6, 11). To calculate RUE scores, Milliman will substitute standardized cost amounts for specific bed-day types (e.g., Maternity, ICU, CCU, NICU beds) for the Allowed Amount of a specified service at any given hospital. Id. This substitution methodology will hold dollar costs constant to enable users to compare the number and type of bed-days that a hospital uses to provide a specified service to patients to the benchmark average amount of resources used by all participating hospitals to provide the same service. (Letter at 11). As with the BCI score, the RUE score will be expressed as a ratio. The numerator for the RUE will be the number of bed days used by that hospital for each procedure that the hospital performed for all five participating payors. Id. The denominator will be the number of bed days for that procedure that the hospital would have used if it used the average number of bed days for that procedure used by all participating hospitals for all five payors. Id.
C. Distribution of Reports
Milliman, through a secure website, will distribute reports that contain the BCI and RUE scores, to the hospitals, payors, and group purchasers participating in the HVI. As you have represented orally to us, and as described below, these recipients will not all receive or have access to the same reports. (Letter at 12).
Each hospital will receive a Hospital-Specific Report, containing BCI and RUE scores that will compare that hospital’s costs and use of resources for treating specified conditions to the average treatment costs of all participating payors at all participating hospitals. (Letter at 13, Appendix C at Attachment B). Specifically, each hospital will receive the All-Payor BCI and RUE scores (described above) for each of its own procedures. (Letter at 13-14). You have represented in your letter and orally that no hospital will receive the Hospital-Specific Report of any other individual hospital nor any Payor-Specific Report. (Letter at 14). The Hospital-Specific Report will contain the number of discharges for each score reported. (Letter at 13).
The Hospital Specific Report will also include a Regional BCI score, which is the average of the BCI scores for all hospitals within the region of California in which the hospital is located. (Letter at 13). The HVI has designated 15 HVI Regions; the HVI Region with the fewest number of participating hospitals is San Francisco, which is expected to have seven participating hospitals. (Letter, Appendix B).
Each payor will receive two types of reports. First, payors will receive an All-Payor Report, which will present the BCI and RUE scores for each participating hospital, based on the claims data of all payors. (Letter at 13, Appendix C at Attachment A). The All-Payor Report will present data in two formats that will be identical for all five payors: (1) a Hospital-Specific Summary; and (2) a Hospital-Comparison Summary. (Letter at 12-13). In the Hospital-Specific Summary, a payor will be able to specify the hospital that it wants the report to display. (Letter at 13). The Hospital-Comparison Summary will allow for the comparison of index scores among hospitals. (Letter at 12-13). There is no difference between the Hospital-Specific Summary and the Hospital-Comparison Summary components in terms of how the statistics are calculated, only in the way the statistics are presented. Id.
If a payor that participated in the HVI owned a participating hospital, a concern could arise about that hospital’s potential access to the data of other hospitals contained in the payor’s All-Payor Report. You have addressed this concern, however, through your representations that no participating payor currently owns a hospital and that if such ownership by any participating payor arises in the future, that payor would no longer be eligible to receive an All-Payor Report.
Second, payors will also receive a Payor-Specific Report. (Letter at 14). The Payor-Specific Report does not provide any information to the payor that the payor did not already have because the report contains BCI and RUE scores based only on that payor’s data. (Letter,
3. Group Purchasers
The group purchasers participating in the HVI are the employers, public agencies, unions and health and welfare funds that are members of the three coalition organizations (PBGH, CalPERS and CHCC) that have sponsored the HVI. These group purchasers will receive the All-Payor Report but not any Payor-Specific Report. (Letter at 13-14).
The following data is excerpted from a sample report that the HVI represents accurately depicts the substance of the BCI and RUE scores that Milliman would disseminate through its All-Payor Hospital Comparison Summary. (Letter at 12).
Excerpts from HVI’s Sample All-Payor Hospital Comparison Report
Hospital A; Line of Business: PPO
(Letter, Appendix C at Attachment A).
D. Other Representations
You represent that, in developing their proposal the sponsors of the HVI have taken into account the objectives of the key principles and purposes of the “Safety Zone” for data exchanges among health care providers set forth in Statement 6 of the Department of Justice’s and Federal Trade Commission’s Statements of Antitrust Enforcement Policy in Health Care (1996) (“Health Care Statements”). (6) (Letter at 16). You also represent that all payors and group purchasers participating in the HVI will sign commitments to conform their participation in the HVI with the antitrust laws, and that the HVI will implement an antitrust compliance program. (Letter at 8, n. 3).
II. Legal Analysis
Based on the information that you have provided, the Antitrust Division concludes that the HVI’s operation is not likely to produce any anticompetitive effects. The central inquiry in analyzing a data sharing arrangement is whether it is likely to facilitate express or tacit collusion resulting in increased prices or reduced quantity or quality, or otherwise reduce competition among the recipients of the data. See Health Care Statements at 6 A. For the reasons discussed below, the Department believes that it is unlikely that the HVI would reduce competition. Rather, the most likely effect of the HVI is that greater information about the relative costs and utilization rates of hospitals in California will lead payors and employers to make more informed decisions when purchasing hospital services.
The Department’s conclusion that the HVI is not likely to produce anticompetitive effects is based on two related analytical steps. First, the Department has analyzed whether the HVI meets the requirements of the Safety Zone in Statement 6 of the Health Care Statements (7). For the reasons outlined below, the Department believes that the HVI is unlikely to satisfy all of the Safety Zone criteria under all circumstances. Second, after further examining the proposal and applying standard antitrust analysis to determine whether the HVI is likely to reduce competition, the Department has concluded that it is not likely to do so.
A. Health Care Statements
The Health Care Statements provide that, absent extraordinary circumstances, the Antitrust Division will not challenge provider participation in written surveys of prices for health care services if the following conditions are met:
The HVI meets the first two criteria of the Safety Zone because: (1) the HVI will use a third-party (Milliman) to collect, aggregate, and analyze claims data; and (2) when the HVI distributes its reports, the claims data upon which they will be based will be between ten and twenty-two months old. However, the HVI probably will not entirely satisfy the third criterion of the Safety Zone, because (1) it is likely that for some All-Payor BCI scores an individual payor’s data will account for more than 25%, on a weighted basis, of either the numerator or the denominator of that statistic, or both; and (2) it is possible that the numerator of some All-Payor BCI scores might not include the data of five payors.
B. Competitive Effects Analysis
Because the HVI does not satisfy all of the Statement 6 Safety Zone criteria, the Department has analyzed the other factors that directly bear on whether the HVI is likely to produce anticompetitive effects. As Statement 6 provides:
After careful analysis of the data that the HVI plans to collect and the form of the statistics that it plans to disseminate, the Department has concluded that the HVI is unlikely to produce anticompetitive effects for several reasons. First, no participating hospital, payor or group purchaser will have access to the raw claims data submitted by any other participant or to any of the other disaggregated data that is collected. (Letter at 8). Second, the HVI’s reports will not disclose the prices that any participating hospital charges for its services. (Letter at 2). Third, and most importantly, it is unlikely that any recipient of the HVI reports will be able to “reverse engineer” the statistics to determine the rates paid by any particular payor to any hospital or the rates charged by any hospital to any particular payor. The most likely mechanism by which a recipient could attempt to reverse engineer the data is to compare its own BCI statistics to the market-wide BCI scores and try to “back-out” the prices paid by a competing payor to a particular hospital. As explained below, it is unlikely that the data will permit this type of analysis.
The HVI’s proposed distribution of the RUE scores does not present any risk of producing anticompetitive effects because the RUE scores are not based on the actual pricing data of any participating hospital and therefore there is no possibility that the scores could be reverse engineered to derive hospital-specific service prices. As previously stated, to calculate RUE scores, Milliman will substitute standardized cost amounts for specific bed-day types (e.g., Maternity, ICU, CCU, NICU beds) for the Allowed Amount of a specified service at any given hospital. (Letter at 6, 11-12). This substitution methodology means that the data used to calculate the RUE scores will not correspond to the actual prices that a hospital charges for a given service.
Each participating hospital will receive its own BCI score and a Regional BCI score that is the average of the BCIs for all hospitals in their HVI Region. Because the denominator of its own BCI is calculated from data from all participating hospitals, it is unlikely that a particular hospital would be able to reverse engineer its competitors' prices from its own BCI. Additionally, it is unlikely that a hospital could learn any substantial information to facilitate collusion from comparing its BCI to the Regional BCI. Even in an unlikely extreme circumstance, such as if there were only two hospitals in an HVI Region for a particular APR-DRG, a hospital would only learn if it is paid more or less than a competing hospital is paid on average by all payors, but it would not learn how its reimbursement from one specific payor compares with the reimbursement that a competitor hospital received from that payor. In short, the complexity of the data used to calculate the BCI statistics and the high degree of aggregation of the statistics make it unlikely that a hospital would learn precise enough information about a competitor to facilitate collusion.
Payors will receive BCI scores for each participating hospital. It is unlikely that payors will be able to reverse engineer the All-Payor BCI to obtain data about a competitor payor because the denominator uses data from all participating payors and hospitals statewide, a huge data set which is not known to any one payor. Additionally, it is unlikely that a payor can learn anything from comparing the Payor-Specific BCI to the All-Payor BCI for a particular hospital, even in an unlikely extreme circumstance, such as if only two payors paid a hospital for a particular APR-DRG, because the two BCI statistics are based on different denominators. This makes a comparison uninformative because differences between these two BCIs are not necessarily due to a difference in the Allowed Amount that was paid to that particular hospital by the two payors, but could also be due to other factors, including possible differences in the levels of one or both payors’ reimbursements to other hospitals, which are included in the denominator of the All-Payor BCI.
Additionally, HVI’s goal of making relevant information available to purchasers of hospital services is procompetitive in nature because it should contribute to more informed purchasing decisions and encourage competition among hospitals. To the extent that the HVI allows payors, employers and other group purchasers of hospital services to better evaluate differences in the costs and resources that hospitals use to treat comparable conditions, it could enhance competition and efficiency in hospital markets and permit more informed decisions to be made in choosing between alternative hospitals. However, the Department has not engaged in any “balancing” of potential anticompetitive and procompetitive effects because, as explained above, the Department has concluded that it is unlikely that the HVI will produce any anticompetitive effects.
For all the foregoing reasons, based on the representations made in your request, the Department concludes that the HVI is unlikely to produce anticompetitive effects. Accordingly, the Department has no present intention to challenge the formation or operation of the HVI. This letter expresses the Department’s current enforcement intentions and is predicated on the accuracy of the information and assertions that you have presented to us, both in your June 19, 2009 letter and in your oral representations, including your projected number of payor and hospital participants in the HVI. In accordance with its normal practice, the Department reserves the right to bring an enforcement action in the future if the actions of the HVI or its participants produce anticompetitive effects in any market.
This statement is made in accordance with the Department’s Business Review Procedure, 28 C.F.R. §50.6. Pursuant to its terms, your business review request and this letter will be made publicly available immediately, and any supporting data will be made publicly available within thirty (30) days of the date of this letter, unless you request that any part of the material be withheld in accordance with Paragraph 10©) of the Business Review Procedure.
1. By its terms your June 19 letter (as clarified by your September 10, 2009 letter), with Appendices A-G, supercedes in its entirety the original business review proposal, as submitted on November 17, 2007, and subsequently revised. (Letter at 1, n. 1). You also have made oral representations to Department officials, as needed to clarify representations made in your June 19 letter. The Department's assessment of your proposal is predicated on the accuracy of both your written and oral representations.
2. PBGH is an association of some of the nation’s largest public and private group purchasers of health care services, representing more than 3 million people. CalPERS purchases health care services for more than 2,500 employers and approximately 1.2 million public employees, retirees, and their families. CHCC is a coalition of forty-two employers, unions, health and welfare funds (Taft-Hartley Trusts) and local government agencies, representing approximately 3 million Californians, that seeks to improve access to affordable, quality health care for its members. (Letter at 3-4).
3. “APR-DRG” means All Patient Refined-Diagnosis Related Group.
4. APR-DRGs are components of a diagnosis-based classification system that differs from the DRG classification system that is used by the Medicare program. (Letter at 9, n. 6). The approximately 500 Medicare DRGs were developed to establish a systematic means of classifying differences in the level of reimbursement and resources used by hospitals to treat various medical conditions. (Letter, Appendix C at 3). APR-DRGs, however, were developed by 3M Health Information Systems to classify medical conditions based on differences in patients’ severity of illness and risk of mortality. Id. In analyzing claims data for each of the approximately 350 APR-DRGs, the HVI will apply a “Severity of Illness” ("SOI") analysis to break each APR-DRG down into several severity levels, resulting in a total of more than 1,200 APR-DRG/SOI categories. (Letter at 9-10). Such analysis will allow the HVI to calculate more precisely the differences in treatment costs among hospitals for a given APR-DRG because its calculations will control for patient mix within an APR-DRG. Id. However, the HVI will report the BCI scores only at the APR-DRG level, by taking a weighted average of the costs at the more detailed APR-DRG/SOI level. (Letter at 9, n. 7).
5. Each of the 25 MDC categories corresponds to a single organ system; they are created to aggregate APR-DRGs in a clinically coherent manner. (Letter at 14-15).
6. Available at http://www.usdoj.gov/atr/public/guidelines/1791.htm.
7. Statement 6 covers exchanges of information among health care providers and does not refer to payor participation in data exchanges, while your data exchange proposal involves participation by both payors and hospitals. Nevertheless, the Department views Statement 6's framework for evaluating exchanges among providers as relevant for assessing the competitive significance of both payors’ and hospitals’ participation in the HVI because there is a significant substantive overlap in the antitrust issues in the two types of data exchanges.