Among the episodes that were related to the nine study conditions, there was substantial variation in average standardized payments,both across episode types and among different episodes of the same type. Per episode average payments for ETGs related to the nine study conditions ranged from an average of $1,306 (episodes related to diabetes) to $21,976 (episodes related to AMI). This is unsurprising, since these conditions have very different resource requirements - AMI requires hospitalization, and perhaps surgery, and rehabilitation, while diabetes typically is managed on an ambulatory basis.
Even among episodes of the same type, as defined by the ETG and MEG grouper tools, there was substantial variation in average payments per episode. The coefficient of variation for an episode type was inversely related to the average cost of that episode type, ranging from 72 percent (episodes related to hip fracture) to 269 percent (episodes related to diabetes). Large variation in average payments per episode highlights the need to understand the extent to which episodes are homogeneous in their construction (i.e., are they measuring the same type of care for the same type of patient or are there different subpopulations of patients within the episode category which accounts for the variation?). There is a need to understand the key sources of variation in payments and which sources need to be accounted for in the episode construction versus which sources could be minimized through application of episodes for performance measurement and/or payment. Variations due to underlying differences in the severity of patients would need to be controlled for in the construction and application of the episode; otherwise, unintended consequences could occur such as avoidance of more difficult cases if the financial risk exposure or challenges in managing the patient to the performance indicators is too great.
The degree of variation in average payments per episode has implications for performance measurement and payment. For example, performance measures that focus on resource use will require a large number of episodes to develop reliable estimates of performance if there is a large amount of variation, since reliability is inversely related to variation. In approaches that tie some portion or all reimbursement to an episode, a high coefficient of variation could suggest financial risk for the accountable entity unless the entity has a large number of episodes to absorb the variation. There are several potential approaches to managing the risk associated with variation in average episode payments that were proposed in the literature review and expert discussions. Risk mitigation techniques include the exclusion of outliers, risk adjustment, and narrower episode definitions.