In Part 1, we tackled the shortcomings of social risk intervention as a reactive model based on survey data or, when available, a limited set of z-codes. Today, we’re going to dive into the exciting next steps of proactive intervention through better datasets, the difference between the two, and how Socially Determined can help clients go from behind the curve to ahead of the game.
Imagine an individual with transportation risk reports via screening that she does not have any issues with transportation because every Wednesday her daughter comes to run errands and take her to all her appointments. She likes spending time with her daughter and does not want to accept help or bother coordinating with a service to find a ride.
Despite not having sustained need, she will inevitably:
Similarly, an individual with diabetes and high food insecurity risk reports having no trouble feeding their family healthy food over the last 30-days. However, they live in a “food swamp,” work long hours, and don’t have a lot of time or interest in cooking.
Despite not having sustained need, this means they:
In both of the above instances, despite under-reported risk, these common behaviors still qualify as being at a level of social risk around transportation and food insecurity. Both also lead to adherence gaps, use of the emergency department/urgent care over primary care, worsened disease management outcomes, and increased costs. Put more simply, the fundamental flaw in surveys is that the circumstances that place an individual at high SDOH risk do not necessarily create an immediate self-perceived need. It will, however, impact outcomes.
True granular, actionable data also narrows the scope when the complete picture still feels elusive. Even in instances where our partners are unable to corroborate social risk with the self-perceived need directly with their members, the relationship between our risk scores and high cost, adherence gaps, and less efficient utilization patterns remain apparent. The relationship is particularly powerful when used as part of a health outcomes model that incorporates clinical or claims data. This is, ultimately, the rubber meeting the road on the gaps between reactive screening and proactive, multi-source analytics. If a patient is so tragically used to their life operating at a certain level of difficulty, or feel shame for “complaining,” they often under-report social risk. With our dataset and platform, our customers will see the patterns that show the true picture and know how to intervene.
For example, an analysis conducted recently with one of our ACO partners successfully predicted patients likely to become “frequent flyers” and/or high-cost cases using predictive models via Random Forest and XGBoost. While both Random Forest and XGBoost models showed positive results, with both area under the curve and F1 scores indicating strong predictive power, they found, “XGBoost demonstrated the highest accuracy (AUC: 0.98 for frequent flyers, 0.94 for high-cost patients). Medical risk, chronic conditions, financial strain, and healthcare access barriers were key predictors. Social determinants, particularly transportation barriers and health literacy, influenced ER visits, while oncology and pharmacy risk were primary cost drivers”.
In another internal study using several years of clinical data, we demonstrated that members with high social risk that had a single major disease incident progressed to major chronic diseases several years faster than those not at high risk. Here again, an XGBoost model showed the highest accuracy through both the AUC, at .86, and an F1 Score of .82. Results showed, “The efficacy of Socially Determined’s SDoH risk features in defining personas and predictors based on HbA1c levels...” underscoring “...the potential of socio-economic indicators to anticipate HbA1c control levels, even without referencing an individual’s medical history.” With the model displaying, “85% accuracy in classifying out-of-control diabetic cohorts based on the drivers of SDoH.”
In each case, these results demonstrate that the sustained conditions contributing to high-risk profiles may not necessarily manifest as reported, discrete instances of need, but they do impact long-term health behaviors and outcomes, in turn driving costs. When it comes to assumed risk under VBC populations, not just managing, but tracking, chronic disease progression is critical for quality of life and managing costs.
At Socially Determined, we didn’t just pioneer a next-generation, data-driven approach to SDOH intervention. We came together with a promise to deliver social risk analytics that are truly actionable, helping organizations target specific individuals and communities and deploy concrete interventions and improve business performance, clinical outcomes, and healthcare for all.
Social determinants of health have been a “white whale” in healthcare for decades. The influences they have on outcomes and cost are known, as are the methods that can change the entire game. What’s missing is knowing who, where, what, and why.
With Socially Determined, those questions are answered. If you’d like to learn more, we’re eager and ready to connect.