Today many tools exist to look at social risk based on census tract, for example. Companies are also utilizing the broad array of publicly available personal level data to build more individualized risk profiles. These tools are being incorporated into rate models for both insurance companies and providers. With new CMS requirements for systematic SDOH screening, screening data may also be incorporated into future rate models, as well as used to populate the SDOH modules in leading electronic health record products to influence clinical care.
Augintel is innovating in this area in an entirely new way. Augintel is incorporating the social risk information provided directly by the patient in their conversations with their caregivers in the course of their typical clinical sessions. This narrative information documented in patient notes, when pooled with other data—linked administrative data about the person, publicly available individual data from the web, and neighborhood-based data that likely represents the person—can provide a much more dynamic portrait of individual risk.
For providers of clinical services venturing into value-based care, the value of this data is two-fold. At the clinical level, it is critical to providing proactive care to that person, being responsive to their social risk, in the same way that care would be tailored to a clinical risk rating. In both cases, an elevated risk index for an individual provides a mandate to dig deeper into the reasons for that rating, to conduct outreach if the person is not getting regular care, and to build a realistic care plan that addresses that person’s vulnerabilities that are likely to result in both poor health and a high health care spend. The latter issue – high cost of care – is the impetus for the second reason risk ratings are critical to the provider organization. Risk and the estimated resources needed to provide appropriate care based on social, as well as medical complexity, form the basis of capitated rate negotiation. Similarly, in systems still relying at least in part on fee-for-service, these social risk ratings are critical to ensuring that the necessary services are reimbursable. Just as the diabetes prevention program (DPP) or cardiac and pulmonary rehabilitation are cost effective covered services, it is increasingly recognized that payment for community health workers and other patient navigators, as well as time-based billing, are critical for working successfully with clinically and socially complex patients.
At present, an organization’s understanding of its patients’ social risks is largely based on geography, payor type, and a handful of formal or informal case studies. Many stories get passed around and filter up to those responsible for negotiating with payors. Perhaps a Medical Director with firsthand knowledge is involved. Screening programs, using tools such as the AHC or PRAPARE tools are also increasingly of value, but only reach a small proportion of patients at this time. What is available, but untapped, is a compilation of social risks based on all of the patient’s stories, not just the select few that get used to illustrate a point.
Natural language processing (NLP) makes this compilation and quantification of risk factors possible. For example, using Augintel’s NLP model to extract these needs from the clinical notes of a random sample of 1000 patients of a community mental health center, 40% had evidence of inadequate housing, 42% of food insecurity, and 35% of domestic violence.
Armed with this information, providers and payors themselves—looking at employers and government purchasers of their managed care products—can better understand patient needs and ensure that adequate payment is in place for the right interventions, including addressing social factors, across the alternative payment model (APM) framework spectrum.