Social Frailty in Older Adults: Proposal and Application of an Original Measurement Index

Study Design and Participants

This study employed a quantitative, observational, and analytical cross-sectional design to explore the prevalence of social frailty in older adults. The research process followed a structured sequence, beginning with the identification of a knowledge gap through a literature review and the formulation of research objectives. Data analysis was grounded in rigorous statistical methods to ensure the validity and reliability of findings, thereby enabling a comprehensive understanding of the phenomenon under investigation. The analysis was based on secondary data from the 2016 Survey on Health, Well-being and Aging (SABE Colombia), a nationally representative dataset that includes a wide range of indicators on the health, functionality, and social conditions of older adults.4) The current study focused on participants aged 60 years or older residing in the departments of Antioquia, Caldas, Risaralda, and Quindío. Proxy responses were excluded to preserve the accuracy of data on perceptions and subjective experiences.

The methodological document of the National Survey on Health, Well-being, and Aging (SABE Colombia) can be consulted at the website here.

Measures

To address the study objectives, variables were operationalized across three main dimensions: demographic, health-related, and social. Demographic indicators included age, sex, marital status (dichotomized as partnered/unpartnered), area and department of residence, ethnicity (minority vs. mestizo), access to technology, access to recreational and transportation services, history of forced displacement, and current employment status. Health-related variables included self-rated health, self-reported vision quality, perceived quality of healthcare, limitations in basic and instrumental activities of daily living (ADLs and IADLs), and presence of multimorbidity (two or more chronic diseases).

The social dimension comprised variables directly contributing to the construction of the SFI the study’s main outcome. These included experiences of discrimination and abuse, social participation, feelings of loneliness, living alone, engagement in volunteer work, and perceived emotional support. All variables were dichotomized to facilitate statistical modeling, with categories defined to capture the presence or absence of risk factors relevant to social vulnerability. The nine variables selected for the construction of the SFI were: self-perception of health, perceived quality of healthcare, dependence in basic ADLs, dependence in IADLs, low social participation, low volunteer participation, perceived emotional support, feelings of loneliness, and living alone.

Social Frailty Index (SFI)

The theoretical foundation of this study draws upon the social production function theory proposed by Ormel et al.10) which posits that an individual’s overall well-being depends on both physical and social dimensions. Based on this framework, social frailty is conceptualized as the loss of essential social resources such as support networks, community participation, and sense of belonging. Crucially, the SFI development aimed to model both direct social deficits (affect, participation) and the structural and functional restrictions that limit an individual's capacity to achieve instrumental goals (affect, comfort, stimulation) within the social production function framework.

Therefore, functional dependency (ADL/IADL) was included as a measure of restriction on the instrumental goal of comfort and stimulation, reflecting the essential intrinsic capacity required for social functioning. Likewise, subjective perceptions of health and quality of care (component 4) were included as indicators of resource environment and contextual constraints on well-being production. In addition, this condition involves limited compensatory mechanisms in the face of such losses, often due to multimorbidity, functional dependence, or sensory impairments.11)

To construct the emergent variable of social frailty, principal component analysis (PCA) was employed. PCA is an unsupervised multivariate exploratory technique that enables the extraction of latent information by reducing the dimensionality of the dataset. This method transforms the original variable coordinates into a new Cartesian system, maximizing the explained variance to facilitate more efficient data interpretation.12) To assess the statistical adequacy of the PCA, the Kaiser-Meyer-Olkin (KMO) measure and Bartlett’s test of sphericity were applied. The KMO index evaluates the proportion of shared variance among the variables, while Bartlett’s test determines whether the correlation matrix significantly differs from an identity matrix, thereby supporting the appropriateness of applying PCA.

The final dataset used for calculating the SFI included a total of 3,506 older adults. To ensure the validity of data related to perceptions and subjective experiences, proxy responses were excluded from the analysis.

Statistical Analyses

All statistical analyses were performed using Jamovi and Google Colab (Python). Nominal variables were summarized using absolute and relative frequencies, while continuous variables were reported as means with standard deviations or as medians with interquartile ranges, depending on their distribution. Associations between normal variables were evaluated using the chi-squared statistical test.

The project was approved by the Research and Innovation Committee of the Faculty of Medicine at Universidad CES (Code: Acta333Proy011).

Sociodemographic and Health Characteristics of the Study Population

The median age of participants was 68 years (interquartile range 63–74). Women predominated, representing approximately 60% of the sample, while 52% of participants reported having a partner. Most individuals resided in the department of Antioquia (43.5%) and in urban areas (83.3%). Socioeconomically, 70% of the participants belonged to the lowest strata (SES 1 and 2), and only 4.4% self-identified as belonging to an ethnic minority. Regarding health characteristics, 15.9% of participants had some level of dependency in basic ADLs, and 33.3% were dependent in IADLs. Multimorbidity, defined as the presence of two or more chronic conditions was present in 43.5% of participants. Around 10% of respondents rated their health as fair or poor, and nearly one-third reported perceiving the quality of healthcare received as inadequate. Additionally, 21.1% experienced visual impairment. For detailed sociodemographic and health characteristics, see Table 1.

Table 1.

Distribution of sociodemographic and health-related variables (n=3,506)

Social Characteristics of the Study Population

Based on the analysis of social variables (Table 2), a sex-stratified approach was used. The analysis included aspects such as group participation, volunteering, experiences of mistreatment and discrimination, perceived loneliness, and access to communication and transportation resources. Participation in social groups was higher among female (46.3%) compared to male. Volunteering rates were similar between sexes, with 12.7% of female and 11.1% of male reporting engagement. Regarding mistreatment, 14.2% of female and 12.9% of male reported experiencing some form of abuse. Loneliness affected 5.1% of the total sample. Notably, perceived discrimination was more frequent among male. Living alone was reported by 9.5% of female and 13% of male.

Table 2.

Distribution of social variables among older adults, by sex

Construction of the Social Frailty Index (SFI)

The construction of the SFI involved several sequential steps. Initially, variables were selected based on theoretical and empirical support from the literature. Normal variables were then subjected to optimal quantification procedures. Following this, statistical assumptions were verified to determine the appropriate number of principal components. Variables with factor loadings greater than 0.4 were retained. The final number of components was established through parallel analysis and explained variance evaluation.

The PCA conducted in this study yielded four components, which together explained 54.2% of the total variance across the selected variables. This multivariate technique enabled the identification of latent dimensions underlying social frailty in older adults by reducing the complexity of the data into interpretable factors (Table 3).

Table 3.

Principal component analysis

To confirm the suitability of the dataset for PCA, Bartlett’s test of sphericity was performed and found to be significant (χ²=929, df=36, p<0.001), indicating that correlations among variables were sufficient for factor analysis. The KMO measure of sampling adequacy yielded an overall value of 0.503, suggesting an acceptable level for PCA. Individual KMO values ranged from 0.499 to 0.571, with the highest values observed for self-rated health (0.571) and quality of care (0.521), reflecting greater adequacy of these variables within the proposed factorial structure.

The final SFI was constructed using the continuous weighted factor scores derived from the four principal components. A “varimax” rotation was employed to maximize the interpretability of these four conceptually distinct components. A threshold based on the 75th percentile of the index score distribution (1.14 points) was used to classify individuals as socially frail. These components capture key dimensions of social frailty among older adults, including functional dependency, limited social participation, emotional isolation, and perceived health and care quality.

Prevalence of Social Frailty

The prevalence of social frailty, calculated using a latent variable derived from PCA, was 25.3% (95% CI, 23.8–26.7) in the study population. A statistically significant difference was observed between male and female (p=0.001); among male, 29.2% (n=400) were classified as socially frail, compared to 22.8% (n=486) of female. The prevalence was also significantly higher among individuals aged 75 years and older (32.4%, n=269) compared to those under 75 years (23.0%, n=617) (p<0.001). Additionally, a statistically significant difference in the prevalence of social frailty was observed across departments, with the highest prevalence reported in Antioquia (27.6%), followed by Quindío (25.3%), Caldas (24.0%), and Risaralda (21.6%).

Social Frailty Index (SFI)

The study was conducted in a region undergoing rapid population aging, which presents significant challenges related to infrastructure, access to healthcare services, and the availability of community support networks.

The SFI was developed using four principal components that together reflect the multidimensional nature of social frailty. These components include functional dependence, reduced social participation, emotional and social isolation, and negative self-perception of health and healthcare. Each dimension captures critical aspects of vulnerability in older adults, including limited support networks, loneliness, and poor access to quality care. By integrating both objective and subjective factors, the index offers a comprehensive understanding of social frailty.

The rationale for this multidimensionality is rooted in the social production function theory, which demands the inclusion of both direct social deficits (component 2 and 3) and the fundamental constraints on well-being production, specifically functional capacity and perceived resources (components 1 and 4). While functional variables (ADL/IADL) may also be considered outcomes of physical frailty, within the social production function framework, they are crucial inputs reflecting intrinsic capacity necessary for maintaining social roles and participation.

From a theoretical standpoint, social frailty is a multidimensional construct characterized by antecedents such as adverse lifestyles, limited resources, poor healthcare access, and health conditions, all contributing to social disintegration. Core attributes include living alone, diminished social ties, and reduced support, heightening the risk of exclusion and isolation. Its consequences span physical decline (falls, disability, hospitalizations) and psychological effects (depression, anxiety, cognitive impairment).13) Additionally, based on social production function theory, social frailty results from an imbalance between the loss of essential social resources and reduced capacity to compensate, often due to multimorbidity, dependency, or sensory impairment. This framework highlights its complexity and health relevance.10)

Although a direct psychometric validation against international tools was not the primary objective of this study, the SFI demonstrates consistency with global social frailty frameworks, the conceptual alignment is confirmed by the latent dimensions that emerged empirically (functional dependence, social engagement, socioemotional isolation, and perception of health/care), which are coherent with the core domains assessed by instruments like the Tilburg Frailty Indicator (TFI) and the Makizako Scale.14)

A recent scoping review was conducted to identify tools and scales used to assess social frailty describe their psychometric properties, and analyze the contexts in which these instruments were applied. The review included 58 studies and identified nine distinct tools, most of which were derived from the conceptual frameworks proposed by Bunt et al.11) Social frailty concept11) offers a theoretical framework rather than a quantifiable index. Similar to the index developed in the present study, it emphasizes key domains such as social resources (e.g., living alone), social behaviors (e.g., participation), and social support. However, general resources such as socioeconomic status an element included in Bunt’s model were not incorporated in this study’s index due to the overall socioeconomic homogeneity of the sample, which was predominantly low-income.

Makizako et al.’s five-item scale assesses social activity, social roles, and relationships.15) While the current index does not include specific items such as frequency of outings or daily contact with others, it captures comparable dimensions by assessing variables such as living alone, social participation, volunteer activity, loneliness, and perceived lack of emotional support. The Social Frailty Scale (SFS-8) includes eight items focusing on specific aspects of interpersonal interaction such as visiting friends, having confidants, and eating with others.16) Although the present index does not assess these detailed behaviors, it does include core components such as living alone and perceived healthcare quality, which are also covered by SFS-8.

The HALFT and HALFE Scales each include five items covering domains like loneliness and living arrangements.17,18) The current index overlaps by incorporating perceived loneliness and living alone but does not consider financial hardship, which was deliberately excluded due to its widespread presence across the sample.

The Seven-Item SFI shares conceptual origins with Bunt et al.11) and includes variables such as living alone, social activity, and deprivation—components reflected in this study’s index as well. Similarly, the SFI for the Indian population aligns with the present tool in its emphasis on community-level participation and social engagement.19) Lastly, the TFI assesses multidimensional frailty through physical, psychological, and social domains. Its social component includes indicators such as living alone, low support, and reduced social participation.20)

Prevalence of Social Frailty

The prevalence of social frailty observed in this study was 25.3%, which aligns with estimates reported in previous research. A meta-analysis conducted among community-dwelling older adults reported a pooled prevalence of 21.1%.21) Similarly, another meta-analysis published in JAMDA found an overall prevalence of 18.8% (95% CI, 14.9–22.7%),9) though notable variation was observed depending on the assessment tool used. For instance, the TFI estimated a prevalence of 32.3%, while the Makizako SFI and the Social Frailty Screening Index yielded estimates of 27.7% and 13.4%, respectively.9)

The prevalence identified in this study falls within an intermediate range, higher than that reported with the Makizako tool.22) This difference may be partially attributed to the characteristics of the study sample and the methodological approach used. Importantly, heterogeneity in the definitions and measurement tools for social frailty remains a key challenge when comparing findings across studies. In this research, a PCA approach was applied to construct a multidimensional index incorporating functional dependence, social participation, emotional isolation, and perceived quality of healthcare. This broader conceptualization may have contributed to a more sensitive detection of social frailty, thereby explaining the slightly higher prevalence found.

In the Colombian context, a recent study investigating the association between social frailty and falls in older adults reported a prevalence of 36.6%. However, that study did not involve the construction of a dedicated measurement tool; instead, it used selected items from the Makizako index, which has not yet been validated in the local setting.23)

Social frailty was more prevalent among men in this study, a finding that contrasts with some international evidence. For instance, Qi et al.17) reported higher prevalence among women in urban China, while rural areas showed greater risk among men. Such differences may reflect cultural norms, gender roles, and methodological approaches. In Colombia, men may have reduced access to social support and be less inclined to seek help, increasing their vulnerability. Additionally, indicators like social participation or emotional support may capture male experiences differently. These findings highlight the importance of sex-specific analyses and the need for culturally grounded research on social frailty.

Social frailty was also more prevalent among individuals aged 75 years and older. Advanced age is a well-established predictor of social frailty. Studies by Zhang et al.9) and Qi et al.17) reported a progressive increase in social frailty prevalence with advancing age. The underlying mechanisms are multifactorial, including physiological decline,24,25) reduced physical functioning, cognitive impairment, and the gradual loss of close social connections.1) In addition, structural changes in family composition and social networks, often exacerbated by urbanization and migration, may further weaken support systems for the oldest-old. These findings reinforce global evidence and emphasize the importance of prioritizing individuals aged 75 and over in public health strategies aimed at addressing social vulnerability in older populations.

The highest prevalence of social frailty was observed among older adults residing in the department of Antioquia. Although no comparative studies have systematically evaluated the living conditions of older populations across the included departments, a possible explanation may be drawn from a 2016 study that assessed the health status of older adults in Antioquia. The authors reported that mental health risks had a substantial impact on overall health in this population. Additionally, nearly six out of 10 older adults exhibited moderate to severe deterioration in their social resources, and 7.6% reported having experienced some form of abuse.26) These findings highlight the high level of social vulnerability among older adults in this region.

Strengths and Limitations

There is a limited body of research focused on the development of social frailty assessment tools, with the majority of studies concentrated in Asian populations.14) This geographic imbalance underscores the significance of the present study, which contributes to the Latin American context by incorporating the region’s distinct cultural, social, economic, political, and historical influences on aging.

From an operational perspective, the index developed in this study offers several advantages. Unlike instruments that focus exclusive on social domains, this multidimensional index reflects the complexity of aging and is grounded in the theory of social production function. This provides a holistic profile of vulnerability that enhances understanding by explicitly integrating physical, emotional, and social constraints. This theoretical framework emphasizes the interplay between resource constraints and compensatory mechanisms. Notably, the index incorporates limitations in basic and instrumental activities of daily living highlighting the role of intrinsic capacity in social functioning as well as subjective health perception and healthcare satisfaction as fundamental components of older adults’ social experience.27)

Furthermore, the inclusion of perceived quality of healthcare and access to technology enhances the tool’s capacity for risk stratification and contextual relevance in resource-limited settings, contributing to a better understanding of social frailty beyond mere social deficits.

This holistic approach offers a comprehensive understanding of how physical, emotional, and social factors interact in the development of frailty. Using publicly available data from SABE Colombia enhances the index’s applicability in public health settings without requiring additional data collection. However, certain limitations must be acknowledged, including reliance on secondary data, potential loss of detail due to variable recoding, and limited geographic coverage, which may affect generalizability. Methodologically, limitations include the use of dichotomous variables for index construction and the initial reliance on a Pearson correlation matrix for PCA, although this was validated by sphericity tests and is supported by some specialized literature. Although the survey used validated instruments, the cross-sectional design captures only a single point in time. Future longitudinal studies are needed to assess the index’s stability and predictive value over time.

Public Health Implications

The development of a multidimensional SFI provides substantial public health benefits by enabling early identification of vulnerability across functional, emotional, social, and healthcare-related domains. This facilitates personalized and targeted interventions aimed at preserving autonomy and improving the quality of life in older adults. The inclusion of variables such as perceived healthcare quality enhances risk stratification for adverse outcomes. Incorporating social frailty screening into geriatric care and community programs promotes more equitable care models and supports efficient resource allocation, aligning with global aging strategies focused on social inclusion and the maintenance of intrinsic capacity.

Particular attention should be given to adults aged 75 and older, who showed significantly higher levels of social frailty in this study. Public health policies must prioritize this group with interventions addressing both social and functional vulnerability. A gender-sensitive approach is also essential, as social frailty may differ between men and women due to variations in social roles, behavior, and access to support networks. In addition, the use of population-based surveys enhances the scalability and applicability of the index for policy development. Identifying social frailty patterns at the population level enables more context-sensitive planning and helps address regional disparities in infrastructure, social support, and healthcare. These findings reinforce the importance of incorporating social frailty metrics into geriatric health monitoring and lay the groundwork for future longitudinal studies and public health strategies.

Conclusion

This study developed and applied a multidimensional SFI using data from older adults in four Colombian departments, identifying key domains such as functional dependence, limited social participation, emotional isolation, and perceived healthcare quality. The index revealed a prevalence of 25.3%, underscoring the importance of addressing social vulnerability in the context of Latin America’s demographic and socioeconomic transitions. The findings support the index’s utility for early detection and targeted public health interventions. Further research is needed to standardize definitions, assess predictive validity, and explore associations with health outcomes through longitudinal studies, especially in low- and middle-income settings.