Previous articles in this series showed why social determinants of health (SDoH) are the hottest recent addition to data used by the medical industry. We’ve explored how SDoH data is defined and shared. This article shows how organizations are collecting it.
Andrew Eye, CEO of the healthcare data science company ClosedLoop, estimates that about 15% of physicians collect SDoH-related data from patients and use it to assess their needs. The percentage is much higher in some areas.
Data can be spotty, according to Eye, who says for instance that data from Centers for Medicare & Medicaid Services (CMS) often lacks racial identifications.
He also says that collecting and processing the data is labor-intensive, because it can be hard to persuade patients to talk openly about financial and logistical problems. At the institutions served by ClosedLoop, case managers or care managers usually reach out to patients by phone. Some of the patients who need support for SDoH don’t visit a PCP. When they do, the doctor may be pressed for time and unable to consider SDoH.
Because of the time required to process SDoH, Eye says, institutions need to prioritize which patients to contact. ClosedLoop applies machine learning to risk-stratify patients. It uses data collected from patients as well as ZIP code or census block data on populations.
Personal and Aggregated Data
The companies I talked to for this series rely mostly on two kinds of SDoH data. The first is personal: information collected from the patient about their needs for food, transportation, in-home support, etc. The second is aggregate data collected by the census or other institutions, and provided by ZIP code or census block.
The HHS roundtable listed in my previous article cited lack of standardization, poor data sharing, and lack of financial incentives as problems in collecting and sharing data (pages 21- 22). But because the report was issued four years ago, we have seen much improvement since then; for instance, Gravity has matured. The standards discussed in the previous article, including Gravity, represent progress but are not yet widely used because there’s no formal enforcement of their use.
The CDC has assembled a directory of state and local data sets about SDoH.
Strategic Solution Group, LLC (SSG) has a low-code software platform called Casetivity. According to vice president Ted Hill, it can modernize how public health agencies serve today’s needs. Casetivity can ingest, standardize, and share health data. Applications of Casetivity are far-ranging in public health, including case management for Early Intervention Part C, Electronic Disease Surveillance, Immunization Information Systems, and addressing SDoH problems such as lead poisoning and family health.
Hill believes public agencies (and those that support them) should utilize modern national/international standards and guidelines for how public health data can be collected and formatted to provide more consistency and better data quality for reporting and analysis. The variety of organizations that provide data (clinical settings, labs, community-based organizations, etc.) do so in different ways and often have incomplete data. Some sophisticated organizations offer APIs for access to data, but others still fax or mail in a scanned document.
Dr. Victor Lee, vice president of clinical informatics at Clinical Architecture, points out that there are no guidelines about who collects SDoH data (a doctor, a medical assistant, a care manager, etc.). In researching this article, I’ve seen that institutions collecting the data find a variety of ways to integrate it with various staff workflows.
In a health data quality survey of 83 healthcare professionals, conducted by Clinical Architecture, SDoH quality came out at the bottom of seven types of data they rated (pages 14, 24).
Sedgwick is a third-party claims administrator that works with 78 of the Fortune 100 companies. Judiann McCrone Romeo, director clinical at Sedgwick, notes, “We’re beginning to see companies employ AI and machine learning to identify significant combinations of words in unstructured data that could signal social determinants of health in a claim. The hope is that through AI, companies will be able to identify red flags earlier in the claim so that any additional support needed would be addressed before a larger concern arises.”
Non-Health Sources For Data
A few companies use data from insurance companies or LexisNexis Risk Solutions. According to Jessica Probst, senior specialist, real world evidence at OM1, insurance data can tell you a person’s employment status, average household income, credit scores, and more. OM1 combines this SDoH information with more traditional sources of health data, like medical records and insurance claims, to generate insights about treatment patterns, patient care, and healthcare outcomes in the real world.
LexisNexis Risk Solutions is often consulted by banks and other institutions evaluating an individual’s financial risk, and many of the datapoints used for those assessments constitute SDoH as well. I talked to Diana Zuskov, associate vice president for healthcare strategy at LexisNexis Risk Solutions, about uses for their data. One use appeals to payers or providers using a value-based reimbursement model, which underlies most of the innovation in chronic healthcare. The other appeals to researchers in the life sciences and clinical settings.
Value-based systems need to find out what’s driving costs and outcomes: daunting medication costs, lack of transportation, etc. The answers can help the organizations decide where to start new programs and deploy their resources generally. Pharmacies can also use the information to choose which services to launch. Researchers are increasingly concerned with diversity and equity for clinical trials and behavioral interventions.
Data from LexisNexis Risk Solutions is always stored at the individual level. The source of the data is mostly public records, The company validates its relevance for questions such as the person’s educational level, how stable their housing is, and how good their access is to food and healthcare providers. They take several steps to improve data accuracy.
SV Krishnan of Sagility says that healthcare institutions can get useful SDoH from many companies. Such data includes income level, the number of people in a household, what kind of car they drive, and so on. Such data has to be combined with clinical data to determine the patient’s risks and how to avert them. And some institutions with responsibility for patients—such as long-term care—don’t have all the clinical data.
Zuskov points out that hospital data doesn’t represent the full population. Payers need to understand everyone they cover—especially ones who fall outside the healthcare system.
Hands-On Data Collection
I talked to Dr. Geoff Rutledge, chief medical officer of HealthTap, a company that came early to telehealth and that I covered several times as far back as 2011. He believes that the doctor should discuss SDoH with patients, collecting data and discussing its implications. He doesn’t believe that passive data collection provides the necessary information or helps the doctor find solutions with the patient.
Dr. Sandhya Gardner, CMO of HealthEdge Clinical Solutions, emphasized the importance of gaining patient trust in order to get good information on SDoH. She says that it’s still hard to collect SDoH in a standardized manner. But there are clinically validated tools to ask about some problems.
Rutledge and Tina Burbine of Healthlink Advisors say that in-patient visits can turn up aspects of SDoH that you wouldn’t get in an office visit or even a virtual conference. A visiting nurse can look in the kitchen cabinets and refrigerator, for instance, to find out whether the patient has enough food and what its quality is. Furthermore, patients tend to be more relaxed and more comfortable discussing personal and sensitive issues in their homes than in a clinical office.
Burbine says that a variety of medical specialties can be certified for home visits: RNs, medics, and others.
Various organizations have created standard questionnaires to collect data. You might well have been handed one of these during a recent doctor’s visit. The popular PHQ9 questionnaire checks for symptoms of depression, ranging from trouble sleeping to thoughts of suicide.
The results of these standardized assessments are often documented in unstructured notes within patients’ medical records, complicating their access for research. OM1 uses medical language processing (MLP) technology to extract these patient-reported outcomes (PROs) and convert them into structured variables useful to researchers. One example cited by Probst from one of their recent studies found that, among patients diagnosed with major depressive disorder, black patients disproportionately experienced more severe disease.
Many organizations are issuing patient surveys using the Protocol for Responding to and Assessing Patients’ Assets, Risks and Experiences (PRAPARE). Translated into several languages, the survey contains 21 questions about race, finances, personal safety, and more.
Rutledge cited a study (Basu S, Berkowitz SA, Davis C, Drake C, Phillips RL, Landon BE. Estimated Costs of Intervening in Health-Related Social Needs Detected in Primary Care. JAMA Intern Med. 2023;183(8):762–774. doi:10.1001/jamainternmed.2023.1964) finding that incorporating SDoH into clinical visits costs $55-$65 per member per month.
Matt Hollingsworth, CEO of Carta Healthcare, says SDoH is easier to capture than clinical data through natural language processing (NLP). Carta Healthcare uses AI to automate some of the data entry associated with patient data. In addition to clinical clients, they work with registries that accept information from different clinical sites in order to produce comparative quality information. Due to the huge legacy data sets and complex processes used at registries, adding new fields to their data definitions is difficult, but some registries have started to store data on SDoH.
Tracking the Elderly
One important type of information, concerning the locations and services offered by senior care facilities, is provided by NIC MAP Vision. The service pulls together the profiles maintained by facilities and local governments and combines relevant information to display to providers.
To help seniors find housing or long-term care, CEO Arick Morton told me that each state has a standardized intake instrument for a senior’s data, covering their medical history, personal care needs, and SDoH components. Each residence also maintains an additional document that helps their staff integrate the senior into social activities, ensure proper nutrition, etc.
Jeff Dorkin, vice president of new product development, said that they base analytics on data drawn from Medicare claims, which cover the whole senior population of the U.S. For strategizing, clients can get analytics at the building level or at a larger level.
The next article in this series will show how some institutions are employing SDoH to make improvements in their service and in their patients’ lives.
