Predicting the surge: Forecasting Ontario's changing mental health needs
Keywords: time-series, machine learning, mental health, forecasting
TL;DR: We assessed machine learning methods to predict traffic for a Canadian mental health helpline, incorporating curated external datasets to enhance forecasting accuracy and offer insights for optimized resource management in mental health services.
Abstract: We implemented several state-of-the-art Machine Learning (ML)-based forecasting techniques to predict traffic to the Connex Ontario platform, a government-funded helpline and referral service for individuals struggling with mental health and addiction in Ontario, Canada. This research aims to serve as a proof of concept to demonstrate the value of developing forecasting models that incorporate alternative datasets collected from platforms other than electronic health records. Preliminary findings for the test period of March 1, 2020 to Feb 1, 2023 suggest that large ML models are able to harness covariates to improve accuracy for the tasks of producing 1-, 4-, or 12-week forecasts for the volume of mental health outreach. These forecasts have the potential to provide clarity regarding fluctuations in mental health service needs and allow policy-makers to make data-driven decisions, such as proactively allocating resources to reduce the strain on Ontario’s mental health care system.
Submission Number: 33
Loading