Back

Developing a Tiered Machine Learning Alert System for Real-Time Suicide Risk Detection in a Digital Mental Health Setting

Donegan, M. L.; Srivastava, A.; Peake, E.; Swirbul, M.; Ungashe, A.; Rodio, M. J.; Tal, N.; Margolin, G.; Benders-Hadi, N.; Padmanabhan, A.

2026-03-30 psychiatry and clinical psychology
10.64898/2026.03.26.26349452 medRxiv
Show abstract

The goal of this work was to leverage a large corpus of text based psychotherapy data to create novel machine learning algorithms that can identify suicide risk in asynchronous text therapy. Advances in the field of natural language processing and machine learning have allowed us to include novel data sources as well as use encoding models that can represent context. Our models utilize advanced natural language processing techniques, including fine-tuned transformer models like RoBERTa, to classify risk. Subsequent model versions incorporated non-text data such as demographic features and census-derived social determinants of health to improve equitable and culturally responsive risk assessment, as well as multiclass models that can identify tiered levels of risk. All new models demonstrated significant improvements over our previous model. Our final version, a multiclass model, provides a tiered system that classifies risk as "no risk," "moderate," or "severe" (weighted F1 of 0.85). This tiered approach enhances clinical utility by allowing providers to quickly prioritize the most urgent cases, ensuring a more accurate and timely intervention for clients in need.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

1
Frontiers in Psychiatry
83 papers in training set
Top 0.1%
25.8%
2
Journal of Medical Internet Research
85 papers in training set
Top 0.3%
12.3%
3
BMC Medical Informatics and Decision Making
39 papers in training set
Top 0.3%
8.4%
4
Frontiers in Digital Health
20 papers in training set
Top 0.1%
4.8%
50% of probability mass above
5
npj Digital Medicine
97 papers in training set
Top 0.9%
4.8%
6
Acta Psychiatrica Scandinavica
10 papers in training set
Top 0.1%
4.3%
7
PLOS ONE
4510 papers in training set
Top 40%
3.6%
8
PLOS Digital Health
91 papers in training set
Top 1%
2.1%
9
BioData Mining
15 papers in training set
Top 0.2%
1.9%
10
Scientific Reports
3102 papers in training set
Top 54%
1.9%
11
Frontiers in Artificial Intelligence
18 papers in training set
Top 0.3%
1.7%
12
Nature Medicine
117 papers in training set
Top 2%
1.7%
13
JAMIA Open
37 papers in training set
Top 0.9%
1.7%
14
JMIR Formative Research
32 papers in training set
Top 0.9%
1.5%
15
Journal of the American Medical Informatics Association
61 papers in training set
Top 1%
1.5%
16
Journal of Affective Disorders
81 papers in training set
Top 1%
1.2%
17
BJPsych Open
25 papers in training set
Top 0.5%
1.2%
18
Journal of Biomedical Informatics
45 papers in training set
Top 1%
1.2%
19
IEEE Journal of Biomedical and Health Informatics
34 papers in training set
Top 1%
1.2%
20
Psychiatry Research
35 papers in training set
Top 1%
1.1%
21
JMIRx Med
31 papers in training set
Top 1%
0.9%
22
European Psychiatry
10 papers in training set
Top 0.6%
0.9%
23
International Journal of Medical Informatics
25 papers in training set
Top 1%
0.9%
24
Life
27 papers in training set
Top 0.5%
0.7%
25
Bioengineering
24 papers in training set
Top 2%
0.7%
26
JAMA Pediatrics
10 papers in training set
Top 0.2%
0.7%
27
Proceedings of the National Academy of Sciences
2130 papers in training set
Top 47%
0.6%
28
Computational Psychiatry
12 papers in training set
Top 0.2%
0.6%