Abstract
Research Objective: Over 70% of Medicare beneficiaries today are in some form of managed care, whether they are assigned to one of Medicare’s shared savings programs or are enrolled in a Medicare Advantage Plan. Developing case management programs for high-cost, high-need Medicare beneficiaries is expensive, but critical for successful financial risk and quality performance. Today’s case management programs are generally one-size-fits-all; however, within the high-cost, high-need Medicare beneficiary population, there exist multiple cost profiles that could lead to population segmentation in case management programs. This dissertation seeks to understand the differences between cost-based Medicare beneficiary categories and to understand if the publicly available Centers for Medicare & Medicaid Services (CMS) Medicare Advantage Hierarchical Condition Categories (HCC) risk adjustment model (CMS-HCC) and DH Kim Claims-Based Frailty Index algorithms can help predict which cost categories a Medicare beneficiary will fall in over subsequent years. As well, this research seeks to identify if there are CMS-HCC conditions and interactions that are characteristic of different categories of high-cost beneficiaries and which can be used to develop targeted case management programs.
Study Design: This dissertation consists of observational studies that use Medicare Parts A and B claims for beneficiaries over a period of three years, 2012-2014. The studies used binomial logistic regressions to understand the differences between four cost categories: always high-cost, persistently high-cost (high-cost two out of three years), transiently high-cost (high-cost one out of three years), and never high-cost. Binomial logistic regressions were used to predict a beneficiary’s future cost category. The dependent variable in the respective regressions was the beneficiary’s cost category, and independent variables include Medicare beneficiary demographics, Medicaid/Medicare dual status, original reason for Medicare enrollment, their 2012 CMS-HCC risk score, DH Kim Frailty score, and their 2012 CMS-HCC conditions. A beneficiary was defined as high-cost if they were in the top 90th percentile of cost in a given year.
Principal Findings: These studies found that it was possible to use the publicly available Centers for Medicare & Medicaid Services (CMS) Hierarchical Condition Categories (HCC) risk adjustment model and the DH Kim Claims-Based Frailty Index, to prospectively predict a Medicare beneficiary’s cost category over the subsequent two-year period. The results identified three major co-morbidity patterns to be potentially used in case management programs, including (1) high-cost chronic conditions, (2) high-cost conditions because of high Part D costs, and (3) high-cost conditions stemming from an episode of care, such as a hip fracture. This thesis also found that costs for persistently and transiently high-cost beneficiaries have distinct cost patterns from the always and never high-cost categories.
Implications for Policy or Practice: These studies indicate that commonly available, and free, risk algorithms can be used to segment high-cost, high-need Medicare beneficiaries into different case management programs tailored to their individual needs. This low-cost approach makes it easier for smaller Medicare Advantage plans or Medicare shared savings programs to implement a segmented case management program.