As the Centers for Medicare & Medicaid Services continues to champion value-based payment models, it must gather key data on the social risk factors that Medicare beneficiaries face, according to a new report.

The report, the fourth in a series of five from the National Academies of Sciences, Engineering and Medicine on how social risk factors can be accounted for in Medicare payouts, includes five “guiding principles” that CMS should consider when choosing sources for gathering sociodemographic data:

  • First scrutinize data it has already gathered
  • Look for ways to use data from other federal agencies, including other groups under the Department of Health and Human Services
  • To the degree that it can, gather additional data when a patient enrolls in Medicare
  • For those social determinants that may change over time, access data reported by providers or included in electronic health records
  • For factors that are related to a patient’s home environment, employ local measures for data collection

From that point, the report delves into ways CMS can identify data sources under those five principles. The agency must improve data that individually measures trends among people of different races, ethnicity and sexualities. Where race and ethnicity data does currently exist, CMS must research ways to make it more accurate, according to the report. The document points to only three areas where CMS now gathers data effectively enough for it to be applicable: if patients live in rural or urban areas; if they are native to the area they live; and if they are dual eligible.

Because some social risk factors, such as race and ethnicity, will not change, these are perfect targets for data collection at the time of enrollment, according to the report. The type of data needed will change depending on the exact type of improvements that are being looked at; for example, data will impact quality measures differently than it would impact cost-cutting measures.

The report also noted that data collection on subjects that can vary so widely may require a multimodal approach to better unite information on multiple related topics.