EXHIBIT 3.
Change in disenrollment and contract switching rates, use of higher- and lower-rated hospitals and nursing homes, and readmissions associated with a 1-star increase in Medicare Advantage (MA) contract star rating
| Baseline rate in 2014 (%) |
Change in outcome associated with 1-star increase in star rating (percentage points)a |
Relative change (%) |
No. of observations |
|
|---|---|---|---|---|
| Switching or disenrollment | 13.0 | −2.7**** | −20.8 | 30,104,761 |
| Admission to hospital with 4+ stars | 23.4 | 0.8*** | 3.4 | 4,132,632 |
| Admission to hospital with <3 stars | 33.0 | −0.9*** | −2.7 | 4,132,632 |
| MedPAR 30-day readmissions | 17.0 | −0.3 | −1.8 | 4,132,632 |
| MedPAR 90-day readmissions | 19.5 | −0.5** | −2.6 | 4,132,632 |
| Admission to nursing home with 4+ stars | 42.4 | 1.0** | 2.4 | 1,342,226 |
| Admission to nursing home with <3 stars | 32.3 | −1.1** | −3.4 | 1,342,226 |
SOURCE Authors’ analysis of data from the 2014 and 2015 Master Beneficiary Summary Files and Medicare Provider Analysis and Review (MedPAR) files. NOTES Admission to nursing home is derived from Minimum Data Set records. Admission to hospital and hospital readmission rates are derived from the MedPAR files. We consider a hospital or a nursing home to be higher quality if rated four or more stars on Hospital Compare or Nursing Home Compare, respectively, and lower quality if rated below three stars. The readmission rates are unadjusted unplanned readmissions using the Healthcare Cost and Utilization Project unplanned readmissions algorithm. An enrollee is considered to have disenrolled if, at the start of the following year, they are still alive but are enrolled in traditional Medicare. An enrollee was considered to have switched plans if they moved from one MA contract to another voluntarily. Disenrollment or switching was considered as a single outcome in this analysis. All regression values are percentage-point differences associated with a one-star rating increase. The baseline column is the mean outcomes among enrollees who did not consolidate. The number of observations changes because of the availability of the outcome variable.
Instrumental variable approach.
p < 0.05
p < 0.01
p < 0.001