THE PERMANENS PROJECT

Suicide constitutes a major, yet preventable, public health issue, representing an annual loss of 34.6 million years of life worldwide. For each suicide death, there are at least 20 suicide attempts, amounting annually to 16 million attempts and 160 million individuals with suicidal ideation and associated risk for suicide globally, in addition to at least five close family members or friends bereaved through each suicide.

People with self-harm are at elevated risk for repetition of self-harm, death by suicide, and for inadequate treatment delivery (i.e., care not consistent with clinical recommendations). Only 1 out of 6 patients with self-harm are directly referred to psychiatric services, and only half receive some psychiatric treatment in the months following self-harm. Those directly referred but without attendance have 3 times higher risk for suicide, highlighting the importance of increasing continuity of care among those patients.

To address this, emergency departments are key healthcare settings to detect high-risk patients and to intervene. They often represent the first medical contact after self-harm and can offer specialized risk assessment and referral to psychiatric intervention. Effective suicide prevention interventions exist (e.g., brief contact interventions following discharge after a suicide crisis, short-term psychotherapy), although it is unclear which interventions are most effective for particular suicide risk profiles, suggesting the need for personalised approaches.

The PERMANENS project aims to develop a Clinical Decision Support System software prototype that assists clinicians in the assessment and management of patients with self-harm at the emergency department. Trained on evidence accumulated in clinical settings and based on the patient’s particular clinical history, the CDSS will provide the clinician with personalized risk profiles for relevant adverse outcomes, including self-harm, method escalation, death by suicide and other causes, and not following up with proposed treatment. The CDSS will provide an overview of the most important risk factors, and propose an evidence-based treatment plan, tailored to the patient’s specific risk profile.

To develop the risk prediction models, population-representative registry data from three countries (Ireland, Norway and Sweden) and one region (Catalonia, Spain) will be used. The OMOP Common Data Model will ensure data interoperability across sites, and a federated analysis approach will eliminate the need for remote access to individual-level data. Data preparation for predictive modelling will include the development of case validation algorithms, the delineation of adverse healthcare trajectories post-discharge, and the creation of a series of clinically relevant predictor variables. Machine learning-based algorithms will be used to develop clinically interpretable prediction models, including state-of-the-art techniques to deal with class imbalance, feature selection, and prediction bias.

The proposed CDSS will enable structured professional judgement, standardization of care, increased patient satisfaction, and higher treatment compliance among patients with self-harm. It will consist of user-friendly and fully functional software, scalable for implementation in entire healthcare systems, enabling the timely and tailored delivery of effective treatment among large populations of patients with suicide risk. Future routine implementation of Clinical Decision Support System for self-harm management at the emergency department has a high potential for effectively reducing suicide mortality in the population.