The No Surprises Act (NSA) created a final-offer arbitration process called the independent dispute resolution (IDR) process to settle disagreements about what insurers must pay providers for certain out-of-network services. On June 13, the Centers for Medicare and Medicaid Services (CMS) released a second tranche of detailed data on outcomes under the IDR process. This tranche encompassed disputes resolved during the third and fourth quarters of 2023. CMS released data for the first half of 2023 earlier this year.
This analysis uses the new data to update our prior analysis of IDR outcomes, which examined disputes resolved in the first half of 2023, to examine the second half of 2023. In doing so, we largely follow the methods we used previously, with some exceptions described in the appendix, and we refer readers to our prior analysis for a full description of our methods. In brief, however, we examine disputes involving three broad categories of professional services—emergency care, imaging, and neonatal/pediatric critical care—and examine the same 49 service codes falling within those three categories that we examined previously. These services accounted for 72% of the claim line items adjudicated in IDR during the second half of 2023, similar to the share they accounted for during the first half of the year.
Following our prior analysis, we express the prices emerging from IDR as a multiple of the prices Medicare would pay for the same services. We do so for three reasons. First, this approach offers readers a familiar unit of measurement, as provider-insurer contracts for physician services commonly specify prices as a multiple of Medicare’s, and researchers commonly report estimates in this form. Second, it offers a simple and tractable way to compare the prices emerging from IDR to pre-NSA prices. Namely, we compare our estimates to previously published Medicare ratios for similar (though not identical) collections of services, which we adjust to reflect changes in the overall level of the prices paid by commercial insurers relative to Medicare over time. This approach takes advantage of the fact that related services tend to be priced at similar Medicare ratios, which likely reflects the role that Medicare’s prices often play in private contracts. Third, the comparison to Medicare may be of direct policy interest, as setting out-of-network prices as a multiple of Medicare’s would be an administratively simpler alternative to the current IDR process.
Our findings for the second half of 2023 are similar, but not identical, to those for the first half the year:
These estimates offer a picture of how IDR has operated to date, but, as we noted previously, they leave at least two important questions unanswered. The first is how arbitrators are reaching decisions. The answer to this question may help inform ongoing debates about what, if any, guidance the federal government should give arbitrators and shed light on why the Congressional Budget Office (CBO) prediction that IDR decisions would hew close to the QPA has not been borne out.
A second key question is what these results portend for the average prices of the affected services and, in turn, insurance premiums. The answer continues to depend on multiple factors, including: whether the IDR outcomes observed so far are representative of what typical provider-insurer pairs would see in IDR; how the parties’ behavior and arbitrators’ decisionmaking evolve in the future; and how IDR outcomes will affect negotiations over in-network prices, all of which remain uncertain. Nevertheless, given that IDR decisions have remained well above prior in-network prices and comparable to or higher than prior out-of-network prices, we continue to see little chance that the law will meaningfully reduce prices and premiums, as CBO had predicted at enactment, and see a realistic possibility that the law will actually raise prices and premiums, in contrast to both CBO’s predictions and the stated goals of the law’s drafters.
This analysis uses the same methodology as our prior analysis, with some minor exceptions.
To maximize comparability between our new estimates and our prior estimates, we analyzed the same 49 CPT/HCPCS codes as in our prior analysis rather than selecting codes using the same criteria we used previously. Of these 49 codes, 9 codes would not have satisfied the selection criteria used in the prior analysis, all because they failed to satisfy the criterion of accounting for at least 200 line items. Applying the selection criteria used in the prior analysis in the new data would also have selected 1 additional code that was not included in the prior analysis.
A limitation of the IDR public use files released for the first half of 2023 was that the file that reports payment amounts (the “payment file”) includes limited information about each line item. Thus, we could not directly exclude two types of line items that were outside the scope of our analysis: (1) line items pertaining to facility, rather than professional, services; and (2) line items that are part of bundled disputes that include multiple services from a single episode of care. Instead, we used the file that did provide information on these characteristics (the “characteristics file”) to estimate the share of the line items for each service code that fell in one of these categories and then excluded the corresponding share of line items from the top of the distribution of payment amounts when analyzing the payment file.
We modified this procedure in a few ways for this analysis:
As in our prior analysis, we compare the prices emerging from IDR to previously published estimates of pre-NSA Medicare ratios for similar services (see Figure 1). In our prior analysis, we presented the estimated Medicare ratios without adjustment. However, the prices that commercial insurers pay for professional services have risen somewhat more quickly than Medicare’s prices since the years examined in these studies. Adjusting for these changes may better approximate the Medicare ratios that would have arisen in a counterfactual world without the NSA, the comparison of primary interest here.
To make such an adjustment, we relied on estimates of average commercial prices for physician services relative to Medicare published annually by the Medicare Payment Advisory Commission (MedPAC). MedPAC’s estimates only extend through 2022, so we used the payer-specific Producer Price Indices for physician services published by the Bureau of Labor Statistics to project MedPAC’s 2022 estimate forward to 2023. We then multiplied each pre-NSA estimate by the ratio of the (projected) MedPAC estimate for 2023 to the MedPAC estimate for the year of the data used to produce the pre-NSA estimate.
The resulting Medicare ratios are presented in Figure 1. These estimates are somewhat higher than the unadjusted estimates we used previously. For emergency care, the adjusted in-network ratios range from 2.6 to 3.0, whereas the unadjusted ratios ranged from 2.5 to 2.6; for out-of-network services, the adjusted range is 4.2 to 5.3, compared to the unadjusted range of 3.9 to 4.7. For imaging services, the adjusted in-network ratios range from 1.9 to 2.5, whereas the unadjusted ratios ranged from 1.8 to 2.4; for out-of-network services, the adjusted range is 3.1 to 3.7, compared to the unadjusted range of 2.9 to 3.3.
We note that even these adjusted estimates of pre-NSA Medicare ratios are an imperfect guide to the prices that would have prevailed in 2023 in a counterfactual world without the NSA, as they do not account for changes over time that are specific to the services that we study here. For example, they do not account for changes in the relative prices Medicare sets for different services (although the fact that changes in relative prices in Medicare tend to spur commensurate changes in relative prices in the commercial market may limit the degree to which this is a concern). Similarly, they do not account for changes in market conditions or provider and insurer negotiating tactics specific to these services that may have affected commercial prices. The direction of any bias is unclear, but we doubt that accounting for these factors would meaningfully alter any of the qualitative conclusions we reach here.
