Project 3 - Integration of Care and Selection: Medicare
Advantage Compared with Traditional Medicare

Joseph Newhouse, PhD, principal investigator

Relative to traditional Medicare (TM), clinically integrated
health plans that participate in Medicare Advantage (MA) may be able to
treat a given patient more efficiently, using fewer resources and
delivering care of equal or superior quality, through their flexibility
in benefit structure, network contracting, and ability to coordinate and
manage care. These very features of clinical integration, however, may
differentially appeal to beneficiaries with different types of health
risks, creating potential incentives on the part of both beneficiaries
and plans for selection. For example, the efficiencies from integration
may not be uniform across diseases, and therefore plans may be able to
achieve more savings for some conditions than others.

The heterogeneity in beneficiary preferences for the various
instruments of integration and the heterogeneity across diseases in
opportunities for integration efficiencies would not matter if risk
adjustment balanced incentives across diseases. But because the current
Medicare risk adjustment algorithms are based on both the administered
prices and the practice patterns of TM in treating a given disease, the
resulting relative prices may depart from MA relative costs. Conditions
that are amenable to greater cost savings from integration may be
relatively more profitable for MA plans, generating incentives for MA
plans to try to attract enrollees with these conditions by engaging in
selection behavior. Importantly from an economics perspective, plans
attempting to influence enrollment patterns may distort their delivery
of services to attract profitable beneficiaries, thereby not making full
use of integration possibilities and making the delivery of care
inefficient. For example, if the risk adjustment scheme fails to give
patients with a given disease sufficient weight, plans may try to
exclude providers with a strong reputation for treating that disease
from their network. Thus, integration and selection are intimately
intertwined.

This project estimates the size of integration efficiencies and the
effects of Medicare reimbursement policy on both integration
efficiencies and on selection. Its specific aims are:

Aim 1
Assess the value of clinical integration in MA plans by
comparing the quality and intensity of care for MA and TM enrollees.
This will be done using measures from the Healthcare Effectiveness Data
and Information Set (HEDIS) measures of MA quality of care and resource
use, corresponding measures to be developed from Part B claims data for
TM enrollees, and MA and TM enrollees' reports of quality of care in
the Consumer Assessment of Healthcare Providers and Systems (CAHPS)
survey.

Aim 2
Assess the predictive accuracy and explained variance of
current risk adjustment methods using individual-level MA plan data from
Kaiser Permanente of Northern California (KP-NC). Currently,
Medicare's risk adjustment is calibrated with TM data. Because TM may
have significantly different patterns of care than MA plans, plans may
have corresponding incentives to favor or disfavor enrollees with
certain diagnoses.

This aim has been largely completed with results that show a substantial divergence between relative costs across the CMS-HCCs using internal health plan data and the relative weights in the CMS-HCCs that are based on utilization patterns in traditional Medicare.  The results substantiate the premise that utilization reductions in Medicare Advantage relative to traditional Medicare do not occur equiproportionately across the various CMS-HCC's, thereby creating incentives to favor or disfavor certain CMS-HCCs.  These results were presented at the 2010 conference of the American Society of Health Economists and have been submitted for publication.

Aim 3
Measure selection using data from MA rather than TM.
Specifically, estimate any changes in the amount of selection (favorable
or unfavorable) into KP-NC from 1998-2011 as risk adjustment that used
diagnostic information was phased in starting in 2000 and as lock-in
periods changed starting in 2006. Using results from Aim 2, determine
if flows into and out of KP-NC are related to pricing errors from using
TM data to calibrate risk adjustment.

Aim 4
Apply the results from Aim 2 to evaluate whether incentives for
selection and integration lead to differences in drug plan design
between MA-Part D plans (MA-PD) and standalone Part D drug plans (PDPs).