The Two-Stage Model of Competition in Hospital Merger Analysis May be Due for an Update

Cutting-edge economic theories and econometric methods are likely to play an increasingly important role in healthcare antitrust analysis. In recent hospital merger litigation, the FTC and courts have looked to the “two-stage model of competition” as a framework for evaluating competitive effects. While this model sometimes performs better than HHIs or other share-based indices, it has several limitations. New research and empirical findings highlight some of those limitations and demonstrate the existence of mechanisms that market participants can use to counter post-merger pricing increases. These findings suggest that cutting-edge modeling may need to extend beyond the basic two-stage construct.

In the past few years, circuit courts have relied on the tenets of the two-stage model of competition and the model’s corollary − the willingness-to-pay construct − in their decisions to reverse the district court opinions in the Penn State Hershey-Pinnacle and Advocate-NorthShore mergers between providers of general acute care hospital services. More recently, a federal court in North Dakota granted a preliminary injunction against Sanford Health’s proposed acquisition of Mid Dakota Clinic, relying on arguments that were based on the two-stage model in the context of physician services.

The two-stage model relies on the premise that competition between healthcare providers occurs, as the name suggests, in two stages. In the first stage, health plans and providers negotiate to determine the prices at which each provider will be included in the health plan’s network. The model assumes that each party’s bargaining leverage is a primary determinant of the negotiated rates between the provider and the health plan. As for bargaining leverage itself, the model assumes that it is a function of each party’s respective threat point, or how well each party would fare if it walked away from the negotiations. In the second stage, in-network providers compete for the health plan’s enrollees. The model assumes that competition in the second stage is based primarily on non-price factors, since patients pay only a small portion of the provider’s rates and out-of-pocket payments tend to be invariant to their choice of in-network provider. The attractiveness of a provider to an individual patient depends on the patient’s location relative to that of the provider, as well as other provider characteristics such as reputation, quality, and range of services.

The model itself first generates estimates of consumer preferences or demand for providers which are revealed through patient choices among in-network providers. The necessary data inputs for the first step of the modeling exercise (which focuses on the second stage of competition between in-network providers) pertain to patients’ choice of provider as well as individual patients’ demographics, type of insurance coverage, medical needs, and the distance from each patient’s home to the locations of each provider in the patient’s choice set. In addition, the model incorporates controls for characteristics of the providers, including size, ownership and proxies for quality or reputation when available. In the case of inpatient hospital services, the requisite data often are available from various state agencies.

The estimates from the first step of the modeling exercise can be used to deduce the incremental value that each provider contributes to the health plan’s network, or the “willingness to pay” for the provider. Importantly, the model assumes that patients’ actual choices can be used to reveal the extent to which providers are substitutes for one another. When in-network substitutes for a provider are readily available, that provider will contribute less incremental value to the network and, hence, both its bargaining leverage and the plan’s willingness to pay for its services will be lower. When two close substitutes merge, the model estimates that the willingness to pay for the two providers combined will be significantly greater than the sum of the willingness to pay for the two providers individually. This results in higher bargaining leverage for the combined entity post-merger.

The second step of the modeling exercise attempts to predict the price effects of the merger. This step requires pricing data which are available from health plans’ claims databases. To generate predicted post-merger price changes, the model uses regression analysis to estimate the relationship between willingness to pay and prices. Once this relationship is estimated, the change in willingness to pay from the first modeling step can be used to estimate predicted post-merger price changes.

Proponents of the two-stage model claim that the model captures bargaining dynamics between payors and healthcare providers and that the model permits more accurate antitrust analyses by placing less weight on patient-travel patterns. Critics of the model point to reasons that the model only imperfectly captures competitive dynamics: it does not incorporate the ability of health plans to use in-network steerage, does not capture competitive responses by rival healthcare providers, and does not explicitly model the ability of consumers to switch health plans in the face of network restrictions or premium increases, among other reasons. For example, a study by Christopher Garmon, using real-world data, finds that the two-stage model does not lead to a correct prediction in about one of every three merger cases included in the study.

Additionally, a recent article in Econometrica by Kate Ho and Robin Lee supports the proposition that the two-stage model does not always provide a complete picture of the competitive dynamics in a healthcare marketplace. These authors extend the two-stage framework to consider the impact of health plan competition and large employers’ ability to influence health plans’ bargaining leverage with providers. Employers control the number of health plans that they decide to offer to their employees. The authors show how employers can exercise such control to influence the degree of bargaining leverage that their chosen health plans have vis-à-vis providers in their regions. For example, if the employer’s enrollee volume is concentrated across two plans instead of three, the remaining two plans may gain an ability to negotiate better rates from providers. These findings suggest that the two-stage model may need to be extended to better predict price effects resulting from horizontal provider mergers.

Lona Fowdur has worked on numerous hospital mergers and other antitrust issues pertaining to healthcare. This article is based on written materials for a panel discussion held during the 2018 Antitrust in Healthcare Conference.