Skip to main content
Download PDF

Identification and prediction of frailty among community-dwelling older Japanese adults based on Bayesian network analysis: a cross-sectional and longitudinal study

BMC Public Health volume 24, Article number: 2141 (2024) Cite this article

Abstract

Background

Frailty is a multifactorial syndrome; through this study, we aimed to investigate the physiological, psychological, and social factors associated with frailty and frailty worsening in community-dwelling older adults.

Methods

We conducted a cross-sectional and longitudinal study using data from the “Community Empowerment and Well-Being and Healthy Long-term Care: Evidence from a Cohort Study (CEC),” which focuses on community dwellers aged 65 and above in Japan. The sample of the cross-sectional study was drawn from a CEC study conducted in 2014 with a total of 673 participants. After excluding those who were frail during the baseline assessment (2014) and at the 3-year follow-up (2017), the study included 373 participants. Frailty assessment was extracted from the Kihon Checklist, while social relationships were assessed using the Social Interaction Index (ISI). Variable selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression and their predictive abilities were tested. Factors associated with frailty status and worsening were identified through the Maximum-min Hillclimb algorithm applied to Bayesian networks (BNs).

Results

At baseline, 14.1% (95 out of 673) participants were frail, and 24.1% (90 out of 373) participants experienced frailty worsening at the 3-years follow up. LASSO regression identified key variables for frailty. For frailty identification (cross-sectional), the LASSO model’s AUC was 0.943 (95%CI 0.913–0.974), indicating good discrimination, with Hosmer–Lemeshow (H–L) test p = 0.395. For frailty worsening (longitudinal), the LASSO model’s AUC was 0.722 (95%CI 0.656–0.788), indicating moderate discrimination, with H–L test p = 0.26. The BNs found that age, multimorbidity, function status, and social relationships were parent nodes directly related to frailty. It revealed an 85% probability of frailty in individuals aged 75 or older with physical dysfunction, polypharmacy, and low ISI scores; however, if their social relationships and polypharmacy status improve, the probability reduces to 50.0%. In the longitudinal-level frailty worsening model, a 75% probability of frailty worsening in individuals aged 75 or older with declined physical function and ISI scores was noted; however, if physical function and ISI improve, the probability decreases to 25.0%.

Conclusion

Frailty and its progression are prevalent among community-dwelling older adults and are influenced by various factors, including age, physical function, and social relationships. BNs facilitate the identification of interrelationships among these variables, quantify the influence of key factors. However, further research is required to validate the proposed model.

Peer Review reports

Background

Frailty is a common condition among older adults, characterized by the decline of multiple systems including physical, social, and cognitive functions [1, 2]. In Japan, the prevalence of frailty among community-dwelling older adults ranges from 8 to 17% and is expected to increase as the population ages [3,4,5,6]. Several factors contribute to an increased risk of developing frailty, including ageing, low physical activity, polypharmacy, multimorbidity, and insufficient social support [7,8,9]. Frailty reduces an individual’s ability to cope with external stressors [10, 11], making them more vulnerable to adverse health outcomes such as falls, hospitalization, increased disability, and higher mortality rates [12,13,14,15].

Evidence suggests that frailty is dynamic and may progress from robustness to functional decline [16, 17]. Studies from different countries have reported varying rates of frailty progression. A previous study [18] found that 23.7% older adults showed deterioration in frailty status after a year in a Chinese sample. Robust older adults have a 4%–7% chance of developing frailty each year In Korea sample [19]. Another study reviewed [20] 16 studies with an average follow-up of 3.9 years and found that 29.1% of participants experienced frailty worsening. Moreover, for every 0.01 increase in frailty score, the risk of death increased by 5% [21], with rapidly progressing individuals facing a higher risk of adverse outcomes [22]. Compared to physical measures used in these studies, the multidimensional measurement is commonly used to measure frailty in Japan [23,24,25,26], including physical function, nutritional status, oral health, housebound status, and cognitive function and provide a more comprehensive assessment of frailty.

Frailty is a complex condition resulting from multiple interconnected factors. Traditional statistical models typically show relationships between dependent and independent variables but fail to capture comprehensive impacts [27]. Bayesian networks (BNs), a graphical statistical tool, can display probabilistic relationships between variables through a directed acyclic graph (DAG) [28, 29]. To enhance the construction of BNs, we use Least Absolute Shrinkage and Selection Operator (LASSO) regression, a method that helps select relevant features and reduce the number of variables [30]. These methods are particularly well-suited for studying frailty, as they can capture the complex, interconnected nature of multiple factors contributing to this syndrome.

The current study applies Bayesian networks with LASSO regression at both cross-sectional and longitudinal levels and aimed to provide insights that can inform preventive interventions and therapeutic approaches for community-dwelling older adults.

Methods

Design and participants

This study used a mixed design that permitted both cross-sectional and longitudinal analyses. Participants were recruited from the “Community Empowerment and Care for Well-being and Healthy Longevity: Evidence from a Cohort Study (CEC)” that started in 1991 and was conducted every 3 years among all residents living in the suburban of central Japan to examine factors related to well-being and longevity. This study used data from the 2014 and 2017 waves.

The inclusion criterion was community-dwelling older adults aged 65 years and over. A total of 1004 participants from the 2014 sample responded to the baseline survey. Of them, 673 participants were included in the cross-sectional study analysis after excluding 331 participants with missing frailty data. In the longitudinal study, we excluded 95 participants with frailty at baseline, 52 who were lost to follow-up after 3 years of follow-up, and 153 with missing frailty data, 373 participants were included in the analysis.

Frailty status assessment

We assessed participants’ frailty using 1–20 items from the Kihon Checklist (KCL), which is widely used in Japan to screen for frailty in older adults [23, 24] and has been shown to have better predictive ability for adverse outcomes than the total 25-item version [25, 26]. It includes physical strength, nutritional status, oral function, housebound status, and cognitive function domains. A score of 7 or more out of 20 was defined as frailty (sensitivity: 77.0%; specificity: 75.6%) [25], and we defined frailty worsening as non-frailty developing to frailty.

Socio-demographic characteristics

Socio-demographic characteristics included age (65–74 years; 75 years or older), sex (male and female), living status (alone; together), and subjective economic status (poor; better off).

Lifestyle characteristics

In terms of lifestyle behaviors, the questionnaire assessed participants’ smoking habits (non-smoker, current smoker, and ex-smoker), heavy alcohol intake (daily and non-daily drinking), and physical activity (inactive and active). The dietary diversity score was calculated based on the sum of frequencies of each food item. The seven types of food included: vegetables, fruits, meat, fish, eggs, beans, and dairy products. The frequency of eating was defined as: 1, hardly eating; 2, 1–2 times per week; 3, 3–4 times per week; and 4, daily. The total dietary diversity scores ranged 7–28.

Disease characteristics

Self-reported medical history data on having diagnosed comorbidities (hypertension, heart disease, hyperlipidemia, diabetes, respiratory disease, liver-stomach disease, kidney disease, musculoskeletal disorders, blood and immune system disorders, cancer, dementia, stroke, and Parkinson’s disease) were recorded. Multimorbidity was defined as having two or more chronic diseases, and participants who had taken five or more drugs were considered to have polypharmacy.

Physical condition

According to Verbrugge and Jette’s model, decreased functional status is a disability-related process that can be defined as difficulty performing activities of daily living [31]. We defined people who needed help or care daily as lacking functional status [32]. Function status was assessed using the item, “Do you need some help or care in your daily life?” If the answer was “yes,” then the participant experienced physical dysfunction. After 3 years of follow-up, if the answer changed from not needing help to needing help, the participant’s functional status was said to have declined.

Psychological condition

The risk of depression in the participants was assessed using items 20–25 of the KCL; scores of 2 or higher were indicative of a depressed mood. Life satisfaction was assessed using one item: “Are you satisfied with your present life?” (yes or no). When participants’ answers changed from yes to no after the follow-up, life satisfaction was said to have declined.

Social relationships

Social relationships were measured using the Social Interaction Index (ISI) [33], which evaluates different aspects of social relationships and includes a total of 18 items and five subscales, including social curiosity, independence, interaction, participation, and feeling safe [34]. The scale ranges from 0 to 18, with higher scores indicating stronger social relationships and has been proven to be valid and reliable [35].

Statistical analysis

Qualitative data are presented as numbers (\(n)\) and percentages (%). The chi-square and Fisher’s exact tests were used for inter-group comparisons. Data complying with normal distribution were described by mean ± standard deviation (SD). For quantitative data that did not conform to a normal distribution, the medians ± 25th–75th percentiles were used to describe the data, and the rank-sum test was used for inter-group comparisons. Without special instructions, the test level α was set at 0.05. Missing values in less than 20% of the covariates were imputed using values from five datasets created with random forests for multivariate imputation, with the parameters set as follows: number of imputations (m) = 5, maximum number of iterations (maxit) = 50, method (meth) = ‘rf’ (random forest), and random seed (seed) = 1220. The optimal imputed values were determined based on the density plot distribution comparison between the original and imputed data.

Factors associated with cross-sectional frailty and longitudinal frailty worsening were selected using LASSO regression. The optimal tuning parameter (λ) was identified through a tenfold cross-validation. Selected factors were then included in multivariate logistic regression. Model performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC) for discrimination and calibration plots with Hosmer–Lemeshow goodness-of-fit test for calibration. LASSO regression was used to address multicollinearity [36]. It incorporates a penalty term to minimize estimated parameters, setting coefficients below a threshold to zero. This method identifies variables with stronger impacts on the dependent variable [30].

BN analysis was employed to examine interactions and probability dependencies between variables [37]. The maximum min hill climb algorithm was used for structural learning, creating the network topology [38]. Parameter learning used maximum likelihood estimation to estimate conditional probabilities for each node [39]. The study flowchart is shown in Fig. 1.

Fig. 1
figure 1

Flowchart of the core steps of the study

Full size image

In addition, we conducted sensitivity analysis to identify key variables of frailty identification and worsening. Our analysis focused on the ‘frailty’ and ‘frailty worsening’ node as the target variable. Analyses were performed using R version 4.2 (packages: “glmnet” for LASSO regression, “mice” for imputation), GeNIe 2.3 for BNs sensitivity analysis.

Results

A total of 673 participants (age: 73.65 ± 7.10 years, female 49.7%) were included at baseline. Of these, 14.1% individuals were classified as frail. After 3 years, 373 participants completed the follow-up, and 24.1% individuals experienced frailty worsening. Table 1 presents the characteristics of the participants. At the cross-sectional level, older adults who were older, female, daily drinkers, and physically inactive were more likely to fail; furthermore, they had lower dietary diversity, functional status, social relationships, and life satisfaction, and higher proportions of multimorbidity, polypharmacy, stroke, heart disease, dementia, musculoskeletal disorders, and depression.

Table 1 Participants characteristics (N = 673)
Full size table

Table 2 shows the participants’ characteristics according to frailty worsening. The proportion of individuals aged 75 years or older, with diabetes, musculoskeletal disorders, and depression in people with frailty worsening was higher. Furthermore, the functional status, life satisfaction, and social relationships of these people were significantly decreased (p < 0.05).

Table 2 Participants characteristics by the transition of frailty (N = 373)
Full size table

At baseline, 673 participants were screened using LASSO regression and tenfold cross-validation, and seven variables that were significantly correlated with frailty were screened in Fig. 2A and B. The horizontal axis represents the logarithm of the λ, and the vertical axis represents the model error, with each curve corresponding to a variable. It shows status changes, the coefficient relative to the coefficient vector of the path of the ℓ 1 norm. The best λ is the lowest point of the red curve and the number of variables is 15. The dotted line on the right represents a clean model (seven variables) within 1 SE. Since the model corresponding to the λ.min and λ.1SE has little variation in error, we choose optimal when λ was log(x) = -3.39 as it maintains a low model error while including fewer variables. And then, we further incorporate these potential factors related to frailty into the multivariate logistic regression model. Ultimately, being 75 years or older (odds ratio [OR] 1.93, 95% confidence interval [CI] 0.97–3.84), having physical dysfunction (OR 19.04, 95%CI 7.86–46.16), polypharmacy (OR 3.04, 95%CI 1.47–6.27), dementia (OR 25.54, 95%CI 2.22–293.81), musculoskeletal disorders (OR 5.07, 95%CI 2.23–11.52), depression (OR 4.93, 95%CI 2.46–9.90), and social relationships (OR 0.71, 95%CI 0.62–0.81) were associated with frailty among older adults (Supplementary Table 1). The model’s AUC value was 0.943 (95%CI 0.913–0.974), and the H–L test p = 0.395, indicating good fit (see Supplementary Fig. A and B curve and calibration plot).

Fig. 2
figure 2

Results of LASSO regression with tenfold cross-validation. A, B LASSO regression for frailty identification (cross-sectional study); C, D LASSO regression for frailty transition (longitudinal study). LASSO, least absolute shrinkage and selection operator

Full size image

The results of LASSO regression with 10-fold cross-validation for frailty worsening is shown in Fig. 2C and D. To achieve the highest predictive accuracy and more important variables, we selected λ.min, where log(λ) = -3.61. The final predictive model (Supplementary Table 2) included being 75 years or older (OR 2.05, 95%CI 1.17–3.58), female (OR 1.46, 95%CI 0.85–2.50), physically active (OR 0.74, 95%CI 0.43–1.28), having musculoskeletal disorders (OR 1.90, 95%CI 0.76–4.72), depression (OR 1.87, 95%CI 0.94–3.75), function status decline (OR 5.84, 95%CI 2.52–13.56), diabetes (OR 2.65, 95%CI 1.28–5.47), life satisfaction decline (OR 1.76, 95% CI 0.88–3.50), and ISI decline (OR 1.54, 95%CI 0.88–2.71). The model’s AUC value was 0.722 (95%CI = 0.656–0.788), indicating moderate discrimination. The H–L test p = 0.26, suggesting good model fit (Supplementary Figure C and D for ROC curve and calibration plot).

According to the variables identified by LASSO regression, we constructed two BNs models. The first model, focusing on frailty-related factors (Fig. 3A), consists of eight nodes and 16 directed edges. It shows that age, functional status, and ISI are directly related to frailty. Musculoskeletal disorders, dementia, polypharmacy and depression are indirectly link to frailty. The second model, predicting frailty progression (Fig. 3B) comprises 10 nodes and 22 directed edges. This model reveals that age, decline in functional status, and ISI scores are directly related to frailty worsening. Other factors such as sex, musculoskeletal disorders, diabetes, physical activity, and depression are indirectly linked to frailty worsening through a complex network relationship.

Fig. 3
figure 3

BNs topology of factors relating to frailty and frailty worsening. A BNs topology of factors relating to frailty status; B BNs topology of predictors relating to frailty worsening. Young: 65–75 years, older: ≥ 75 years. FS: function status. ISI: Index of Social Interaction score; divided by median; low: ISI score < 17, high: ISI score ≥ 17. BNs, Bayesian networks; ISI, Index of Social Interaction

Full size image

Figure 4 presents the conditional probability distribution of frailty and its progression, quantifying the relationship between parent nodes and the probability of frailty and worsening. Fig. 4A reveals an 85% probability of frailty in individuals aged 75 or older with physical dysfunction, polypharmacy, and low ISI scores. Conversely, participants aged 65–74 years with normal physical function and high ISI scores have the lowest conditional probability of frailty at 3.1%. Moreover, Fig. 4B shows an 83.3% probability of frailty worsening in individuals aged 75 or older with physical dysfunction, polypharmacy, and decreased ISI scores. In contrast, the lowest conditional probability of frailty worsening, at 16.1%, is observed in individuals aged 65–74 years with normal physical functioning and high ISI scores.

Fig. 4
figure 4

Conditional probability distribution of frailty and frailty worsening. A Conditional probability distribution of frailty with age, function status, and ISI as parent nodes; B Conditional probability distribution of frailty worsening with age, decline of function status (FS) and ISI as parent nodes. FS, function status; ISI, Index of Social Interaction

Full size image

Based on the participant’s physical, psychological, and social factors, BNs can infer the individual probability of frailty and its progression in community-dwelling older adults. For instance, a participant aged 75 years or older with polypharmacy, physical dysfunction, and low ISI scores has an estimated frailty probability of 85.0%. If their social relationships and polypharmacy status improve, this probability reduces to 50.0% (Fig. 5A and B). Similarly, the probability of frailty worsening is 75.0% for participants aged 75 years or older with reduced physical function and social relationship scores. However, if their physical function and social relationships improve, the probability decreases to 25.0% (Fig. 5C and D).

Fig. 5
figure 5

BNs under known evidence variables. A Conditional probability distribution of frailty for individuals who are 75 years old or older, with physical dysfunction, polypharmacy, and low ISI scores; B Conditional probability distribution of frailty with age 75 or older, physical dysfunction, high ISI scores, and without polypharmacy. C Conditional probability distribution of frailty worsening with age 75 or older, with FS and ISI decline. D Conditional probability distribution of frailty worsening with age 75 or older, without FS and ISI decline. FS, function status; ISI, Index of Social Interaction

Full size image

The results of BNs sensitivity analysis revealed the key factors influencing frailty identification and its worsening. For frailty identification, the top three influential variables were functional status, dementia, and age, with functional status having the greatest impact (sensitivity value: 0.57), followed by dementia (sensitivity value: 0.30) and age (sensitivity value: 0.10). Regarding frailty worsening, the most influential variables were functional status decline, age, and ISI decline. Functional status decline had the greatest effect (sensitivity value: 0.22), followed by age (sensitivity value: 0.15) and ISI decline (sensitivity value: 0.06). Additionally, musculoskeletal disorder and diabetes were also identified as significant factors.

Discussion

In the current study, we found that the prevalence of frailty among older adults in Japanese communities was assessed as 14.1%. Furthermore, at the longitudinal level, 24.1% of the participants transitioned from a non-frail to a frail state after 3 years. These results are comparable to previous findings [18,19,20] indicating that frailty and its progression are relatively common among community-dwelling older adults. Our study combined LASSO regression and BNs to offer a comprehensive analysis of these factors, providing valuable insights for future research.

Frailty is a multifaceted condition influenced by various factors. Our findings revealed that age, function status, and social relationships were directly related to frailty and frailty worsening. Age-related changes, particularly those affecting the musculoskeletal system, are common in older adults. Our study found that musculoskeletal disorders are associated with frailty through function status. The underlying mechanisms can be linked to factors identified in previous studies, such as decreased muscle mass and strength, reduced bone density, inflammation, immune dysfunction, and metabolic changes [40,41,42,43]. These physiological changes can contribute to musculoskeletal issues through alterations in muscle protein turnover, hormonal imbalances, and decreased physical activity [44]. Consequently, these changes may affect older adults’ physical function and potentially contributing to frailty and its worsening [45].

Our study also highlighted the indirect role of social relationships and depression in frailty through their impact on function status. Evidence suggests significant interactions between social factors, mental health, and frailty [46]. Social relationships, assessed through the ISI, provide access to health-related information and social support, which can buffer stress and promote healthy behaviors [47]. These mechanisms act as protective factors against frailty. Cohen et al. [48] suggested that social support influences health behaviors and functional states by reducing stress and encouraging positive health behaviors, thereby mitigating the risk of frailty and its progression.

Dementia, musculoskeletal disorders, polypharmacy, and depression were indirectly related to frailty by affecting physical function and social relationships. The relationship between dementia and frailty is particularly noteworthy as they often coexist and synergistically impact overall health in older adults [49]. Our logistic regression results showed a significant association between dementia and frailty, though with a wide confidence interval, indicating uncertainty likely due to the small number of dementia cases or significant differences between the frail and non-frail groups. Nevertheless, the BNs model developed supports this relationship, reinforcing the importance of considering dementia as a key factor in frailty assessments. Additionally, the results of the BNs indicate that reducing polypharmacy and improving social relationships among individuals aged 75 years or older can lower frailty risk. Our BNs model suggests that polypharmacy, defined as taking five or more medications, is associated with frailty indirectly through depression and social relationships. Previous studies have validated the mechanisms behind this association, showing that certain medications, such as tranquilizers, anticholinergics, and corticosteroids, can cause psychological side effects leading to depressive symptoms [50,51,52,53,54,55]. These symptoms can reduce social interactions, thereby increasing frailty risk [52, 53].

Our BNs model suggests that age, function status decline, and ISI decline are directly associated with frailty worsening. In contrast, sex, diabetes, musculoskeletal disorders, physical activity, depression, and decline in life satisfaction do not show direct associations with frailty worsening but may be indirectly associated through their relationships with other factors in the model. Among these, the most significant impacts were from the function status decline, age, ISI decline, musculoskeletal disorders, and diabetes. Our study indicated that the relationship between diabetes and worsening frailty is associated with decreased physical activity and functional status. This connection may be due to complications in patients with diabetes, such as peripheral neuropathy (nerve damage) and peripheral arterial disease (reduced blood flow to the extremities) [56, 57], which further impair their physical capabilities and functional status. While life satisfaction reflects an individual’s overall contentment with life, and chronic diseases or declining physical function can decrease life satisfaction, leading to depression and reduced social relationships, thereby increasing the risk of frailty. Conversely, individuals with high life satisfaction typically have stronger social networks and support systems [58]. These social relationships can provide emotional and practical support, promote engagement in healthy behaviors (e.g., physical activity and healthy diet), and help individuals cope with the aging and frailty challenges [59].

Our study’s BNs model can predict the probability of frailty and its progression, helping to identify individuals at risk and prioritize interventions. This study had several strengths. First, frailty occurrence is complex and involves several factors. We combined LASSO regression with BNs to identify and predict frailty status in community-dwelling older Japanese adults. LASSO effectively handles high-dimensional data and multiple variables while avoiding overfitting [60, 61]. BNs characterize interactions between factors and their effects on frailty through DAG [28, 29]. Second, rather than using a single measure of social participation, we used the ISI to comprehensively assess various daily social activities that may influence frailty.

Limitations

Our study has some limitations. First, we used a common frailty screening tool in Japan and data from a Japanese community in a population-based study. The results may vary when different assessment tools are used. Second, we did not collect information on income and education, which may influence frailty. Future studies should include these variables. Third, we relied on self-reported data, which may not accurately reflect actual physical capabilities. Future studies could incorporate objective measures like walking speed and grip strength. Fourth, a critical limitation is the lack of external validation. While we performed cross-validation of the LASSO model and sensitivity analysis on the BNs, we did not validate our findings in an independent cohort. Future research should prioritize external validation, k-fold cross-validation, and comparison with larger, external cohorts to enhance the model’s accuracy and generalizability. Finally, with a larger sample size, more detailed stratification of frailty factors could lead to a more robust model.

Conclusions

This study revealed the prevalence of frailty and its progression among Japanese community-dwelling older adults while examining the variations in the conditional probability of experiencing frailty and its worsening based on physical, psychological, and social factors. Our findings indicated a direct association between age, function status, and social relationships in older adults with frailty and its worsening. Moreover, our study highlights the potential benefits of the early identification of factors contributing to functional decline and social relationships decline, as it can aid in the prevention and delay of frailty development. BNs facilitate a deeper comprehension of the intricate interdependencies among variables, necessitating further investigation to consistently validate and optimize this framework for the early detection of risk factors and enhanced preventive interventions for frailty.

Availability of data and materials

The datasets are available from the corresponding author upon reasonable request.

References

  1. Gobbens RJ, Luijkx KG, Wijnen-Sponselee MT, Schols JM. Toward a conceptual definition of frail community dwelling older people. Nurs Outlook. 2010;58:76–86.

    Article  PubMed  Google Scholar 

  2. Gobbens RJ, van Assen MA, Luijkx KG, Schols JM. The predictive validity of the tilburg frailty indicator: disability, health care utilization, and quality of life in a population at risk. Gerontologist. 2012;52:619–31.

    Article  PubMed  Google Scholar 

  3. Murayama H, Kobayashi E, Okamoto S, Fukaya T, Ishizaki T, Liang J, et al. National prevalence of frailty in the older Japanese population: findings from a nationally representative survey. Arch Gerontol Geriatr. 2020;91:104220.

    Article  PubMed  Google Scholar 

  4. Yamada Y, Nanri H, Watanabe Y, Yoshida T, Yokoyama K, Itoi A, et al. Prevalence of frailty assessed by Fried and Kihon checklist indexes in a prospective cohort study: design and demographics of the Kyoto-Kameoka longitudinal study. J Am Med Dir Assoc. 2017;18(733):e7-733.e15.

    Google Scholar 

  5. Satake S, Senda K, Hong YJ, Miura H, Endo H, Sakurai T, et al. Validity of the Kihon checklist for assessing frailty status. Geriatr Gerontol Int. 2016;16:709–15.

    Article  PubMed  Google Scholar 

  6. Satake S, Shimokata H, Senda K, Kondo I, Toba K. Validity of total Kihon checklist score for predicting the incidence of 3-year dependency and mortality in a community-dwelling older population. J Am Med Dir Assoc. 2017;18:552.e1-552.e6.

    Article  PubMed  Google Scholar 

  7. Chu WM, Tange C, Nishita Y, Tomida M, Shimokata H, Otsuka R, et al. Effect of different types of social support on physical frailty development among community-dwelling older adults in Japan: evidence from a 10-year population-based cohort study. Arch Gerontol Geriatr. 2023;108:104928.

    Article  PubMed  Google Scholar 

  8. Jarach CM, Tettamanti M, Nobili A, D’avanzo B. Social isolation and loneliness as related to progression and reversion of frailty in the Survey of Health Aging Retirement in Europe (SHARE). Age Ageing. 2021;50:258–62.

    Article  PubMed  Google Scholar 

  9. Ding YY, Kuha J, Murphy M. Multidimensional predictors of physical frailty in older people: identifying how and for whom they exert their effects. Biogerontology. 2017;18:237–52.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet. 2013;381:752–62.

    Article  PubMed  Google Scholar 

  11. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci. 2001;56:M146–56.

    Article  CAS  PubMed  Google Scholar 

  12. Ensrud KE, Ewing SK, Taylor BC, Fink HA, Cawthon PM, Stone KL, et al. Comparison of 2 frailty indexes for prediction of falls, disability, fractures, and death in older women. Arch Intern Med. 2008;168:382–9.

    Article  PubMed  Google Scholar 

  13. Kojima G. Frailty as a predictor of disabilities among community-dwelling older people: a systematic review and meta-analysis. Disabil Rehabil. 2017;39:1897–908.

    Article  PubMed  Google Scholar 

  14. Theou O, Sluggett JK, Bell JS, Lalic S, Cooper T, Robson L, et al. Frailty, hospitalization, and mortality in residential aged care. J Gerontol A Biol Sci Med Sci. 2018;73:1090–6.

    Article  PubMed  Google Scholar 

  15. Bartley MM, Geda YE, Christianson TJ, Pankratz VS, Roberts RO, Petersen RC. Frailty and mortality outcomes in cognitively normal older people: sex differences in a population-based study. J Am Geriatr Soc. 2016;64:132–7.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Lang PO, Michel JP, Zekry D. Frailty syndrome: a transitional state in a dynamic process. Gerontology. 2009;55:539–49.

    Article  PubMed  Google Scholar 

  17. Bessa B, Coelho T, Ribeiro Ó. Social frailty dimensions and frailty models over time. Arch Gerontol Geriatr. 2021;97:104515.

  18. Yuan Y, Lin S, Huang X, Li N, Zheng J, Huang F, et al. The identification and prediction of frailty based on Bayesian network analysis in a community-dwelling older population. BMC Geriatr. 2022;22:847.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Kim E, Sok SR, Won CW. Factors affecting frailty among community-dwelling older adults: a multi-group path analysis according to nutritional status. Int J Nurs Stud. 2021;115:103850.

    Article  PubMed  Google Scholar 

  20. Kojima G, Taniguchi Y, Iliffe S, Jivraj S, Walters K. Transitions between frailty states among community-dwelling older people: a systematic review and meta-analysis. Ageing Res Rev. 2019;50:81–8.

    Article  PubMed  Google Scholar 

  21. Evans SJ, Sayers M, Mitnitski A, Rockwood K. The risk of adverse outcomes in hospitalized older patients in relation to a frailty index based on a comprehensive geriatric assessment. Age Ageing. 2014;43:127–32.

    Article  PubMed  Google Scholar 

  22. Lewis C, O’Caoimh R, Patton D, O’Connor T, Moore Z, Nugent LE. Risk prediction for adverse outcomes for frail older persons with complex healthcare and social care needs admitted to a community virtual ward model. Clin Interv Aging. 2020;15:915–26.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Cui M, Jiao D, Liu Y, Zhu Y, Li X, Zhu Z, et al. Longitudinal assessment of the relationship between frailty and social relationships among Japanese older adults: a random intercept cross-lagged panel model. BMC Public Health. 2024;24:706.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Ishikawa N, Katsura T, Hara M. Changes in Kihon checklist items and new certification of long-term care needs among Japanese community-dwelling elders. J Rural Med. 2021;16:270–9.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Tomata Y, Hozawa A, Ohmori-Matsuda K, Nagai M, Sugawara Y, Nitta A, et al. Validation of the Kihon checklist for predicting the risk of 1-year incident long-term care insurance certification: the Ohsaki cohort 2006 study [Nihon Koshu eisei zasshi]. Nihon Koshu Eisei Zasshi. 2011;58:3–13.

    PubMed  Google Scholar 

  26. Ogawa K, Fujiwara Y, Yoshida H, Nishi M, Fukaya T, Kim M, et al. The validity of the "Kihon Check-list" as an index of frailty and its biomarkers and inflammatory markers in elderly people. Jpn J Geriatr. 2011;48:545–52.

  27. Wang X, Pan J, Ren Z, Zhai M, Zhang Z, Ren H, et al. Application of a novel hybrid algorithm of Bayesian network in the study of hyperlipidemia related factors: a cross-sectional study. BMC Public Health. 2021;21:1375.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Scutari M. Learning Bayesian networks with the bnlearn R package. J Stat Softw. 2010;35:1–22.

    Article  Google Scholar 

  29. Fuster-Parra P, Vidal-Conti J, Borràs PA, Palou P. Bayesian networks to identify statistical dependencies. A case study of Spanish university students’ habits. Inform Health Soc Care. 2017;42:166–79.

    Article  CAS  PubMed  Google Scholar 

  30. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33:1–22.

    Article  PubMed  PubMed Central  Google Scholar 

  31. Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38:1–14.

    Article  CAS  PubMed  Google Scholar 

  32. Watanabe K, Tanaka E, Watanabe T, Chen W, Wu B, Ito S, et al. Association between the older adults’ social relationships and functional status in Japan. Geriatr Gerontol Int. 2017;17:1522–6.

    Article  PubMed  Google Scholar 

  33. Anme T. Study to develop “index of social interaction.” J Natl Rehab Cen Disabled. 1992;13:1–7.

    Google Scholar 

  34. Watanabe K, Tanaka E, Watanabe T, Tomisaki E, Ito S, Okumura R, et al. Social relationships and functional status among Japanese elderly adults living in a suburban area. Public Health. 2020;179:84–9.

    Article  CAS  PubMed  Google Scholar 

  35. Jiao D, Watanabe Miura K, Sawada Y, Tanaka E, Watanabe T, Tomisaki E, et al. Changes in social relationships and physical functions in community-dwelling older adults. J Nurs Res. 2022;30:e228.

    Article  PubMed  Google Scholar 

  36. Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc Series B. 1996;58:267–88.

    Article  Google Scholar 

  37. Scutari M, Denis JB. Bayesian networks: with examples in R. Boca Raton: Chapman and Hall/CRC; 2014.

  38. Tsamardinos I, Brown LE, Aliferis CF. The max-min hill-climbing Bayesian network structure learning algorithm. Mach Learn. 2006;65:31–78.

    Article  Google Scholar 

  39. Heckerman D, Geiger D, Chickering DM. Learning Bayesian networks: the combination of knowledge and statistical data. Mach Learn. 1995;20:197–243.

    Article  Google Scholar 

  40. Larsson L, Degens H, Li M, Salviati L, Lee YI, Thompson W, et al. Sarcopenia: aging-related loss of muscle mass and function. Physiol Rev. 2019;99:427–511.

    Article  PubMed  Google Scholar 

  41. Gielen E, Dupont J, Dejaeger M, Laurent MR. Sarcopenia, osteoporosis and frailty. Metabolism. 2023;145:155638.

    Article  CAS  PubMed  Google Scholar 

  42. Soto ME, Pérez-Torres I, Rubio-Ruiz ME, Cano-Martínez A, Manzano-Pech L, Guarner-Lans V. Frailty and the interactions between skeletal muscle, bone, and adipose tissue-impact on cardiovascular disease and possible therapeutic measures. Int J Mol Sci. 2023;24:4534.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Allen SC. Systemic inflammation in the genesis of frailty and sarcopenia: an overview of the preventative and therapeutic role of exercise and the potential for drug treatments. Geriatrics (Basel, Switzerland). 2017;2:6.

    PubMed  Google Scholar 

  44. Leng S, Chaves P, Koenig K, Walston J. Serum interleukin-6 and hemoglobin as physiological correlates in the geriatric syndrome of frailty: a pilot study. J Am Geriatr Soc. 2002;50:1268–71.

    Article  PubMed  Google Scholar 

  45. Chou CH, Hwang CL, Wu YT. Effect of exercise on physical function, daily living activities, and quality of life in the frail older adults: a meta-analysis. Arch Phys Med Rehabil. 2012;93:237–44.

    Article  PubMed  Google Scholar 

  46. Tough H, Siegrist J, Fekete C. Social relationships, mental health and wellbeing in physical disability: a systematic review. BMC Public Health. 2017;17:414.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Hsu HC, Chang WC. Trajectories of frailty and related factors of the older people in Taiwan. Exp Aging Res. 2015;41:104–14.

    Article  PubMed  Google Scholar 

  48. Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol Bull. 1985;98:310–57.

    Article  CAS  PubMed  Google Scholar 

  49. Kulmala J, Nykänen I, Mänty M, Hartikainen S. Association between frailty and dementia: a population-based study. Gerontology. 2014;60:16–21.

    Article  PubMed  Google Scholar 

  50. Nwadiugwu MC. Frailty and the risk of polypharmacy in the older person: enabling and preventative approaches. J Aging Res. 2020;2020:6759521.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Cheng C, Bai J. Association between polypharmacy, anxiety, and depression among Chinese older adults: evidence from the Chinese longitudinal healthy longevity survey. Clin Interv Aging. 2022;17:235–44.

    Article  PubMed  PubMed Central  Google Scholar 

  52. Qato DM, Ozenberger K, Olfson M. Prevalence of prescription medications with depression as a potential adverse effect among adults in the United States. JAMA. 2018;319:2289–98.

    Article  PubMed  PubMed Central  Google Scholar 

  53. López-Álvarez J, Sevilla-Llewellyn-Jones J, Agüera-Ortiz L. Anticholinergic drugs in geriatric psychopharmacology. Front Neurosci. 2019;13:1309.

    Article  PubMed  PubMed Central  Google Scholar 

  54. Gutiérrez-Valencia M, Izquierdo M, Cesari M, Casas-Herrero Á, Inzitari M, Martínez-Velilla N. The relationship between frailty and polypharmacy in older people: a systematic review. Br J Clin Pharmacol. 2018;84:1432–44.

    Article  PubMed  PubMed Central  Google Scholar 

  55. Soysal P, Veronese N, Thompson T, Kahl KG, Fernandes BS, Prina AM, et al. Relationship between depression and frailty in older adults: a systematic review and meta-analysis. Ageing Res Rev. 2017;36:78–87.

    Article  PubMed  Google Scholar 

  56. Choi JH, Kim HR, Song KH. Musculoskeletal complications in patients with diabetes mellitus. Korean J Intern Med. 2022;37:1099–110.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Zhao RR, O’Sullivan AJ, Fiatarone Singh MA. Exercise or physical activity and cognitive function in adults with type 2 diabetes, insulin resistance or impaired glucose tolerance: a systematic review. Eur Rev Aging Phys Act. 2018;15:1.

    Article  PubMed  PubMed Central  Google Scholar 

  58. Papi S, Cheraghi M. Multiple factors associated with life satisfaction in older adults. Prz Menopauzalny. 2021;20:65–71.

    PubMed  PubMed Central  Google Scholar 

  59. Etman A, Kamphuis CB, van der Cammen TJ, Burdorf A, van Lenthe FJ. Do lifestyle, health and social participation mediate educational inequalities in frailty worsening? Eur J Public Health. 2015;25:345–50.

    Article  PubMed  Google Scholar 

  60. Li Y, Chen X, Wang Y, Hu J, Shen Z, Ding X. Application of group lasso regression based Bayesian networks in risk factors exploration and disease prediction for acute kidney injury in hospitalized patients with hematologic malignancies. BMC Nephrol. 2020;21:162.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Hepp T, Schmid M, Gefeller O, Waldmann E, Mayr A. Approaches to regularized regression - a comparison between gradient boosting and the Lasso. Methods Inf Med. 2016;55:422–30.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We appreciate all the study participants, the staff of Tobishima, and their investigators.

Funding

This work was supported by grants from the Funds for Cooperation Project of Nanchong City and North Sichuan Medical College (grant number: 22SXQT0189), and the China Scholarship Council (grant number: 202208510018).

Author information

Author notes

  1. Mengjiao Yang and Yang Liu are equal first authors.

Authors and Affiliations

  1. Graduate School of Comprehensive Human Science, University of Tsukuba, Tsukuba, Japan

    Mengjiao Yang, Yang Liu, Zhu Zhu, Xiang Li, Mingyu Cui, Jinrui Zhang, Meiling Qian & Lujiao Huang

  2. Department of Cardiovascular Surgery. Affiliated Hospital of North Sichuan Medical College, Nanchong, 637000, China

    Mengjiao Yang

  3. RIKEN Center for Advanced Intelligence Project, Tokyo, 1030027, Japan

    Kumi Watanabe Miura

  4. University of Reiwa Health Sciences, Fukuoka, 8110213, Japan

    Munenori Matsumoto

  5. Department of Nursing, The First Affiliated Hospital and College of Clinical Medicine, Henan University of Science and Technology, Luoyang, 471003, China

    Dandan Jiao

  6. School of Nursing, Hangzhou Normal University, Hangzhou, 311121, China

    Zhu Zhu

  7. College of Child Development and Education, Zhejiang Normal University, Hangzhou, 311231, China

    Xiang Li

  8. Department of Clinical Nutrition, Sichuan Provincial People’s Hospital, University of Electronic Science and Technology of China, Chengdu, 610000, China

    Lujiao Huang

  9. Faculty of Medicine, University of Tsukuba, Tsukuba, Ibaraki, 3058577, Japan

    Tokie Anme

Authors
  1. Mengjiao Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yang Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Kumi Watanabe Miura
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Munenori Matsumoto
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Dandan Jiao
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Zhu Zhu
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Xiang Li
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Mingyu Cui
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Jinrui Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Meiling Qian
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Lujiao Huang
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Tokie Anme
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

YMJ and LY contributed to the conception, design, analysis, drafting and revision of this paper. KM, MM, JDD, CMY, ZJR, QML, and HLJ contributed to the interpretation, drafting and revision of this paper. KM, MM, JDD, ZZ, and LX contributed to the collection of the data. TA was the primary researcher of the original data. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Tokie Anme.

Ethics declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the University of Tsukuba (approval number 1331–5). The research utilized anonymized data provided by the local government. According to the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan, the requirement for written informed consent was waived. Participants retained the right to opt out of the study for any reason. All procedures were conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yang, M., Liu, Y., Miura, K.W. et al. Identification and prediction of frailty among community-dwelling older Japanese adults based on Bayesian network analysis: a cross-sectional and longitudinal study. BMC Public Health 24, 2141 (2024). https://doi.org/10.1186/s12889-024-19697-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12889-024-19697-y

Keywords

BMC Public Health

ISSN: 1471-2458

Contact us