Medicare increasingly uses various “value-based” payment models which often rely on some cost prediction or risk-adjustment methodology. There has been significant controversy over those adjustment methods, with many opining that they may penalize providers and plans who serve poorer populations, which may have unidentified risk factors out of the control of the provider or plan. A study in Health Affairs finds this to be the case in regard to poorer, often dual-eligible beneficiaries, due to greater functional, cognitive and social issues affecting health and health care needs. (HA Article) The authors looked at differences in these areas between dual-eligible and non-dual-eligible enrollees, using survey, claims and administrative data for fee-for-service Medicare beneficiaries in the time period 2006-2013. The dependent variable was Medicare spending, total and divided into common categories. The independent variables included age, sex, Medicare-eligibility basis, institutionalization, HCC score and comorbidities, as well as income, education, cognitive status and activities of daily living limitations. Dual enrollees were younger than non-duals, were more likely on Medicare due to disability or ESRD, more likely to be female, more likely to be institutionalized, and had higher risk scores. 93% of duals had income less than $25,000; compared to only 33% of non-duals. About 50% did not have a high school education, compared to 17% of non-duals. They were twice as likely to be unmarried, which removes a social support. They also had much higher rates of cognitive disorders and limitations in ADLs.
Total average annual costs for dual-eligible beneficiaries were $11,928 versus $8,310 for non-duals. Other that outpatient physician costs, every other expense category showed substantially higher costs for dual-eligibles. The HCC model used alone, as it commonly is for value-based payments, including Medicare Advantage, underpredicted actual costs. Including the cognitive and functional factors substantially improved prediction accuracy; adding social factors only gave a very minor boost to prediction. But in some sub-categories of costs, such as inpatient and skilled nursing facility, social factors did aid in improving accuracy. Interestingly, adding these factors also improved cost prediction for non-dual beneficiaries, lowering predicted costs to closer to actual ones, and implying that Medicare may be overpaying for some non-dual eligibles. While there are administrative and cost difficulties in gathering this cognitive, social and ADL data, it clearly would make payment methodologies more accurate and effort should be spent to find solutions to do this.