Development of a predictive model for long-term care and support eligibility certification associated with dementia-related functional impairments: the IRIDE cohort study

The Long-Term Care Insurance (LTCI) system in Japan provides a framework for assessing dementia-related functional decline, enabling the development of predictive models based on daily functional independence. This study aimed to develop and validate risk prediction models for dementia-related functional impairments using data from the Integrated Research Initiative for Living Well with Dementia Cohort Study (IRIDE-CS), integrating multiple cohort studies conducted in urban Japan. Data, including participant characteristics and dementia-related functional impairments, were collected from 3678 participants in the IRIDE-CS aged 65 and older. These data were separated into a derivation set (n = 2764) and a validation set (n = 914) for model development and validation, based on the independent cohorts that constitute the IRIDE-CS. Dementia-related functional impairments was defined as the first certification of cognitive impairment at Level I or higher in the LTCI system, with follow-up from April 2018 to August 2023. Two models were developed: Model 1, focused on dementia-related functional impairments factors in Japan, whereas Model 2, prioritized social implementation by limiting predictors to self-reportable items. The predictive performance of each model was evaluated using the concordance index (C-index). During follow-up, dementia-related functional impairments occurred in 16.9% of the derivation set and 16.3% of the validation set. Model 1, which included eight predictors (age, history of stroke, visual impairment, hearing impairment, exercise habits, gait speed, grip strength, and albumin), achieved a C-index of 0.75 (95% confidence interval [CI]: 0.73–0.77) in the derivation dataset and 0.78 (95% CI: 0.74–0.82) in the validation dataset. Model 2, which included six predictors (age, history of stroke, history of heart disease, visual impairment, hearing impairment, and exercise habits), achieved a C-index of 0.73 (95% CI: 0.70–0.75) in the derivation dataset and 0.77 (95% CI: 0.73–0.81) in the validation dataset. We developed and validated two predictive models for assessing the risk of dementia-related functional impairments using cohort data from urban Japan, demonstrating comparable performance to existing models. Further studies are needed to assess the applicability of these models in non-urban areas and middle-aged populations to enhance generalizability.