AI and Behavioral Data to Predict Mental Health Treatment Outcomes in Young Adults
核心洞察
Mount Sinai and IBM Research (搜索) are collaborating on the PREDiCTOR study to predict mental health treatment outcomes using AI and behavioral data.
The study will analyze clinical interviews, smartphone data, and cognitive testing to identify patterns indicative of treatment discontinuation and hospitalization.
Researchers aim to develop objective, scalable clinical signatures for personalized prediction and decision-making in treating mental health disorders (搜索) in 15-30 year olds.
Mount Sinai Health System and IBM Research (搜索) have initiated a collaborative research project, Phenotypes Reimagined to Define Clinical Treatment and Outcome Research (PREDiCTOR), aimed at enhancing the objectivity of psychiatric evaluations through the integration of artificial intelligence (AI) and behavioral data. The study seeks to predict outcomes such as treatment discontinuation, hospitalizations, and emergency room visits for young individuals seeking mental health care.
The PREDiCTOR study, supported by a $20 million grant from the National Institute of Mental Health (NIMH), will employ objective, scalable measurements to define novel clinical signatures. These signatures are intended for individual-level prediction and to aid clinical decision-making in the treatment of mental health disorders (搜索). The research will be conducted in partnership with researchers from Harvard, Johns Hopkins, Columbia, and Carnegie Mellon universities, as well as Deliberate AI.
Leveraging Behavioral Data
Cheryl Corcoran, M.D., Program Leader in Psychosis (搜索) Risk for Icahn Mount Sinai, emphasized the potential of behavioral data collected during clinical visits. "Every clinical visit provides a wealth of untapped behavioral data that includes spoken language, eye contact, and facial expressions from both the patient and clinician," said Corcoran. She further explained that advancements in computational approaches enable the quantification of these behaviors through audiovisual data analysis. This data, combined with behavioral information from smartphones—tracking physical activity, geolocation, social interactions, sleep patterns, and audio diaries—can help develop clinical signatures indicative of key outcomes.
Study Design and Patient Population
The study will enroll new patients aged 15 to 30 years seeking initial treatment at six outpatient mental health clinics within the Mount Sinai Health System. These clinics serve a diverse community from various socioeconomic backgrounds. This age range is critical because it represents a developmental period during which many disturbances of thought, emotion, and behavior emerge, and diagnoses and prognoses are often unclear.
Rene Kahn, M.D., Ph.D., Chair of Psychiatry at the Icahn School of Medicine at Mount Sinai, highlighted the potential impact of individualized prognosis and clinical decision-making during this period. "Individualized prognosis and clinical decision-making during this critical period may have a profound impact on the lifetime trajectory of these young patients," Kahn stated.
Data Collection and Analysis
The research team will invite all new patients to participate in audio and visual recordings of their clinical visits over one year. Cognitive function will be assessed at baseline, with ongoing assessments throughout the year. By analyzing these behavioral datasets from routine clinical visits and smartphones, the team aims to develop clinical signatures for clinically relevant events, such as treatment disengagement, emergency room visits, and hospitalizations.
Goals and Expected Outcomes
Guillermo Cecchi, Ph.D., Director of the Computational Psychiatry and Neuroimaging groups in IBM Research (搜索), explained the study's objectives. "Our goal is to gain a better understanding of what predicts whether young people stay in mental health treatment or drop out, and what predicts whether their symptoms worsen such that they need acute care in an emergency crisis center or hospital," Cecchi said. He added that while AI has shown promise in predicting outcomes in controlled settings, the team believes current advancements are powerful enough to be applied in routine clinical practice.
This research is supported by the NIMH’s Individually Measured Phenotypes to Advance Computational Translation in Mental Health program, which focuses on using behavioral measures and computational methods to define novel clinical signatures for individual-level prediction and clinical decision-making in treating mental disorders.