Development, Feasibility and Effectiveness of a Digital Support Platform for Mental Health in Primary Care (PRESTO) Based on a Machine Learning Approach.

The prevalence of mental health disorders in Primary Care (PC) largely exceeds the limited resources available. The main aim of this project is to develop a comprehensive machine learning (ML) digital support platform (PRESTO) to approach people with mental health symptoms in PC. PRESTO will offer a transdiagnostic triage of those cases needing specialized care while most of the mild and moderate cases with anxiety and depressive symptoms will be allocated through ML models to either: 1.a periodic follow-up, 2.symptoms monitoring and brief psychological intervention with a smartphone app, or 3.a specific psychopharmacological treatment. To reach this objective, first, a ML predictive severity model will be build based on all the cases referred to the PC mental health support programme during the last 5 years retrieved from electronic health records from 5 PC centres(PCC) in Barcelona. Simultaneously, a smartphone app (PRESTOapp) monitoring symptoms and delivering a psychological intervention for non-severe anxious and depressive symptomatology will be developed and tested in a feasibility study and in a randomized clinical trial. Finally, the ML models obtained from the first phase of the project and the data from the PRESTOapp study will be integrated in a comprehensive self-learning web platform which will triage and assign to each case a specific intervention based on the predicted outcome. The effectiveness of PRESTO to reduce waiting times in receiving appropriate and specific care of mental health problems will be tested by means of a stepped-wedge randomized controlled trial in 5 PCCs in Barcelona.

Here we register a Randomized controlled clinical trial with PRESTOapp 2.0.

• Design: Single-blind randomized controlled clinical trial.

• Sample: Referrals to mental health support programme (PCMHSP) by GPs from the 5 primary care centres included in the study.

• Sample size calculation: Considering the priority (primary outcome) the reduction of depressive symptoms assessed by PHQ-9 and taking into account two groups (PRESTOapp 2.0 vs. treatment as usual) in addition to previous results of effect sizes between 0.30 - 0.35 from similar studies (similar intervention, same scale), we have established a power of 0.80 and a α of 0.05. Considering the current numbers of visits by all members of the PCMHSP who can potentially be offered inclusion in the study in 6 months (1000 individuals), the total sample should have at least 122 participants. However, experience in similar studies indicates an expected 25-30% drop-out. Therefore, it was decided to add 15 more subjects per branch for preventive purposes and to ensure that at the end of the study there would be a sufficient sample to guarantee the strength of the data.

• Intervention Group (PRESTOapp 2.0): 76 participants

• Control Group (Treatment as usual): 76 participants TOTAL: 152 participants.

Considering the number of PCMHSP members involved in the project as well as the high number of referrals (which is the main problem this project is trying to solve), reaching these numbers is fully feasible within the stipulated time.