Investigating the differential impact of psychosocial factors by patient characteristics and demographics on Veteran suicide risk through machine learning extraction of cross-modal interactions
Accurate prediction of suicide risk is crucial for identifying patients with elevated risk burden, helping ensure these patients receive targeted care. The US Department of Veteran Affairs' suicide prediction model primarily leverages structured electronic health records (EHR) data. This approach largely overlooks unstructured EHR, a data format that could be utilized to enhance predictive accuracy. This study aims to enhance suicide risk models' predictive accuracy by developing a model that incorporates both structured EHR predictors and semantic NLP-derived variables from unstructured EHR. XGBoost models were fit to predict suicide risk- the interactions identified by the model were extracted using SHAP, validated using logistic regression models, added to a ridge regression model, which was subsequently compared to a ridge regression approach without the use of interactions. By introducing a selection parameter, α, to balance the influence of structured (α=1) and unstructured (α=0) data, we found that intermediate α values achieved optimal performance across various risk strata, improved model performance of the ridge regression approach and uncovered significant cross-modal interactions between psychosocial constructs and patient characteristics. These interactions highlight how psychosocial risk factors are influenced by individual patient contexts, potentially informing improved risk prediction methods and personalized interventions. Our findings underscore the importance of incorporating nuanced narrative data into predictive models and set the stage for future research that will expand the use of advanced machine learning techniques, including deep learning, to further refine suicide risk prediction methods.
Abstract: Introduction: Persistent inequities exist in obstetric and neonatal outcomes in military families despite universal health care coverage. Though the exact underlying cause has not been identified, social determinants of health may uniquely impact military families. The purpose of this study was to qualitatively investigate the potential impact of social determinants of health and the lived experiences of military individuals seeking maternity care in the Military Health System. Materials and methods: This was an Institutional Review Board-approved protocol. Nine providers conducted 31 semi-structured interviews with individuals who delivered within the last 5 years in the direct or purchased care market. Participants were recruited through social media blasts and clinic flyers with both maximum variation and homogenous sampling to ensure participation of diverse individuals. Data were coded and themes were identified using inductive qualitative research methods. Results: Three main themes were identified: Requirements of Military Life (with subthemes of pregnancy notification and privacy during care, role of pregnancy instructions and policies, and role of command support), Sociocultural Aspects of the Military Experience (with subthemes of pregnancy as a burden on colleagues and a career detractor, postpartum adjustment, balancing personal and professional requirements, pregnancy timing and parenting challenges, and importance of friendship and camaraderie in pregnancy), and Navigating the Healthcare Experience (including subthemes of transfer between military and civilian care and TRICARE challenges, perception of military care as inferior to civilian, and remote duty stations and international care). Conclusions: The unique stressors of military life act synergistically with the existing health care challenges, presenting opportunities for improvements in care. Such opportunities may include increased consistency of policies across services and commands. Increased access to group prenatal care and support groups, and increased assistance with navigating the health care system to improve care transitions were frequently requested changes by participants.