JMIR Publications

ABSTRACT

Background:

Traditional cognitive screening tools for mild cognitive impairment (MCI) often have limited ecological validity and may impose substantial burden on both clinicians and older adults. Scenario-based serious games that simulate everyday activities offer a promising digital alternative for engaging and accessible cognitive assessment.

Objective:

This study aimed to develop a high-fidelity, scenario-based serious game for screening mild cognitive impairment and to evaluate its diagnostic accuracy in comparison with a conventional neuropsychological assessment battery.

Methods:

A scenario-based serious game incorporating daily living tasks was developed to assess multiple cognitive domains relevant to MCI. A total of 67 older adults were recruited, including 20 individuals with MCI and 47 cognitively normal participants. Participants completed the game-based assessment, and a Random Forest machine learning model was trained to classify cognitive status using in-game performance features. Clinical diagnosis based on the Korean version of the Consortium to Establish a Registry for Alzheimer’s Disease neuropsychological battery (CERAD-K) served as the reference standard.

Results:

The machine learning model demonstrated good diagnostic performance, achieving an area under the receiver operating characteristic curve (AUC) of 0.776. The model achieved a sensitivity of 80.0% and a specificity of 84.6%, while maintaining an overall classification accuracy of 83.6% in distinguishing individuals with MCI from cognitively normal participants. Feature importance analysis indicated that performance metrics from the Sale Items task—specifically the total accuracy score reflecting visual-verbal memory and reaction time—were identified as the most influential predictors for MCI classification. Performance on the game-based assessment showed convergence with CERAD-K outcomes while requiring substantially less administration time than the conventional neuropsychological battery.

Conclusions:

The proposed scenario-based serious game demonstrates acceptable diagnostic accuracy and ecological validity for MCI screening. By capturing subtle deficits in memory and processing speed through everyday scenarios, this digital assessment approach may serve as a scalable and efficient screening tool for early cognitive impairment in both clinical and community settings.