Large study finds selection bias in Medicare shared savings erases savings and quality improvements – advocates saw this coming

Researchers from the University of Michigan found that the modest savings and quality improvements reported by Medicare’s extensive shared savings program (MSSP) are likely due to adverse selection. High cost clinicians and beneficiaries were far more likely than others to exit the program. When adjusted for the selective bias in MSSP exit, reported savings and quality improvements from the program evaporate.  Advocates and providers predicted this but were dismissed.

Medicare’s program, like CT Medicaid’s and commercial shared savings programs, relies on Accountable Care Organizations (ACOs) – health systems that are paid part of the savings generated on their patients. Shared savings has been offered as the solution to lower healthcare costs but also raises concerns that ACOs (like HMOs of the past) will generate savings by denying appropriate care to members (underservice or stinting on care) and/or shifting less lucrative members out of the pool (adverse selection).

The new study, published in the Annals of Internal Medicine, uses a rigorous methodology to track clinicians and patients moving in and out of Medicare ACOs between 2008 and 2014. The study included a random sample of claims for 835,100 Medicare beneficiaries attributed to 337 ACOs across the US. Between 2013 and 2014, researchers found that 4,054 clinicians stopped participating in MSSP. Researchers found that clinicians at the 95^th and 99^th percentiles of spending had a 22.3% and 30.4% chance, respectively of leaving MSSP, compared to only 13.8% for median spending clinicians. In addition, high cost beneficiaries (at 99^th percentile) were far more likely to leave MSSP even if their clinician stayed in the program, than median cost beneficiaries (17.7% vs. 8.9%). Adjusting for this selective exit of costly beneficiaries and clinician changes reported savings of $118 per beneficiary per quarter to spending $5 more per beneficiary per quarter.

The researchers also found that adjusting for more costly clinicians and beneficiaries exits also erased reported quality improvements in reductions in all-cause hospitalizations, preventable hospitalizations, LDL cholesterol testing, diabetes hemoglobin testing and retinal exams.

The authors note that ACOs have good information on the spending by clinicians in their health system that allows “pruning” to generate misleading savings payments and improve quality performance scores. The study follows more limited examples of adverse selection and calls into question even the poor results from the first year of PCMH Plus, Connecticut Medicaid’s shared savings experiment covering over 100,000 people. The study authors conclude, “These findings suggest caution in extending ACOs to other settings and patients without stronger evidence that the program saves money or improves quality of care.” DSS is promoting expansion of the troubled PCMH Plus program to fragile seniors and people with disabilities eligible for both Medicare and Medicaid.