Research Training Program in Mental Health Biostatistics and Data Science

Project Summary Mental health disorders are a major cause of disability in the US and more broadly throughout the rest of the world. The impact of mental illness extends to other aspects of human health and disease and significantly reduces life expectancy. As new technologies emerge for assessing mental health disorders, and as research designs become increasingly complex, the Department of Biostatistics at Columbia University, in partnership with leading scholars at Columbia’s Department of Psychiatry and New York State Psychiatric Institute (NYSPI), has developed an innovative program to train predoctoral scholars in biostatistical methods, machine learning, data science, and interdisciplinary research in order to meet the emerging challenges brought by cutting edge technologies. Our training program, built on the expertise of the program mentors and resources at the Department of Psychiatry, NYSPI, and Columbia University Irving Medical Center (CUIMC) has three goals tailored toward mental health research. The first training goal focuses on core, foundational skills of biostatistics and data science. Many modern tools have become especially invaluable in mental health research, including causal inference, innovative designs of clinical trials (e.g., sequential multiple assignment randomized trials; SMARTs), and observational studies (e.g., electronic health records; EHRs), psychometrics, latent variable modeling, and machine learning (supervised and unsupervised learning, reinforcement learning, deep learning). Foundational training in these methods is essential to next-generation mental health statisticians. The second training goal provides the best possible individualized training, where each trainee will create a personalized training plan (PTP) that specifies their choice among three elements. These include (a) computational psychiatry, which encompasses advanced machine learning methods for high-dimensional, complex, or big data; (b) methods for precision medicine in mental health; and (c) methods under the Research domain criteria (RDoC) framework. The third goal trains candidates to gain interdisciplinary research experiences. The best biostatistical collaborators are those with broad knowledge of mental health issues (e.g., familiar with RDoC concepts; Diagnostic and Statistical Manual diagnoses) who are also effective and experienced communicators fluent in the language of mental health. Trainees will thus be prepared to partner with other researchers in mental health to make meaningful contributions towards achieving the goals of improving basic understanding of mental health conditions, optimizing treatment strategies, and reducing the incidence of mental health disorders.