Speaker
Keith Marsolo, PhD
Associate Professor
Department of Population Health Sciences
Duke University School of Medicine
Keywords
PCORnet, Common Data Model, EHR, Social Determinants of Health
Key Points
- There are many different definitions of social determinants of health. The World Health Organization defines social determinants of health as non-medical factors that influence health outcomes and conditions in which people are born, grow, work, live, and age, and the wider set of forces and systems shaping the conditions of daily life.
- The PCORnet Common Data Model (CDM) includes data available from Clinical Research Networks. Some data, such as basic clinical data and demographics, are ready for research. Other data, such as immunizations, social determinants of health (SDOH), patient-generated data, and others, may or may not be in the PCORnet Common Data Model and require additional work for use in research.
- In 2021-2022, PCORI contracted with NORC at the University of Chicago to undertake a series of convenings to consider data infrastructure enhancements to PCORnet. A social determinants of health convening built upon efforts of prior PCORnet SDOH workgroup and included survey development, key informant interviews, and public webinars.
- When we talk about patient-level SDOH measures, the CDM has some general purpose tables that can store that data. Adding these data to the CDM generally involves several steps. Identifying whether there are codes to represent the measures in standard terminologies. Partners then must find the relevant measures within their EHRs and harmonizing them to the appropriate code. In many EHRs, data may be captured using various workflows over time, which can also affect the overall data completeness.
- For example, consider food security. 22 sites within the network were able to load some record of food security. There was a wide spread of information that was available. Insurance status is captured in an encounter level with the payor name. It is often that you have to take the raw values and harmonize to a particular insurance type (source of payment typology). Partners within the network have to take raw names and try to harmonize those to a specific type of insurance. It can be complicated to tease out by the insurance name.
- PCORnet has demonstrated that patient-level SDOH data can be incorporated to the CDM. Data availability is dependent on adoption and utilization by health systems. It may be suitable for studies on targeted populations but will depend on collection practices at a given health system.
- Area-level measures can provide population-level SDOH insights. 5-digit zip and county can be included in Limited Data Sets and are more easily used in distributed analytics. Capabilities for geocoding exist at many institutions but will require involvement of local personnel to generate values based on census tract or latitude/longitude. May be be best suited for specific studies.
Discussion Themes
-Have the data been used by PCORnet studies? It is new and has not been used widely across studies.
–What are you measuring with insurance status and can you capture in a model individual effect and social effect? We are working to get the best information we can. Having no insurance information about the patient can be problematic. We are trying to work with sites to find the right level of granularity when describing insurance. Insurance status does not highlight whether a patient lives in a food desert or in unsafe housing. It is important to look at what is available to people, what interventions are able to be done by the cluster data – by housing authorities, federal groups, and the health system. First steps are getting the data in such a state that would allow us to learn about some of those social determinants of health.
Tags
#pctGR, @Collaboratory1