Personalized cognitive training: Protocol for individual-level meta-analysis implementing machine learning methods

Abstract

Accumulating evidence suggests that cognitive training may enhance well-being. Yet, mixed findings imply that individual differences and training characteristics may interact to moderate training efficacy. To investigate this possibility, the current paper describes a protocol for a data-driven individual-level meta-analysis study aimed at developing personalized cognitive training. To facilitate comprehensive analysis, this protocol proposes criteria for data search, selection and pre-processing along with the rationale for each decision. Twenty-two cognitive training datasets comprising 1544 participants were collected. The datasets incorporated diverse training methods, all aimed at improving well-being. These training regimes differed in training characteristics such as targeted domain (e.g., working memory, attentional bias, interpretation bias, inhibitory control) and training duration, while participants differed in diagnostic status, age and sex. The planned analyses incorporate machine learning algorithms designed to identify which individuals will be most responsive to cognitive training in general and to discern which methods may be a better fit for certain individuals.