Extending the usefulness of the verbal memory test: The promise of machine learning

The evaluation of verbal memory is a core component neuropsychological assessment in wide range clinical and research settings. Leveraging story recall to assay neurocognitive function could be made more useful if it were possible administer frequently (i.e., would allow for the collection patient data over time) automatically assess recalls with machine learning methods. In present study, we evaluated novel test 24 parallel forms that was deployed using smart devices 94 psychiatric inpatients 80 nonpatient adults. Machine vector-based natural language processing methods employed automate scoring, performance these their incremental validity, criterion validity convergence trained human raters), reliability. Our results suggest moderate high consistency across forms, raters (r values ~ 0.89), discriminating between groups. While much work remains, findings are critical implementing an automated, deployable remote technologies multiple frequent administrations.