https://orcid.org/0000-0003-4391-4286
Figures
Abstract
Background
The construct of social vulnerability attempts to understand social circumstances not merely as a descriptor, but as a predictor of adverse health events. It can be measured by aggregating social deficits in a social vulnerability index (SVI). We describe a standard procedure for constructing a multi-level SVI using two working examples.
Methods
First, we describe a six-step approach to constructing a SVI. Then, we conducted a secondary analysis of a clinical dataset (Canadian Immunization Research Network’s Serious Outcomes Surveillance Network (SOS)) and a population-based dataset (Canadian Longitudinal Study on Aging (CLSA)). In both datasets, we construct SVIs, use descriptive statistics to report distributions by age and sex, and perform a multivariable linear regression of social vulnerability on frailty.
Results
Procedures for drafting a list of candidate social items, selecting deficits for inclusion, and screening deficits to meet inclusion criteria were applied to yield a 18-deficit SVI for the SOS and 74-deficit SVI for the CLSA. Deficits in each SVI were re-scored between 0 and 1, where 1 indicates the greater risk. Finally, the sum of all deficits is calculated into an index. In the SOS, SVI was associated with age only for females and was weakly associated with frailty (r = 0.26, p<0.001). In the CLSA, SVI was associated with age for both sexes and moderately associated with frailty (r = 0.41, p<0.001).
Citation: Mah JC, Theou O, Perez-Zepeda MU, Penwarden JL, Godin J, Rockwood K, et al. (2024) A standard procedure for constructing a multi-level social vulnerability index using CLSA and SOS data as working examples. PLoS ONE 19(12): e0315474. https://doi.org/10.1371/journal.pone.0315474
Editor: Francesco Curcio, University of Naples Federico II, ITALY
Received: May 30, 2024; Accepted: November 26, 2024; Published: December 13, 2024
Copyright: © 2024 Mah et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data are available from the Canadian Longitudinal Study on Aging (www.clsa-elcv.ca) for researchers who meet the criteria for access to de-identified CLSA data. This research has been conducted using the CLSA dataset Baseline Tracking Dataset version 3.3 and the Baseline Comprehensive Dataset version 3.2, under Application Number 1906015. The CLSA is led by Drs. Parminder Raina, Christina Wolfson and Susan Kirkland. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging. Data requests can be made at https://www.clsa-elcv.ca/data-access/ for the CLSA data and to the Canadian Immunization Research Network’s Serious Outcomes Surveillance Network at https://cirnetwork.ca/cirn-call-for-proposals/ for SOS data. The data from the Canadian Immunization Research Network’s Serious Outcomes Surveillance Network have restrictions regarding availability of these data.
Funding: This research received no specific grant or funding from any funding agency. However, KR, MKA, JG and JCM are part of the Canadian Consortium on Neurodegeneration in Aging (CCNA) Team 14, which investigates how multi-morbidity, frailty and social context modify risk of dementia and patterns of disease expression. The CCNA receives funding from the Canadian Institutes of Health Research (CNA-137794) and partner organizations (www.ccna-ccnv.ca). This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces, Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia.
Competing interests: JCM is an Internal Medicine resident with Nova Scotia Health and receives scholarships supporting her PhD research from the Department of Medicine at Dalhousie University, Dalhousie Medical Research Foundation, Dr. Patrick Madore Foundation and the Pierre Elliott Trudeau Foundation. KR is President and Chief Science Officer of DGI Clinical, which in the last five years has contracts with pharma and device manufacturers on individualized outcome measurement. In 2019 he attended an advisory board meeting with Nutricia. Otherwise any personal fees are for invited guest lectures and academic symposia, received directly from event organizers, chiefly for presentations on frailty. He is Associate Director of the Canadian Consortium on Neurodegeneration in Aging, which is funded by the Canadian Institutes of Health Research, and with additional funding from the Alzheimer Society of Canada and several other charities. He receives career support from the Dalhousie Medical Research Foundation as the Kathryn Allen Weldon Professor of Alzheimer Research, and research support from the Canadian Institutes of Health Research, the QEII Health Science Centre Foundation, the Capital Health Research Fund and the Fountain Family Innovation Fund of the QEII Health Science Centre Foundation. KR has asserted copyright of the Clinical Frailty Scale through Dalhousie University. Use is free for research, education or not-for-profit care (users are asked not to change it or charge for its use). MKA reports grants from Canadian Consortium on Neurodegeneration in Aging (CCNA), with funding from Canadian Institutes of Health Research (CIHR). MKA reports grant funding and honoraria from Sanofi, GSK, Pfizer, Seqirus and the Canadian Frailty Network for work relating to frailty and vaccine preventable illness. OT (with Dr. Kenneth Rockwood) have asserted copyright of the Pictorial Fit-Frail Scale, which is made freely available for education, research, and not-for-profit health care. Licenses for commercial use are facilitated through the Dalhousie Office of Commercialization and Industry Engagement. All other authors declare that they have no competing interests. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Background
The conditions in which people are born, live, work and age collectively influence their ability to anticipate, cope, resist and recover from an adverse event [1]. Social vulnerability provides a way of understanding the social environment not merely as a descriptor, but as an attempt to quantify an individual’s or community’s relative vulnerability to changes in their environment, social circumstances, health, or functional status [2]. In short, when all other non-social factors are equal, how do disadvantageous social circumstances lead to a community being disproportionally devastated by an epidemic or to an individual being unable to recover in the expected timeframe following an adverse health event?
Social circumstances are complex; there are many social factors, multiple layers of social factors from personal supports to neighbourhood dynamics, and numerous potentially unforeseen interactions among such factors. While measuring social vulnerability may be perceived as challenging, conveniently, social vulnerability can readily be measured using available data. One way to estimate social vulnerability is through an index aggregating social factors. An index approach has several benefits. It can provide a holistic picture of social circumstances by including different categories of social factors (e.g., socioeconomic status, social engagement, social capital), by avoiding arbitrarily separating related factors into distinct categories and by accounting for gradations in social vulnerability [2, 3].
Social vulnerability indices (SVIs) are used to measure complex social circumstances associated with health outcomes. The SVI employed by the Agency for Toxic Substances and Disease Registry / Centers of Disease Control and Prevention [4] is widely used, and in the United States, associated with many adverse outcomes, such as in SARS-Cov-2 [5], surgery [6], and heart failure re-admissions [7]. Another well-known SVI by Cutter and colleagues [8] has been adapted and shown to be associated with cancer risk [9] and Lyme disease incidence [10]. However, neither of these routinely used social vulnerability tools were initially developed for use in health or medical fields. Further, a recent scoping review suggested a SVI might be strengthened if composed of social factors which reflect vulnerabilities at the individual, household, and community levels [11]. To address these gaps, we aim to describe a standard procedure for constructing a multi-level SVI with relevance to the health of individuals using a social deficit approach. We provide two working examples of constructed SVIs using this approach.
Methods
Theory
Our operationalization of social vulnerability builds upon social capital theory, especially that such capital can be deployed in times of need. Broadly, social capital is the organization of social structures and how these structures facilitate actions of stakeholders in the society [12]. Social capital is a collective resource; like economic capital, harnessing social capital brings advantages for the wealthy or powerful through their entrenched networks and institutionalized relationships [13]. It consists in social support, social engagement and access to resources (including economic capital) [14]. Adequate social capital is productive; in its absence, achievement of desired ends would not be possible [12]. To illustrate, following a hip fracture, economic capital to purchase a wheelchair or renovate a home with a ramp improves function, but so will social capital in the form of free exchanges of food, time or company provided by friends, family and community members. Whether social capital exists at the level of the individual or the collective has been the subject of much debate. We see social capital as related both to characteristics of the individual (e.g., educational), and also to other aspects at the level of culture and the environment (e.g., neighbourhood safety). These related but distinct non-medical factors—the social determinants of health [15]—exist with social capital on a continuum from individual to collective [14].
A socio-ecological framework [16], wherein an individual is nested within expanding spheres of social influence, offers a useful way to think about how multiple social factors influence health [2]. Social factors exist on a continuum of multiple levels of social influence—from the individual to family and friends, neighbourhoods, and communities, and society at large—and contribute to overall social vulnerability [2]. For an adult seeking healthcare, at the micro level, social factors are their own health behaviours and their closest links with family and caregivers. We define the meso level according to Newman and Newman as the interrelations among two or more microsystems that then impact the individual [17]. Examples includes family-friends interactions, or friends-healthcare interactions (e.g., the health literacy of friends and relations and their access to resources and supports). The exo level refers to the available community supports available such as home care services or access to rehabilitation programs. The macro level encompasses the attitudes towards older adults in broader policy reflected in pension plans or universal health care [16]. A measure of social vulnerability must account for this complexity and include social factors across the continuum from an individual to a group level. Similarly, a lone marker of social circumstances cannot adequately reflect the multifaceted interactions between social factors; therefore, a global index of social vulnerability must incorporate multiple social determinants of health.
Constructing the SVI
We describe six steps, and recommendations and considerations, in the construction of a SVI.
1. Draft a list of candidate items for the SVI.
Recommendations. Choose social factors, for potential inclusion as items in the index, with high face validity and comprehensibility supported by a good theoretical base, with evidence of the potential to adversely impact health outcomes. Lists of candidate social deficits may be procured based on existing data availability [18], but have also been generated through consensus with experts [19].
Considerations. For controversial social factors (e.g., retirement or rurality where the deficit state may be beneficial to some and detrimental to others), consult experts who are familiar with the population of interest.
2. Select deficits for inclusion in the SVI.
Recommendations. Items selected will become deficits in the index. Collectively, they must include a range of factors across multiple social domains and across multiple levels of social influence representing a holistic view of a person’s social circumstances.
- Include social deficits across multiple social domains. Examples of social domains are: access to material resources (e.g. income or socioeconomic status), social support (e.g. links to family, friends or community) and social engagement (e.g. participation in collective society) [14]. Social domains can also include examples from the social determinants of health such as income & social status, employment & working conditions, education & literature, childhood experiences, physical environments, social supports & coping skills, and access to health services [15]. Social domains can also be chosen by including previously reported items in measurement tools such as a social support index [20] or social isolation index [21], among others.
- Include social factors from multiple spheres or levels of influence across the continuum from individual to group levels [22]. A recent scoping review summarizing the composition of items in SVIs noted more than half of all SVIs included items reflective of individual socioeconomic status, but also prevalence of at-risk populations in a geographic region [11].
- Include factors that are both objective (e.g. living alone, number of close friends) and subjective (e.g. loneliness, availability of emotional support).
Considerations. Selecting deficits for inclusion in the SVI necessitates a balance between creating a robust measure (generally optimized with a greater number of items) and data availability (in the case of secondary analysis of existing datasets) or participant burden (in the case of prospective data collection). Some SVIs use personality factors (e.g., neuroticism) and lifestyle factors (e.g., exercise and diet); we see these as independent aspects of wellbeing and representing another dimension of health, not to be used in a SVI.
Choosing deficits based on statistical correlation to one another is not recommended. Deficits are not required to be correlated with one another. Items in the SVI are more appropriately considered causal variables rather than indicator variables [23]. For example, education and feelings of loneliness may not be correlated, but they both contribute to a person’s social vulnerability. Both the causal direction and the lack of correlation between the deficits renders factor analysis inappropriate. The selection of deficits rests on their potential contribution to social vulnerability rather than their intercorrelations.
3. Code deficits for the SVI.
Recommendations. The coding for each item in the SVI depends on its scale of measurement. Regardless, all social deficits receive a score from 0 to 1. A value of 1 indicates the greatest state of relative vulnerability to damage (e.g., education: never completed high school = 1 and completed high school or greater = 0). For intermediate responses, deficits may take a value of 0.5 (e.g. never completed high school = 1, completed high school only = 0.5, post-secondary education = 0). Ordinal items may rank into a score according to number of levels. For example, a deficit with four levels (e.g., never completed high school, completed high school, college or university, post-graduate education) would be coded 0 for the social factor characteristic that is most protective, 0.33 and 0.66 for middle states of vulnerability and 1 for the most detrimental characteristic for vulnerability. Continuous items may be categorized according to pre-established cut points or coded into a continuous score between 0 and 1. In our education example, if measured in years, >20 years of school would be coded 0 and decreasing number of years allocated a score according to the equation 1—(# years of school / 20 max years of school).
Considerations. The theoretical basis of coding deficits for the SVI combines a deficit accumulation principle [24] with social capital underpinnings. Living alone, for example, is not always considered a deficit (i.e., an adverse social circumstance increasing the risk of damage or prolonging recovery time following an adverse health event) and individuals can be content living alone. However, social capital refers to resources one can draw upon should a crisis occur. Through this lens, living alone does confer an increased risk of social vulnerability and can be coded as the highest deficit state.
Another challenge with social deficit coding is the lack of self-evident cut points. We suggest two options. First, cut offs for vulnerable states could depend on expert consensus drawing on literature and experience. Second, cut offs can be determined by mapping distribution of the social variable using basic statistical techniques. Individuals with the deficit beyond the 75th percentile of a specific dataset could be coded as having the most vulnerable state for that variable. This latter is for datasets that are truly representative of a population. The coding represented here is similar to previous work [24, 25].
4. Screen deficits that meet inclusion criteria.
Recommendation. Screen the chosen social deficits for missingness. The threshold for missing values is contingent on the total number of social deficits available for the construction of the index, although >5% missing data is generally acceptable as a cut off to exclude the deficits in similar indices [26]. Allowing for high levels of missingness at the item level may result in losing observations when calculating the SVI scores, assuming no other methods of dealing with missing data is employed such as multiple imputation [27].
Screen deficits for prevalence. A rare social deficit in the population (i.e., <1%) could be combined with another deficit to avoid exclusion in the final calculations or inflating the denominator.
5. Calculate the final SVI.
Recommendations. The overall SVI score per individual is calculated by summing the coded values for all social items (reflection of the deficits) and dividing by the absolute count of items included in the SVI. Therefore, the final SVI also takes a value between 0 and 1, allowing for standardization and development of a common language as it pertains to social vulnerability.
Considerations. Calculate a final SVI only for individuals with sufficient data. Calculating a final SVI for individuals missing more than 20% of SVI items may not accurately reflect their social circumstances and may underestimate true social vulnerability. Using statistical methods of dealing with missing data such as multiple imputation can be considered [27].
Similar to the frailty index, our SVI builds in natural weighting. For example, individuals who are not married are more likely to score social deficits for living alone and having less social support. While there are benefits of weighting items in a SVI (e.g., gaining performance or separability measurement), we aim for the SVI to be highly generalizable across contexts.
6. Report the SVI.
Recommendations. Report the number of items and domains. List the deficits and their coding. Report distribution of the SVI. Report correlation with age and sex/gender. If a SVI is to be used in the same dataset or across multiple time series, it should consist of the same variables from one iteration to the next.
Worked examples
Samples
To demonstrate the standard approach described above, we calculate and compare SVIs in two separate datasets.
The first is a clinical dataset with a minimal number of social variables. The Canadian Immunization Research Network’s Serious Outcomes Surveillance (SOS) Network is a prospective dataset of Canadians hospitalized with acute respiratory illness in six Canadian provinces. Within the SOS dataset, we selected all individuals over the age of 65 years old admitted to hospital during the 2011–2012 influenza season. Ethics approval for the use of SOS data was obtained from the Capital Health Research Ethics Board (CDHA-RS/2010-123).
The second dataset is a weighted population-based dataset: the Canadian Longitudinal Study on Aging, a national, stratified, prospective study of over 50,000 community-dwelling Canadian women and men aged 45 to 85 years old at time of recruitment [28]. We draw on the complete sample of the CLSA using both the comprehensive and tracking cohort from the baseline CLSA assessment in 2011. This research has been conducted using the CLSA dataset Baseline Tracking Dataset version 3.3, the Baseline Comprehensive Dataset version 3.2 and CLSA Sample Weights Version 1.2. We classify each item in the SVI by the domains included in the CLSA (i.e., socio-demographic, home ownership, education, social networks, social support availability, social participation, income, built environment and psychosocial). For performing a secondary analysis of data, a waiver of ethics application was approved by the Nova Scotia Health Authority Research Ethics Board. Data was obtained from the CLSA agreement #1906015.
Statistical methods.
We use descriptive statistics to report SVI distributions by age and sex and report Pearson correlation coefficients. We performed a multivariable linear regression of social vulnerability on frailty. Frailty is measured using frailty indices (FI) from previously published papers [29, 30]. We calculated Receiver Operating Characteristics (ROC) Curves [31] and area under the ROC curves (AUC) for SVI with frailty (using a FI cut off of 0.21) and mortality [32].
Constructing the SVIs
The SVI calculated in the SOS Network dataset is an index composed of 18 items. Nine domains are represented by these 18 deficits when the CLSA social domain classifications are applied as shown in Table 1 and Fig 1. Of the 18 deficits, 13 were coded as dichotomous and the remainder were ordinal. The SOS SVI is an example of an index constructed based on data availability (see step 2 considerations).
The population-based SVI is constructed from 74 items that met criteria in the CLSA. The CLSA SVI is an index constructed based on the theories proposed in Steps 1, 2 and 3 given the abundance of social items for in the dataset. Initially, 84 items were drafted from the CLSA. Seven items did not meet screening criteria due to poor face validity (e.g. personality or lifestyle factors rather than social factors) or because the author group could not come to consensus on the item’s deficit state (e.g. retirement) The CLSA SVI is composed of deficits from ten social domains that were already established by the CLSA data collection process; deficits from the social support domain were the most prevalent (20/74) as seen in Table 2 and shown in Fig 1. Fig 1 also shows how each item relates to the socio-ecological framework and each level of social influence. Most deficits were coded as ordinal variables (55/74) and the remaining were coded as dichotomous. Screening of CLSA deficits (Step 4) for prevalence and missingness was possible due to large numbers and is available in S1 File. Other recreational activities, church activities and club and fraternal services were combined due to low individual prevalence. No items were excluded due to missing data due to the completeness of the CLSA dataset. Finally, one item was removed (i.e., social inequity) because it was not available in all waves of the CLSA and a SVI that can be used longitudinally should have the same items (Step 6).
Characteristics of the Indices
The mean SVI score in the SOS network dataset is 0.30(SD 0.13) and 0.22(SD 0.10) in the CLSA dataset. Descriptive statistics for the two datasets are presented in Table 3 showing that the SOS network dataset contains individuals who are older and living with a greater degree of frailty than the individuals in the CLSA cohort, which is not surprising given that all SOS participants were hospitalized. The average SVI score for women is higher on average than for men in both cohorts. Fig 2 shows the association of SVI with age by sex. In SOS, SVI increases with age only in women (r = 0.11, p = 0.04). In CLSA, SVI slowly increases by age in the total cohort (r = 0.28, p<0.001) and in men (r = 0.15, p<0.001) and women (r = 0.29, p<0.001). The distribution of the SVIs in both cohorts are shown in Fig 3. Internal consistency calculated using Cronbach’s alpha was 0.65 for the SOS SVI and 0.89 for the CLSA SVI.
Social vulnerability and frailty and mortality
Frailty was measured using a 39-deficit frailty index and 52-deficit frailty index, each already validated in the SOS Network [33] and CLSA datasets [30], respectively. In the SOS, SVI is weakly correlated with frailty (r = 0.26, p<0.001). This relationship was stronger for women (r = 0.36, p<0.001) than men (r = 0.40, p<0.001) for frailty (Fig 4). In the CLSA, correlation with frailty was r = 0.37 (p<0.001) with a stronger correlation for women (r = 0.41, p<0.001) than men (r = 0.33, p<0.001). In a multivariable linear regression model adjusted for age and sex, a 0.1 increase in the FI was associated with a 0.28 (95% CI 0.23–0.42, p<0.001) increase in SVI in the SOS and a 0.68 (95% CI 0.63–0.72, p<0.001) increase in SVI in the CLSA.
Fig 5 shows the ROC curves of the discrimination performance of the SVI and frailty and mortality. The AUC of the SOS SVI for baseline frailty (two weeks prior to admission) was 0.64 (95% CI: 0.59–0.69) and mortality was 0.60 (95% CI: 0.51–0.70). The AUC of the CLSA SVI was 0.77 (95% CI 0.76–0.79) for frailty and 0.66 (95% CI: 0.65–0.68) for mortality.
Discussion
Our research group has over a decade of experience constructing SVIs. We aim to measure social vulnerability so as to capture a rich description of an individual’s social deficits (or problems), utilize data that is readily and practically measurable in population and clinical settings, respond to changing social circumstances, and predict important health outcomes.
Here we present a method of constructing a SVI that takes into consideration the whole person in society. We have highlighted the social theories underpinning the construction of this SVI and presented two examples, one in a clinical dataset with few variables and one in a larger population dataset with many variables. Here, the SVI-SOS allows us to demonstrate how to calculate the index with a smaller, homogenous population, and fewer variables. Further, we illustrate how inclusion of a SVI can be feasible with prospective clinical data collection. The SVI-CLSA provides an example of screening and selecting variables for inclusion in the index (S1 File). Furthermore, the CLSA SVI here demonstrates properties of previously calculated SVIs. Women tend to be more socially vulnerable than men, and importantly, and unlike a perfect state of health (frailty index = 0), almost no older adult has zero social vulnerability [34]. Most importantly, both SVIs are composed of deficits representing multiple domains of social circumstances and different levels of social influence from the individual to the community. In this way social vulnerability reflects how frailty can be conceptualized as being the expression of problems across multiple body systems from the cellular to organ to systems (e.g., cardiovascular) level.
Only in the last decade have SVIs become popular in the medical literature [11]. SVIs developed in non-medical fields include: Cutter et al created a SVI to environmental hazards [8], The Centres of Disease Control’s SVI initially developed for emergency management and disaster planning [4] and, from Brazil, a SVI as a tool for urban management and development [35]. All these SVIs are composed of only geographical or census level deficits. Whereas these measures subsequently were adapted to measure health outcomes, we approach social vulnerability through a health lens from the start. We especially consider the impact of social factors on the ability to resist the adverse consequences of any adverse medical event (or procedure) or to repair or cope with it. In deficit accumulation terms, the distinction between resisting a stress as “robustness” and recovery or maintenance (“resilience”) appears to be useful.
Social vulnerability measured using an index has several strengths. It gives a quantitative overview of an individual’s social circumstance, which would not be possible by examining each social variable or domain in isolation. Calculation of the SVI allows for flexibility and gradations. Small increases in social vulnerability can be captured and studied in relation to frailty and other outcomes of interest. Social vulnerability as a gradient can be useful to better differentiate risk and vulnerability. In our experience, there is value in using the SVI in tandem with the FI. Since both constructs capture heterogenous aspects of a person’s clinical picture, adding the SVI builds a better model to predict health related outcomes. This is supported by previous literature suggesting frailty indices may be strengthened by the inclusion of socioeconomic deficits [36]. Nonetheless, we view social vulnerability and frailty as distinct concepts as prior analyses show SVI is independently associated with mortality and disability even after controlling for frailty [37], and is associated with mortality in older adults without frailty [38]. Furthermore, there is great utility for policymakers and clinicians to be able to adapt or replicate the SVI in any database or population, regardless of whether the data come from surveys, clinical sources, or administrative records. The SVI can help to pinpoint vulnerable populations or regions and to target social and health resources as demonstrated by the Covid-19 pandemic [5]. Any such SVI can be constructed with a variety of social variables, so long as the basic tenant of encompassing multiple broad social domains and levels is met.
Many questions remain unanswered in this area. One limitation is that the temporal aspect of social vulnerability, or the chronosystem as described by the ecological model, remains an aspect of complex social environments not fully captured by this approach. For example, the SVI may be highly influenced by cohort effects over time–social factors that were protective in the past might not be anymore (e.g., in the CLSA reading the newspaper nowadays might be substituted by social media use, and reading the newspaper in certain cohorts may indicate a vulnerable trait). Previous work with the SVI has also suggested that social vulnerability plays a bigger role in the fittest individuals [29, 38]. Does the lack of time dependence of social vulnerability arise only when the frailest have already died? In comparison to our work on frailty, social vulnerability is not as strongly associated with age. Future research is encouraged to examine the dynamics of social vulnerability, and how they change with age, and whether they show temporal trends. Additionally, we build our SVI using a deficit accumulation approach, but is this the same as taking a resilience approach? Social vulnerability should ideally encompass two concepts articulated by Ukraintseva, Yashin and Arbeev: robustness (the ability to resist deviation from the healthier state) and resilience (recovery to the healthier state aided by a well connected and supportive social situation) [39]. Our SVI does not distinguish whether the absence of a deficit is the same as the presence of a resilience factor. Conceivably these would not confer the same degree of (dis)advantage yet only the former is captured by our index. Such considerations are motivating additional inquires by our group.
Another area for further research surrounds the measurement properties of SVIs. Further work should continue to evaluate validity and reliability of these indices. SVIs may have features of both formative and reflective models, however, future research may also focus on replicating features of the deficit accumulation approach seen in the frailty index [40]. Frailty indices, even if they do not consist of the same items or use the same databases, show similar accumulation patterns (i.e. 0.03/year) and have a maximal limit accumulation [24, 41]. A limitation of this paper is the varied number of deficits in the SVI–the SOS may have too few items and the CLSA SVI may have redundant items. Determining if SVI performance is similar after a certain number of items is another area of exploration.
Conclusion
We present a standard method of constructing a SVI. In our holistic approach to understanding social circumstances, our SVI incorporates factors from multiple social domains and levels in a social-ecological model. We demonstrate construction of the SVI and its feasibility in two different datasets, with the potential for operationalization in many other datasets. Social vulnerability may have reproducible associations with age, sex and frailty. This SVI can be used to improve our understanding of social vulnerability and its impacts on the health of communities and individuals.
References
- 1. Singh SR, Eghdami MR, Singh S. The Concept of Social Vulnerability: A Review from Disasters Perspectives. International Journal of Interdisciplinary and Multidisciplinary Studies 2014; 1: 12.
- 2.
Andrew MK. Social vulnerability in old age. In: Brocklehurst’s Textbook of Geriatric Medicine and Gerontology. Saunders, 2010.
- 3. Mah JC, Andrew MK. Social vulnerability indices: A pragmatic tool for COVID-19 policy and beyond. The Lancet Regional Health—Europe 2022; 14: 100333. pmid:35194590
- 4. Flanagan BE, Gregory EW, Hallisey EJ, et al. A Social Vulnerability Index for Disaster Management. Journal of Homeland Security and Emergency Management; 8. Epub ahead of print 5 January 2011.
- 5.
Agency for Toxic Substances and Disease Registry (ATSDR). Innovative Uses of SVI During COVID-19, https://www.atsdr.cdc.gov/placeandhealth/project_snapshots/svitool_covid.html (2021, accessed 12 December 2022).
- 6. Hyer JM, Tsilimigras DI, Diaz A, et al. High Social Vulnerability and “Textbook Outcomes” after Cancer Operation. Journal of the American College of Surgeons 2021; 232: 351–359. pmid:33508426
- 7. Regmi MR, Tandan N, Parajuli P, et al. Social Vulnerability Indices as a Risk Factor for Heart Failure Readmissions. Clin Med Res 2021; 19: 116–122. pmid:34531268
- 8. Cutter SL, Boruff BJ, Shirley WL. Social Vulnerability to Environmental Hazards. Social Science Quarterly 2003; 84: 242–261.
- 9. Georgantopoulos P, Eberth JM, Cai B, et al. Patient- and area-level predictors of prostate cancer among South Carolina veterans: a spatial analysis. Cancer Causes Control 2020; 31: 209–220. pmid:31975155
- 10. Ratnapradip D, McDaniel JT, Barger A. Social vulnerability and Lyme disease incidence: a regional analysis of the United States, 2000–2014. Epidemiology, Biostatistics and Public Health; 14. Epub ahead of print 2017.
- 11. Mah J, Penwarden J, Pott H, et al. Social vulnerability indices: A scoping review. BMC Public Health. pmid:37380956
- 12. Coleman JS. Social Capital in the Creation of Human Capital. American Journal of Sociology 1988; 94: S95–S120.
- 13.
Bourdieu P. THE FORMS OF CAPITAL. In: Richardson J (ed) Handbook of Theory and Research for the Sociology of Education. Westport, CT: Greenwood, 1986, pp. 241–58.
- 14. Andrew MK. Le capital social et la santé des personnes âgées. RETRAITE ET SOCIÉTÉ 2005; 46: 129–143.
- 15.
WHO Commission on Social Determinants of Health, World Health Organization (eds). Closing the gap in a generation: health equity through action on the social determinants of health: Commission on Social Determinants of Health final report. Geneva, Switzerland: World Health Organization, Commission on Social Determinants of Health, 2008.
- 16.
Bronfenbrenner U, Morris P. The ecology of developmental processes. In: Damon W, Lerner RM (eds) Handbook of child psychology: Theoretical models of human development. John Wiley & Sons Inc., 1998, pp. 993–1028.
- 17.
Newman B, Newman P. Newman B.M. & Newman P.R. (2020) Ecological theories In Newman B.M. & Newman P.R. (Eds.) Theories of Adolescent Development, (pp. 313–335) Elsevier.
- 18. Andrew MK, Mitnitski A, Rockwood Kenneth. Social Vulnerability, Frailty and Mortality in Elderly People. PLoS ONE; 3. pmid:18493324
- 19. Prabhu S, Wanje G, Oyaro B, et al. Adaptation of a social vulnerability index for measuring social frailty among East African women. BMC Public Health 2022; 22: 167. pmid:35073906
- 20. Mazzella F, Cacciatore F, Galizia G, et al. Social support and long-term mortality in the elderly: Role of comorbidity. Archives of Gerontology and Geriatrics 2010; 51: 323–328. pmid:20153534
- 21. Steptoe A, Shankar A, Demakakos P, et al. Social isolation, loneliness, and all-cause mortality in older men and women. Proc Natl Acad Sci USA 2013; 110: 5797–5801. pmid:23530191
- 22.
Glanz K, Rimer BK, Viswanath K (eds). Chapter 20: Ecological Models of Health Behavior. In: Health Behavior and Health Education: Theory, Research, and Practice. San Francisco, California: Jossey-Bass, 2008, p. 591.
- 23. Fayers PM, Hand DJ. Causal variables, indicator variables and measurement scales: an example from quality of life. J Royal Statistical Soc A 2002; 165: 233–253.
- 24. Searle SD, Mitnitski A, Gahbauer EA, et al. A standard procedure for creating a frailty index. BMC Geriatr 2008; 8: 24. pmid:18826625
- 25. Stubbings G, Rockwood K, Mitnitski A, et al. A quantile frailty index without dichotomization. Mechanisms of Ageing and Development 2021; 199: 111570. pmid:34517019
- 26. Rockwood K, Song X, Mitnitski A. Changes in relative fitness and frailty across the adult lifespan: evidence from the Canadian National Population Health Survey. Canadian Medical Association Journal 2011; 183: E487–E494. pmid:21540166
- 27. Pridham G, Rockwood K, Rutenberg A. Strategies for handling missing data that improve Frailty Index estimation and predictive power: lessons from the NHANES dataset. GeroScience 2022; 44: 897–923. pmid:35103915
- 28. Raina PS, Wolfson C, Kirkland SA, et al. The Canadian Longitudinal Study on Aging (CLSA). Can J Aging 2009; 28: 221–229. pmid:19860977
- 29. Godin J, Theou O, Black K, et al. Long-Term Care Admissions Following Hospitalization: The Role of Social Vulnerability. Healthcare 2019; 7: 91. pmid:31311101
- 30. Pérez-Zepeda MU, Godin J, Armstrong JJ, et al. Frailty among middle-aged and older Canadians: population norms for the frailty index using the Canadian Longitudinal Study on Aging. Age and Ageing 2021; 50: 447–456. pmid:32805022
- 31. Mandrekar JN. Receiver Operating Characteristic Curve in Diagnostic Test Assessment. Journal of Thoracic Oncology 2010; 5: 1315–1316. pmid:20736804
- 32. Hoover M, Rotermann M, Sanmartin C, et al. Validation of an index to estimate the prevalence of frailty among community-dwelling seniors. Health Rep 2013; 24: 10–17. pmid:24258362
- 33. Andrew MK, Shinde V, Ye L, et al. The Importance of Frailty in the Assessment of Influenza Vaccine Effectiveness Against Influenza-Related Hospitalization in Elderly People. The Journal of Infectious Diseases 2017; 216: 405–414. pmid:28931244
- 34.
Pineau E. The Influence of Social Vulnerability on Cognitive Decline and Hospitalization among Older Assisted Living Residents with and without Dementia. Master, University of Waterloo, 2016.
- 35. Nahas M, Ribeiro C, Esteves O, et al. The map of social exclusion in Belo Horizonte: Methodology of building an urban management tool. Cad Cienc Soc; 7.
- 36. Abete P, Basile C, Bulli G, et al. The Italian version of the “frailty index” based on deficits in health: a validation study. Aging Clin Exp Res 2017; 29: 913–926. pmid:28688080
- 37. Andrew MK, Keefe JM. Social vulnerability from a social ecology perspective: a cohort study of older adults from the National Population Health Survey of Canada. BMC Geriatr 2014; 14: 90. pmid:25129548
- 38. Andrew MK, Mitnitski A, Kirkland SA, et al. The impact of social vulnerability on the survival of the fittest older adults. Age and Ageing 2012; 41: 161–165. pmid:22287038
- 39. Ukraintseva S, Yashin AI, Arbeev KG. Resilience Versus Robustness in Aging. GERONA 2016; 71: 1533–1534.
- 40. Stolz E, Mayerl H, Godin J, et al. Reliability of the Frailty Index Among Community-Dwelling Older Adults. The Journals of Gerontology: Series A 2024; 79: glad227. pmid:37738215
- 41. Theou O, Haviva C, Wallace L, et al. How to construct a frailty index from an existing dataset in 10 steps. Age and Ageing 2023; 52: afad221. pmid:38124255