299. SENSORS IN-HOME FOR ELDER WELLBEING (SINEW): DIGITAL PHENOTYPING AND ARTIFICIAL INTELLIGENCE FOR EARLY DETECTION OF DEMENTIA

Abstract Background With rapid population ageing, the prevalence of geriatric conditions, including dementia and cognitive frailty is increasing. Early identification of individuals at increased risk of mild cognitive impairment (MCI), a pre-dementia state, can provide a critical window for prompt intervention to prevent or reverse disease progression. Use of digital technology to capture behavioural patterns as digital biomarkers for predictive modelling is of immense promise given its ubiquity. Aims & Objectives Our study utilizes a continuous, home-based monitoring sensor system for older adults in real-world community settings to distinguish those exhibiting normal ageing from those with MCI or early dementia and to predict their transition from normal ageing to one of these conditions. Method A longitudinal cohort design (target sample size =200) was utilised to continuously gather information on behaviours associated with the performance of daily tasks and activities related to cognitive and physical function over three years. A sensor system comprising passive infrared sensors, door contact sensors, bed sensors, medication box sensors, wearable activity bands, and proximity beacons was used. Digital phenotypes were analysed with data from annual comprehensive cognitive and clinical assessments. The collected data and machine learning models were used to classify cognitive states. Results We present preliminary findings from 87 participants (37 MCI, 50 HC; Mage=76.15, SDage=6.20; 74.71% female). Participants with MCI had significantly poorer scores on the MoCA (MMCI=20.62, SDMCI=4.31; MHC=24.08, SDHC=4.23, p<.001) as compared to those with HC. There were no significant differences for the MMSE (MMCI=26.40, SDMCI=2.61; MHC=27.14, SDHC=2.21, p=.160), Geriatric Depression Scale (MMCI=1.35, SDMCI=1.45; MHC=0.84, SDHC=1.20, p=.086), and Pittsburg Sleep Quality Index Scores (MMCI = 10.42, SDMCI =2.73; MHC = 10.40, SDHC =2.55, p=.974). We observed that MCI participants were less active, had fewer daily steps, had a higher frequency of leaving personal belongings at home, and had a shorter total sleep time. Predictive models were built using various machine learning techniques including K Nearest Neighbor (KNN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), and fusion ART (FusionART), in identifying MCI cases based on the digital biomarker features. We found that FusionART produces the highest level of performance across all three measures of precision, recall and F1 scores (0.91). Discussion & Conclusions We aim to develop a reliable and effective sensor system for in-home use that will facilitate the early detection of cognitive and physical decline. It is common for older adults to seek clinical intervention only when their cognitive impairment has already reached an advanced stage. The implementation of readily deployable sensor systems within community settings presents us with opportunities for prompt intervention, which holds the potential for delaying or reversing disease progression and allowing for a healthier lifespan.