Bundled Payment Analysis

The cost of healthcare continues to be a topic of intense discussion from kitchen tables, to board rooms to congressional hearings. Spurred by these may conversations and allegations, insurance companies and entities such as Medicare and Medicaid have striven to conceive alternate payment approaches besides the traditional fee for service methods on the basis of reducing costs. 

Capitation (a system in which providers are given a defined sum per patient regardless of how many services are rendered in a defined period of time) has been around for many years and has been fraught with complaints among providers for inconsistencies between responsibility for and authority over patient interactions. The HMO model has also been attempted in various iterations with its focus on preventive care, but most have not survived, Kaiser Permanente being an exception. Medicare has since developed a method falling between that of full capitation and fee for service—the use of bundled payments or put more descriptively, episode-based payments. In this method, reimbursement of healthcare providers (both hospitals and physicians) is based on the expected costs for clinically defined episodes of care. Since 1984, bundling payment methods have been tried and as of 2012 almost one third of medical reimbursement is now from a bundling system.

In 2018, Medicare introduced a variation called Bundled Payment for Care Improvement Advanced (BPCI Advanced).  In this model there are 48 episodes, 45 inpatient and three outpatient. Again, there are four payment models, with model two being the most common.  Several of the medical episodes included in BPCI Advanced include CHF, COPD, Sepsis, Acute MI and Pneumonia. As every physician would know, there is a wide variation in the degree of illness and the course of therapy within these diagnoses, and there is nothing uniform about each case.  

For example, Sepsis includes three DRGs which range from uncomplicated to septic shock.  Unlike elective surgery, providers cannot screen these patients to avoid complications, and in fact many of these patients develop complications and/or have significant comorbidities.  All of these variables create significant variations of length in stay (LOS) and other costs associated with each hospitalization. In these cases, where there is a wide variation in costs, a need exists to employ a method to better predict these costs.

That methods exists but seems too seldom used. It is decision tree modeling that offers both the flexibility and complexity of interaction to more accurately predict costs than just a linear model which is commonly used.  Since diseases such as sepsis, COPD and CHF can become a complex affair with many possible outcomes that would affect costs, this type of modeling lends itself to such an analysis. Armed with better predictability results, providers are better able to defend both real costs and to negotiate for more fairly applied payment schemes.