Physicians have two classic responses to quality or performance reports:
- My patients are sicker, or
- Your data is wrong.
But there is a third concern, very significant and real:
- …are poorer, less educated
- …live in challenging neighborhoods
- …don’t have access to healthy food, or for other reasons don’t eat healthy food
- …don’t exercise, breathe unhealthy air or drink unhealthy water
- …in short, are more “difficult”
Particularly when quality measures are used to set incentive payments or penalties, it’s very important to apply social risk adjustment. Without it, these incentive programs might work like a Robin Hood in reverse scheme: rewarding practices in strong socio-economic areas, at the expense of those who practice in challenging neighborhoods. Remember, MIPS is budget neutral, and I understand that to mean that rewards given to high-performing practices are offset by equal amounts in penalties to other practices.
Medicare adjusts payments to providers based on how sick a provider’s population is – using Hierarchical Condition Categories. (This is why thorough and complete documentation and coding, plus seeing patients regularly to update their diagnoses and problems, can increase payments to the practice). But HCC-adjustment doesn’t level the playing field between practices that take care of wealthy, highly educated patients vs. poor and disadvantaged communities.
How would social risk adjustment work?First, get a data source for a broad socio-economic risk indicator. Most of these will be at a zipcode or census track level – and that’s OK. Once aggregated to a practice, the risk indicator will properly reflect the social risk mix for that practice. There are excellent publicly available sources for this data – see this Community Vital Signs paper, with a wonderful list of data sources. Or use a data set from a commercial data aggregator – Centraforce is an example.
From this data set (or sets), pick a broad set of “orthogonal” indicators to build your social risk index – include elements that reflect income, home ownership, education, environmental factors, food availability – the more socio-economic domains your risk indicator touches, the better. The NQF study used a limited set of social indicators (insurance product, education, race, ethnicity, language, income) and was able to demonstrate the effect of risk adjustment. Include a broad set of social determinants to create a strong social risk index.
With this data, calculate a zip-code level risk indicator – and use this for calculating “observed” and “expected” quality scores. Exactly like we currently risk-adjust with a utilization-based risk index. I’ll go into more detail on this in a future post.
In short: Select a source for socio-economic data – preferably with a varied set of data elements. With this data, compose a social risk index. Use the index to risk-adjust quality and performance measures. Then compare a practice in an underserved neighborhood to a practice in an upscale area, on one of the measures you risk-adjusted – for example NQF 59, diabetes control. Did the suburban practice have a higher score? Probably. Did your risk-adjusted data bring the practices closer together? That would be the result you’re looking for.
Value-based care, and Population Health, is about providing financial incentives for high-quality, efficient care, at the population level. To do this right, and fairly, we need to include social determinants in the calculation of the rewards and penalties – or we might inadvertently penalize physicians and clinicians who provide care to the people who need it most.
We know that social determinants are a stronger driver of someone’s health status than the level of healthcare they receive. It’s important that we reflect this reality in the way we measure, rate, and reward healthcare providers.