It’s been a slow evolution, but Americans have finally begun to appreciate that patient health depends on much more than the healthcare that patients receive, with non-clinical factors such as social determinants of health (SDoH) playing a critical role.
Even Congress has come to realize the importance of SDoH, as evidenced by the recently proposed Social Determinants Accelerator Act, a bipartisan bill that would allocate tax dollars toward grants and technical assistance for community groups that help patients address SDoH.
That social determinants are enjoying a moment-in-the-sun undeniably represents progress toward a more wholistic approach to healthcare. However, advances toward better use of social determinants information threaten to be stalled by a common healthcare industry problem: messy data.
To effectively address SDoH, healthcare organizations (HCOs) must first gather data on patients’ social determinants, then analyze it and most importantly find ways to translate that data into actionable insights that lead to improvements in population health management.
The problem for many clinicians is that they don’t know what they don’t know about their patients’ SDOH. Often, clinicians make treatment decision without the benefit of this information and don’t gain awareness of social determinants until patients have already experienced negative health outcomes.
Generally, the problem is not that HCOs fail to gather data on social determinants. More often, they have most of the information they need in the EHR, but experience difficulty in accessing, analyzing and leveraging the data for improved outcomes. The culprit frequently is unstructured data, usually in the form of clinical notes that is essentially hiding in clinical information systems.
For example, it’s estimated that 80% of clinical data in the EHR is unstructured and, as a result, difficult to analyze. Data such as clinician narratives, nurse notes, radiology reports, discharge summaries and patient-reported information have the ability to contribute a wealth of useful clinical information, but are often unusable, depriving HCOs of a valuable means of improving population health.
NLP’s role in improving population health
To overcome problems caused by unstructured data, many HCOs are investing in Natural Language Processing (NLP) technology to support extraction of SDoH and other clinical concepts, which automates the identification and extraction of key concepts from large volumes of clinical documentation, effectively unlocking unstructured data. With the benefit of accurate and comprehensive SDoH information, HCOs are able to make better-informed decisions about population health.
To address population health, HCOs must first understand the health risks that their patients face. After analyzing and stratifying their patient populations by risk, HCOs may then develop evidence-based care plans and safety net programs tailored to each patients’ conditions and level of risk.
However, these plans and programs cannot be effective without factoring in patients’ SDoH, largely because only 10% of the factors affecting premature death are related to clinical care, according to the Center for Disease Control. Another 30% of factors relate to genetics, meaning that 60% of factors impacting premature death are based on a combination of social/environmental factors (20%) and behavior (40%). Numerous studies have shown how addressing SDoH benefits health outcomes, but to highlight a recent example: Medicaid members who received SDoH screenings from community-based organizations experienced a nearly 30% percent decline in inpatient hospital admission rates, according to AmeriHealth Caritas.
A major challenge for HCOs when launching population health programs involves analyzing heterogenous and unstructured data. Without employing advanced technologies, HCOs may face little choice but to turn to expensive and time-consuming manual chart reviews. Alternatively, NLP’s ability to automatically extract precise data customized to the HCOs needs from unstructured text can quickly and efficiently enable clinicians to gain insight into individual patients’ SDoH, such as smoking and alcohol consumption, living arrangements, access to care and mobility status.
NLP in action: Improving diabetes population health
To understand how NLP can help deliver insights that lead to improvements in population health, consider how an accountable care organization (ACO) could leverage the technology to improve care for its diabetes patients. Diabetes and prediabetes affect more than 100 million Americans, with the cost of care for people with diabetes averaging $16,752 per year and accounting for approximately 25% of U.S. healthcare spending, according to a study in Diabetes Care.
Social and economic factors are closely related to diabetes risk, and while an analysis of structured data is likely to list risk factors related to weight and age, the analysis is likely to miss important information contained in clinicians’ notes. In contrast, the ACO could leverage NLP to analyze patient records to reveal many other additional risk factors, such as limited access to proper medications and healthy foods, barriers to physical activity, high stress levels and social isolation.
Further, several of the signs of early diabetes – mentions of excessive thirst or hunger, frequent urination, fatigue, and blurred vision – generally appear in free text, while information pertaining to laboratory values such as hemoglobin A1c and blood glucose levels also may be missed when relying on structured data alone. As a result of these deeper insights enabled by NLP, the ACO has a roadmap for developing a population health program that is more personalized and responsive to the needs of its diabetes patients.
Although the healthcare industry has come a long way in recognizing the importance of SDoH, many HCOs lag in their ability to transform SDoH data into actionable intelligence that augments patient care. NLP technology represents an antidote that can deliver new insights from previously hidden information, enabling HCOs to improve population health by optimizing the data they already possess.