Case-Mix Adjustment Guidance

Guidance on Standard Covariate Adjustment for SEER-CAHPS Analyses

Background: Evaluations of patient experience surveys, including Medicare CAHPS, have identified respondent characteristics not under control of the health or drug plan but consistently related to the sampled member’s survey responses, even among beneficiaries under the same contract. Such associations may occur because beneficiaries with some characteristics are more likely to encounter problems in health care (e.g., patients requiring frequent treatment for chronic conditions) and/or because some characteristics are associated with differences in the use of response scales. Public reports of Medicare Advantage (MA) and Prescription Drug Plan (PDP) CAHPS Survey results are adjusted for the predictable effects of such characteristics. This process of case-mix adjustment helps to control for variability in patient experience ratings due to different distributions of patient characteristics known to be associated with patient experience scores. More information on case-mix adjustment as performed regularly by the MA & PDP CAHPS Project Team can be found on the Medicare Advantage and Prescription Drug Plan CAHPS® SurveyExternal Web Site Policy site.

Case-mix adjustment variables for MA and PDP CAHPS Survey results*

  1. Age
  2. Education
  3. Health Status
  4. Mental Health Status
  5. Received Help Responding
  6. Proxy Answered Questions for Respondent
  7. Medicaid
  8. Low Income Subsidy
  9. Chinese Language

In addition, the scores of PDPs, which have multi-state areas of operation, are adjusted for state of residence.

*Please see the Medicare Advantage and Prescription Drug Plan CAHPS® SurveyExternal Web Site Policy site for the recommended parameterization of these variables.

Guidance

The standard case-mix variables have been shown to predict individuals’ reports on their health care experiences and are generally acknowledged in the literature on patient experience reporting to represent characteristics largely defined prior to the period of care reported on. Inclusion of these covariates is appropriate for most analyses in which CAHPS measures are the outcomes (dependent variables), including multivariable analyses that use SEER-CAHPS data (i.e. studies that use MA, PDP, and Fee for Service CAHPS data). We suggest using these set of covariates as a default setting for all analyses and always when making comparisons among health plans or other health care units.

Considerations

  • Case-mix adjustment variables have been added over time, and can be found on the Medicare Advantage and Prescription Drug Plan CAHPS® SurveyExternal Web Site Policy site. When pooling data over multiple years, investigators are encouraged to include the covariates common to all years, which will often be the covariates used in the first year of CAHPS survey data requested. For example, if requesting CAHPS survey data from 2005-2013, investigators should use case-mix adjustment variables recommended in 2005. There may be certain situations where researchers choose to combine case-mix adjustment variables for analysis (e.g., dual eligibility and low-income subsidy; proxy response status variables).
  • Because analyses conducted using SEER-CAHPS data will typically involve cancer-related population subgroups and/or cancer-specific variables not used in standard CAHPS reporting analyses, coefficients of the estimates regressing CAHPS ratings and composites on standard case-mix variables will generally differ from those published in relation to the MA and PDP reports for the corresponding years.
  • Investigators should also consider additional covariates (e.g., cancer-specific variables) for inclusion in analyses. It is important to assess, however, whether these additional covariates are collinear with the standard case-mix adjustment variables. In addition, investigators are cautioned that inferences and interpretation of unadjusted CAHPS results are not appropriate.
Last Updated: 12 May, 2020