Automated Mental Health Referral System

SUMMARY This proposal addresses a significant barrier to obtaining treatment for college-age youth with mental disorders. Many college-age youth with impairing mental disorders remain untreated because of concerns about stigma and privacy, inconvenience and wait times, and because universities are often unable to service all such students. Also, of critical importance, when referral for treatment is implemented, it is without regard to the person's pathology, because of the erroneous assumption that treatment need not be tailored lo the individual. This proposal aims to address this critical clinical issue. We advance that a sophisticated automated online referral system would resolve all of these problems, but there is no expert-trained system for psychiatric referrals. We propose to automate the referral process, designed for college-age youth, by bridging online, mental health assessments and curated, up-to-date, mental health provider networks. To this end, the non-profit Child Mind Institute is partnering with the for-profit MiResource. Assessment expertise is provided by the Child Mind Institute, which treats children and adolescents with mental health disorders, conducts mental health research, has acquired large assessment datasets, has in-house expertise in mental health assessment, and through its MATTER Lab has developed novel assessment technologies such as the MindLogger data collection and assessment platform. Referral infrastructure is provided by MiResource, a software-as-a-service solution designed to help universities connect students to local mental health providers. The MATTER Lab and MiResource will develop an automated online assessment and referral platform that uses expert-trained machine learning to provide users with personalized referrals for mental health care. Expert referrals will be based on the six dimensions of the Level of Care Utilization System (risk of harm, functional status, comorbidity, environment, treatment history, and attitude) applied to college students' responses to mental health assessments. In Phase I, we will (1-1) build mental health assessments into the MindLogger platform, (1-2) build an expert referral collection interface, and (1-3) set up a machine learning pipeline for training and testing an updatable classification model for automated clinically appropriate, personalized referrals. In Phase II, we will build, refine, and clinically validate our product for commercialization. Specifically, we will (11-1) validate the Phase I framework on a university population, (11-2) integrate Mindlogger's assessments into MiResource, and (11-3) conduct usability and quality assurance tests of the new Mindlogger plus MiResource platform, to get feedback about issues related to accessibility, relevance, accuracy, and esthetics, and incorporate solutions in response to this feedback into a final version.