A study published in January concluded that even with limited medical record data, it is possible to stratify patients based on social determinants of health (SDoH) factors using artificial intelligence (AI). The paper, published in The American Journal of Managed Care, highlights the growing trend of the use of AI to predict SDoH risk – but still with some gaps noted.[i]
“Current analyses, predictive models, and prevention initiatives focus on addressing SDOH at the population level or the zip code level. The shortcoming of this approach is a gap in addressing the individual patient’s needs, such as defining clinical action steps that are relevant to the patient as opposed to an overall population approach,” wrote the researchers.
The far-reaching SDoH factors play a significant role as healthcare and social services providers work to meet the standards for new financial and quality expectations. These include such evolving models of Accountable Care Organizations (ACOs) and other risked-based contracts that pay for value (outcomes), rather than volume outcomes (fee-for-service). However, a recent report published in Health Affairs found that fewer than half of clinical delivery providers working under ACO arrangements had adequate data tools or a platform to gauge SDoH factors adequately.[1]
Further, the authors found that the majority relied on partnerships with community-based organizations to refer patients for social services. The study also concluded that:
“Most ACOs were not yet planning to share data with community partners regarding their shared patients, let alone actually sharing such data. Critically, innovation among the ACOs in our sample was constrained by the lack of financial support. ACOs that were considering social service integration from an ROI perspective were not yet confident in their capacity to calculate the ROI for this work.”
AI Powers SDoH-Solution for Health System With Nearly 900 locations
Despite such challenges, one large Iowa-based health system, MercyOne, is satisfied that an AI-powered platform is very useful in managing SDoH issues to lower costs and improve outcomes.[2] MercyOne cares for just more than three million patients, with more than 20 value-based contracts, and operates 870 locations in collaboration with 181 organizations. MercyOne thinks of a SDoH solution as “moving upstream” to better manage outcomes.
“We started our journey focused on clinical data elements from claims and the EHR,” said Nathan Riggle, division director of analytics for the MercyOne population health services organization. “We have spent time on that data to understand our patients’ risk and how we can serve them better. But as we have evolved as a population health organization, we have moved to other sources of public data, such as ZIP code-level data from the Iowa Department of Health. But what is really transformative is collecting data from patients and having them tell us about their social needs. We are finding new uses of social data and patient-reported data.”[3]
Riggle and his team developed an iPad app to screen patients for social needs in the systems waiting rooms. In a year, after screening 11,000 patients, the app identified 700 patients with SDoH needs. Emily Fletcher, ambulatory care program manager for MercyOne, admitted that she was initially naïve about the impact of social factors on health.
One-in-Five Patients Identified With SDoH Needs
“I would never have guessed that one in five patients has social needs, such as challenges with feeding their family or paying rent. When you have those things going on, how do you focus on your A1c or taking your medications correctly? We often wonder why a patient isn’t following through on our guidance,” she said. “This brought a new understanding, and a new lens for providers to see the total health of the patient and what is going on outside the clinical setting.”
Fletcher stresses that community health workers are crucial to their program’s success. She gave an example of a patient, despite food insecurity, who was unable to travel to a food pantry by public transportation and wait in line. Local service providers were able to arrange alternative transportation to take the patient to the pantry during off-hours.
CDC Identifies Five SDoH Categories
The CDC defines five broad SDoH categories;[4]
1: Genetics: Related to physical and cognitive disabilities, as well as onset diseases.
2: Behavior: Use of alcohol and drugs, smoking, and dietary choice.
3: Physical influences: Housing and food inadequacy, as well as the effects of violence.
4: Medical care: Access to care and insurance coverage.
5: Social factors: Income and education levels.
In addition to patient surveys, the MercyOne development team also integrated data from more than 100 disparate clinical systems, including 15 different proprietary electronic health records. Other data sources included payer claims, billing files, scheduling data, public data, and admission discharge transfers files from more than 35 hospitals.
Blog Sources:
https://www.healthaffairs.org/doi/full/10.1377/hlthaff.2019.01266
