Milliman has been selected by the Society of Actuaries (SOA) to conduct an extensive research study on accountable care organizations (ACOs). The goal is to identify more actionable benchmarks for an ACO to assess its performance by looking at episodes of care rather than the traditional actuarial view of benchmarks based on types of services.

Wide variation currently exists in costs per patient for numerous episode types. Many factors contribute to this variation, including differences in regional provider pricing, clinical practice patterns, and the underlying risks of the patients included in the sample. While many studies have examined the variation in per-episode costs related to price and practice differences, all struggle to normalize costs per episode across populations with different levels of underlying risk. We will be developing methods to address this need.

From a broad perspective, this study will analyze the impact of evaluating and rewarding ACOs based on performance at the level of individual episodes of care. This episode-based study will measure the risk-adjusted relative use of services, by service types, within an episode. From this information, more informed assessments of medical management impacts to financial results can be made. Statistical fluctuation can also be evaluated.

Risk adjustment mechanisms are widely used in certain markets today. Generally these methods address the aggregate "total cost" risk for an individual. While this is appropriate for payments from Medicare to health plans, or from an insurance company to an individual practice association (IPA), risk adjustment methods do not explain the differences in costs and utilization that are due to episode-specific risk factors, which may or may not impact total cost for the patient.

Based on the MarketScan commercial claims database, we will be developing risk adjustment algorithms for a select group of episode types. These risk adjustment methods will be based upon criteria that have been shown to be significantly correlated to the costs and levels of utilization within each type of episode. The study will develop the mean and statistical distribution for costs and utilization of selected services within the episode, breaking total episode costs into five service categories and including utilization benchmarks for risk adjustment that span these categories. The result will offer a benchmark tool for comparison of ACO claims experiences to episode-specific, risk-adjusted expected utilization and costs, by service type, for each episode type considered.

The episode types selected will be meaningful to the total cost of care, either as a low-incidence, high-cost component of total medical costs, or as a high-incidence, mid-level cost component with wide variation in episode costs. While the number of episode types evaluated will be small initially, we hope that additional industry-led research will expand the list of episodes dramatically.

ACO contracts generally include rewards and/or penalties for financial performance within a specified time period, typically a calendar year. Episodes, on the other hand, vary in length and may include services in multiple calendar years. Thus, a method of adapting episode-level evaluations into a financial model is needed. This is the final stage of our engagement.

We will develop a financial model to quantify the change in overall financial results that can be expected from specific changes in the episode benchmarks. We will evaluate the impact of population size and statistical fluctuation on these results. We will also evaluate the statistical significance of year-over-year changes in the risk-adjusted benchmarks.

Our goal is to offer insight to both providers and insurers as to where episode-level costs or utilization of service types exist outside normative bands. As well, we hope to create an understanding of the changes over time to episode-level costs as providers assume risks and strive to reduce costs.

This study will not incorporate a specific component to monitor continued or improving quality of care. We expect this approach to ACO evaluation to be performed in concert with other programs that focus on assuring quality outcomes.

We anticipate that the report and accompanying model will be finished before the end of the year.

Bill O'Brien, FSA, MAAA, is a consultant in the Houston office of Milliman.