The Most Important Factor in BPCI Advanced Target Prices

The BPCI Advanced program, the successor to Bundled Payments for Care Improvement (BPCI), involves Medicare’s most sophisticated strategy for setting episode target prices among its bundled payment initiatives to date. Previously, Medicare acute care bundled payment programs utilized more facile target price adjustment strategies, adjusting target prices by episode, through Medicare Severity Diagnostic Related Group (MS-DRG), as seen in BPCI classic, and fracture status, as seen in Comprehensive Care for Joint Replacement (CJR). These strategies have been subject to scrutiny due to the lack of adjustment for other significant patient characteristics. While developing BPCI Advanced target process, CMS created a patient case mix adjustment (PCMA) factor to remediate the limitations of those early approaches—controlling for additional patient level characteristics within a clinical episode category.

This methodological advancement led the team of data scientists at DataGen to ask the question: how much of the PCMA factor is driven by MS-DRG/APC and fracture status versus other patient characteristics?

Despite the attempt by CMS to improve the accuracy of risk adjustment in target setting by broadening the inclusion of elements defining case mix adjustment, the MS-DRG/Ambulatory Payment Classification (APC) assignment is the most important factor in estimating an episode’s overall patient adjusted spending, regardless of clinical episode category. All episodes demonstrated that the MS-DRG/APC had meaningful impact on overall case mix adjusted spending, but, the level of influence on overall patient adjusted spending varies by clinical episode category. The introduction of APC to the case mix adjustment in BPCI Advanced is the result of the outpatient episodes in the program.

In about a third of the clinical episode categories, the variation in the overall case mix adjusted spending is mostly explained by the MS-DRG/APC. Another third of the clinical episode categories had more than half of the variation in the overall case mix adjusted spending explained by MS-DRG/APC. Among the remaining third, MS-DRG/APC is important, but other patient characteristics may be more influential than seen in the other clinical episode categories.

To better understand these findings, first we must understand BPCI Advanced target prices. Target prices for BPCI Advanced account for a provider’s historic payment efficiency, peer group-related payment trends, along with the PCMA. Target prices are derived through a series of regression model risk adjustments.

Patient case mix in the BPCI Advanced program is defined as follows:

• Beneficiaries’ age
• Dual eligibility
• Disability
• Long-term institutional care
• Specific comorbidities via hierarchal condition categories (HCC)
• MS-DRG/APC
• Fracture status and location for major joint replacement of the lower extremity episodes
• Recent hospitalizations prior to episode start.

To adjust targets for patient case mix, CMS includes a PCMA factor which compares an episode initiator’s patient case mix to the case mix of the national population for a specific clinical episode category. The PCMA factor is calculated using estimates of patient adjusted episode spending, which is the focus of this analysis.

Evaluating the impact of the known elements of patient case mix with those newly introduced required that DataGen data scientists assess the 32 clinical episodes of BPCI Advanced (29 inpatient-based and three outpatient-based episodes). The episodes anchor dates fell within the period January 1, 2014, and December 31, 2016. Risk adjustment parameter information was obtained from preliminary baseline target files for 2018 as published by CMS.

To begin, DataGen calculated the overall case mix adjusted spending for all episodes using the complete risk adjustment parameter estimates applied by CMS in the target price methodology. The data scientists then followed that calculation by creating a measure DataGen calls the MS-DRG/APC adjusted spending. This measure was calculated using risk adjustment parameter estimates applied by CMS in the target price methodology for only a subset of characteristics—accounting for the MS-DRG/APC for all episodes and fracture status and location for major joint replacement of the lower extremity episodes only. MS-DRG/APC adjusted spending (MDAAS) was calculated for all episodes.

To approximate the influence of MS-DRG/APC on the overall patient adjusted spending in the BPCI Advanced target price, a ratio between the MDAAS and the overall case mix adjusted spending for each episode was calculated. This resulting ratio is referred by DataGen to as the MDAAS ratio. DataGen’s data science team applied statistical methods and calculated the interquartile range and coefficient of determination by clinical episode category, to understand variances in the resulting MDAAS ratios.

MS-DRG/APC assignment is the most important factor in estimating an episode’s overall patient adjusted spending, regardless of clinical episode category. However, the level of influence on overall patient adjusted spending varies by clinical episode category. In certain clinical episode categories, other patient characteristics still play an important role in determining overall case mix adjusted spending. These are the clinical episode categories that may benefit the most from initiatives focused on improving coding practices for comorbid conditions. Furthermore, the findings have direct implications on how comparable preliminary target prices and final target prices may be for the actualized performance period episodes.

In follow up to this analysis, DataGen’s data science team is hard at work developing analyses for each clinical episode category to determine the impact of static patient level characteristics such as age, disability status and dual-eligibility, and to identify which actionable factors are important.