Episode #1: Contrasting BPCI and PROMETHEUS Methodologies

Aver Insights recently spoke with Steve Kohlmann and Rory Weier, Aver’s in-house experts on the PROMETHEUS Payment model and Bundled Payments for Care Improvement (BPCI), respectively. The following transcript has been edited for length and clarity.

Aver Insights (AI): At a very basic level, what are BPCI and PROMETHEUS payments?

Rory Weier (RW): BPCI is a bundled payment arrangement developed by the Centers for Medicare & Medicaid Services (CMS) for episodes of care that involve a hospital inpatient stay. Even if the episode doesn’t include the inpatient stay, as under Model 3, which I’ll describe later, the episode is still tied to an inpatient admission. Aver has built out the BPCI methodology so that we can apply the episode logic to both commercial and CMS datasets.

Steve Kohlmann (SK): PROMETHEUS is actually an acronym - Provider Payment Reform for Outcomes, Margins, Evidence, Transparency, Hassle-reduction, Excellence, Understandability and Sustainability. It is an initiative sponsored by the Health Care Incentives Improvement Institute (HCI3) to design episodes of care. PROMETHEUS offers its version of a playbook for how to build an episode. These episode playbooks outline the detailed requirements that Aver follows to become a Certified Partner with Claims and Provider Analysis Certification.

AI: How do BPCI and PROMETHEUS define an episode?

RW: BPCI offers 4 different models for episode design. Model 1, which is no longer active, included only inpatient hospital services. Model 2 includes inpatient services as well as concurrent professional services delivered during the admission and post-acute care services provided for either 30, 60, or 90 days following discharge (at the participant’s choice). Model 3 includes only post-acute care services for 30, 60, or 90 days following initiation of those services. And Model 4 includes inpatient services, concurrent professional services, and post-acute care provided for 30 days following discharge, but this model utilizes prospective payments, with the episode-initiating hospital managing payments to physicians and post-acute care providers. My previous post provides more information about the 3 BPCI models still underway.

SK: One goal of the PROMETHEUS project was to categorize as much of the medical spend as possible into an episode. To do this, they created different episode classes or types. For example, PROMETHEUS offers several procedural episodes, which involve a surgical procedure and windows of time around the procedure during which care is delivered. Similarly, acute episodes are related to a single event with windows of time around the event during which care is delivered, but these events do not include a surgery. Good examples are acute myocardial infarction (AMI) or pneumonia. PROMETHEUS also has chronic disease episodes, which aim to capture costs of chronic care management that fall outside procedural or acute episodes. They recently released several cancer care episodes involving chemotherapy. The chronic and cancer episodes are examples that do not feature a single anchoring event with windows of time before and after, but instead may have extended or annualized windows of time divided throughout the long duration of care. Finally, there are “other” episodes that do not fit into these classes neatly, such as pregnancy or preventive services like colonoscopies or vaccinations.

AI: How many episode definition are available?

RW: BPCI has defined 48 different episodes of care, with each episode including one or more Medicare Severity-Diagnosis Related Groups (MS-DRGs). The MS-DRG system classifies a Medicare patient’s inpatient hospital stay into a particular “group” to facilitate payment. With BPCI, participating providers are able to choose the episodes in which they want to participate.

SK: The current version of PROMETHEUS has 97 distinct episodes, but many of these episodes are clinically similar and can be grouped as related care of the same disease.

AI: How do these episode models determine a payment amount or rate?

RW: Payment amounts are calculated for each episode initiator using 3 years of historical claims data, which are then trended forward to the participation year. A discount is applied to ensure Medicare savings.

SK: The PROMETHEUS model does not establish a payment amount or rate for each bundle, but it does offer advice on how one might come to a rate. Rather than being prescriptive about variables such as geographic variation, PROMETHEUS allows the user to establish the rate. Initially, payers utilizing PROMETHEUS bundle definitions may let a rate be established by what the data show and may set the rate at a certain percentage of historic costs. One unique feature of PROMETHEUS is that it identifies and categorizes services deemed potentially avoidable complications. So, for example, if two providers have similar historic costs but one provider has substantially more complications than the other, a payer may offer a more favorable rate to the provider with a history of fewer complications.

AI: What about risk adjustment? Does either model incorporate risk?

RW: Because BPCI episodes are comprised of different MS-DRGs, the risk adjustment relies on the DRG system to adjust for patient severity. As one example, the episode for acute myocardial infarction (AMI) is composed of 3 different MS-DRGs; one for AMI without complication or comorbidity, one for AMI with complication or comorbidity, and one for AMI with major complication or comorbidity. Each MS-DRG triggers a different episode payment rate and helps to adjust for patient severity and risk.

SK: PROMETHEUS has its own, proprietary method for risk adjustment using primarily clinical indicators that are present at the anchor or triggering event. The model utilizes two years of historical claims data to gather patient diagnoses, which helps to inform risk adjustment and establishes a predicted budget to compare against actual costs in these episodes. In addition, the risk adjustment can also be applied prospectively. In lieu of a patient’s historical claim sample, the surgeon or other provider would likely have the patient complete a questionnaire to further establish their medical history and inform the risk factors used in predicting the expected risk adjusted budget.

AI: In addition to differences in risk adjustment, what are some other key differences between the two models?

RW: I think BPCI’s basis on the MS-DRG system is a key differentiator and, in practice, means the model can really only be used for episodes of care involving an inpatient stay in some form, while the PROMETHEUS model offers episodes not tied to an inpatient stay. BPCI also has fewer episodes, 48 to PROMETHEUS’ 97. Finally, while both models can be customized, the PROMETHEUS model is a bit more prescriptive and allows for a more refined customization experience for the user.

SK: Right, I agree with Rory. BPCI’s use of MS-DRGs, compared to PROMETHEUS’ use of procedure and diagnosis codes, is a key difference. The issue of timing is another - most PROMETHEUS episodes include windows of time both before and after the triggering or anchor event, while BPCI episodes focus on care delivered after admission (and in some cases, after discharge). PROMETHEUS also uses claim assignment, which is a complex system of determining which claims belong to each episode, whereas BPCI relies more on one list of excluded conditions to determine which claims do not belong in the episode. In addition, there may be times when one claim is related to two different PROMETHEUS episodes, in which case PROMETHEUS will split the claim, assigning half to each episode. Lastly, there is a concept in PROMETHEUS wherein episodes that are distinct events for the same patient within a short period of time might be considered clinically related and will be clustered together for analysis and possibly for payment. A good example is if a patient who undergoes bariatric surgery suffers a complication like an AMI, a new AMI episode is triggered. After the AMI episode closes, it is linked back to bariatric surgery episode already open for that patient.

AI: Thank you both so much for laying out these two episode payment models. In our next blog, we will focus on how and why payers might use these models and how Aver works with payers to implement an episode payment program.

Senior Analytic Designer
Aver, Inc.


Director of R&D
Aver, Inc.

Want fresh insights on Value-Based Care delivered right to your Inbox? Become a member of the Aver Insights Community