JMIR Publications

ABSTRACT

Background:

Neuropsychological tools for diagnosing neurodegenerative diseases in the older population are becoming more relevant and increasingly complemented by digital capabilities. Digital tools are also increasingly used. In this context, mainly explicit memory is examined. The diagnosis of implicit memory still takes place to a lesser extent. One common measure to do so is the Serial Reaction Time Task (SRTT).

Objective:

The objective of our study was to develop and examine a digital tablet-based SRTT with older subjects with cognitive impairments (CI) and healthy control (HC) subjects. Based on the parameters response accuracy, reaction time, and learning curve, we measure implicit learning and compare groups HC and CI.

Methods:

45 individuals (27 HC, 18 CI – diagnosed by a multi-professional team) completed a tablet-based serial reaction time task. They were presented with 4 blocks of stimuli in sequence and a fifth block were stimuli appeared in random order. Statistical and machine learning modelling approaches were used to investigate how healthy and cognitively impaired individuals differed in their task performance and implicit learning.

Results:

Linear mixed-effects models showed that impaired individuals had significantly higher error rates, higher reaction times, and lower implicit learning, measured via the response increase between sequence and random block(s). Further, machine learning models based on these findings were able reliably and accurately to predict if an individual was in the HC or CI group.

Conclusions:

Our results showed that the HC and CI group differed substantially in their performance in the serial reaction time task. This highlights the promising potential of implicit learning paradigms in the detection of cognitive impairments. The short paradigm based on these results is easy to use in clinical practice.