Medicare risk adjustment for the frail elderly

Several analyses have shown that diagnosis-based risk adjusters do not fully predict the expenditures of the frail elderly, where frailty is generally defined in terms of functional impairments (Pope et al., 1998, 2000; Kautter and Pope, 2001, 2004; Kautter et al., 2007). Diagnosis-based models do predict the expenditures of the frail elderly substantially better than demographic models, but some residual expenditures statistically associated with functional impairment remain unexplained. CMS has thus had a continuing interest in exploring ways to incorporate frailty adjustment into the CMS-HCC risk adjustment methodology for Medicare Advantage and other Medicare private organizations (Pope et al., 2004). The goal of frailty adjustment is to account for the costs not explained by diagnosis-based risk adjustment.


Predicting expenditures accurately for subgroups of Medicare beneficiaries is desirable. Accurate prediction for the frail elderly is especially significant because they do not comprise a uniform proportion of the enrollment of all Medicare capitated organizations, and their expenditures are considerably higher than the average beneficiary. This is a particularly important issue for organizations whose models of care focus disproportionately on the frail elderly, for example PACE organizations. (1) A payment factor to account for potentially higher expenditures for the frail elderly is important in ensuring the viability of these organizations, and access for beneficiaries to the care they provide. Therefore, since 2004, CMS has applied a frailty adjustment to payments for enrollees in PACE organizations (Kautter and Pope, 2004-2005). (2) CMS adopted the approach taken by many researchers and clinicians of defining frailty as functional impairment, and using counts of difficulty in performing activities of daily living (ADLs) as the core measure of functional impairment. The original frailty adjuster model was estimated using ADL information in the Medicare Current Beneficiary Survey (MCBS). The frailty adjuster is prospective, meaning that Medicare expenditures in a given year are predicted by ADL information in the prior year.

As reported here, the frailty adjustment factors have recently been updated and refined. Effective 2008, CMS is applying these new frailty factors to PACE organization payments on a 5-year phase-in schedule (Centers for Medicare & Medicaid Services, 2007a,b; 2008). (3)

We present research results for Medicare risk adjustment of the frail elderly since the adoption of frailty adjustment for PACE organizations in 2004 (Kautter, Ingber, and Pope, 2007). In particular, we describe the development of a frailty adjuster estimated on the Medicare Fee-For-Service Consumer Assessment of Healthcare Providers and Systems (CAHPS[R]) Survey. Medicare is transitioning PACE organization payments to 100 percent of the revised frailty adjuster over the 5-year period 2008-2012.

ORIGINAL FRAILTY ADJUSTER

The original frailty adjuster was calibrated using 1994 to 1997 data from the MCBS for the community-residing, age 55 or over population enrolled in fee-for-service (FFS) Medicare (Kautter and Pope, 2004). The MCBS is a nationally representative sample of Medicare beneficiaries. (4) We found that frailty factors are quite different for community-residing versus long-term institutionalized (nursing home) beneficiaries, and concluded that the appropriate frailty adjuster for the long-term institutionalized should be a factor of zero. (5)

At the time the initial frailty model was created, the MCBS data was the only comprehensive data available that allowed linkage of individual-level functional impairment data (ADLs) to Medicare claims data. Information from the MCBS was used to predict expenditures related to frailty that were unexplained by the CMS-HCC risk adjustment model. The ADLs may not relate to the incremental expenditures causally, but are strongly correlated with additional expenditures. Actual frailty scores for health organizations are calculated at the contract level (rather than the plan benefit package level) (6) using these frailty factors and an estimate of the ADL limitations of enrollees reported in the Health Outcomes Survey (HOS) sent to a sample of enrollees in each organization. These frailty scores are added to the risk adjustment factors in payment. The original frailty factors calibrated on the 1994-1997 MCBS were 1.094, 0.340, 0.172, and -0.143 respectively, for, counts of ADL difficulty 5-6, 3-4, 1-2, and 0 (Kautter and Pope, 2004).

UPDATE AND REFINEMENT OF FRAILTY ADJUSTER

The source of data used to calibrate the frailty factors was changed so that the methods used to gather ADL-related data for both calibration and payment would be similar, avoiding measurement disparities that come from using different data collection methods. As previously noted, the original frailty factors were calibrated using ADL limitation information gathered from MCBS in-person surveys. CAHPS[R] data, which were used to update and refine the frailty factors, and HOS data, which are used to calculate frailty scores for payment, both collect ADL information via mail surveys with telephone followup.