Season-specific associations between electricity-based frailty risk and life-space mobility, health-related quality of life, and physical function in older adults

Introduction

Background

Frailty has been reported to be associated with all-cause mortality (1). Individual-level factors, such as depression, loneliness, and quality of life (QoL), and relational-level factors, including living alone and household income, are also associated with frailty (1,2). In Japan, the combined prevalence of frailty and prefrailty among older adults is ~50% (frailty: 8.7%; prefrailty: 40.8%), and this prevalence increases with age (3-5). Accordingly, preventing frailty and prefrailty represents an urgent challenge in Japan, as the country continues to become an increasingly super-aged society.

Frailty comprises physical, cognitive, psychological, social, and economic factors. Commonly used frailty assessment instruments include Fried’s Frailty Phenotype, the Clinical Frailty Scale, and the electronic frailty index (eFI) (6-10). These assessments are generally based on one of two conceptual approaches: the phenotype model proposed by Fried or the deficit accumulation model (11,12). Phenotype-based measures often fail to capture cognitive, psychological, or social domains, whereas deficit accumulation models require diverse data sources, making them burdensome for time-series assessments.

Electricity consumption data allow artificial intelligence (AI) algorithms to infer daily routines and detect behavioral deviations that may indicate early signs of physical or cognitive decline. This passive approach may reduce the burden on older adults compared with assessments requiring active participation (13). Electricity-based frailty risk was estimated using e-Frail Navi (Chubu Electric Power Co., Inc.), a system that noninvasively detects signs of frailty by using AI to analyze household electricity consumption data and infer changes in the daily living patterns of older adults (14). Conventional questionnaire-based frailty assessments are useful for evaluating health status at a given time point; however, because they are typically administered intermittently, they may be affected by respondent burden and recall bias. In contrast, electricity-based monitoring may offer additional value by capturing temporal changes in daily living patterns without requiring active responses from participants. Therefore, electricity-based monitoring should be regarded as a complementary approach rather than a replacement for established questionnaire-based frailty assessments. Osaka City, where the present study was conducted, is classified as having a humid subtropical climate under the Köppen-Geiger climate classification (Cfa), with generally hot and humid summers and relatively mild winters with little snowfall (15). Residential electricity consumption is responsive to changes in ambient temperature, particularly in regions with high penetration of air conditioning or electric heating (16). Electricity-based monitoring enables the collection of time-series data, such as seasonal variations in daily living patterns. Furthermore, a recent editorial has suggested that older adults with frailty may be characterized as a vulnerable group with respect to seasonal influences because physiological decline reduces their ability to cope with seasonal fluctuations (17). Therefore, it is critically important to examine the health of older adults while accounting for seasonal variations.

In recent years, global warming has contributed to the emergence of extreme temperature patterns, such as shorter spring, autumn, and winter seasons and prolonged periods of high summer temperatures (18). Seasonal variation has been reported to be associated with changes in physical and mental functioning, including decreased muscle strength, decreased physical activity, and poorer mental health outcomes such as depressive symptoms (19-23). These declines in physical and mental functioning have also been reported to be particularly pronounced in winter.

A recent report from the United States described an association between social isolation and temperature variation. Specifically, for each 1-℃ increase in temperature, the proportion of residents who stayed at home all day decreased by 0.17 percentage points (24). As the season transitions from winter to summer, opportunities to go outside may increase, potentially reducing social isolation. Moreover, such declines in physical and mental functioning are associated with frailty and prefrailty and overlap with risk factors linked to seasonal variations (25-27). Indeed, previous studies have reported that lower average winter temperatures are associated with higher odds of frailty (28). Frailty has also been linked to life-space mobility and health-related quality of life (HRQoL) (29,30). Previous research examining the association between seasonal changes in outdoor temperature and outdoor behavioral patterns using the Life-Space Assessment (LSA) reported that LSA scores measured in winter were lower than those measured in spring among older adults in Finland (31). This finding suggests that life-space mobility may be more restricted in winter than in spring. Seasonal variation has also been associated with HRQoL. Seasonal changes in mood and behavior have been shown to be related to HRQoL, and physical HRQoL has been reported to be lowest in winter and highest in summer (32).

Based on these findings, we can infer that winter conditions may reduce opportunities for going outside, restrict life-space mobility, and negatively affect physical and psychological health among older adults. As electricity-based frailty risk is intended to capture aspects of daily living patterns, such as wake-up time, bedtime, and time spent outside the home, it may be sensitive to winter-related behavioral changes. Accordingly, we hypothesized that the associations between electricity-based frailty risk and life-space mobility and HRQoL would be more pronounced in winter than in other seasons.

Rationale and knowledge gap

Despite the growing interest in seasonal influences on frailty, no previous study has examined within-person, season-specific associations between monthly frailty risk estimates and health-related outcomes using objective and noninvasive monitoring methods. Previous research has reported that a decline in the mean January temperature during winter has been associated with increased odds of frailty, even after adjusting for demographic characteristics, socioeconomic status, and health behaviors (28). However, in that study, frailty was assessed using a frailty index based on the deficit accumulation model, and frailty was measured only once per participant, precluding the evaluation of within-person seasonal changes. Other studies have examined seasonal differences by assessing frailty during spring and winter using the Kihon Checklist (KCL), a comprehensive frailty screening tool, and the Japanese version of the Cardiovascular Health Study (CHS) phenotype criteria (33). These studies reported no seasonal differences in frailty defined by the CHS criteria, whereas frailty assessed using the KCL was higher in spring than in winter. These findings indicate that the relationship between frailty and seasonality may vary depending on the frailty assessment tool used. Although the eFI can capture month-to-month changes in frailty status (11), it has not been used to specifically investigate frailty as a seasonal phenomenon. Moreover, few studies have examined, within the same individuals, the association between season-specific frailty and health-related outcomes.

Objective

This study aimed to describe seasonal differences in electricity-based frailty risk among community-dwelling older adults and to examine exploratory associations between seasonally averaged frailty risk derived from a 12-month monitoring period and subsequently assessed health-related outcomes. Clarifying these relationships while considering seasonal variations in frailty risk may contribute to the development of preventive strategies aimed at improving health-related outcomes by incorporating seasonal variations into frailty management. We present this article in accordance with the STROBE reporting checklist (available at https://mhealth.amegroups.com/article/view/10.21037/mhealth-2026-0013/rc).

Methods

Participants

This study employed an observational design that combined retrospectively collected longitudinal electricity consumption data with a cross-sectional, face-to-face questionnaire survey. Participants were recruited through community association bulletin boards, flyer distribution, and in-person announcements. Recruitment took place during the face-to-face assessment period (October to November 2024) among individuals who attended an information session on the day of assessment. Individuals were enrolled if they agreed to undergo frailty assessment using e-Frail Navi and provided written informed consent to participate in the study. Participants were a volunteer sample of older adults who agreed to participate in both the 12-month electricity-based monitoring study and the face-to-face assessment.

This study included men and women aged ≥60 years who were living alone in apartment complexes in Joto Ward, Osaka City, Japan. The study area is an urban district covering 8.38 km², with a total population of 169,144 residents. Osaka City is located in a temperate region of Japan, where winters are generally mild and snowfall is rare. Eligible participants were individuals whose residence in the apartment complexes between October 2023 and September 2024 could be confirmed and who provided informed consent for the use of their household electricity consumption data for analysis using electricity-based frailty risk. Exclusion criteria included difficulty in ambulation without assistance; the presence of cardiovascular or respiratory diseases that limit participation in physical activity; and a Mini-Cog© score ≤2. A Mini-Cog© score of 0–2 has been used as a positive screen for possible cognitive impairment or dementia in previous validation studies (34). Therefore, in this study, a cutoff of ≤2 was used to exclude participants for whom reliable questionnaire responses and informed participation in the monitoring study might have been difficult.

This study was conducted in accordance with the Ethical Guidelines for Life Science and Medical Research Involving Human Subjects (March, 2023) and the Declaration of Helsinki and its subsequent amendments. The study was approved by the Research Ethics Committee of the Graduate School of Rehabilitation Science, Osaka Metropolitan University (No. 2024-111). Written and oral informed consent was obtained from all participants.

Data collection

This study combined retrospectively collected 1-year electricity consumption data, used to estimate electricity-based frailty risk, with physical and mental function data collected through a single face-to-face assessment. The face-to-face assessment included questionnaires on demographic characteristics, including age, sex, educational attainment, smoking status, floor of residence, length of residence, body mass index (BMI), comorbidities, and the use of assistive devices for outdoor walking. Health-related indicators were assessed across four domains: activity (life-space mobility and physical activity), mental/psychological status (HRQoL and depressive symptoms), physical function (objective physical performance), and social networks (social isolation). To minimize information bias, face-to-face assessments were conducted by assessors trained in standardized procedures using the same measurement equipment and questionnaires. Health-related outcomes were assessed once between October and November 2024, after the 12-month electricity monitoring period. Therefore, the present analyses examined exploratory associations between seasonally averaged Frailty Risk Index values derived from the preceding monitoring period and health-related outcomes assessed at a single subsequent time point, rather than strictly concurrent season-specific associations. Monthly mean ambient temperature data for the Osaka meteorological station were obtained from historical weather data provided by the Japan Meteorological Agency.

Frailty Risk Index

“e-Frail Navi” is an AI-based system developed by Chubu Electric Power Co., Inc. that analyzes frailty risk using household electricity consumption data from individuals living alone (14). The system detects frailty risk based on patterns of daily living behaviors, such as wake-up time, bedtime, and time spent outside the home. The electricity-based frailty risk score range is 0–100, with higher scores indicating higher frailty risk. In the present study, the score was analyzed as a continuous Frailty Risk Index to examine seasonal variations in electricity-based frailty risk. The detailed algorithm, feature weighting, and training dataset used by e-Frail Navi are proprietary and have not been publicly disclosed. Therefore, in the present study, the Frailty Risk Index should be interpreted as an exploratory electricity-based risk indicator reflecting daily living patterns rather than as a validated diagnostic measure of frailty. The data collection period comprised 12 months of household electricity consumption data calculated on a monthly basis from October 2023 to September 2024. Seasonal variations in the Frailty Risk Index were classified into four seasons—spring, summer, autumn, and winter—based on the seasonal definitions provided by the Japan Meteorological Agency. The mean Frailty Risk Index values were calculated for each season (Figure 1).

Figure 1 Monthly Frailty Risk Index values across the 12-month study period. Monthly Frailty Risk Index values are shown from March to February. Bars are color-coded by season: spring, summer, autumn, and winter.

Activity

Life-space mobility

Life-space mobility was assessed using the LSA (35). The LSA is a questionnaire that evaluates life-space mobility across five levels: Level 1, within the home beyond the bedroom; Level 2, within the home premises; Level 3, within the neighborhood; Level 4, within the town; and Level 5, outside the town. For each level, both the frequency of movement and the degree of assistance required are assessed. The LSA is self-administered, and the total score is calculated. The LSA score ranges from 0 (completely bedridden) to 120 (independent daily mobility beyond the residential area without assistance). The LSA has been reported to have excellent test–retest reliability (35).

Physical activity and sedentary time

Physical activity was evaluated using the International Physical Activity Questionnaire (IPAQ) Short Version (36), which comprehensively assesses physical activity across multiple domains, including work-related activities, transportation (e.g., commuting and shopping), household and gardening activities, and leisure-time exercise and recreation. In this study, the total physical activity [metabolic equivalent of task (MET)—minutes per week] and sedentary time (minutes per day) were calculated.

Mental and psychological status

HRQoL

HRQoL was assessed using the Japanese version of the EuroQol 5-Dimension 5-Level questionnaire (EQ-5D-5L) (37). The EQ-5D-5L evaluates five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension is rated on five levels: no problems, slight problems, moderate problems, severe problems, and inability to perform the activity. Health utility values were calculated based on participants’ responses. When applying the value set derived from the general Japanese population, EQ-5D-5L utility scores range from −0.0255 to 1.000, with higher values indicating better HRQoL.

Depressive symptoms

Depressive symptoms were screened using the 5-item Geriatric Depression Scale (GDS-5) (38), a self-administered questionnaire designed for older adults. Each item is scored as 1 point for a depressive response based on yes/no answers, yielding a total score ranging from 0 to 5. A score of ≥2 indicates possible depression (39).

Physical function

Physical function was assessed using the Timed Up and Go (TUG) test, the 5-m walking test, and the Five Times Sit-to-Stand Test (SS-5) (40,41). The TUG test measures the time required for a participant to stand up from a seated position, walk 3 m, turn around at a marker, return to the chair, and sit down again. For the 5-m walking test, a 5-m measurement zone was established at the center of an 8-m walking path. Participants were instructed to walk at a comfortable pace for both the TUG and the 5-m walking tests, and the mean value of the two trials was calculated. For the SS-5 test, participants were instructed to stand up fully until the knees were completely extended and to sit down until the buttocks firmly contacted the seat as quickly as possible. Two trials were performed, and the faster time was adopted for analysis.

Social networks

Social networks were assessed using the Lubben Social Network Scale-6 (LSNS-6) (42), which is calculated by summing the scores of six items. The total scores range from 0 to 30, with higher scores indicating stronger social networks.

Statistical analysis

Monthly Frailty Risk Index values were summarized into seasonal averages for the primary analyses. This approach was chosen because health-related outcomes were assessed only once, and month-specific analyses in this very small sample would have produced unstable estimates and increased the risk of multiple testing. Given the exploratory nature of this pilot study and its small sample size, Bayesian analyses were adopted as the primary analytical framework, with frequentist correlations performed as supplementary sensitivity analyses. Bayesian correlations were treated as the primary analysis because Bayes factors quantify the relative evidence for the alternative hypothesis compared with the null hypothesis (43). This framework was used to describe the degree of evidence rather than to provide definitive hypothesis testing. Spearman’s rank correlation coefficients were calculated to examine whether the direction of the associations was consistent when assessed using a conventional nonparametric method. When Bayesian and Spearman results differed, interpretation was based primarily on the Bayesian results, whereas Spearman correlations were used to assess the consistency and direction of the associations. Bayesian correlation and multiple linear regression analyses were conducted. Bayes factors (BF₁₀) were calculated to compare each model with the null model. A BF₁₀ >3 was interpreted descriptively as evidence favoring the alternative model. Spearman analyses were performed using the same variables examined in the Bayesian correlation analyses, including the Frailty Risk Index, age, BMI, LSA, HRQoL, physical function, and other measurements across the four seasonal categories.

Linear mixed-effects models were fitted as exploratory analyses using monthly Frailty Risk Index values as repeated observations. Participant ID was included as a random intercept, and season, each health-related outcome, the season-by-outcome interaction, and monthly mean ambient temperature were included as fixed effects. Monthly mean ambient temperature was included as a fixed effect because residential electricity consumption may be influenced by ambient temperature. Because health-related outcomes were assessed only once, they were treated as person-level covariates. Given the very small sample size, separate models were fitted for each outcome, and the results were interpreted cautiously. For significant season-by-outcome interactions, exploratory post hoc simple-slope analyses and pairwise comparisons of slopes were performed using the Bonferroni correction. For the Spearman analyses and linear mixed-effects models, statistical significance was set at P<0.05. All analyses were performed using JASP (version 0.95.4; JASP Team).

As this study was planned as an exploratory pilot study, no a priori sample size or power calculations were performed. All participants who provided informed consent during the study period and for whom electricity consumption data were available were included in the analysis. Owing to the very small sample size, the analyses were not intended to provide definitive hypothesis testing or precise effect estimates. The absence of sufficient statistical evidence should therefore not be interpreted as evidence of no association, because the study may have been underpowered and susceptible to type II error.

Results

Participants

Of the 18 individuals who agreed to participate, 14 were included in the final analysis after excluding those with missing data and those who did not meet the criterion for the length of residence [median age: 81.0 (10.3) years; 86% women] (Figure 2). Four participants were excluded owing to incomplete Frailty Risk Index data due to insufficient residence duration (n=1), missing LSA data (n=1), no face-to-face assessment data (n=1), and a Mini-Cog^© score ≤2 (n=1). As only four participants were excluded, whether missingness was systematic could not be formally evaluated. The demographic and clinical characteristics of all participants are summarized in Table 1. Monthly Frailty Risk Index values are shown descriptively in Figure 1 to illustrate within-year fluctuations during the 12-month monitoring period. Because health-related outcomes were assessed only once, the primary inferential analyses used seasonal averages rather than month-specific values. The median Frailty Risk Index varied across seasons, with median [interquartile range (IQR)] values of 9 (7.2) in spring, 14 (11.3) in summer, 8 (9.3) in autumn, and 12 (12.8) in winter. The median LSA total score was 91.0 (IQR, 29.5), and the median EQ-5D-5L utility value was 0.92 (IQR, 0.20).

Figure 2 Flowchart of participant selection. Among the 18 participants aged 60 years or older who met the inclusion criteria, 4 were excluded because of incomplete FRI data due to insufficient residence duration, missing LSA data, no face-to-face assessment data, or Mini-Cog© score ≤2, leaving 14 participants in the final analytic sample. FRI, Frailty Risk Index; LSA, Life-Space Assessment.

Season-specific relationships with life-space mobility, HRQoL, and physical function

In winter, the Frailty Risk Index showed its most distinctive pattern. Bayesian analyses provided evidence favoring negative associations between the Frailty Risk Index and both life-space mobility [LSA: r=−0.656, 95% confidence interval (CI): −0.857 to −0.146, BF₁₀ =6.285] and HRQoL (EQ-5D-5L utility: r=−0.715, 95% CI: −0.885 to −0.237, BF₁₀ =13.950) (Table 2). By contrast, evidence was insufficient for associations with age (r=0.347, 95% CI: −0.208 to 0.696, BF₁₀ =0.648), BMI (r=−0.010, 95% CI: −0.489 to 0.476, BF₁₀ =0.329), and objective physical function (TUG: r=0.337, 95% CI: −0.217 to 0.690, BF₁₀ =0.622; 5-m walk test: r=0.319, 95% CI: −0.234 to 0.680, BF₁₀ =0.580; SS-5: r=0.161, 95% CI: −0.365 to 0.587, BF₁₀ =0.378). Similarly, evidence was insufficient for physical activity and sedentary behavior (total physical activity: r=−0.424, 95% CI: −0.738 to 0.133, BF₁₀ =0.937; sedentary time: r=0.112, 95% CI: −0.402 to 0.556, BF₁₀ =0.352), social networks (LSNS-6: r=0.067, 95% CI: −0.436 to 0.527, BF₁₀ =0.337), and depressive symptoms (GDS-5: r=−0.022, 95% CI: −0.497 to 0.467, BF₁₀ =0.330). Supplementary Spearman correlations in winter were significant for LSA (ρ=−0.588, P=0.03) and EQ-5D-5L utility (ρ=−0.637, P=0.01). However, correlations with age (ρ=0.305, P=0.29), BMI (ρ=−0.055, P=0.86), physical function (TUG: ρ=0.205, P=0.48; 5-m walk test: ρ=0.168, P=0.57; SS-5 ρ=−0.040, P=0.89), total physical activity (ρ=−0.451, P=0.11), sedentary time (ρ=0.093, P=0.75), social networks (LSNS-6: ρ=−0.019, P=0.95), and depressive symptoms (GDS-5: ρ=0.022, P=0.94) were not significant.

In spring, summer, and autumn, Bayesian correlation analyses indicated insufficient to limited evidence across all assessed outcomes, with all 95% CIs spanning zero (Table 2). Bayes factors remained low for demographic characteristics (age: all BF₁₀ ≤1.041; BMI: all BF₁₀ ≤0.375), activity (LSA: all BF₁₀ ≤1.655; total physical activity [IPAQ MET-min/week]: all BF₁₀ ≤0.450; sedentary time: all BF₁₀ ≤0.614), mental/psychological status (EQ-5D-5L utility: all BF₁₀ ≤1.413; GDS-5: all BF₁₀ ≤0.368), physical function (TUG/5-m walk test/SS-5: all BF₁₀ ≤1.326), and social networks (LSNS-6: all BF₁₀ ≤0.747). Supplementary Spearman rank correlations were significant for LSA and EQ-5D-5L utility in spring and autumn and for TUG in autumn (P<0.05), whereas correlations with the other assessed outcomes were not significant. In summer, none of the Spearman correlations reached statistical significance (all P>0.05).

Bayesian multiple linear regression for the winter Frailty Risk Index

In the Bayesian multiple linear regression analysis with the winter Frailty Risk Index as the dependent variable, the model including both the LSA and EQ-5D-5L showed the numerically highest posterior support (BF_M =2.089, R²=0.603) (Tables 3,4). This was followed closely by the model including the EQ-5D-5L alone (BF_M =2.035, R²=0.512) and the model including the LSA alone (BF_M =0.829, R²=0.431). These regression results should be interpreted cautiously given the very small sample size, the minimal difference in BFM between the combined and EQ-5D-5L-only models, and the uncertainty of the regression coefficients.

Exploratory linear mixed-effects models using monthly Frailty Risk Index values

Linear mixed-effects models using monthly Frailty Risk Index values showed significant interactions between season and LSA (F=3.833, P=0.01) and between season and EQ-5D-5L (F=3.271, P=0.02) after adjustment for monthly mean ambient temperature and accounting for repeated measurements within participants (Tables 5-7). These findings suggest that the associations of life-space mobility and HRQoL with monthly Frailty Risk Index values differed by season. In contrast, the interaction between season and TUG was not significant (F=0.124, P=0.95). Monthly mean ambient temperature was not significantly associated with monthly Frailty Risk Index values in any model. Exploratory post hoc simple-slope analyses showed that the negative slopes for both LSA and EQ-5D-5L were largest in winter. Pairwise comparisons of slopes with Bonferroni correction indicated that the winter slope was significantly more negative than the summer slope for both LSA (P=0.002) and EQ-5D-5L (P=0.01), whereas differences between winter and spring or autumn were not statistically significant (Tables 5-7).

De-identified participant-level data used in the analyses are provided in Table S1.

Discussion

Key findings

This study investigated seasonal differences in frailty risk among community-dwelling older adults and examined the associations between frailty risk and health-related outcomes in each season. Season-specific analyses demonstrated that, in winter, the Frailty Risk Index was negatively associated with both life-space mobility (LSA) and HRQoL (EQ-5D-5L), with comparatively greater Bayesian support observed for the association with EQ-5D-5L. In the Bayesian multiple linear regression analysis for winter frailty risk, the model including both the LSA and EQ-5D-5L showed the numerically highest posterior support, although the difference from the EQ-5D-5L-only model was minimal. These findings suggest that winter Frailty Risk Index values may be related to reduced life-space mobility and poorer HRQoL, although the results should be interpreted cautiously given the exploratory nature and small sample size of this study. The exploratory linear mixed-effects models further supported the possibility that the associations between electricity-based frailty risk and health-related outcomes varied by season. After accounting for repeated measurements within participants and adjusting for monthly mean ambient temperature, significant interactions were observed between season and LSA and between season and EQ-5D-5L, whereas the interaction between season and TUG was not significant. These findings are consistent with season-specific correlation analyses and post hoc simple-slope analyses, which revealed that associations were most evident for winter Frailty Risk Index values, when compared with summer. However, because health-related outcomes were assessed only once after the electricity-monitoring period and the sample size was very small, these results should be interpreted as exploratory evidence rather than definitive proof of winter-specific effects.

Comparison with similar research

This is the first study to clarify the association between seasonal variations in frailty risk—an aspect not addressed in previous research—and life-space mobility and HRQoL. Previous studies, including systematic reviews, have consistently shown that frailty is associated with HRQoL and that seasonal sensitivity is also closely related to HRQoL (44,45). Several studies have also reported associations between seasonal variation and HRQoL, as well as between frailty and life-space mobility (32,46). Moreover, frailty, life space, and HRQoL have been reported to be interrelated (29,30). Some studies have examined the association between outdoor temperature and outdoor behavioral patterns using the LSA (31). For example, a study on older adults in Finland reported that LSA scores measured in winter were lower than those measured in spring, suggesting that life-space mobility may be more restricted in winter than in spring. By contrast, a case study conducted in the northern Netherlands reported that, although individuals spend more time engaging in social interaction and exercise in summer than in autumn or winter, life space may expand more in autumn and winter than in summer because snow and related conditions increase travel by car (47). Because evidence regarding seasonal variation in life-space mobility is inconsistent across studies and may be strongly influenced by regional and climatic factors, the present findings should be interpreted in the context of Osaka, an urban area with relatively mild winters and little snowfall. In contrast to settings such as Finland, where severe cold and snow may directly restrict outdoor mobility, winter in Osaka is less likely to limit mobility through these mechanisms alone. Therefore, seasonal weather conditions as well as context-specific behavioral and environmental reactions to winter in a temperate Japanese city may be reflected in the associations that were most evident in the winter.

Explanations of findings

The winter Frailty Risk Index showed a negative association with LSA scores, indicating that participants with a more restricted life-space mobility tended to have a higher frailty risk during winter. Winter has been associated with increased mortality during cold spells, and prior studies have linked winter conditions to age-related frailty as well as cardiovascular and ischemic heart diseases (48,49). Previous research has also demonstrated that the frequency and duration of long-distance outdoor trips decline substantially during mid-winter (around February) compared with early winter (around November) (47). Furthermore, several studies have shown that physical activity and muscle strength among older adults tend to be higher from summer to autumn than in winter (19,50). Cold seasons such as winter have also been shown to be associated with more depressive states and greater social isolation (22,24), often characterized by staying home throughout the day.

Collectively, these findings suggest that winter may be a period in which physical, psychological, and social factors related to frailty tend to deteriorate, particularly compared with summer. In addition, life-space mobility has also been reported to be more restricted in winter than in spring (31). Therefore, the present finding that winter frailty risk is associated with life-space mobility is consistent with previously reported seasonal changes in behavior, psychological status, and environmental conditions during winter.

The winter Frailty Risk Index also showed a negative association with EQ-5D-5L utility values, indicating that participants with lower HRQoL tended to exhibit a higher winter frailty risk. In the Bayesian multiple linear regression analysis with the winter Frailty Risk Index as the dependent variable, the model including both the LSA and EQ-5D-5L showed the highest posterior support. However, this result should be interpreted cautiously because the model included two predictors in a very small sample of 14 participants. Therefore, the stability of the regression coefficients may be limited, and the possibility of overfitting cannot be excluded.

Regarding seasonal variation and HRQoL, a study on healthy adults reported that physical HRQoL among healthy adults is highest in summer and lowest in winter (32), whereas mental HRQoL is highest in summer and lowest in spring and autumn. By contrast, studies of individuals with affective disorders attributed to seasonal changes in circadian rhythms, such as seasonal affective disorder, have shown that HRQoL—assessed using the SF-20—is lowest in winter than in summer (51).

In the present study, Frailty Risk Index in spring and autumn showed only negative trends with HRQoL, and the evidence supporting these associations was insufficient. Additionally, we also did not observe any association with depressive symptoms. One possible explanation relates to the characteristics of the EQ-5D-5L. Although the instrument assesses physical health across four dimensions, mental health is represented by only one dimension, which may limit sensitivity to psychological aspects (52,53). Moreover, the participants in the present study scored below the cutoff for depressive symptoms, which may have reduced the likelihood that mental health status was reflected in the measure outcomes (39). Furthermore, although the LSA has been shown to be associated with depressive symptoms, this relationship is largely explained by physical function, physical activity, and activities of daily living (54). Consequently, winter frailty risk in the present study may have been more readily explained by the LSA and EQ-5D-5L, both of which more strongly reflect physical aspects.

By contrast, the associations with physical function measures, which are generally considered key determinants of frailty, were not clearly supported in the present study. The lack of clear associations with objective physical function measures may reflect the nature of electricity-based AI monitoring. Frailty is a multidimensional construct that encompasses not only physical but also psychological and social aspects (6). In addition, electricity-based AI monitoring does not directly measure physical function itself; rather, it primarily captures daily living behaviors and activity patterns, such as changes in wake-up time, evening activity, appliance use, and kitchen-related activity (13). Previous research has reported that the frequency of going outdoors is associated with HRQoL among older adults (55). Moreover, the LSA is calculated based not only on the spatial extent of mobility but also on the frequency of movement and the level of assistance required (35). Therefore, the electricity-based Frailty Risk Index may have been more closely related to broader daily-life indicators, such as HRQoL and life-space mobility, than to physical function measures such as the TUG test. However, because this study did not directly evaluate which components of frailty were reflected in electricity consumption patterns, and because the detailed algorithm and feature weighting of e-Frail Navi have not been publicly disclosed, this interpretation should be regarded as exploratory.

Previous studies have noted that participants in health surveys often report better self-rated health, suggesting the presence of selection bias whereby individuals who are more health-conscious are more likely to participate in health programs (56). In the present study, individuals with a relatively low frailty risk may have been more likely to attend voluntary assessments, which could have produced a ceiling effect in the measured indicators. Moreover, because physical function evaluations were conducted only once during the face-to-face assessment conducted in autumn, these measures may likely reflect the performance specific to that season rather than capturing variations in physical function corresponding to frailty risk throughout the year.

Study limitations

This study has some limitations. First, the final analytic sample was very small and imbalanced, with only 14 participants and a high proportion of women, which likely increased the uncertainty of the estimates and reduced the ability to obtain season-specific associations. Therefore, the possibility of type II error should be considered, particularly for outcomes for which no sufficient evidence of association was observed. In addition, the examination of multiple season-specific associations may have increased the risk of chance findings. Therefore, the findings should be interpreted as exploratory and hypothesis-generating rather than robust confirmatory evidence. In addition, the Bayesian multiple regression model for the winter Frailty Risk Index included two predictors despite the very small sample size of 14 participants; therefore, the stability of the regression coefficients may be limited, and the possibility of overfitting cannot be excluded. Larger studies with adequate sample sizes and repeated health-related outcome assessments are needed to validate these findings. Second, participants were restricted to older adults living alone in apartment buildings within a single urban district in Osaka City, which may limit the generalizability of the findings to individuals living in cohabiting households, rural areas, detached housing environments, and regions with different climatic and neighborhood characteristics. Electricity-use patterns and life-space mobility may differ substantially according to housing type, household composition, regional climate, transportation environments, and access to community resources; therefore, the external validity of the present findings is limited. In addition, because Osaka has relatively mild winter conditions and little snowfall compared with colder or snowier regions, the observed winter-related associations may not be generalizable to regions where severe cold or snowfall directly restricts outdoor mobility. Furthermore, participants were a volunteer sample of older adults who agreed to participate in both the 12-month electricity-based monitoring study and the face-to-face assessment. Therefore, selection bias may have occurred, with individuals who were more health-conscious or had lower frailty risk being more likely to participate. Accordingly, the findings should be generalized with caution and should not be assumed to apply to all community-dwelling older adults living alone, particularly those with poorer health status or higher frailty risk. Participants also showed relatively high LSA and EQ-5D-5L scores and low GDS-5 scores, suggesting that the final analytic sample may have represented a relatively healthy subgroup of older adults living alone. Therefore, the observed associations may not be generalizable to older adults with poorer health status, lower mobility, lower HRQoL, or higher depressive symptoms. Third, although the Frailty Risk Index was computed monthly over a 12-month period, life-space mobility, HRQoL, and physical function were assessed only once between October and November 2024, after the electricity monitoring period. Therefore, the winter Frailty Risk Index was not compared with health-related outcomes measured during the same winter season. This temporal mismatch limits internal validity and precludes conclusions regarding strictly concurrent or causal season-specific relationships. The observed associations should therefore be interpreted as exploratory relationships between past seasonally averaged electricity-based frailty risk and subsequent health status. This temporal misalignment represents a fundamental design limitation of the present study. Although the exploratory linear mixed-effects models accounted for repeated measurements within participants and adjusted for monthly mean ambient temperature, the health-related outcomes were measured only once. Therefore, these models could not fully evaluate concurrent month-to-month relationships between frailty risk and health-related outcomes. This limitation is particularly relevant to physical function measures, because TUG, the 5-m walk test, and SS-5 were assessed only once during the autumn face-to-face assessment. Therefore, the absence of associations with physical function should not be interpreted as evidence that electricity-based frailty risk is unrelated to physical function. In addition, because health-related outcomes were assessed at a single time point, the direction of the observed associations cannot be determined. Lower HRQoL or restricted life-space mobility may have contributed to behavioral changes detected from electricity consumption patterns. Therefore, these associations should not be interpreted as causal or predictive relationships. Fourth, although monthly Frailty Risk Index values were available and were used in exploratory linear mixed-effects models, health-related outcomes were assessed only once. Therefore, the present study still could not fully exploit the high-frequency nature of electricity-based monitoring to examine concurrent month-to-month relationships. Future studies should collect repeated health-related outcomes and examine month-to-month fluctuations using longitudinal models. Fifth, because this was an observational study, residual or unmeasured confounding factors cannot be ruled out. Although we collected background information such as the floor of residence, comorbidities, and use of outdoor walking aids, we could not adequately adjust for economic status, energy poverty, heating costs, indoor thermal environments, detailed comorbidity profiles, transportation availability, or access to community resources. These factors should not be regarded as secondary influences, because they may be strongly associated with both electricity-use patterns and health-related outcomes. In particular, cold-weather behavior and electricity consumption patterns may be influenced not only by physical frailty but also by the affordability of heating and indoor thermal conditions. Therefore, these unmeasured or insufficiently measured factors may have confounded the observed associations. Sixth, the detailed algorithm, feature weighting, training data, and validation information for e-Frail Navi have not been publicly disclosed. Therefore, the Frailty Risk Index should be interpreted as an exploratory electricity-based risk indicator rather than as a validated diagnostic measure of frailty. Further studies are needed to validate the e-Frail Navi algorithm against established frailty measures and clinically meaningful outcomes in relevant target populations. Owing to the proprietary nature of these algorithmic details, the extent to which the Frailty Risk Index aligns with established frailty constructs remains uncertain, and the reproducibility of the present findings and the application of this method in other settings are limited.

Implications and required actions

The novelty of this study is that frailty-related risk was estimated using household electricity consumption data. This approach may complement conventional frailty assessments by passively capturing daily living patterns without requiring questionnaires, physical performance tests, wearable devices, or active responses from older adults. However, it should not be considered a replacement for established frailty assessments, because it does not directly measure physical function or multidimensional frailty. Further validation against established frailty measures and clinically meaningful outcomes is needed. Future research should examine the reproducibility and generalizability of the observed associations between electricity-based frailty risk, life-space mobility, and HRQoL in larger studies involving populations across diverse regional contexts, including urban, suburban, and rural environments. Study designs should also align the monthly frailty-risk time series with repeated season-specific assessments of life-space mobility, HRQoL, physical function, and psychosocial measures to enable within-person analyses. Finally, implementation-oriented research that combines winter-focused interventions—such as support for outdoor activities, indoor exercise programs, and initiatives to promote social participation—with monitoring of electricity-based frailty risk may help establish frailty prevention strategies that explicitly account for seasonal variability.

Conclusions

The electricity-based Frailty Risk Index varied across seasons and negative associations with life-space mobility and HRQoL were most evident in winter. These findings suggest that electricity-based frailty monitoring may have potential utility as a complementary approach for identifying potential seasonal vulnerabilities, particularly during winter. However, given the exploratory pilot nature of this study and the very small sample size, larger longitudinal studies with repeated health assessments are needed to validate its clinical usefulness. Future studies should also establish practical alert thresholds for determining when changes in electricity-based frailty risk should trigger clinical follow-up or intervention.

Acknowledgments

We thank Osaka Metro Co., Ltd., Chubu Electric Power Co., Inc., and Omichikai Social Medical Corporation for providing the relevant data and for their support in study operations. We are also grateful to the community residents who participated in the study.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://mhealth.amegroups.com/article/view/10.21037/mhealth-2026-0013/rc

Data Sharing Statement: Available at https://mhealth.amegroups.com/article/view/10.21037/mhealth-2026-0013/dss

Peer Review File: Available at https://mhealth.amegroups.com/article/view/10.21037/mhealth-2026-0013/prf

Funding: This work was supported in part by a competitive research grant from the Ministry of Land, Infrastructure, Transport and Tourism of Japan, through the “Smart Mobility × Smart Aging City Co-Creation Demonstration Project”.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://mhealth.amegroups.com/article/view/10.21037/mhealth-2026-0013/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Ethical Guidelines for Life Science and Medical Research Involving Human Subjects (March, 2023) and the Declaration of Helsinki and its subsequent amendments. The study was approved by the Research Ethics Committee of the Graduate School of Rehabilitation Science, Osaka Metropolitan University (No. 2024-111). Written and oral informed consent was obtained from all participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Ekram ARMS, Woods RL, Britt C, et al. The Association between Frailty and All-Cause Mortality in Community-Dwelling Older Individuals: An Umbrella Review. J Frailty Aging 2021;10:320-6. [Crossref] [PubMed]
2. Boucham M, Salhi A, El Hajji N, et al. Factors associated with frailty in older people: an umbrella review. BMC Geriatr 2024;24:737. [Crossref] [PubMed]
3. Kojima G, Iliffe S, Taniguchi Y, et al. Prevalence of frailty in Japan: A systematic review and meta-analysis. J Epidemiol 2017;27:347-53. [Crossref] [PubMed]
4. Murayama H, Kobayashi E, Okamoto S, et al. National prevalence of frailty in the older Japanese population: Findings from a nationally representative survey. Arch Gerontol Geriatr 2020;91:104220. [Crossref] [PubMed]
5. Makizako H, Nishita Y, Jeong S, et al. TRENDS IN THE PREVALENCE OF FRAILTY IN JAPAN: A META-ANALYSIS FROM THE ILSA-J. J Frailty Aging 2021;10:211-8. [Crossref] [PubMed]
6. Levers MJ, Estabrooks CA, Ross Kerr JC. Factors contributing to frailty: literature review. J Adv Nurs 2006;56:282-91. [Crossref] [PubMed]
7. Gaston S, Billman E, Han L, et al. Frailty assessment utilization around the globe-a systematic review. J Frailty Aging 2025;14:100088. [Crossref] [PubMed]
8. Faller JW, Pereira DDN, de Souza S, et al. Instruments for the detection of frailty syndrome in older adults: A systematic review. PLoS One 2019;14:e0216166. [Crossref] [PubMed]
9. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146-56. [Crossref] [PubMed]
10. Rockwood K, Song X, MacKnight C, et al. A global clinical measure of fitness and frailty in elderly people. Can Med Assoc J 2005;173:489-95.
11. Stow D, Matthews FE, Hanratty B. Frailty trajectories to identify end of life: a longitudinal population-based study. BMC Med 2018;16:171. [Crossref] [PubMed]
12. Rasiah J, Gruneir A, Oelke ND, et al. Instruments to assess frailty in community dwelling older adults: A systematic review. Int J Nurs Stud 2022;134:104316. [Crossref] [PubMed]
13. Deng Y, Karako K, Yamauchi K, et al. Implementation and current status of frailty assessment in Japanese hospitals: Processes, epidemiology, and future directions. Biosci Trends 2026;20:139-48. [Crossref] [PubMed]
14. Uenishi M, Song P. Responding to a super-aged society: A community-based model for early frailty detection using AI and smart meter data -Insights from Japan. Glob Health Med 2025;7:439-43. [Crossref] [PubMed]
15. Peel MC, Finlayson BL, McMahon TA. Updated world map of the Köppen-Geiger climate classification. Hydrol Earth Syst Sci 2007;11:1633-44.
16. Davis LW, Gertler PJ. Contribution of air conditioning adoption to future energy use under global warming. Proc Natl Acad Sci U S A 2015;112:5962-7. [Crossref] [PubMed]
17. Dent E, Ambagtsheer RC, Beilby J, et al. Editorial: Frailty and Seasonality. J Nutr Health Aging 2020;24:547-9. [Crossref] [PubMed]
18. Wang J, Guan Y, Wu L, et al. Changing Lengths of the Four Seasons by Global Warming. Geophys Res Lett 2021;48:e2020GL091753.
19. Hasegawa J, Suzuki H, Yamauchi T. Impact of season on the association between muscle strength/volume and physical activity among community-dwelling elderly people living in snowy-cold regions. J Physiol Anthropol 2018;37:25. [Crossref] [PubMed]
20. Garriga A, Sempere-Rubio N, Molina-Prados MJ, et al. Impact of Seasonality on Physical Activity: A Systematic Review. Int J Environ Res Public Health 2021;19:2. [Crossref] [PubMed]
21. Bernard P, Chevance G, Kingsbury C, et al. Climate Change, Physical Activity and Sport: A Systematic Review. Sports Med 2021;51:1041-59. [Crossref] [PubMed]
22. Øverland S, Woicik W, Sikora L, et al. Seasonality and symptoms of depression: A systematic review of the literature. Epidemiol Psychiatr Sci 2019;29:e31. [Crossref] [PubMed]
23. Cianconi P, Betrò S, Janiri L. The Impact of Climate Change on Mental Health: A Systematic Descriptive Review. Front Psychiatry 2020;11:74. [Crossref] [PubMed]
24. Vachuska K. Cold weather isolation is worse in poor and non-white neighborhoods in the United States. Prev Med Rep 2024;38:102541. [Crossref] [PubMed]
25. Zhao W, Hu P, Sun W, et al. Effect of physical activity on the risk of frailty: A systematic review and meta-analysis. PLoS One 2022;17:e0278226. [Crossref] [PubMed]
26. Vaughan L, Corbin AL, Goveas JS. Depression and frailty in later life: a systematic review. Clin Interv Aging 2015;10:1947-58. [Crossref] [PubMed]
27. Katsurasako T, Murata S, Goda A, et al. Relationship between Mild Cognitive Impairment, Pre-Frailty, Physical and Psychological Functioning, and Functional Capacity among Community-Dwelling Older Adults. Healthcare (Basel) 2023;11:2542. [Crossref] [PubMed]
28. Zhou F, Zhou W, Wang W, et al. Associations between Frailty and Ambient Temperature in Winter: Findings from a Population-Based Study. Int J Environ Res Public Health 2022;20:513. [Crossref] [PubMed]
29. Portegijs E, Rantakokko M, Viljanen A, et al. Is frailty associated with life-space mobility and perceived autonomy in participation outdoors? A longitudinal study. Age Ageing 2016;45:550-3.
30. Chitalu P, Tsui A, Searle SD, et al. Life-space, frailty, and health-related quality of life. BMC Geriatr 2022;22:646. [Crossref] [PubMed]
31. Portegijs E, Iwarsson S, Rantakokko M, et al. Life-space mobility assessment in older people in Finland; measurement properties in winter and spring. BMC Res Notes 2014;7:323. [Crossref] [PubMed]
32. Jia H, Lubetkin EI. Time trends and seasonal patterns of health-related quality of life among U.S. adults. Public Health Rep 2009;124:692-701.
33. Hasegawa J. Elderly Individuals’ Satisfaction With Own Physical Activity Predicts Their Need for Care 3 Years Later. Gerontol Geriatr Med 2024;10:23337214241249917. [Crossref] [PubMed]
34. Borson S, Scanlan JM, Chen P, et al. The Mini-Cog as a screen for dementia: validation in a population-based sample. J Am Geriatr Soc 2003;51:1451-4. [Crossref] [PubMed]
35. Baker PS, Bodner EV, Allman RM. Measuring life-space mobility in community-dwelling older adults. J Am Geriatr Soc 2003;51:1610-4. [Crossref] [PubMed]
36. Craig CL, Marshall AL, Sjöström M, et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003;35:1381-95. [Crossref] [PubMed]
37. Oppe M, Devlin NJ, van Hout B, et al. A program of methodological research to arrive at the new international EQ-5D-5L valuation protocol. Value Health 2014;17:445-53. [Crossref] [PubMed]
38. Hoyl MT, Alessi CA, Harker JO, et al. Development and testing of a five-item version of the Geriatric Depression Scale. J Am Geriatr Soc 1999;47:873-8. [Crossref] [PubMed]
39. Brañez-Condorena A, Soriano-Moreno DR, Navarro-Flores A, et al. Accuracy of the Geriatric Depression Scale (GDS)-4 and GDS-5 for the screening of depression among older adults: A systematic review and meta-analysis. PLoS One 2021;16:e0253899. [Crossref] [PubMed]
40. Podsiadlo D, Richardson S. The timed “Up & Go”: a test of basic functional mobility for frail elderly persons. J Am Geriatr Soc 1991;39:142-8. [Crossref] [PubMed]
41. Bohannon RW, Bubela DJ, Magasi SR, et al. Sit-to-stand test: Performance and determinants across the age-span. Isokinet Exerc Sci 2010;18:235-40. [Crossref] [PubMed]
42. Lubben JE. Assessing social networks among elderly populations. Fam Community Health 1988;11:42-52.
43. Wagenmakers EJ, Marsman M, Jamil T, et al. Bayesian inference for psychology. Part I: Theoretical advantages and practical ramifications. Psychon Bull Rev 2018;25:35-57.
44. Mabe-Castro D, Gomez KT, Castillo-Aguilar M, et al. Frailty, Seasonal Sensitivity and Health-related Quality of Life in Older People Living in High Southern Latitudes: a Bayesian Analysis. Can Geriatr J 2024;27:56-62. [Crossref] [PubMed]
45. Kojima G, Iliffe S, Jivraj S, et al. Association between frailty and quality of life among community-dwelling older people: a systematic review and meta-analysis. J Epidemiol Community Health 2016;70:716-21. [Crossref] [PubMed]
46. Grimaldi S, Partonen T, Saarni SI, et al. Indoors illumination and seasonal changes in mood and behavior are associated with the health-related quality of life. Health Qual Life Outcomes 2008;6:56. [Crossref] [PubMed]
47. Meijering L, Osborne T, van Doorne M, et al. Seasonal differences in mobility and activity space in later life: a case study of older adults in the Northern Netherlands. Mobilities 2024;19:736-55. [Crossref] [PubMed]
48. Nakaji S, Parodi S, Fontana V, et al. Seasonal changes in mortality rates from main causes of death in Japan (1970--1999). Eur J Epidemiol 2004;19:905-13. [Crossref] [PubMed]
49. Ma C, Yang J, Nakayama SF, et al. Cold Spells and Cause-Specific Mortality in 47 Japanese Prefectures: A Systematic Evaluation. Environ Health Perspect 2021;129:67001. [Crossref] [PubMed]
50. Kimura T, Kobayashi H, Nakayama E, et al. Seasonality in physical activity and walking of healthy older adults. J Physiol Anthropol 2015;34:33. [Crossref] [PubMed]
51. Michalak EE, Tam EM, Manjunath CV, et al. Quality of life in patients with seasonal affective disorder: summer vs winter scores. Can J Psychiatry 2005;50:292-5. [Crossref] [PubMed]
52. Herdman M, Gudex C, Lloyd A, et al. Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L). Qual Life Res 2011;20:1727-36. [Crossref] [PubMed]
53. Brazier J. Is the EQ-5D fit for purpose in mental health? Br J Psychiatry 2010;197:348-9. [Crossref] [PubMed]
54. Polku H, Mikkola TM, Portegijs E, et al. Life-space mobility and dimensions of depressive symptoms among community-dwelling older adults. Aging Ment Health 2015;19:781-9. [Crossref] [PubMed]
55. Harada K, Masumoto K, Katagiri K, et al. Frequency of going outdoors and health-related quality of life among older adults: Examining the moderating role of living alone and employment status. Geriatr Gerontol Int 2018;18:640-7. [Crossref] [PubMed]
56. Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population. Am J Epidemiol 2017;186:1026-34. [Crossref] [PubMed]