Modeling Recurrent Suicide Attempts Using Probabilistic Hawkes Processes

A history is more than a count.

Recurrent events contain information in their timing, not only their number. This study investigates how probabilistic temporal models can describe the dynamics of recurrent suicide attempts in a longitudinal cohort.

Modeling event history.

A Hawkes process is a self-exciting temporal point process: an event raises the conditional intensity of future events, and its effect can decay over time. The study conditions process parameters on clinical features through a neural network, generating individualized intensity and survival curves. A Gaussian mixture model is used to investigate temporal profiles.

Findings and limits.

The analysis finds temporally clustered recurrence and shows how Hawkes-based modeling can represent changing intensity. Prospective validation is needed to establish clinical integration, scalability and utility. This is research, not a clinical prediction tool or medical advice. The demonstration below uses synthetic events and is not fitted to clinical data.

My role.

I am the first author of this collaborative paper in the Spanish Journal of Psychiatry and Mental Health, published online in 2026.