Dementia transition in early cognitive decline trajectories (DETECT) study: a prospective longitudinal study protocol Authors: Hyunju Ji, Aeyoung Cho, Harim Lee, Gyeoul Jeong, Eun Hye Hong, Kyung Hee Lee Published: 2026 ABSTRACT BackgroundMild cognitive impairment is widely recognized as a high-risk state associated with a progression to dementia. Although previous studies have reported several predictors of the transition from MCI to dementia, many of these predictive factors are non-modifiable, and evidence on modifiable factors remains limited and inconsistent. Moreover, to date, social factors and their role in this transition have not been sufficiently examined. Guided by the Biopsychosocial Model of Dementia, this protocol describes a study which will examine the transition to dementia as a multidimensional process shaped by biological, psychological, and social factors. This study aims to identify the determinants associated with the transition to dementia and to develop machine learning-based prediction models for this transition among older adults with mild cognitive impairment.MethodsThis is a 3-year prospective longitudinal study which will be conducted among older adults with mild cognitive impairment in South Korea. The participants will be recruited from community-based settings, and annual home visits will be conducted to collect data on biological, psychological, and social factors. This data will be obtained through self-reported questionnaires, physical measurements, wrist-worn actigraphy, sweat patch sampling, indoor environmental sensing, and public records. The outcome will be transition to dementia, defined by clinical diagnosis or cognitive screening criteria. Machine-learning algorithms will be used to develop the prediction models, and model performance will be evaluated using classification metrics and area under the receiver operating characteristic curve analysis.DiscussionThis study is expected to provide evidence on the determinants associated with the transition to dementia among older adults with mild cognitive impairment. The use of home-based assessments and multiple data sources may offer a broader understanding of factors related to this transition. The findings may be useful for early risk identification and for developing community-based interventions and policies to support dementia prevention. Background Dementia is a progressive clinical syndrome characterized by cognitive deterioration and functional decline [1]. In 2021, an estimated 57 million people worldwide were living with dementia, and its incidence and prevalence are projected to rise markedly by 2040, reaching approximately 145 and 822 per 100,000 population, respectively [1, 2]. This growing burden is accompanied by substantial economic impacts, with global societal costs estimated at $1.31 trillion United States (US) dollars in 2019, or approximately $23,800 US dollars per affected person [3]. Given the absence of definitive treatments, early identification of dementia at the stage of mild cognitive impairment (MCI) has become central to prevention [4]. MCI is widely recognized as a high-risk transitional state, with an estimated annual progression rate to dementia of 10% to 15% [5]. Longitudinal evidence has suggested that the risk of transition increases substantially with age, particularly among individuals aged 70 years and older [6]. However, as the clinical course of MCI is heterogeneous, identifying factors associated with progression is essential to distinguish individuals at greatest risk and inform targeted interventions [7]. The transition from MCI to dementia has been associated with multiple factors. A meta-analysis of cohort studies revealed that the progression from MCI to dementia was associated with neurodegenerative and biological markers, including medial temporal lobe atrophy, abnormal cerebrospinal fluid biomarkers, and genetic susceptibility such as the presence of the apolipoprotein E genotype [7]. Similarly, a systematic review found that magnetic resonance imaging findings, apolipoprotein E genotype, older age, and Mini-Mental State Examination (MMSE) scores were strong predictors of the transition from MCI to dementia [8]. However, many of these predictors are non-modifiable or not readily amenable to intervention, which limits their usefulness for the development of preventive strategies. Some studies have identified potentially modifiable predictors of the transition from MCI to dementia, including anxiety [9], depression [6], physical inactivity [6, 10] and alcohol consumption [6, 11, 12]. Nevertheless, studies focusing on modifiable factors remain limited and the findings have been inconsistent. For example, some studies have shown that heavy alcohol intake is associated with an increased risk of a progression from MCI to dementia [11], whereas other studies have linked alcohol consumption to a lower risk of this transition [6, 12]. Moreover, dementia risk should be examined within a broader social context, as cognitive decline may be shaped by contextual conditions that influence exposure to risk and access to protective resources throughout the life course [4]. However, these dimensions remain insufficiently examined and inconsistently integrated into dementia risk research [13]. To address these gaps, the study described in this protocol will employ a prospective longitudinal design to develop a predictive model for the transition from MCI to dementia among older adults. Guided by the Biopsychosocial Model of Dementia [14], this study will examine dementia risk as a multidimensional process shaped by the interaction of biological, psychological, and social factors. Figure 1 shows the theoretical framework for the study. To operationalize the social dimension, we will incorporate social determinants of health (SDOH), allowing a more granular assessment of social conditions potentially associated with the transition to dementia [15]. This multidimensional approach is expected to improve risk prediction for a transition to dementia among older adults with MCI by incorporating factors that reflect both individual vulnerability and social context. Fig. 1Theoretical framework of this study Theoretical framework of this study Methods/design Aims This study will prospectively examine the transition from MCI to dementia among older adults and to develop a predictive model for dementia transition. The specific aims are: (1) to identify the biological, psychological, and social factors associated with the progression from MCI to dementia and (2) to develop and evaluate a predictive model for the transition to dementia using these multidimensional factors. Study design This study is a 3-year prospective longitudinal study designed to develop a predictive model for the transition from MCI to dementia. Theoretical framework This study will be guided by the Biopsychosocial Model of Dementia [14]. Within this framework, dementia is viewed as a multifactorial and dynamic condition shaped by the interplay of biological, psychological, and social factors. The model depicts dementia as a process extending from aging to death, in which these interacting factors affect symptom severity and the rate of disease progression. Accordingly, the model provides a comprehensive theoretical basis for identifying predictors of a transition to dementia within a multidimensional context. Participants The participants will be recruited from community-based settings in Seoul, South Korea, including dementia care centers, senior welfare centers, and day and night care facilities commonly utilized by older adults with MCI. Additional participants will be recruited using a snowball sampling approach. Participants are eligible for inclusion if they meet the following criteria: (1) aged 65 years or older and (2) have a clinical diagnosis of MCI. Exclusion criteria are as follows: (1) a diagnosis of major neurological or psychiatric disorders; (2) a diagnosis of dementia; (3) a score of ≤ 19 on the Korean Mini-Mental State Examination (K-MMSE); and (4) inability to complete self-reported questionnaires owing to communication or comprehension difficulties. Sample size The required sample size was estimated using G*Power version 3.1.9.4 for repeated-measures analysis of variance with a within–between interaction. Assuming a medium effect size, a significance level of 0.05, and a statistical power of 0.80, the minimum sample size was calculated as 60 participants. Considering an anticipated attrition rate of approximately 30%, the target sample size was increased to 86 participants. Those participants enrolled in the first year of the study will be followed up annually for 3 years. Additional participants will be recruited in the second year as needed to compensate for loss to follow-up to maintain the planned longitudinal sample size. Measures The study variables are organized according to the Biopsychosocial Model of Dementia [14], encompassing biological, psychological, and social factors. The outcome variable is the transition from MCI to dementia during the follow-up period. Detailed variables, instruments, and data collection methods are shown in Table 1. Table 1Table of variables and measurementsDomainVariableInstrument/MeasureAssessment methodBiological factors General characteristicsAgeaQuestionnaireSexaQuestionnaireYear of MCI diagnosisaQuestionnaire Aging and homeostasisCurrent medicationbName, class, route, dosage, and scheduleQuestionnaireFrailtyFried frailty phenotype [16]Questionnaire + Physical measurements Immune systemCytokinesInterleukins, tumor necrosis factor-alpha, etc.Sweat patch Metabolic factorsWaist circumferenceFlexible tape measurePhysical measurementsBody mass indexWeight and heightQuestionnaireBlood pressureAutomated sphygmomanometerPhysical measurementsBlood glucose levelsPortable glucometerPhysical measurements Physical capabilityComorbidity burdenCIRS-G [18]QuestionnaireIADLK-IADL [19]QuestionnaireADLK-ADL [19]QuestionnaireGrip strengthHand dynamometerPhysical measurementsGait speedUsual walking speedPhysical measurementsPhysical activityEnergy expenditure, step counts, time spent in different intensity levels, etc.ActigraphySleepSleep duration, sleep efficiency, and sleep–wake patterns, etc.ActigraphyPsychological factors AffectAnxietyK-GAI [20]Questionnaire Mood stateDepressionGDS-15 [21]Questionnaire Neuropsychiatric symptomsMild behavioral impairmentbMBI-C [22]Questionnaire Needs statePerceived needs for healthcare servicesRegular health examinations, health education and counseling, etc.QuestionnairePerceived needs for home safety equipmentGrab bars, non-slip mats, gas detectors, etc.QuestionnairePerceived needs for aging in place services following functional declineDaily living assistance, home-based medical services, etc.QuestionnaireSocial factors Economic stabilityOccupationQuestionnaireHousehold incomeQuestionnaireUse of meal support servicesQuestionnaireNutritional statusNSI [24]Questionnaire Education access and qualityEducational levelaQuestionnaireParticipation in learning activitiesType of learning and frequency of participation, etc.QuestionnaireDigital device useOwnership and use of devices, etc.Questionnaire Healthcare access and qualityHealth insuranceQuestionnaireHealthcare service utilizationReceipt of health examinations, reason for not receiving health screening, etc.QuestionnaireUsual source of carePresence of a regular healthcare institution or physician for illness, Services received from the usual source of care, etc.QuestionnaireeHealth literacyKeHEALS [26]Questionnaire Neighborhood and built environmentCrime ratePublic dataSocial disorderPublic dataAir pollutionPublic dataSatisfaction with community environmentConvenience of access to commercial facilities/medical facilities/public institutions, Opportunities for interaction with neighbors, etc.QuestionnaireIndoor environmentAir quality, temperature, humidity, etc.Indoor environmental sensor Social and community contextMarital statusaQuestionnaireCohabiting family membersQuestionnaireSocial networkK-LSNS [28]QuestionnaireOutcomeTransition to dementiaClinical diagnosis of dementiaQuestionnaireK-MMSE [29]QuestionnaireMoCA-K [30]cQuestionnaireADL Activities of daily living, CIRS-G Cumulative Illness Rating Scale–Geriatrics, GDS-15 15-item Geriatric Depression Scale, IADL Instrumental Activities of Daily Living, K-ADL Korean Activities of Daily Living, K-GAI Korean version of the Geriatric Anxiety Inventory, K-IADL Korean Instrumental Activities of Daily Living, K-LSNS Korean version of the Lubben Social Network Scale, MBI-C Mild Behavioral Impairment Checklist, K-MMSE Korean Mini-Mental State Examination, KeHEALS Korean eHealth Literacy Scale, MoCA-K Korean version of the Montreal Cognitive Assessment, NSI Nutritional Screening InitiativeaVariables will be assessed only at baseline; all other variables will be assessed in years 1, 2, and 3bData will be obtained from family members where available; otherwise, participants will provide self-reportscThe MoCA-Basic will be administered in place of the MoCA-K for illiterate participants Table 1Table of variables and measurementsDomainVariableInstrument/MeasureAssessment methodBiological factors General characteristicsAgeaQuestionnaireSexaQuestionnaireYear of MCI diagnosisaQuestionnaire Aging and homeostasisCurrent medicationbName, class, route, dosage, and scheduleQuestionnaireFrailtyFried frailty phenotype [16]Questionnaire + Physical measurements Immune systemCytokinesInterleukins, tumor necrosis factor-alpha, etc.Sweat patch Metabolic factorsWaist circumferenceFlexible tape measurePhysical measurementsBody mass indexWeight and heightQuestionnaireBlood pressureAutomated sphygmomanometerPhysical measurementsBlood glucose levelsPortable glucometerPhysical measurements Physical capabilityComorbidity burdenCIRS-G [18]QuestionnaireIADLK-IADL [19]QuestionnaireADLK-ADL [19]QuestionnaireGrip strengthHand dynamometerPhysical measurementsGait speedUsual walking speedPhysical measurementsPhysical activityEnergy expenditure, step counts, time spent in different intensity levels, etc.ActigraphySleepSleep duration, sleep efficiency, and sleep–wake patterns, etc.ActigraphyPsychological factors AffectAnxietyK-GAI [20]Questionnaire Mood stateDepressionGDS-15 [21]Questionnaire Neuropsychiatric symptomsMild behavioral impairmentbMBI-C [22]Questionnaire Needs statePerceived needs for healthcare servicesRegular health examinations, health education and counseling, etc.QuestionnairePerceived needs for home safety equipmentGrab bars, non-slip mats, gas detectors, etc.QuestionnairePerceived needs for aging in place services following functional declineDaily living assistance, home-based medical services, etc.QuestionnaireSocial factors Economic stabilityOccupationQuestionnaireHousehold incomeQuestionnaireUse of meal support servicesQuestionnaireNutritional statusNSI [24]Questionnaire Education access and qualityEducational levelaQuestionnaireParticipation in learning activitiesType of learning and frequency of participation, etc.QuestionnaireDigital device useOwnership and use of devices, etc.Questionnaire Healthcare access and qualityHealth insuranceQuestionnaireHealthcare service utilizationReceipt of health examinations, reason for not receiving health screening, etc.QuestionnaireUsual source of carePresence of a regular healthcare institution or physician for illness, Services received from the usual source of care, etc.QuestionnaireeHealth literacyKeHEALS [26]Questionnaire Neighborhood and built environmentCrime ratePublic dataSocial disorderPublic dataAir pollutionPublic dataSatisfaction with community environmentConvenience of access to commercial facilities/medical facilities/public institutions, Opportunities for interaction with neighbors, etc.QuestionnaireIndoor environmentAir quality, temperature, humidity, etc.Indoor environmental sensor Social and community contextMarital statusaQuestionnaireCohabiting family membersQuestionnaireSocial networkK-LSNS [28]QuestionnaireOutcomeTransition to dementiaClinical diagnosis of dementiaQuestionnaireK-MMSE [29]QuestionnaireMoCA-K [30]cQuestionnaireADL Activities of daily living, CIRS-G Cumulative Illness Rating Scale–Geriatrics, GDS-15 15-item Geriatric Depression Scale, IADL Instrumental Activities of Daily Living, K-ADL Korean Activities of Daily Living, K-GAI Korean version of the Geriatric Anxiety Inventory, K-IADL Korean Instrumental Activities of Daily Living, K-LSNS Korean version of the Lubben Social Network Scale, MBI-C Mild Behavioral Impairment Checklist, K-MMSE Korean Mini-Mental State Examination, KeHEALS Korean eHealth Literacy Scale, MoCA-K Korean version of the Montreal Cognitive Assessment, NSI Nutritional Screening InitiativeaVariables will be assessed only at baseline; all other variables will be assessed in years 1, 2, and 3bData will be obtained from family members where available; otherwise, participants will provide self-reportscThe MoCA-Basic will be administered in place of the MoCA-K for illiterate participants Table of variables and measurements ADL Activities of daily living, CIRS-G Cumulative Illness Rating Scale–Geriatrics, GDS-15 15-item Geriatric Depression Scale, IADL Instrumental Activities of Daily Living, K-ADL Korean Activities of Daily Living, K-GAI Korean version of the Geriatric Anxiety Inventory, K-IADL Korean Instrumental Activities of Daily Living, K-LSNS Korean version of the Lubben Social Network Scale, MBI-C Mild Behavioral Impairment Checklist, K-MMSE Korean Mini-Mental State Examination, KeHEALS Korean eHealth Literacy Scale, MoCA-K Korean version of the Montreal Cognitive Assessment, NSI Nutritional Screening Initiative aVariables will be assessed only at baseline; all other variables will be assessed in years 1, 2, and 3 bData will be obtained from family members where available; otherwise, participants will provide self-reports cThe MoCA-Basic will be administered in place of the MoCA-K for illiterate participants Biological factors The biological factors considered in this study include general characteristics, aging, homeostasis, immune system, metabolic factors, and physical capability. General characteristics will include age, sex, and year of MCI diagnosis. Aging and homeostasis will be evaluated based on current medication use and frailty status. Current medication includes prescribed medication and dietary supplements and will be confirmed with family members when available. Frailty will be measured using the Fried frailty phenotype [16]. The immune system will be assessed using cytokines extracted from sweat samples, a method that has been validated as a non-invasive measure in older adults with dementia [17]. Metabolic factors will include waist circumference, body mass index, blood pressure, and blood glucose levels. Physical capability will encompass comorbidity burden [18], instrumental activities of daily living (IADL) [19], activities of daily living (ADL) [19], grip strength, gait speed, physical activity, and sleep. Comorbidity burden will be measured using the Cumulative Illness Rating Scale–Geriatrics (CIRS-G) [18]. Gait speed will be determined by the time taken for the participants to walk 4.5 m at their usual pace. Physical activity and sleep will be assessed using wrist-worn actigraphy. Psychological factors Psychological factors will incorporate affect, mood state, neuropsychiatric symptoms, and needs state. Affect is operationalized as anxiety and will be assessed using the Korean version of the Geriatric Anxiety Inventory (K-GAI) [20]. Mood state refers to depression and will be measured using the 15-item Geriatric Depression Scale (GDS-15) [21]. Neuropsychiatric symptoms are operationalized as mild behavioral impairment and will be assessed using the Mild Behavioral Impairment Checklist (MBI-C) [22]. Given that the instrument is designed to be completed by patients, close informants, or clinicians, responses will be obtained from close informants when the participant is not living alone and an available close informant is present. Needs state will be assessed using items from the 2023 National Survey of Older Koreans (NSOK) [23], which is a large-scale, nationally representative survey conducted every 3 years among adults aged 65 years or older in South Korea. The selected items include perceived needs for healthcare services (e.g., regular health examinations), perceived needs for home safety equipment (e.g., grab bars), and perceived needs for ‘aging in place’ services following functional decline (e.g., daily living assistance). Responses for each item are rated on a 5-point Likert scale ranging from “not needed at all” to “very needed.” Social factors In Healthy People 2030, a national initiative designed to improve population health and reduce health disparities in the US, the SDOH is organized into five domains: (1) economic stability; (2) education access and quality; (3) health care access and quality; (4) neighborhood and built environment; and (5) social and community context [15]. In the proposed study, these domains will be used to structure the assessment of the social factors associated with the transition from MCI to dementia. Economic stability will be evaluated using occupation, household income, use of meal support services, and nutritional status. Nutritional status will be assessed using the Korean-translated Nutrition Screening Initiative checklist (NSI), originally developed by Posner et al. [24]. Education access and quality include educational level, participation in learning activities, and digital device use. The items assessing participation in learning activities (e.g., type of learning and frequency of participation) and digital device use (e.g., ownership and use of devices) have been adapted from the 2023 NSOK [23]. Healthcare access and quality will be examined using health insurance status, healthcare service utilization [23], usual source of care [25], and eHealth literacy [26]. Healthcare service utilization will be assessed using items from the 2023 NSOK [23], including receipt of health examinations. The usual source of care will be determined using items from the 2nd Korea Health Panel Survey [25], including whether participants have a regular healthcare institution or physician for routine illness. The Korea Health Panel Survey is a nationally representative household panel survey that collects data on healthcare utilization, medical expenditures, and health behaviors [25]. eHealth literacy will be measured using the Korean eHealth Literacy Scale (KeHEALS) [26]. The neighborhood and built environment are defined by crime rate, social disorder, air pollution, and satisfaction with the community and indoor environments. The data on the crime rate, social disorder, and air pollution will be compiled by linking publicly available administrative and geospatial data from national public data sources [27]. Satisfaction with the community environment will be assessed using items from the 2023 NSOK [23], and indoor environmental conditions will be measured using an indoor environmental sensor. The social and community context will be assessed using marital status, cohabiting family members, and social network [28]. The social network will be measured using the Korean version of the Lubben Social Network Scale (K-LSNS) [28]. Outcome Our outcome variable is the transition to dementia, defined as a clinical diagnosis of dementia or a K-MMSE score ≤ 19 [29] and a MoCA-K score ≤ 17 [30]. For participants who are illiterate, the MoCA-Basic [31] will be administered instead of the MoCA-K [30]. Data collection Data will be collected annually through home visits conducted by trained research assistants (RAs) over a 3-year follow-up period. Standard questionnaire-based assessments will be administered using tablet devices. To minimize potential barriers related to age-associated visual or functional limitations, the RAs will read each item aloud and record the participants’ answers. Physical examinations including measurements of waist circumference, blood pressure, and blood glucose levels will be conducted according to standardized procedures. Physical activity and sleep will be monitored using a wrist-worn actigraphy device (ActiGraph wGT3X-BT [ActiGraph LLC, Pensacola, FL]). Sweat samples will be collected using a sterile adhesive sweat patch (PharmChek, PharmChem Inc., Fort Worth, TX, USA) as a noninvasive method to assess circulating inflammatory biomarkers. Environmental data will be collected using an indoor environmental sensor (People Multi Co., Ltd., Seoul, Republic of Korea) installed in the room where the participants spend most of their time. Procedure Study recruitment will be conducted in collaboration with participating community-based institutions. Recruitment notices will be posted where permitted. At dementia care centers, outpatient nurses will facilitate the identification of potential candidates diagnosed with MCI. The RAs will approach these patients to explain the study and confirm willingness to participate. When community-based institutions hold programs or group activities for older adults with MCI, the institutional staff will introduce the study to attendees, and trained RAs will provide further explanation to those who express interest in participation. Older adults who are willing to participate will be contacted to schedule a home visit, during which written informed consent will be obtained. On the first day of data collection, two trained RAs will visit each participant’s home. During this visit, the RAs will administer questionnaire-based measures, perform physical examinations, apply the sweat patch to the upper back using aseptic techniques, attach the actigraphy device, and install the indoor environmental sensor. The PharmChek Sweat Patch is designed to be worn for a minimum of 24 h to ensure sufficient sweat absorption for laboratory testing [32]. Participants will be instructed to wear the actigraphy device continuously on the non-dominant wrist for at least 7 consecutive days, except while bathing or showering, and to avoid removing or disturbing the sweat patch and indoor environmental sensor during the monitoring period. The indoor environmental sensor will be installed in the room where the participant spends most of their time. Written instructions on device use, related precautions, and emergency contact information will also be provided. At a follow-up home visit conducted at least 24 h after the sweat patch application, an RA will remove the patch using aseptic techniques and store the collected patches in sterile containers at low temperatures to preserve the sample integrity. During this visit, the RA will determine if the actigraphy device is being worn as instructed and confirm that the indoor environmental sensor is functioning properly. In the laboratory, sweat will be extracted from the patches by centrifugation and stored under appropriate freezing conditions until analysis. After 7 full days of monitoring, the RAs will make a final home visit to retrieve the actigraphy device and indoor environmental sensor. Actograms will be visually inspected to identify device malfunction and confirm compliance with wear-time instructions. To reduce loss to follow-up, each participant will receive a brief written report by mail summarizing the assessment results after each assessment wave to support participant engagement and retention throughout the longitudinal study period. Data analysis The longitudinal data collected over the 3-year study period will be used to develop predictive models for transition from MCI to dementia. Machine-learning approaches will be applied to construct the prediction models, including Gradient Boosting Machine, Extreme Gradient Boosting (XGBoost), CatBoost, Support Vector Machine, and logistic regression model. To improve model interpretability and efficiency, feature selection techniques will be employed. The model performance will be evaluated using multiple metrics, including accuracy, recall, precision, specificity, F1 score, F2 score, and the area under the receiver operating characteristic curve (AUC), to ensure a comprehensive assessment of predictive validity. Ethical considerations This study has received Institutional Review Board approval (IRB No. 4-2025-1050), and all procedures will be conducted in accordance with the Declaration of Helsinki. This study will be conducted and reported with reference to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. As the study population consists of older adults with MCI, decisional capacity will be assessed using the Evaluation to Sign Consent (ESC) before enrollment. Only those participants able to provide informed consent will be included. Written informed consent will then be obtained from all eligible participants after they have received explanations and descriptions of the study purpose, procedures, and data collection methods. To protect confidentiality, all data will be managed using unique identification numbers. Data will be stored in an encrypted, password-protected cloud system accessible only to authorized research personnel. Sweat patches and extracted samples will be stored in a locked freezer with access restricted to the research team. The participants will be informed that their participation is voluntary, that they may request data correction or deletion and withdraw at any time without penalty, and that all study data and biological samples will be destroyed after the retention period in accordance with institutional guidelines. Discussion This study protocol describes a 3-year prospective longitudinal study which has been designed to identify determinants of the transition to dementia among older adults with MCI and to develop predictive models for this transition. The study protocol is guided by the Biopsychosocial Model of Dementia [14], which frames the transition to dementia as a process shaped by biological, psychological, and social dimensions. The anticipated findings may contribute to earlier identification of older adults at high risk of dementia and help guide prevention strategies that incorporate both individual-level and broader contextual factors. A major strength of this protocol is its focus on potentially modifiable determinants of the transition to dementia. Although previous studies have identified several predictors of the transition from MCI to dementia, many of these predictors, such as demographic factors, neuroimaging markers, and genetic predispositions, are not amenable to modification [7, 8]. In contrast, this study will incorporate a comprehensive range of modifiable factors related to lifestyle behaviors, nutritional status, healthcare access, social activities, and environmental conditions. Elucidating these factors may provide a practical basis for developing future interventions to reduce the risk of dementia among older adults with MCI. This approach is consistent with the 2024 Lancet Commission’s life-course perspective on dementia prevention, which highlights the importance of addressing modifiable risks [4]. Another strength of our protocol is the use of multimodal assessment methods, including self-report questionnaires, physical measurements, actigraphy, sweat patches, indoor environmental sensors, and publicly available data. This integrative approach will enable the capturing of biological, psychological, and social factors in community-based settings. In particular, the inclusion of social factors may provide further insight into the role of social conditions in the transition to dementia among older adults with MCI. The development of machine learning-based prediction models may further enhance the practical value of this study. Machine learning has become increasingly relevant in healthcare research as it can integrate data from diverse sources and identify meaningful patterns [33]. These benefits are particularly important in dementia research, as the transition from MCI to dementia is unlikely solely associated with a single factor. Rather, this transition likely reflects the combined influence of biological, psychological, and social factors [14]. By comparing multiple algorithms, the proposed study will evaluate which combinations of predictors most effectively distinguish individuals at higher risk of a transition to dementia. Several limitations which could potentially impact the interpretation of the results of this study should be considered. First, the home-based design precludes the inclusion of blood-based laboratory tests and neuroimaging assessments such as magnetic resonance imaging, limiting the ability to capture certain biological and neuroimaging markers of dementia progression. However, the study design allows the collection of real-world data, including actigraphy, sweat patch sampling, and indoor environmental monitoring, which may provide a more comprehensive understanding of the daily behavioral and environmental conditions of the participants. Second, attrition over the 3-year period is a potential concern. To address this limitation, assessment results will be sent to the participants during follow-up to encourage continued participation. In addition, supplementary recruitment will be considered in the second year to maintain the target sample size. Overall, the study findings are expected to provide an empirical basis for future community-based interventions and policies aimed at preventing dementia among older adults with MCI.