How Social, Behavioral Determinants of Health Can Enable Next-Gen Risk Stratification

June 21, 2019
Jean Drouin
Jean Drouin

,
Michelle Chiu
Michelle Chiu

While generally used at a population level, addressing SBDoH factors can and should be personalized to individuals.

Social and behavioral determinants of health (SBDoH) have been part of the healthcare lexicon for over 20 years. The notion that 60% of health outcomes are driven by demographic, socioeconomic, and environmental factors is practically cliché within the health innovation space.

Yet, traditional models of population health and care delivery tend to focus on reactive interventions for individuals who have chosen to engage with the healthcare system. Patients often wait far longer than they should to access services, challenged by issues such as cost, transportation, childcare, time, and lack of knowledge. As a result, the root causes driving worsening health, disease progression, and poor outcomes either are partly or completely unaddressed. 

Payers and providers are still only scratching the surface when it comes to using SBDoH to personalize care. Meanwhile, Amazon routinely predicts exactly what goods an individual needs to re-order and proactively suggests new book titles that might be of interest. Inspired by progress in other industries, recent advances in both data availability and analytics now make it possible to capitalize on the actionable use of SBDoH to deliver better healthcare outcomes and reduce costs.

Using SBDoH data to better understand members 

Payers have historically used SBDoH at the population level, for example in looking at how community and ZIP code differences might impact member risk and highlight environmental needs. While recognizing the importance of environmental challenges is one step in the right direction, ZIP code analytics alone run the risk of treating individuals according to socio-economic and geographic stereotypes.

Let’s take downtown San Francisco for instance, where two people within the same zip code may sit on completely opposite ends of the income scale. Given the discrepancies that can be associated with using population-level SBDoH in individual-level risk stratification, why aren’t individually-attributed SBDoH being used more widely? This is something that retailers and banks do as a routine matter. Can we imagine a bank deciding on whether to approve an individual mortgage or loan with zip code level analysis?

In order to effectively evaluate and manage risk, payers and providers need to personalize their understanding of individuals, rather than group them into faceless cohorts. The promise of SBDoH is to enable earlier identification of at-risk and non-engaged members, increased focus on impactable members with addressable social risk factors, and better matching of interventions to members.

Related: How One Health Plan Is Successfully Addressing SDOH

Applying SBDoH to real-world applications

The ability to precisely sort members to interventions such as mail-order prescriptions, food assistance, and Uber Health rides isn’t a pipe dream-it’s a reality. We can now shift from zip code to individually-attributed SBDoH and create analytics engines that can bring the power of consumer data to bear, powered by the marriage of clinical knowledge with machine learning technology.

Rather than replacing providers, precise individual-level risk assessment augments their effectiveness by targeting the right care resources at underlying root causes.

Two of the most widely used SBDoH questionnaires, AHC and PRAPARE, include 20+ questions, although the average primary care visit is only 16.5 minutes. By contrast, using appropriately curated consumer data can rapidly focus providers and care managers on the likely root causes that really matter.

For instance, knowing that a patient recently lost her job might alert the clinician to the reality that the inability to afford a drug would lead to non-adherence of the recommended treatment. Similarly, research we have conducted at Clarify Health suggests that while low income only increases admission risk by 9% among seniors, this quadruples to a 36% odds increase among senior diabetics.

What’s required to succeed?

1. Comprehensive longitudinal data

Consumer data offers a treasure trove of individualized insights over time, enabling analytics not possible with SBDoH questionnaires’ limited member reach and standardization across providers.

Precise individual-level healthcare risk assessment can borrow from the progress within consumer data analytics, as long as these techniques are coupled with rigorous data evaluation processes to ensure the reliability of information being used in patient care. Quickly understanding and evaluating data issues, from 11-month old credit agency attributes to non-random missingness, is a key first step to leveraging the growing realm of commercial information.

Analytics engines designed to make these assessments in a standardized and repeatable way will be well-positioned to leverage the growth in consumer data towards generating a more holistic understanding of members.

2. The ability to filter signal from noise

The complex network of factors that impact health can make it difficult to understand which factors are important, and for which individuals. For instance, certain occupations increase risk of contracting and exacerbating COPD, and many high-functioning seniors live alone successfully, but those transitioning from living with loved ones to living alone find themselves at increased risk.

In fact, many of these signals are well-studied in academic clinical and consumer behavioral research. Applying these research findings, often catered to specific cohort populations, to high-powered analytics is one path to identifying signals within noise.

The initial thinking on how to identify the right cohorts and right variables already exists and leveraging clinician-supervised machine learning can enable payers to deftly graduate up from simple analytic frameworks or unsupervised algorithms lacking common care provider insights.

3. Predictive power and specificity

Patient comorbidities and historical utilization are indisputably core drivers of future health events, though personalizing individuals’ SBDoH profiles will certainly bring additional predictiveness to the table. Challenges increasing modeling advantage from SBDoH are often limited by training populations that are too diffuse, missing key signals across underlying patient segments.

Leveraging SBDoH to improve risk prediction is only the beginning.

Applying the psychographic segmentation principles more typically associated with the consumer industry will enable better assessment of individuals’ willingness to engage in their health, and the best methods and mediums by which to do so. The era of checking boxes in standard patient questionnaires or relying magically on providers heroically finding additional hours to spend with patients to understand their full context, is coming to a close.

Payers and providers alike are now better placed than ever to identify at-risk members, develop more effective interventions, and ultimately reduce costs and adverse events.

Jean Drouin, CEO, brings 20 years of experience in healthcare management, strategy, operations, finance, and cultural change to Clarify Health. As the CEO, Drouin leads the development and direction of the company strategy and commercial teams. Michelle Chiu is a director of clinical transformation at Clarify Health Solutions and leads product development and commercialization strategy for its Member Risk Management solution.