Physical frailty predicts the development of social frailty: a prospective cohort study

This study uses a prospective cohort analysis design. Baseline information was gathered in the period September 2015 to December 2017 from the study on Frail Elderly in the Sasayama-Tamba Area (FESTA). Tamba-Sasayama City, with a population of 41,490 as of 2015, is situated in the suburban area of the Hyogo prefecture. The average age of the city’s population is higher than the Japanese average (31.4% aged 65 or above). We recruited community-dwelling older adults to participate by using an community advertisement, placing posters at Sasayama Medical Center, and oral announcements by medical staff in the city. The two-year follow-up assessment was conducted between September 2017 and December 2019 to assess the incidence of the decline of the social frailty component.

The present study targeted participants who at least engaged in social activity or had contact with neighbors, which are potentially related to physical frailty, among subdomains of social frailty. The inclusion criteria were: (1) aged 65 years or older, (2) able to walk independently with/without a cane, and (3) not socially frail/pre-frail at baseline. The exclusion criteria were: (1) cognitive impairment, which was determined by a Mini-Mental State Examination (MMSE) score less than 21 [14], and (2) participants containing missing data.

During the initial phase, 625 older adults participated in the baseline assessment. Among them, 376 met the inclusion criteria (out of 625, 201 were pre-socially frail, and 48 were socially frail). One person was excluded due to cognitive impairment. Out of 375 people, 352 participated in the secondary assessment (23 dropped out), and 10 people had missing data. Accordingly, the study comprised the remaining 342 participants.

We used a modified social frailty index [5] founded on Bunt’s social frailty concept, which measures general and social resources, social behaviors, and the satisfaction of basic social requirements [15]. The screening index was developed to briefly assess social frailty status, which can predict future incidents of activity limitation and mortality in community-dwelling older adults [5]. The question regarding general resources (financial difficulties) is: “Do you have a financial problem in your daily life?” Financial difficulty was defined as a “yes” answer. The question regarding social resources (living alone) is: “How many people do you live with?” Inadequate social resources were reflected in the answer “alone.” The question pertaining to social behavior (lack of social activity) is: “Do you participate in any community activities or volunteer activities?” Poor social behavior was defined as an answer “none.” The question assessing the satisfaction of basic social requirements (lack of contact with neighbors) is: “Do you sometimes visit your friends?” Deficiency in basic social needs was identified as a “no” answer. A score of 2 or more was defined as social frailty, 1 as pre-social frailty, and 0 as social robustness.

We assessed physical frailty status according to the Fried phenotype [1, 16]: (1) slow gait speed, (2) weakness, (3) exhaustion, (4) low activity, and (5) weight loss. Participants who did not show any of these five symptoms were considered non-frail, while those with one or two symptoms were defined as pre-frail. For the assessment of gait speed, we asked participants to cover a 12 m walkway at their usual speed. Then, the time for 10 m in the way was assessed [17]. Slowness was determined by a cut-off point less than 1.0 m/s [1]. We measured maximum grip strength by using a grip strength tester (GRIP-A; Takei Ltd., Niigata, Japan). Strength weakness was identified according to established cut-off (< 26 kg for men, < 18 kg for women) [18]. We assessed exhaustion with the following question from the Kihon Check List [19]: “In the last two weeks, have you felt tired without a reason?” Weight loss was assessed with the following question: “Have you lost 2 kg or more in the past six months?” [19]. PA was assessed with a wrist wearable accelerometer (Actiband, TDK Co., Tokyo, Japan) for 14 days. The epoch duration for recording PA was 5 min. The reliability and validity of this accelerometer have been confirmed previously [20]. Data from participants with complete measurements recorded for at least 3 days were included in the analysis. Participants who did not record at least 600 min of PA were excluded [21]. We defined lower than 1 standard deviation away from mean of moderate to vigorous PA (MVPA, ≥3 METs) [22] in the participants as low PA.

Each subject’s age, sex, comorbidity, and education were self-reported. We used the Geriatric Depression Scale (GDS) [23] to assess depressive symptoms. We also assessed the Instrumental Activity of Daily Living (IADL) using the Tokyo Metropolitan Institute of Gerontology Index of Competence (TMIG-IC) [24] and participants’ history of falls in the previous year.

The social frailty score (four domains) was used as the primary outcome. We re-assessed it during the two-year follow-up, and categorized participants into a socially maintained or socially declined group according to their change in social score (four domains) to clarify the temporal relationship between physical frailty and social frailty. As the secondary analysis, we focused on social behavior (social activity) and fulfillment of basic social needs (contact with neighbors) in the social frailty index, which would be affected by the physical components because the two activities require physical movement. We accordingly excluded the remaining two social variables in the secondary analysis, general resources (economic hardship), and social resources (living alone), because they may be difficult to control using the physical aspect in older adults.

The participant’s characteristics assigned to the physically robust and pre-frailty or frailty (with any frailty subdomain) groups at the baseline were compared by using a Student’s t-test, Mann–Whitney U, or chi-square tests according to the type of variables. We also assessed the baseline differences between the socially maintained and socially declined groups (four domains) during the two-year follow-up. A modified Poisson regression model [25] was used to estimate the risk ratio (RR) and 95% confidence intervals (CIs) of physical frailty for the development of social frailty (four domains) in the crude and adjusted model as a primary analysis. First, the physical frailty condition (robust or any-frailty) was included as an independent variable. Second, five subdomains of physical frailty were entered using the forced entry method. Of those, we added walking speed, handgrip strength, and PA as continuous variables. The adjusted model added age, sex, MMSE, GDS, multimorbidity (two or more chronic illnesses), IADL score, and fall experience as covariates. As a secondary analysis, we assessed the incidence risk of development of social frailty focusing on social activity and contact with neighbors in the social frailty index, domains which possibly have a direct relationship with physical components. Data were analyzed using IBM SPSS ver. 24 (IBM Japan Ltd., Tokyo, Japan). Statistical significance was set at p < 0.05.