|Suicide Risk Can Be Predicted Using In-Person Screening and Machine Learning|
In-Person Screening and Machine Learning may help in the prediction of Suicide Risk
A new study shows that combining face-to-face screening with an EHR-based machine learning model can help clinicians accurately predict and classify suicide risk in adults.
The researchers found that suicide risk predictions for adult patients are significantly improved when using an approach that combines in-person screening with an EHR-based machine learning (ML) model.
The study, published in JAMA Network Open earlier this month, finds that 800,000 people worldwide die by suicide each year, made worse during the COVID-19 pandemic.
While suicidal behavior is increasing in the US, the rate at which patients proactively disclose suicidal thoughts and behaviors is not. The researchers also point out that mental health diagnoses are often missing from the medical records of those who die by suicide. This combination of factors makes better risk identification essential to improve outreach and prevention, they noted.
Previous research has shown that universal face-to-face detection and artificial intelligence (AI) prediction models have achieved high performance in predicting suicide risk on their own, so the researchers decided to study the success of an approach which uses both.
The researchers studied an observational cohort of 83,394 patients at Vanderbilt University Medical Center (VUMC) from June 2019 through September 2020. During that time, these patients had 120,398 medical center encounters in the settings of hospitalization, outpatient surgery and emergency department. All patient data from these encounters were extracted from Vanderbilt Research Derivative, the VUMC clinical research repository.
VUMC implemented universal suicide risk assessment in June 2018 using the Columbia Suicide Severity Rating Scale (C-SSRS), a standardized assessment of suicidal ideation and behavior. Since this data was already routinely collected at VUMC, the researchers chose to use it for the face-to-face screening aspect of their study.
For their prediction model, they used the Vanderbilt Suicide Attempt and Ideation Likelihood (VSAIL) model, which has generated real-time suicide risk predictions across the VUMC system since June 2019. The model uses historical EHR data to predict suicide risk. suicide risk, including demographic data. such as age, gender, and race, diagnosis codes, medication data, prior health care use, and the patient’s Zip Code Area Deprivation Index, which is a multidimensional assessment of a region’s socioeconomic conditions and how they may relate to health results.
The researchers extracted C-SSRS response data and corresponding VSAIL risk scores for each patient tested. They then compared the ensemble model, which combined prediction results from C-SSRS and VSAIL, with each method individually.
To predict suicide risk, the researchers measured suicidal ideation (SI) and suicide attempt (SA), defined by self-injurious thoughts and behaviors (SITB) coded in ICD-9-CM and ICD-10-CM, that occurred within seven, 30, 60, 90, and 180 days after the discharge date of each documented visit during the study period.
Overall, SI was reported in 614 (0.51 percent), 1,486 (1.23 percent), 2,036 (1.69 percent), 2,433 (2.02 percent), and 3,126 (2.60 percent). ) cases at seven, 30, 60, 90 and 180 days. respectively. SA was documented in 84 (0.07 percent), 205 (0.17 percent), 272 (0.23 percent), 356 (0.30 percent), and 514 (0.43 percent) cases. the seven, 30, 60, 90 and 180 days, respectively.
At all time intervals, the ensemble model outperformed the C-SSRS and VSAIL models when used individually.
These findings indicate that combining pretrained ML models with structured clinical assessment to create ensemble prediction systems for suicide risk could significantly improve outcomes. However, the researchers note that their study will need to be replicated and validated to achieve greater generalizability.