Associations Between Natural Language Processing–Enriched Social Determinants of Health and Suicide Death Among US Veterans

Abstract

ImportanceSocial determinants of health (SDOHs) are known to be associated with increased risk of suicidal behaviors, but few studies use SDOHs from unstructured electronic health record notes.ObjectiveTo investigate associations between veterans’ death by suicide and recent SDOHs, identified using structured and unstructured data.Design, Setting, and ParticipantsThis nested case-control study included veterans who received care under the US Veterans Health Administration from October 1, 2010, to September 30, 2015. A natural language processing (NLP) system was developed to extract SDOHs from unstructured clinical notes. Structured data yielded 6 SDOHs (ie, social or familial problems, employment or financial problems, housing instability, legal problems, violence, and nonspecific psychosocial needs), NLP on unstructured data yielded 8 SDOHs (social isolation, job or financial insecurity, housing instability, legal problems, barriers to care, violence, transition of care, and food insecurity), and combining them yielded 9 SDOHs. Data were analyzed in May 2022.ExposuresOccurrence of SDOHs over a maximum span of 2 years compared with no occurrence of SDOH.Main Outcomes and MeasuresCases of suicide death were matched with 4 controls on birth year, cohort entry date, sex, and duration of follow-up. Suicide was ascertained by National Death Index, and patients were followed up for up to 2 years after cohort entry with a study end date of September 30, 2015. Adjusted odds ratios (aORs) and 95% CIs were estimated using conditional logistic regression.ResultsOf 6 122 785 veterans, 8821 committed suicide during 23 725 382 person-years of follow-up (incidence rate 37.18 per 100 000 person-years). These 8821 veterans were matched with 35 284 control participants. The cohort was mostly male (42 540 [96.45%]) and White (34 930 [79.20%]), with 6227 (14.12%) Black veterans. The mean (SD) age was 58.64 (17.41) years. Across the 5 common SDOHs, NLP-extracted SDOH, on average, retained 49.92% of structured SDOHs and covered 80.03% of all SDOH occurrences. SDOHs, obtained by structured data and/or NLP, were significantly associated with increased risk of suicide. The 3 SDOHs with the largest effect sizes were legal problems (aOR, 2.66; 95% CI, 2.46-2.89), violence (aOR, 2.12; 95% CI, 1.98-2.27), and nonspecific psychosocial needs (aOR, 2.07; 95% CI, 1.92-2.23), when obtained by combining structured data and NLP.Conclusions and RelevanceIn this study, NLP-extracted SDOHs, with and without structured SDOHs, were associated with increased risk of suicide among veterans, suggesting the potential utility of NLP in public health studies.

Efficient Determination of Social Determinants of Health From Clinical Notes for Timely Identification of Suicidality Among US Veterans

“…Our current approach to identifying individuals who need help is inadequate. Novel technologies that leverage existing information and resources to enable life-saving interventions, such as the approach proposed by Mitra et al, 1 are greatly welcome and should be explored for creating universal screening deployments.…”

“…Thus, while the use of SDOHs to assess SI/SA risk has indeed been explored before, as discussed by the authors, 1 the novelty here lies in the source of this information. SDOH data are typically not available publicly for various privacy reasons, and manually collecting such information can be expensive.…”

“…However, adjusting for population age and sex differences, the current suicide rate for veterans is still 57.3% greater than for nonveteran US adults; the unadjusted rate of suicide in 2020 among US veterans was 31.7 per 100 000, 5 about twice that of the general population (13.5 per 100 000). The work of Mitra et al 1…”

“…SDOH data are typically not available publicly for various privacy reasons, and manually collecting such information can be expensive. Mitra et al 1 suggest that we generate such information from the clinical notes included in standard EHR via NLP. Thus, the authors' approach can indeed provide the foundation for addressing the key hurdles in enacting efficient universal assessment for suicide risk among the veterans and perhaps in the general population.…”

Using natural language processing to study homelessness longitudinally with electronic health record data subject to irregular observations

“…In recent years, Electronic Health Records (EHR) have increasingly been utilized as a data source for studying homelessness and other adverse SDoH. [1][2][3][4][5][6][7][8][9][10] Many of these studies have utilized natural language processing (NLP), a set of techniques for extracting information from unstructured clinical texts. Improved measurement of housing instability offers a number of benefits to researchers and policymakers such as the ability to study risk factors for becoming homeless or the effectiveness of homelessness interventions.…”