Abstract
Objectives
Few studies have examined the association between social frailty and dementia. We aim to assess if social frailty indices predict dementia and evaluate causality.
Methods
Data from 851 community-dwelling participants aged 70 years and older without dementia from the Sydney Memory and Ageing Study were used. Social frailty was assessed using five published indices that incorporated a range of psychosocial factors. Incident dementia was assessed by consensus diagnosis using biennial neuropsychological tests over a 12-year period.
Results
Based on Cox regression, social frailty was predictive of an increased risk of incident dementia using three different indices. After adjusting for a range of confounders, including physical and psychological frailty, evidence of a causal association was found using one of the indices (socially frail vs nonfrail hazard ratio [HR] 1.47, 95% confidence interval [CI] 1.05–2.07). Low financial and family satisfaction, infrequent social contacts, and limited participation in social activities and volunteering were associated with an increased risk of dementia, with low financial satisfaction an independent predictor.
Discussion
Well-designed social frailty indices are an effective way to help identify older adults at increased risk of dementia. Replication of these findings in other cohorts is necessary to validate a screening tool for social frailty. Screening for social frailty using brief measures in primary care is a key next step, alongside social prescribing programs for those at risk of social frailty and cognitive decline.
Keywords: Cognitive decline, Risk prediction, Social frailty index, Social vulnerability
Frailty is a complex multifactorial condition that renders older adults susceptible to adverse health outcomes (Khalil & Gobbens, 2023). Frailty may be regarded as comprising three dimensions, (1) physical, (2) psychological, and (3) social, with social frailty receiving less attention (Bunt et al., 2017). Social frailty is characterized by vulnerability to losing resources that are important to fulfill basic social needs (Bunt et al., 2017). It is associated with numerous adverse health events, including incident disability, mortality, physical frailty, and reduced cognitive function (Li et al., 2023). Prevalence varies by population and assessment method, with estimates ranging from 4.9% in Chinese older adults to 29.2% in European older adults (Zhang et al., 2023).
Despite its importance to cognitive function, research on social frailty is limited. A recent review found only one longitudinal study on social frailty and cognitive impairment (Li et al., 2023). This study of 3,720 Japanese older adults (mean age 71.7 years) found an increased risk of Alzheimer’s disease over 53 months in socially frail versus nonfrail participants. No association was found for socially prefrail (displaying some features of social frailty but insufficient to meet the criteria) participants compared with nonfrail participants (Tsutsumimoto et al., 2019).
There is limited evidence on whether sex differences exist in the association between social frailty and cognitive impairment. However, given that dementia risk and social frailty rates are higher in women (Gong et al., 2023; Zhang et al., 2023), it is plausible that sex may moderate this association. Previous studies have shown sex differences in the relationship between social connections and cognitive health. In a South Korean cohort, social club participation provided more protection against cognitive decline in men than women (Lee & Jean Yeung, 2019).
Various tools incorporating a wide range of psychosocial factors have been developed to assess social frailty (Cohen et al., 2023). Commonly assessed factors include social isolation and the lack of a social network (e.g., frequency of visits with friends and family, living alone, absence of a confidant, frequency of talking to someone), emotional aspects (e.g., feeling helpful to others, experiencing loneliness), and adequacy of resources (e.g., sufficient financial resources and support from others) (Makizako et al., 2015; Teo et al., 2017; Yamada & Arai, 2018).
Although numerous indices have been developed, their application in exploring the association between social frailty and dementia risk is limited, particularly in the Australian population. Early identification of socially frail older adults could lead to interventions that improve their psychosocial environment, potentially reducing the risk of dementia and other adverse health outcomes. Interventions to reduce social vulnerability in older adults have shown beneficial effects on cognition, function, and subjective health (Mah et al., 2022).
This study aims to validate a concise screening tool for social frailty, suitable for primary care settings or self-screening, to identify older adults at increased risk of dementia. Our primary goal is to evaluate existing social frailty indices and their effectiveness in identifying older Australian adults at higher dementia risk. In addition, we will investigate whether social frailty independently predicts dementia risk or if associations are due to physical or psychological frailty or confounding factors. Finally, we will examine whether these associations differ by sex.
Method
Study design and population
We mapped items from social frailty indices against data from the Sydney Memory and Ageing Study (MAS), a longitudinal, observational study of older adults. Details of recruitment methods and exclusion criteria have been described previously (Sachdev et al., 2010). In brief, participants were recruited between 2005 and 2007 from two electoral areas of Sydney using the Australian electoral roll. At baseline, 1,037 English-speaking community-dwelling adults aged 70–90 years without dementia were recruited. Participants were followed up approximately every 2 years for up to 12 years.
The MAS study was approved by the Ethics Committees for UNSW Sydney and the South Eastern Sydney Illawarra Area Health Service. All research was conducted in accordance with the guidelines set out in the Declaration of Helsinki. All participants provided written informed consent.
Social frailty indices
Based on a global review of social frailty literature, we identified five commonly used social frailty indices; the 5-item index by Makizako et al. (2015), the 8-item index by Shah et al. (2023), the 7-item index by Teo et al. (2017), the 3-item social components of the Tilburg Frailty Index by Gobbens et al. (2010), and the 4-item index by Yamada and Arai (2018). Each question from these indices was mapped to the most similar question in the MAS baseline assessment. One question from the index by Shah et al. (How often are you treated with less courtesy or respect than other people?) was excluded due to the lack of a similar question in the MAS assessment (Table 1 indices summary) (Supplementary Tables 1–5 [see online supplementary material] social frailty detailed indices and mapping to MAS questions).
Table 1.
Summary of social frailty indices.
| Component | Makizako | Shah | Teo | Tilburg | Yamada |
|---|---|---|---|---|---|
| Activity engagement | Dropped activities and interests | Frequency volunteering | Frequency social activities and volunteering | Frequency social activities and volunteering | |
| Network size | Number of friends and relatives in regular contact | ||||
| Loneliness | Frequency experienced | Frequency experienced | |||
| Social support | Presence of confidant | ||||
| Self-worth perception | Feelings of worthlessness | ||||
| Living situation | Live alone | Live alone | Live alone | Live alone | |
| Frequency of face-to-face contact | Frequency contact (<11/month) | Frequency contact (<11/month) | Infrequent contact (<1/month) | Frequency contact (<11/month) | |
| Relationship satisfaction | Satisfaction with family | Satisfaction with family and friendships | |||
| Employment situation | Frequency paid work | ||||
| Neighborhood satisfaction | Satisfaction level | ||||
| Financial situation | Satisfaction level | Satisfaction level | Satisfaction level | ||
| Educational attainment | Highest qualification | ||||
| Scoring | 0–5 | 0–7 | 0–6 | 0–3 | 0–4 |
| Categories | |||||
| Frail | 2+ | 3+ | 2+ | 2+ | 2+ |
| Prefrail | 1 | 2 | 1 | 1 | 1 |
| Nonfrail | 0 | 0 | 0 | 0 | 0 |
Each question was scored as 1 or 0, indicating the presence or absence of a risk factor for social frailty. The scores were summed to categorize participants as “frail,” “prefrail,” or “nonfrail” according to predefined cut-offs. As the Shah index uses the lowest deciles or tertiles rather than predefined cutoffs to identify frail participants, cutoffs in the current study were determined to approximate these quantiles. In the main analysis, we used a cutoff score of 3 to approximate tertiles, while a score of 4 was used in sensitivity analysis to approximate deciles. If 20% or more questions were missing, the assessment was invalid, except if the total score already met the frailty threshold based on completed questions.
Dementia assessment
At baseline and each follow-up, dementia status was assessed via consensus diagnosis (details Supplementary Table 6, see online supplementary material) based on DSM-IV criteria (American Psychiatric Association, 1994).
Covariates
Covariates were identified based on the Livingston et al. (2024) modifiable risk factors for dementia and included current smoking (yes/no), alcohol consumption (units/day), vision loss (present/absent), hypertension (present/absent), diabetes (present/absent), hyperlipidemia (present/absent), and traumatic brain injury (present/absent). In addition, physical frailty based on Fried’s 5-item Frailty Phenotype (Fried et al., 2001) and psychological frailty based on the psychological components of the Groningen Frailty indicator (Peters et al., 2012) were included (Supplementary Table 7 [see online supplementary material] assessment methods for all covariates). Education was included in one of the indices and therefore not additionally included as a covariate. Similarly, a measure of the participants’ mood, taken from the Geriatric Depression Scale (GDS; Yesavage et al., 1982), was incorporated within the psychological frailty covariate; the full GDS score was not included because two of its questions were included in one of the indices. Baseline cognition using the Mini-Mental State Examination (MMSE) (Folstein et al., 1975) was included as a covariate in the sensitivity analysis. All covariates were assessed at baseline.
Statistical analysis
Participant characteristics by sex were presented as mean and standard deviation (SD) for continuous data and percentage for categorical data. Continuous data were assessed for normality using skewness and kurtosis statistics and visual inspections of histograms.
The average percentage of missing covariate data was 1.5% with the highest being 8.5% for the timed sit-to-stand variable. Missing covariate data were imputed with 20 imputed datasets using all other covariate data as predictors. Convergence statistics were checked, and the imputed data were compared against original data using summary statistics. Imputation was performed using the mice package version 3.15.0 in R (Van Buuren & Groothuis-Oudshoorn, 2011).
Associations between baseline “frail” and “prefrail” status and incident dementia were analyzed using Cox Proportional Hazards regression with “nonfrail” as the reference group. HR and 95% CIs were calculated accounting for both the average of the coefficients and standard errors across imputations, as well as the variability of the coefficients between imputations (Rubin, 1987). Time-to-event was calculated as the midpoint between the first follow-up where dementia was identified and the previous dementia assessment or the time between baseline and the last assessment for participants without incident dementia. Proportional hazards assumptions were checked. Analysis was performed in R using the survival package version 3.5-5 (Therneau, 2023; Therneau & Grambsch, 2000).
Age and sex were included as covariates in the minimally adjusted model to assess the increased risk associated with social frailty beyond the risk associated with these demographic variables. Physical and psychological frailty were additionally adjusted for in the partially adjusted model to assess if a relationship existed independent of other frailty phenotypes. The fully adjusted model additionally adjusted for smoking, alcohol, vision, hypertension, diabetes, hyperlipidemia, and traumatic brain injury.
The analysis was then repeated, adding an interaction term between social frailty and sex to test for sex differences. Where a sex difference existed, analyses were performed with participants stratified by sex.
To assess the robustness of our findings, two sensitivity analyses were conducted. First, the cut-off scores for each index were increased by one, in line with the two cut-off scores derived from the quantiles used in the Shah index. In the second, baseline cognition, measured continuously using the MMSE, was included as an additional covariate.
Results
Participant baseline characteristics
Of the 1,037 original participants, 186 were excluded due to not completing neuropsychological testing at baseline and at least one follow-up, leaving 851 participants. Their mean age (SD) was 77.9 years (4.7 years), with women comprising 54.6% of the sample. Most participants lived in the community, with male participants typically living with their spouse, while female participants generally lived alone. Compared with men, women had lower education, were less likely to be in paid employment, more likely to perform charity work, and more likely to have someone to confide in (Table 2). Over the 12-year follow-up period, 260 new cases of dementia were identified (30.6% of participants).
Table 2.
Baseline characteristics of participants.
| Characteristic | Total | Female | Male | p-value |
|---|---|---|---|---|
| Age, years (SD) | 77.9 (4.7) | 77.9 (4.7) | 78.0 (4.6) | .684 |
| Sex, n (%) | 851 (100.0%) | 465 (54.6%) | 386 (45.4%) | |
| Education, years (SD) | 11.6 (3.5) | 11.1 (3.0) | 12.3 (3.9) | <.001 |
| Marital status, n (%) | <.001 | |||
| Never married | 105 (12.4%) | 44 (9.5%) | 61 (15.8) | |
| Married/defector | 363 (42.8%) | 130 (28.0%) | 233 (60.5%) | |
| Divorced/separated | 102 (12.0%) | 65 (14.0%) | 37 (9.6%) | |
| Widowed | 279 (32.9%) | 225 (48.5%) | 54 (14.0%) | |
| Where living, n (%) | <.001 | |||
| Community alone | 397 (46.7%) | 285 (61.3%) | 112 (29.0%) | |
| Community with spouse | 340 (40.0%) | 119 (25.6%) | 221 (57.3%) | |
| Community with others | 100 (11.8%) | 55 (11.8%) | 45 (11.7%) | |
| Hostel villa/retirement home | 14 (1.6%) | 6 (1.3%) | 8 (2.1%) | |
| Perform paid work, n (%) | .006 | |||
| Yes | 150 (18.1%) | 62 (13.7%) | 88 (23.4%) | |
| Perform charity work, n (%) | .034 | |||
| Yes | 430 (52.6%) | 240 (54.2%) | 190 (50.8%) | |
| Have someone to confide in, n (%) | .035 | |||
| Yes | 766 (94.6%) | 430 (96.2%) | 336 (92.6%) | |
| Face-to-face contacts per month, n (%) | .053 | |||
| 11 or more | 534 (64.4%) | 313 (68.8%) | 221 (59.1%) | |
| 5–10 | 184 (22.2%) | 90 (19.8%) | 94 (25.1%) | |
| 1–4 | 91 (11.0%) | 43 (9.5%) | 48 (12.8%) | |
| <1 month | 20 (2.4%) | 9 (2.0%) | 11 (2.9%) | |
| Current smoker, number /day (SD) | 31 (3.7%) | 20 (4.3%) | 11 (2.9%) | .004 |
| Alcoholic drinks, number/day (SD) | 2.0 (1.6) | 1.6 (1.2) | 2.5 (1.8) | <.001 |
| Diabetes present, n (%) | 121 (14.2%) | 45 (9.7%) | 76 (19.7%) | <.001 |
| Hypertension present, n (%) | 705 (82.8%) | 367 (78.9%) | 338 (87.6%) | .001 |
| Hyperlipidemia present, n (%) | 513 (60.5%) | 281 (60.6%) | 232 (60.4%) | 1.000 |
| Vision loss present, n (%) | 3 (0.4%) | 2 (0.4%) | 1 (0.3%) | 1.000 |
| Body mass index, kg/m2 | 27.1 (4.4) | 26.7 (4.6) | 27.6 (4.1) | .002 |
| Sit-to-stand, sec (SD) | 16.6 (5.4) | 16.8 (5.9) | 16.4 (4.8) | .291 |
| 6-m walk, sec (SD) | 9.1 (2.8) | 9.3 (3.0) | 8.8 (2.4) | .006 |
| Physical activity, min/week (SD) | 426.5 (455.9) | 345.7 (354.6) | 523.9 (538.5) | <.001 |
| History of head injury, n (%) | 126 (14.8%) | 70 (15.1%) | 56 (14.6%) | .924 |
Note. Values in bold indicate p < .05. p-values based on t tests for continuous data and Chi-squared tests for categorical data.
Compared with included participants, those excluded were older, engaged in less physical activity, and had slower sit-to-stand and 6-m walk times. They also had fewer face-to-face contacts, consumed less alcohol, were more likely to experience vision loss, and were less likely to live with their spouse or be engaged in paid work (Supplementary Table 8, see online supplementary material).
Association between social frailty and incident dementia
Minimally adjusted model
Hazard ratios for the association between social frailty indices and incident dementia are shown in Table 3. The increased risk associated with being socially frail versus nonfrail was highest using the Shah and Teo indices (HR 1.58 [95% CI 1.14–2.19] and HR 1.58 [95% CI 1.08–2.30], respectively). The Yamada index also indicated increased risk (HR 1.46 [95% CI 1.03–2.06]), whereas the Makizako and Tilburg indices did not. For prefrail versus nonfrail participants, only the Shah index showed increased risk (HR 1.59 [95% CI 1.16–2.18]).
Table 3.
HR for associations between social frailty status and incident dementia.
| Social frailty status | n | Prevalence | Cases | Incidence | Minimally adjusteda | Partially adjustedb | Fully adjustedc |
|---|---|---|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||
| Makizako | 848 | 100.0% | 259 | ||||
| Nonfrail | 210 | 24.8% | 64 | 30.5% | Ref. | Ref. | Ref. |
| Prefrail | 391 | 46.1% | 110 | 28.1% | 0.88 (0.65, 1.21) | 0.88 (0.64, 1.20) | 0.84 (0.61, 1.16) |
| Frail | 247 | 29.1% | 85 | 34.4% | 1.22 (0.87, 1.71) | 1.14 (0.81, 1.62) | 1.12 (0.79, 1.58) |
| Shah | 806 | 100.0% | 241 | ||||
| Nonfrail | 278 | 34.5% | 70 | 25.2% | Ref. | Ref. | Ref. |
| Prefrail | 276 | 34.2% | 92 | 33.3% | 1.59 (1.16, 2.18) | 1.56 (1.14, 2.13) | 1.57 (1.14, 2.15) |
| Frail | 252 | 31.3% | 79 | 31.3% | 1.58 (1.14, 2.19) | 1.45 (1.04, 2.03) | 1.47 (1.05, 2.07) |
| Teo | 832 | 100.0% | 252 | ||||
| Nonfrail | 307 | 36.9% | 87 | 28.3% | Ref. | Ref. | Ref. |
| Prefrail | 405 | 48.7% | 123 | 30.4% | 1.11 (0.83, 1.49) | 1.11 (0.83, 1.48) | 1.11 (0.83, 1.48) |
| Frail | 120 | 14.4% | 42 | 35.0% | 1.58 (1.08, 2.30) | 1.47 (1.00, 2.15) | 1.47 (0.99, 2.18) |
| Tilburg | 516 | 100.0% | 153 | ||||
| Nonfrail | 213 | 41.3% | 60 | 28.2% | Ref. | Ref. | Ref. |
| Prefrail | 223 | 43.2% | 71 | 31.8% | 1.13 (0.79, 1.64) | 1.14 (0.79, 1.64) | 1.09 (0.75, 1.58) |
| Frail | 80 | 15.5% | 22 | 27.5% | 1.29 (0.77, 2.17) | 1.18 (0.69, 2.00) | 1.12 (0.65, 1.94) |
| Yamada | 816 | 100.0% | 244 | ||||
| Nonfrail | 218 | 26.7% | 62 | 28.4% | Ref. | Ref. | Ref. |
| Prefrail | 379 | 46.4% | 107 | 28.2% | 0.93 (0.67, 1.27) | 0.92 (0.67, 1.27) | 0.89 (0.64, 1.22) |
| Frail | 219 | 26.8% | 75 | 34.2% | 1.46 (1.03, 2.06) | 1.36 (0.95, 1.94) | 1.35 (0.95, 1.94) |
Note. Differences in n exist between indices as missing/invalid social frailty data differs by index. Values in bold indicate p < .05. HR = hazard ratio; Ref. = Reference group.
Minimally adjusted model adjusted for age and sex.
Partially adjusted model additionally adjusted for physical and psychological frailty.
Fully adjusted model additionally adjusted for smoking, alcohol, vision, hypertension, diabetes, hyperlipidemia, and traumatic brain injury.
Partially adjusted model
The increased risk of incident dementia for socially frail participants, as indicated by the Shah and Teo indices, persisted after adjusting for physical and psychological frailty (HR 1.45 [95% CI 1.04–2.03] and HR 1.47 [95% CI 1.00–2.15], respectively). The Yamada index no longer showed increased risk. For prefrail versus nonfrail participants, the increased risk with the Shah index remained (HR 1.56 [95% CI 1.14–2.13]).
Fully adjusted model
After full adjustment for confounders, only the Shah index showed an increased risk of incident dementia for socially frail participants (HR 1.47 [95% CI 1.05–2.07]). The increased risk for prefrail versus nonfrail participants, based on the Shah index, also persisted (HR 1.57 [95% CI 1.14–2.15]).
Sex differences
The only interaction found was between sex and social frailty using the Tilburg index. Stratified analysis showed an increased risk of incident dementia in men (HR 1.61 [95% CI 1.00–2.60]) but not women, in the partially adjusted model. No other associations were detected (Supplementary Table 9, see online supplementary material).
Sensitivity analysis
In the first sensitivity analysis, with increased cutoff scores, the association between social frailty and dementia was only observed using the Shah index. Effect sizes increased, but confidence intervals widened, leading to fewer associations (Table 4). In the second sensitivity analysis, including baseline cognition as a covariate, the association with the Shah index was no longer observed for socially frail participants (HR 1.39 [95% CI 0.98–1.95]). However, the relationship for prefrail versus nonfrail participants persisted (HR 1.52 [95% CI 1.11–2.09]) (Supplementary Table 10, see online supplementary material).
Table 4.
Sensitivity analysis—HR for associations between social frailty status and incident dementia using higher cutoff scores.
| Social frailty status | n | Prevalence | Cases | Incidence | Minimally adjusteda | Partially adjustedb | Fully adjustedc |
|---|---|---|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||
| Makizako | 848 | 100.0% | 259 | ||||
| Nonfrail | 210 | 24.6% | 64 | 30.5% | Ref. | Ref. | Ref. |
| Prefrail | 584 | 68.9% | 176 | 30.1% | 0.98 (0.73, 1.31) | 0.95 (0.71, 1.28) | 0.92 (0.69, 1.24) |
| Frail | 54 | 6.4% | 19 | 35.2% | 1.25 (0.74, 2.10) | 1.11 (0.66, 1.89) | 1.06 (0.62, 1.82) |
| Shah | 801 | 100.0% | 239 | ||||
| Nonfrail | 278 | 34.7% | 70 | 25.2% | Ref. | Ref. | Ref. |
| Prefrail | 476 | 59.4% | 146 | 30.7% | 1.50 (1.13, 2.00) | 1.45 (1.08, 1.94) | 1.47 (1.09, 1.97) |
| Frail | 47 | 5.9% | 23 | 48.9% | 2.36 (1.46, 3.81) | 2.14 (1.31, 3.49) | 2.20 (1.34, 3.63) |
| Teo | 829 | 100.0% | 250 | ||||
| Nonfrail | 307 | 37.0% | 87 | 28.3% | Ref. | Ref. | Ref. |
| Prefrail | 509 | 61.4% | 158 | 31.0% | 1.18 (0.89, 1.55) | 1.18 (0.89, 1.55) | 1.15 (0.87, 1.52) |
| Frail | 13 | 1.6% | 5 | 38.5% | 2.11 (0.85, 5.24) | 2.11 (0.85, 5.24) | 1.78 (0.71, 4.48) |
| Tilburg | 506 | 100.0% | 152 | ||||
| Nonfrail | 213 | 42.1% | 60 | 28.2% | Ref. | Ref. | Ref. |
| Prefrail | 282 | 55.7% | 88 | 31.2% | 1.16 (0.81, 1.65) | 1.14 (0.80, 1.63) | 1.10 (0.77, 1.57) |
| Frail | 11 | 2.2% | 4 | 36.4% | 2.14 (0.77, 5.93) | 2.00 (0.72, 5.60) | 1.33 (0.39, 4.48) |
| Yamada | 809 | 100.0% | 240 | ||||
| Nonfrail | 218 | 26.9% | 62 | 28.4% | Ref. | Ref. | Ref. |
| Prefrail | 555 | 68.6% | 164 | 29.5% | 1.04 (0.77, 1.40) | 1.02 (0.75, 1.38) | 1.00 (0.74, 1.35) |
| Frail | 36 | 4.4% | 14 | 38.9% | 1.45 (0.81, 2.60) | 1.16 (0.63, 2.13) | 1.08 (0.58, 2.01) |
Note. Differences in n exist between indices as missing/invalid social frailty data differs by index. Values in bold indicate p < .05. HR = hazard ratio; Ref. = Reference group.
Minimally adjusted model adjusted for age and sex.
Partially adjusted model additionally adjusted for physical and psychological frailty.
Fully adjusted model additionally adjusted for smoking, alcohol, vision, hypertension, diabetes, hyperlipidemia, and traumatic brain injury.
Individual components of social frailty indices
In the minimally adjusted model, increased risk of incident dementia was predicted by lower financial satisfaction, lower frequency of social activities and volunteering, social contact less than once per month, and lower satisfaction with family. In the fully adjusted model, only the relationship between low financial satisfaction and dementia remained (Table 5).
Table 5.
Hazard ratio for associations between individual components of social frailty indices and incident dementia.
| Component | Minimally adjusteda
|
Partially adjustedb
|
Fully adjustedc
|
|||
|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | |
| Contact <1/month | 2.10 | (1.03, 4.26) | 2.08 | (1.02, 4.24) | 1.99 | (0.97, 4.07) |
| Low financial satisfaction | 2.08 | (1.33, 3.26) | 1.92 | (1.22, 3.02) | 1.95 | (1.22, 3.12) |
| Low family satisfaction | 1.55 | (1.02, 2.37) | 1.43 | (0.93, 2.20) | 1.38 | (0.88, 2.15) |
| Low social activities and volunteering | 1.44 | (1.03, 2.02) | 1.30 | (0.92, 1.84) | 1.33 | (0.94, 1.89) |
| Low family and friends’ satisfaction | 1.49 | (0.99, 2.25) | 1.39 | (0.92, 2.11) | 1.34 | (0.87, 2.06) |
| Not paid work | 1.39 | (0.96, 2.02) | 1.35 | (0.93, 1.97) | 1.38 | (0.94, 2.00) |
| Low education | 1.38 | (0.73, 2.60) | 1.40 | (0.74, 2.64) | 1.39 | (0.73, 2.64) |
| Small network size | 1.28 | (0.73, 2.24) | 1.23 | (0.70, 2.15) | 1.10 | (0.61, 1.96) |
| Loneliness | 1.27 | (0.84, 1.90) | 1.24 | (0.82, 1.88) | 1.16 | (0.76, 1.77) |
| Contact <11 times/month | 1.21 | (0.93, 1.56) | 1.13 | (0.87, 1.47) | 1.13 | (0.87, 1.47) |
| Dropped activities | 1.16 | (0.86, 1.55) | 1.10 | (0.81, 1.49) | 1.13 | (0.84, 1.53) |
| Absence of confidant | 1.16 | (0.66, 2.04) | 1.16 | (0.66, 2.05) | 1.15 | (0.65, 2.05) |
| Low self-worth | 1.16 | (0.60, 2.26) | 1.12 | (0.58, 2.20) | 1.06 | (0.54, 2.09) |
| Infrequent volunteering | 1.13 | (0.88, 1.46) | 1.09 | (0.85, 1.41) | 1.13 | (0.87, 1.47) |
| Live alone | 0.99 | (0.76, 1.30) | 0.99 | (0.76, 1.30) | 0.97 | (0.74, 1.28) |
| Low neighborhood satisfaction | 0.71 | (0.23, 2.22) | 0.63 | (0.20, 1.98) | 0.62 | (0.20, 1.96) |
Note. Values in bold indicate p < .05. HR = hazard ratio.
Minimally adjusted model adjusted for age and sex.
Partially adjusted model additionally adjusted for physical and psychological frailty.
Fully adjusted model additionally adjusted for smoking, alcohol, vision, hypertension, diabetes, hyperlipidemia, and traumatic brain injury.
Discussion
This study of 851 Australian adults aged 70–90 years found that three of the five social frailty indices predicted increased dementia risk over a 12-year follow-up. Key components associated with increased dementia risk included low financial and family satisfaction, infrequent social contact, and limited participation in social activities and volunteering. After adjusting for physical/psychological frailty and other covariates, social frailty independently predicted incident dementia using only one of the five indices; with the other two indices, the associations were no longer observed. Low financial satisfaction was the only component that independently predicted incident dementia. Little evidence of sex differences was observed.
Our finding that low financial satisfaction, low family satisfaction, infrequent social contacts, and limited participation in social activities are associated with an increased risk of dementia is consistent with prior research (Samtani et al., 2022; Wang et al., 2023; Zunzunegui et al., 2003). A recent meta-analysis reported an increased risk of dementia (RR 1.40 [95% CI 1.12–1.74]) for individuals from low socioeconomic status (SES) groups compared with high SES groups (Wang et al., 2023). Similarly, infrequent social contact and low participation in social activities have been linked to dementia risk in a cohort of Spanish older adults (Zunzunegui et al., 2003). These findings are further supported by longitudinal evidence from a multi-cohort study involving 40,000 older adults, showing that stronger social connections, such as frequent interactions with family and friends, participation in community groups, absence of loneliness, and being married, were all associated with better cognitive outcomes (Samtani et al., 2022).
Although all indices suggested a potential increase in dementia risk for socially frail participants, results were not significant using the Makizako and Tilburg indices. Notably, neither index incorporates financial satisfaction. In addition, the Makizako index does not include relationship satisfaction, and the Tilburg index does not contain social activities. These factors, predictive of dementia in the MAS cohort, may explain the absence of a significant association in our study using these indices. In contrast, Tsutsumimoto et al.’s (2019) study of 3,720 Japanese older adults found that social frailty classified by the Makizako index was associated with an increased risk of incident Alzheimer’s disease (AD) over 4.4 years. Differences in study design (shorter follow-up and Japanese participants were approximately 6 years younger) and the wording of questions may account for differing results. In addition, Tsutsumimoto et al.’s focus on AD risk likely contributed to the lower dementia incidence observed (5.2% vs. 30.6% in the current study). In a younger population with a lower baseline dementia risk, social frailty may have had a stronger impact, and the lower overall incidence may have magnified the effect among those who developed dementia.
The Shah index was the best performing in this context as it was the only index that predicted dementia risk in both frail and prefrail older adults, contained all four components which predicted dementia, and maintained associations after adjusting for physical and psychological frailty and other covariates, even when the cutoff score was increased. The only component in the Shah index not associated with dementia risk was low satisfaction with the neighborhood. However, participants in the MAS study were from two neighboring high SES areas, so differences between neighborhoods were expected to be low; hence, neighborhood satisfaction may be more important in other populations. Future research validating the Shah index in diverse populations and in different settings is warranted. We recommend testing the cut-off scores of 3 and 4 in these varying contexts. In our study, the lower cutoff score resulted in a social frailty prevalence rate of 31.3%, whereas the higher cutoff yielded 5.9%. For comparison, a recent meta-analysis reported a prevalence of social frailty of 18.8% among community-dwelling adults aged 60 years and older (Zhang et al., 2023). These differences highlight the need for further refinement of cutoff scores to ensure accurate assessment of social frailty across different groups. In addition, the acceptability of the index to consumer groups and health professionals needs to be assessed. The Shah index could eventually be used in primary care settings or for self-screening. Social prescribing by healthcare professionals to connect socially frail individuals with community resources, fostering social connections and reducing isolation, should be considered (Bickerdike et al., 2017). Dissemination of the Shah index could raise awareness of social frailty among clinicians, consumers, researchers, and the general public. Communication messages could highlight risk factors and prevention strategies for social frailty, while encouraging support for individuals experiencing social frailty.
Several mechanisms may explain how social frailty influences dementia risk. Although the impact of financial resources on cognition is unclear, theories suggest it may contribute to greater cognitive reserve through higher education, healthier lifestyle choices, better access to cognitively stimulating environments, and reduced social isolation (Wang et al., 2023). In addition, low SES in early life has been associated with lower gray matter volume and surface area of the prefrontal cortex (Rakesh & Whittle, 2021). Stronger social connections may provide a more cognitively enriching environment and buffer against stress (Perry et al., 2022). Social isolation has also been associated with reduced gray matter volume, which has been shown to partially mediate the relationship between social isolation and cognition (Shen et al., 2022). Future research to investigate these mechanisms is warranted.
This study’s results should be viewed considering its limitations. The population examined was generally healthy, well-educated, and Caucasian. Included participants had more social connections, higher physical activity, and better physical performance than those excluded. Therefore, results may not be generalizable to less healthy, lower socioeconomic populations with fewer social connections and more cultural diversity. Another limitation is that we used existing longitudinal data, restricting us to the available questions. There may be other social questions that could have improved the predictive ability of the social frailty indices, and there are differences between our questions and those used in the original indices. For example, we asked about face-to-face contact frequency, while the original indices asked about talking frequency, which could include phone calls. Although we found some evidence that the relationship between social frailty and dementia is causal and excluded participants with baseline dementia, reverse causation remains possible. Associations may be due to participants with lower cognitive skills, but not meeting the definition of dementia, who may have reduced their social interactions and capacity to maintain their financial status. Sensitivity analysis, including baseline cognition as a covariate, revealed that the association between social frailty and dementia was no longer observed in socially frail participants; however, the relationship persisted for prefrail participants. There are arguments for and against adjusting for baseline cognition, as it may reflect the impact of social frailty before baseline, and including it as a covariate partially adjusts for the impact of social frailty.
In summary, this study found that social frailty indices are an effective way to help identify older adults at increased risk of dementia. In particular, the Shah index identified both frail and prefrail participants as having an approximately 50% higher risk of incident dementia versus nonfrail participants. Our work suggests social frailty may increase the risk of dementia independent of other risk factors. Financial satisfaction was a key predictor of incident dementia, which we recommend be incorporated in future indices. We suggest validating these findings in diverse populations with the view to using social frailty indices in primary healthcare to identify older adults at increased risk of dementia. Future research should examine the predictive ability of social frailty indexes for other health outcomes, such as mortality and poor quality of life.
Supplementary Material
Acknowledgments
We thank the participants and their informants for their time and generosity in contributing to this research. We also acknowledge the MAS research team: https://cheba.unsw.edu.au/research-projects/sydney-memory-and-ageing-study.
Contributor Information
Annabel P Matison, Centre for Healthy Brain Ageing (CHeBA), School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Suraj Samtani, Centre for Healthy Brain Ageing (CHeBA), School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Henry Brodaty, Centre for Healthy Brain Ageing (CHeBA), School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Perminder S Sachdev, Centre for Healthy Brain Ageing (CHeBA), School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Simone Reppermund, Centre for Healthy Brain Ageing (CHeBA), School of Clinical Medicine, University of New South Wales, Sydney, New South Wales, Australia.
Supplementary material
Supplementary material is available at The Journals of Gerontology Series B: Psychological and Social Sciences online.
Data availability
Individual-level data from MAS cannot be made publicly available due to ethical and legal restrictions. However, access to this data may be provided subject to approval of the MAS Governance Committee. Further information about data accessibility and study governance may be obtained by contacting the study coordinator, details of whom can be found on our website: https://cheba.unsw.edu.au. The study reported in the manuscript was not preregistered.
Funding
This work was supported by three National Health & Medical Research Council (NHMRC) Program Grants (ID350833, ID568969, and APP1093083); NHMRC Investigator Grant ID1196150 (to P.S.S.) and National Institutes of Health (NIH) Grant 2R01AG057531 (to P.S.S. and A.M.). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflict of interest
S.S. declares funding from the Dementia Australia Research Foundation, although not for the current manuscript. H.B. is or has been an advisory board member or consultant to Biogen, Eisai, Eli Lilly, Medicines Australia, Roche, and Skin2Neuron. PSS has been on the Expert Advisory Panels for Biogen Australia and Roche Australia in 2021–2022. A.P.M. and S.R. have no conflicts of interest to declare.
References
- American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders (4th ed.). American Psychiatric Association. [Google Scholar]
- Bickerdike L., Booth A., Wilson P. M., Farley K., Wright K. (2017). Social prescribing: Less rhetoric and more reality. A systematic review of the evidence. BMJ Open, 7, e013384. 10.1136/bmjopen-2016-013384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bunt S., Steverink N., Olthof J., van der Schans C. P., Hobbelen J. S. M. (2017). Social frailty in older adults: A scoping review. European Journal of Ageing, 14, 323–334. 10.1007/s10433-017-0414-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cohen C. I., Benyaminov R., Rahman M., Ngu D., Reinhardt M. (2023). Frailty: A multidimensional biopsychosocial syndrome. Medical Clinics of North America, 107, 183–197. 10.1016/j.mcna.2022.04.006 [DOI] [PubMed] [Google Scholar]
- Folstein M. F., Folstein S. E., McHugh P. R. (1975). “Mini-mental state.” A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189–198. 10.1016/0022-3956(75)90026-6 [DOI] [PubMed] [Google Scholar]
- Fried L. P., Tangen C. M., Walston J., Newman A. B., Hirsch C., Gottdiener J., Seeman T., Tracy R., Kop W. J., Burke G., McBurnie M. A.; Cardiovascular Health Study Collaborative Research, G. (2001). Frailty in older adults: Evidence for a phenotype. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 56, M146–M156. 10.1093/gerona/56.3.m146 [DOI] [PubMed] [Google Scholar]
- Gobbens R. J., van Assen M. A., Luijkx K. G., Wijnen-Sponselee M. T., Schols J. M. (2010). The Tilburg Frailty Indicator: Psychometric properties. Journal of the American Medical Directors Association, 11, 344–355. 10.1016/j.jamda.2009.11.003 [DOI] [PubMed] [Google Scholar]
- Gong J., Harris K., Lipnicki D. M., Castro-Costa E., Lima-Costa M. F., Diniz B. S., Xiao S., Lipton R. B., Katz M. J., Wang C., Preux P. M., Guerchet M., Gbessemehlan A., Ritchie K., Ancelin M. L., Skoog I., Najar J., Sterner T. R., Scarmeas N., Woodward M.(2023). Sex differences in dementia risk and risk factors: Individual-participant data analysis using 21 cohorts across six continents from the COSMIC consortium. Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association, 19, 3365–3378. 10.1002/alz.12962 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Khalil A. H., Gobbens R. J. J. (2023). What if the clinical and older adults’ perspectives about frailty converge? A call for a mixed conceptual model of frailty: A traditional literature review. Healthcare, 11, 3174. 10.3390/healthcare11243174 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee Y., Jean Yeung W. J. (2019). Gender matters: Productive social engagement and the subsequent cognitive changes among older adults. Social Science and Medicine, 229, 87–95. 10.1016/j.socscimed.2018.08.024 [DOI] [PubMed] [Google Scholar]
- Li X., Gao L., Qiu Y., Zhong T., Zheng L., Liu W., Li G., Chen L. (2023). Social frailty as a predictor of adverse outcomes among older adults: A systematic review and meta-analysis. Aging Clinical and Experimental Research, 35, 1417–1428. 10.1007/s40520-023-02421-y [DOI] [PubMed] [Google Scholar]
- Livingston, G.,, Huntley, J.,, Liu, K. Y., Costafreda, S. G., Selbæk, G., Alladi, S., Ames, D., Banerjee, S., Burns, A., Brayne, C., Fox, N. C., Ferri, C. P., Gitlin, L. N., Howard, R., Kales, H. C., Kivimäki, M., Larson, E. B.,, Nakasujja, N., Rockwood, K.Mukadam, N. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet (London, England). 404, 572–628. 10.1016/S0140-6736(24)01296-0. [DOI] [PubMed] [Google Scholar]
- Mah J., Rockwood K., Stevens S., Keefe J., Andrew M. K. (2022). Do interventions reducing social vulnerability improve health in community dwelling older adults? A systematic review. Clinical Interventions in Aging, 17, 447–465. 10.2147/cia.S349836 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Makizako H., Shimada H., Tsutsumimoto K., Lee S., Doi T., Nakakubo S., Hotta R., Suzuki T. (2015). Social frailty in community-dwelling older adults as a risk factor for disability. Journal of the American Medical Directors Association, 16, 1003.e1007–1011. 10.1016/j.jamda.2015.08.023 [DOI] [PubMed] [Google Scholar]
- Perry B. L., McConnell W. R., Coleman M. E., Roth A. R., Peng S., Apostolova L. G. (2022). Why the cognitive “fountain of youth” may be upstream: Pathways to dementia risk and resilience through social connectedness. The Journal of the Alzheimer’s Association, 18, 934–941. 10.1002/alz.12443 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Peters L. L., Boter H., Buskens E., Slaets J. P. (2012). Measurement properties of the Groningen Frailty Indicator in home-dwelling and institutionalized elderly people. Journal of the American Medical Directors Association, 13, 546–551. 10.1016/j.jamda.2012.04.007 [DOI] [PubMed] [Google Scholar]
- Rakesh D., Whittle S. (2021). Socioeconomic status and the developing brain—A systematic review of neuroimaging findings in youth. Neuroscience and Biobehavioral Reviews, 130, 379–407. 10.1016/j.neubiorev.2021.08.027 [DOI] [PubMed] [Google Scholar]
- Rubin D. B. (1987). Multiple Imputation for Nonresponse in Surveys (1st ed.). John Wiley & Sons. [Google Scholar]
- Sachdev P. S., Brodaty H., Reppermund S., Kochan N. A., Trollor J. N., Draper B., Slavin M. J., Crawford J., Kang K., Broe G. A., Mather K. A., Lux O., Memory, & Ageing Study, T. (2010). The Sydney Memory and Ageing Study (MAS): Methodology and baseline medical and neuropsychiatric characteristics of an elderly epidemiological non-demented cohort of Australians aged 70-90 years. International Psychogeriatrics, 22, 1248–1264. 10.1017/S1041610210001067 [DOI] [PubMed] [Google Scholar]
- Samtani S., Mahalingam G., Lam B. C. P., Lipnicki D. M., Lima-Costa M. F., Blay S. L., Castro-Costa E., Shifu X., Guerchet M., Preux P. M., Gbessemehlan A., Skoog I., Najar J., Rydberg Sterner T., Scarmeas N., Kim K. W., Riedel-Heller S., Rohr S., Pabst A., ... & Brodaty, H. (2022). Associations between social connections and cognition: A global collaborative individual participant data meta-analysis. The Lancet Healthy Longevity, 3, e740–e753. 10.1016/S2666-7568(22)00199-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shah S. J., Oreper S., Jeon S. Y., Boscardin W. J., Fang M. C., Covinsky K. E. (2023). Social Frailty Index: Development and validation of an index of social attributes predictive of mortality in older adults. Proceedings of the National Academy of Sciences of the United States of America, 120, e2209414120. 10.1073/pnas.2209414120 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shen C., Rolls E. T., Cheng W., Kang J., Dong G., Xie C., Zhao X. M., Sahakian B. J., Feng J. (2022). Associations of social isolation and loneliness with later dementia. Neurology, 99, e164–e175. 10.1212/wnl.0000000000200583 [DOI] [PubMed] [Google Scholar]
- Teo N., Gao Q., Nyunt M. S. Z., Wee S. L., Ng T. P. (2017). Social frailty and functional disability: Findings from the Singapore longitudinal ageing studies. Journal of the American Medical Directors Association, 18, 637.e613–637.e619. 10.1016/j.jamda.2017.04.015 [DOI] [PubMed] [Google Scholar]
- Therneau T. (2023). A Package for Survival Analysis in R. R package version 3.5-5. https://CRAN.R-project.org/package=survival
- Therneau T. M., Grambsch P. M. (2000). Modeling survival data: Extending the Cox model. Springer. [Google Scholar]
- Tsutsumimoto K., Doi T., Nakakubo S., Kim M., Kurita S., Ishii H., Shimada H., Kawachi I. (2019). Impact of social frailty on Alzheimer’s disease onset: A 53-month longitudinal cohort study. Journal of Alzheimer’s Disease, 70, 587–595. 10.3233/JAD-181178 [DOI] [PubMed] [Google Scholar]
- Van Buuren S., Groothuis-Oudshoorn K. (2011). Multivariate imputation by chained equations in R. Journal of Statistical Software, 45, 1–67. 10.18637/jss.v045.i03 [DOI] [Google Scholar]
- Wang A. Y., Hu H. Y., Ou Y. N., Wang Z. T., Ma Y. H., Tan L., Yu J. T. (2023). Socioeconomic status and risks of cognitive impairment and dementia: A systematic review and meta-analysis of 39 prospective studies. Journal of Prevention of Alzheimer’s Disease, 10, 83–94. 10.14283/jpad.2022.81 [DOI] [PubMed] [Google Scholar]
- Yamada M., Arai H. (2018). Social frailty predicts incident disability and mortality among community-dwelling Japanese older adults. Journal of the American Medical Directors Association, 19, 1099–1103. 10.1016/j.jamda.2018.09.013 [DOI] [PubMed] [Google Scholar]
- Yesavage J. A., Brink T. L., Rose T. L., Lum O., Huang V., Adey M., Leirer V. O. (1982). Development and validation of a geriatric depression screening scale: A preliminary report. Journal of Psychiatric Research, 17, 37–49. 10.1016/0022-3956(82)90033-4 [DOI] [PubMed] [Google Scholar]
- Zhang X. M., Cao S., Gao M., Xiao S., Xie X., Wu X. (2023). The prevalence of social frailty among older adults: A systematic review and meta-analysis. Journal of the American Medical Directors Association, 24, 29–37.e29. 10.1016/j.jamda.2022.10.007 [DOI] [PubMed] [Google Scholar]
- Zunzunegui M. V., Alvarado B. E., Del Ser T., Otero A. (2003). Social networks, social integration, and social engagement determine cognitive decline in community-dwelling Spanish older adults. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 58, S93–S100. 10.1093/geronb/58.2.s93 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Individual-level data from MAS cannot be made publicly available due to ethical and legal restrictions. However, access to this data may be provided subject to approval of the MAS Governance Committee. Further information about data accessibility and study governance may be obtained by contacting the study coordinator, details of whom can be found on our website: https://cheba.unsw.edu.au. The study reported in the manuscript was not preregistered.