AI-driven telerehabilitation for older adults with mild cognitive impairment: a randomized controlled trial

With the rising prevalence of aging populations, accessible interventions for Mild Cognitive Impairment (MCI) are critical. MCI, a high-risk transitional state to dementia, is considered the most opportune window for intervention to maintain or enhance cognitive function. This study evaluates the clinical efficacy of Zenicog^®, an Artificial Intelligence (AI)-driven, self-guided, home-based cognitive rehabilitation program designed to overcome logistical barriers of conventional therapy.

A total of 70 older adults (aged 65 years or older) with MCI (indicated by a K-MMSE2 score between 18 and 26) were recruited for this delayed-treatment parallel randomized clinical trial with an exploratory crossover phase. Participants were randomized into Group AB (intervention-first) or Group BA (control-first). The intervention consisted of a 5-week home program (24 sessions) using an AI algorithm for autonomous difficulty adjustment. The primary outcome was global cognitive function, measured by the K-MMSE2. Secondary outcomes included domain-specific cognitive tests [Digit Span Forward (DSF), Digit Span Backward (DSB), Trail Making Test-A/B (TMT-A/B)], psychosocial measures and usability and adverse events were also assessed. Primary analysis focused on Period 1 (T1), comparing Group AB (intervention) to Group BA (control). Participants were randomized via a computer-generated block randomization method. Randomization used concealed allocation via sequentially numbered opaque envelopes prepared independently, with group assignment performed by a blinded investigator.

Sixty-two participants were included in the final analysis, randomly allocated to Group AB (n = 35) or Group BA (n = 35). At the end of Period 1, the intervention group showed significantly higher K-MMSE2 scores compared to the control group (median 28.0 vs. 26.0; p < 0.001). Crossover analysis confirmed significant K-MMSE2 gains occurred only during the intervention periods for both groups. Clinical success (K-MMSE2 ≥27) was achieved by 93.9% of the intervention group vs. 0% of the control. No significant differences were found in domain-specific tests. Usability and satisfaction scores were high (≥4.5/5), with zero dropouts due to adverse effects.

AI-driven, self-guided telerehabilitation is a feasible and effective strategy for improving global cognitive function in MCI patients. Its scalability and minimal supervision requirements make it a viable alternative to clinic-based therapy in aging societies.

Clinical Research Information Service (CRIS), KCT0008968, https://cris.nih.go.kr/cris/search/detailSearch.do?seq=31040&search_page=L. Registered on November 21, 2023.