Abstract:Background Current research on firearm violence is largely limited to patients who received care in emergency departments or inpatient acute care settings or who died. This is because standardized disease classification codes for firearm injury only represent bodily trauma. As a result, research on pathways and health impacts of firearm violence is largely limited to people who experienced acute bodily trauma and does not include the estimated millions of individuals who were exposed to firearm violence but did not sustain acute injury. Assessing and collecting data on exposure to firearm violence in ambulatory care settings can expand research and more fully frame the public health issue. Objective The aim of the study is to evaluate the demographic and clinical characteristics of patients who self-reported exposure to firearm violence during a behavioral health visit. Methods This study assessed early data from an initiative implemented in 2022 across a national network of ambulatory behavioral health centers to support trauma-informed care by integrating structured data fields on trauma exposure into an electronic health record behavioral health patient assessment form (SmartForm), as such variables are generally not included in standard outpatient medical records. We calculated descriptive statistics on clinic characteristics, patient demographics, and select clinical conditions among clinics that chose to implement the SmartForm and among patients who reported an exposure to firearm violence. Data on patient counts are limited to positive reports of exposure to firearm violence, and the representativeness of firearm exposure among all patients could not be calculated due to unknown variability in the implementation of the SmartForm. Results There were 323 of 629 (51%) clinics that implemented the SmartForm and reported at least 1 patient exposed to firearm violence. In the first 11 months of implementation, 3165 patients reported a recent or past exposure to firearm violence across the 323 clinics. Among patients reporting exposure, 52.7% (n=1669) were male, 38.8% (n=1229) were Black, 45.7% (n=1445) had posttraumatic stress disorder, 37.5% (n=1186) had a substance abuse disorder (other than nicotine), and 11.7% (n=371) had hypertension. Conclusions Current research on firearm violence using standardized data is limited to acute care settings and death data. Early results from an initiative across a large network of behavioral health clinics demonstrate that a high number of clinics chose to implement the SmartForm, resulting in thousands of patients reporting exposure to firearm violence. This study demonstrates that collecting standardized data on firearm violence exposure in ambulatory care settings is feasible. This study further demonstrates that resultant data from ambulatory settings can be used for meaningful analysis in describing populations affected by firearm violence. The results of this study hold promise for further collection of structured data on exposure to firearm violence in ambulatory settings.

A machine-learning prediction model to identify risk of firearm injury using electronic health records data

Development and validation of a machine learning algorithm for predicting the risk of postpartum depression among pregnant women

A machine learning-based approach for the prediction of periprocedural myocardial infarction by using routine data

Using Electronic Health Records and Machine Learning to Predict Postpartum Depression.

Machine learning classification of new firearm injury encounters in the St Louis region: 2010-2020

Prediction of the mechanism of suicide among Minnesota residents using data from the Minnesota violent death reporting system (MNVDRS)

Improving Cardiovascular Risk Prediction Through Machine Learning Modelling of Irregularly Repeated Electronic Health Records

Development of the SaFETy Score: A Clinical Screening Tool for Predicting Future Firearm Violence Risk

The prediction of asymptomatic carotid atherosclerosis with electronic health records: a comparative study of six machine learning models

Predicting Short Time-to-Crime Guns: a Machine Learning Analysis of California Transaction Records (2010-2021)

Machine Learning and Real-World Data to Predict Lung Cancer Risk in Routine Care

Improved accuracy and efficiency of primary care fall risk screening of older adults using a machine learning approach

Machine learning for suicide risk prediction in children and adolescents with electronic health records

Personalized prediction model specific for sudden cardiac death in the population: a machine learning study based on EHR

Using administrative data to identify U.S. Army soldiers at high-risk of perpetrating minor violent crimes

Predicting suicide death after emergency department visits with mental health or self-harm diagnoses

Early Results of an Initiative to Assess Exposure to Firearm Violence in Ambulatory Care: Descriptive Analysis of Electronic Health Record Data

Machine Intelligence for Outcome Predictions of Trauma Patients During Emergency Department Care

Abstract 347: Artificial Intelligence for Targeted and Personalized Prediction of Sudden Cardiac Death in the General Population: A Machine Learning Study Based on Electronic Health Records

A clinical risk prediction tool for identifying the risk of violent offending in severe mental illness: A retrospective case-control study

Machine learning computational model to predict lung cancer using electronic medical records