The promise of machine learning in predicting treatment outcomes in psychiatry

Abstract

For many years, psychiatrists have tried to understand factors involved in response to medications or psychotherapies, in order to personalize their treatment choices. There is now a broad and growing interest in the idea that we can develop models to personalize treatment decisions using new statistical approaches from the field of machine learning and applying them to larger volumes of data. In this pursuit, there has been a paradigm shift away from experimental studies to confirm or refute specific hypotheses towards a focus on the overall explanatory power of a predictive model when tested on new, unseen datasets. In this paper, we review key studies using machine learning to predict treatment outcomes in psychiatry, ranging from medications and psychotherapies to digital interventions and neurobiological treatments. Next, we focus on some new sources of data that are being used for the development of predictive models based on machine learning, such as electronic health records, smartphone and social media data, and on the potential utility of data from genetics, electrophysiology, neuroimaging and cognitive testing. Finally, we discuss how far the field has come towards implementing prediction tools in real‐world clinical practice. Relatively few retrospective studies to‐date include appropriate external validation procedures, and there are even fewer prospective studies testing the clinical feasibility and effectiveness of predictive models. Applications of machine learning in psychiatry face some of the same ethical challenges posed by these techniques in other areas of medicine or computer science, which we discuss here. In short, machine learning is a nascent but important approach to improve the effectiveness of mental health care, and several prospective clinical studies suggest that it may be working already.

Schizophrenia Detection Using Machine Learning Approach from Social Media Content

“…Machine learning models have drawn great attention in recent years [ 11 , 12 ]. Machine learning techniques have played an important role in the applications of natural language processing including sentiment analysis, chatbot systems, question answering systems, information retrieve systems, and machine translation [ 13 ].…”

Screening of Mood Symptoms Using MMPI-2-RF Scales: An Application of Machine Learning Techniques

“…For many years, classical statistical approaches have been used to confirm or refute certain hypotheses, but modern ML research focuses on the overall predictive power of models, especially on how accurately they predict desired outcomes in new and unencountered datasets. Research in this field is primarily evaluated by its potential clinical impact and whether it can provide reliable information about the prognosis for future patients (especially when patients go unscreened and undiagnosed for many years), leading to treatments and interventions [ 49 ]. It will be a useful predictive technique for detection-resistant mood disorders.…”

“…ML-based predictive models are becoming increasingly popular due to their ability to combine large amounts of data into a single model and their self-evaluation of predictive value for previously unseen patients. Specifically, researchers have used ML methods successfully to predict the persistence, duration, and severity of major depressive disorder [ 48 ], as well as treatment responses [ 49 ], suicide attempts [ 50 ], and first onset of major depressive episodes in US soldiers [ 51 , 52 ]. ML methods are especially useful in fields where evaluation is complex, time-consuming, and expensive, such as when a differential diagnosis is required because the initial diagnosis is unclear [ 53 ].…”

Systematic Review of Functional MRI Applications for Psychiatric Disease Subtyping

“…Even in mostly proof-of-principle settings, 15 out of 20 studies engaged in robustness and generalizability analyses (Tables 2-4). Most of them, however, yielded intra-sample reports (by partitioning one available dataset instead of using truly external data) which can lead to inflated generalizability estimates (37). While nearly all studies (17/20, Tables 2-4) undertook an interpretation of their solution using clinically relevant measures, only four of them (56,58,60,73) attempted to report generalizability across multiple data collection sites, using truly external data.…”

“…As the dimensionality of this data is extremely high (millions of genetic variants per subject) and individual polymorphism contributions are generally small, however, genetic data is rarely useful for unsupervised learning. Supervised approaches, however, which aim to classify individuals among already defined labels, have shown more success (37). Lastly, in addition to questionnaires and more traditional clinical datasets, a data modality that gained momentum over the last few years is digitomics (electronic health records, mobile sensor data) (37,155,156).…”