Do Interventions Reducing Social Vulnerability Improve Health in Community Dwelling Older Adults? A Systematic Review

Background

What is Social Vulnerability

When two older adults have the same medical conditions and the same degree of frailty, the difference between living independently in the community with supports and requiring admission to a hospital or long-term care facility can be explained by social circumstances. Social vulnerability (SV) is defined as the degree to which a person’s overall social circumstances leave them susceptible to further insults (ie health or socially related adverse events).¹

Social vulnerability occurs when individuals have been relatively disadvantaged by the broader conditions in which they were born, lived, worked and aged; also known as the social determinants of health.² Socially vulnerable individuals have variable presentations. For example, an older woman living below the poverty line in a racialized neighbourhood may be vulnerable due to economic and housing reasons but has a robust social network of friends and family. Another example of social vulnerability is the reclusive and wealthy older man who has no one to call in the event of a natural disaster. The commonality is the inability to respond in times of crises due to a social reason (ie poverty or isolation) rather than a health-related reason. Unequal determinants directly impact individuals, but arise from complex political, social and economic structures. For this reason, the social determinants of health (SDOH) and social vulnerability are often conceptualized through ecological models.^3–5

How SV Influences Health & How Reducing SV Might Work

The link between SDOH and health is well documented, accounting for 30–55% of adverse health outcomes globally, such as mortality from chronic diseases.⁶ The construct of social vulnerability as the manifestation of poor social determinants in an individual is less prominent; nonetheless, being socially vulnerable has been associated with greater mortality,^5,7–10 risk of cognitive decline¹¹ and disability,⁹ and likelihood of long-term care admission.¹² Social divergence theory suggests socioeconomic disparities accumulate with time, resulting in worse health inequities in older age, which are compounded by shrinking social circles due to infirmity and death of friends and family members.¹³ Figure 1 shows the associations between SDOH, SV and health outcomes and helps to explain how interventions reducing social vulnerability through the SDOH might work.

Figure 1 Conceptual model linking the social determinants of health to social vulnerability and health outcomes.

Why It is Important to Do This Review

Studying social vulnerability is challenging as complex interrelated pathways by which upstream social factors influence health to do not allow for hypothesis testing through traditional research methods. Moreover, since multicomponent intervention studies rarely distinguish the social components of the intervention, there is little evidence suggesting what can be done to reduce health inequities through the addition of social programs in real life situations. As social determinants research “for the most part exists outside the literature concerned with effectiveness of interventions”,¹⁴ especially in non-hospitalized populations,^15,16 this is one gap this review will try to address. To our knowledge, no previous systematic review has attempted to collate and evaluate complex interventions based on a strong social intervention component in older adults. Additionally, for older adults with multiple medical conditions, it is nearly impossible to separate out those who are purely frail from those who are purely socially vulnerable. By accepting that multiple things can go wrong simultaneously in an individual, interventions aimed at social vulnerability cannot ignore medical frailty and vice versa; to do so would continue to fractionate research areas, moving away from treating individuals holistically. Instead, the lens used here to study social interventions within complex interventions accounts for the broader context where older adults would presumably receive both medical and social supports. Combined with a systematic mixed-methods literature review methodology, the authors attempt to answer the broad research question: is there evidence to support the notion that complex interventions targeting reduction in social vulnerability improve health outcomes for older adults living in the community?

Methods

Review Objectives

To identify, appraise and synthesize evidence on the effectiveness of complex interventions where a main component of the intervention targets reduction in social vulnerability for improving health related outcomes (mortality, function, cognition, subjective health and healthcare use) in older adults living in the community.

Study Design, Search Strategy & Selection Criteria

This systematic review employs a mixed methods approach to synthesizing complex interventions following PRISMA-CI guidelines.^17,18 Five databases were searched for primary studies: Medline, Cochrane Library (CENTRAL), EPPI-Centre’s Trials Registrar of Promoting Health Interventions, Campbell Collaboration, and Social Sciences Citation Index. The search strategy in Medline was approved by a medical librarian using the Peer Review Electronic Search Strategies (PRESS) checklist,¹⁹ then translated to the remaining databases (Supplement A provides the complete search strategy). The last search was conducted on May 18, 2021. Reference lists of included studies were also searched. Due to the nature of the interventions of interest, there was a concurrent search of grey literature to find interventions conducted by governments, non-profit or volunteer organizations.²⁰ This included a targeted website search of known national and international organizations who fund, implement or evaluate social interventions (Administration for Community Living, The Care Policy and Evaluation Centre, Government of Canada, HelpAge International, The International Federation on Ageing and the World Health Organization).

Inclusion criteria is fully described in Table 1 using the PICOT approach. Quantitative, qualitative or mixed studies were included if they investigated complex interventions targeting reductions in social vulnerabilities through the SDOH in a community-dwelling older adult population, and evaluated a priori outcomes of mortality, function, cognition, self-assessed health, hospital use or other healthcare use (primary care visits to doctors or nurses). A common pitfall in complex intervention reviews is the ambiguity of included studies; hence Supplement A3 provides a full list of exclusion criteria with examples. Although closely related, this is not a review of interventions targeting only social isolation, physical fitness or falls prevention, which already have considerable literature bases and excellent systematic reviews.^21–27 The search was limited to English studies from the last ten years, enabling the assessment of practices and challenges relevant in the current political and healthcare climate. This review looked exclusively at active interventions, hereby called social vulnerability interventions, rather than broad governmental policies influencing social vulnerability.

Table 1 Search Strategy: Inclusion Criteria, PICOT Format

Data Collection & Extraction

Citations were imported into EndNote³¹ for de-duplication and title and abstract screening. Then full texts were assessed for relevance to inclusion criteria and reasons for exclusions were recorded. Screening and extraction was conducted by the primary author (JM). Relevant data was collected using a piloted data extraction form (Supplement B) to collect information on: (1) general information, (2) study design (3) population (4) intervention details (5) results and outcomes (6) key author conclusions. Methodological quality was assessed using the Mixed Methods Appraisal Tool (MMAT).³²

Data Synthesis

Evidence was synthesized using an integrated and aggregated synthesis approach following the Joanna Briggs Institute recommendations.³³ Due to the heterogeneity of the studies’ interventions, populations and outcomes/measures, a meta-analysis was not possible as few studies reported sufficient data to calculate standardized effect sizes; nonetheless, three methods of quantitative synthesis were used to answer the primary objective: (1) all studies were synthesized using vote counting based on direction of effect,^34,35 (2) for studies including a control group, p values were combined using Fisher’s method to answer the question: “is there evidence of a positive effect on outcome in at least one study” to demonstrate the intervention is beneficial compared to the control,³⁴ and (3) Albatross plots were constructed for studies with a control group and calculable mean differences, which uses 2 sided p values plotted against sample sizes to estimate a magnitude of intervention effect.³⁶ Full details and calculations are described in Table 2. Data extraction and synthesis was conducted by one author (JM). A second reviewer also extracted, assessed risk of bias and performed the quantitative data synthesis on a randomly sampled 25% of the included papers to check that the data extraction and synthesis was consistent (audit results also shown in Supplement D3 and E3). All analyses were conducted with RStudio and Excel.

Table 2 Methods of Quantitative Data Synthesis, organized by descending amount of information provided by synthesis method

Results

Results of Search

Of the 5918 citations retrieved, 567 were duplicates, 4457 were excluded after title and abstract screening, and 174 full text articles were obtained (see Figure 2). This review included a total of 38 studies, representing 34 distinct interventions. 52.9% of studies were conducted in the United States.

Figure 2 Flowchart of study selection.

Description of Interventions

Table 3 (with full details in Supplement C) summarizes the characteristics of each complex intervention and target population, grouped into five categories of reducing social vulnerability. On average, participants were 74.9 years old, 69.6% female and 50.2% lived alone. Social workers and nurses were the most prevalent trained professionals.

Table 3 Characteristics and Population of Interventions Included in Review, by Five Categories of Reducing Social Vulnerability

Lasting between 10 weeks to 3 years, there were 12 weekly or biweekly programs aimed at strengthening the social supports and communities where older adults live. Eight involved group activities in senior centers, community or public spaces^21,38–45 and 4 involved matched volunteers facilitating social activities.^46–49 The interventions varied from organized group activities^{38,39,41–43,45} to sessions promoting health literacy or knowledge of community assets.^21,39,43

Ranging from 4 to 52 weeks, some interventions helped older adults and their caregivers navigate health and social services. Navigators were community health workers or volunteers,^50–52 social workers^53–56 or consisted of multidisciplinary teams.^57–63 All took place at home, but 4 were initiated during or shortly after a hospital admission.^{50,51,54–56} The Community Aging in Place: Advancing Better Living for Elders (CAPABLE) intervention (and its variations) is noteworthy as it is one of the few that has been implemented in multiple states and cities and has a large evidence base (five studies in 6 citations included for this review).^57–62

Six interventions worked to enhance the neighbourhood or home built environment. Half provided social services, personalized care coordination, and environmental hazard reduction to affordable housing buildings.^64–66 Two offered private home modifications for safety^67,68 and 1 provided accommodations for an adult in need by placing them in a home of an approved Shared Lives carer.⁶⁹

Two interventions provided financial assistance to older adults in the form of supplemental income^70,71 or medication co-payment assistance,⁷² both followed for 18 months. Three interventions mainly focused on education: 2 as college programs designed for seniors^73,74 and 1 for nutrition.⁷⁵

Description of Study Designs and Quality Assessment

Across all studies, 35 reported only quantitative outcomes, 2 studies used a mixed methodology and one study reported qualitative outcomes. Of the quantitative studies, 9 were randomized controlled trials (RCTs),^{38–40,51,58,59,61,70,75} 7 were quasi-experimental studies or longitudinal cohort studies with a control group (hereafter called controlled before after (CBA)),^{41,43,47,62,64,67,68} 3 were retrospective cohort studies (RCS),^46,54,65 and 1 study conducted a cross-sectional survey (XCS).⁶⁹ The remaining quantitative studies did not have a control group. The mean and median sample sizes of all quantitative studies were 6908 and 242 participants, respectively, ranging from 12 to 172,965 participants. The largest studies were CBA or RCS in study design. Interviews were conducted with 6, 11 and 17 participants in the mixed methods or qualitative studies.

A summary of the individual MMAT scores per question and summary scores per study design are available in Supplement D. No patterns emerged when cross tabulating intervention category with study design. Overall, the quality of the included studies were poor or moderate, with only four studies matching all criteria and deemed high quality.^47,54,59,69 Poor quality studies were not excluded from synthesis.

Effects of Interventions

Following quality appraisal, data was synthesized according to intervention category and outcome type (individual results in Supplement E). A visual summary of effect direction by vote counting of all included studies has been tabulated in Table 4. Across all studies, there was a positive direction of effect for function, cognition (borderline), subjective health and decreased hospital use. The sign test for the effect on non-hospital physician or nurse visits was 1, signalling interventions were equally likely to result in increased or decreased healthcare use.

Table 4 Summary of Direction of Health Impacts from Included Studies Using Vote Counting

Mortality

Only one study examined mortality as an outcome in relation to a social intervention. The Long Live the Elderly (LLE) study demonstrated a 13% reduction in mortality (p<0.001) during the summer of 2015 in urban areas with LLE compared to areas without the program, with indirect evidence to also suggest LLE attenuated the pre-existing association of higher mortality and lower socioeconomic status.⁴⁶

Function

Of the 10 studies, 9 used a measure for ADLs or IADLs;^{42,43,52,57–59,61,62,67,70} all but two had a positive direction of effect.^52,59 Across 8 studies with a comparison group, the combined one-sided p value was <0.001, suggesting a positive effect on function favouring the intervention in at least one study. The studies reporting outcomes with calculable mean differences are visualized in Figure 3A with small estimated effect sizes (less than a 0.2 point difference in any measure used). The study of highest quality and design in this review examined the CAPABLE intervention. At its primary endpoint of 5 months, there was a medium improvement in the CAPABLE group’s ADL (aOR: 0.70, CI: 0.54–0.93, p=0.01) and IADL scores (aOR: 0.83, CI: 0.65–1.06, p=0.13) but not at 12 months (secondary endpoint).⁵⁹

Figure 3 Albatross plots depicting the association between interventions and five health outcomes (A – function, B – cognition, C – subjective health, D – hospital utilization, E – other healthcare utilization). Effect size contours represent mean differences of 0.2 (small), 0.5 (medium) and 0.8 (large). The left wings represent a negative association, the right wings represent a positive association and a p value of 1.00 represents a null effect.

Cognition

Five out of 6 studies demonstrated benefit on cognition as an outcome of interest.^{41,49,55,70,74} Three of the interventions focused on strengthening social and community participation.^38,41,49 Of the studies with a control group, there was strong evidence of cognitive benefit in at least one study using eight distinct measures of cognition (p<0.001, 3 studies). One RCT reported mixed results with statistically significant benefits on spatial memory only out of 6 cognitive measures.³⁸ Another demonstrated improved verbal recall in the intervention group at 6 and 12 months.⁷⁰ The estimated effect sizes from the two RCTs are presented in Figure 3B.

Subjective Health

Subjective health (self-perceived or self-reported) or health related quality of life was the most commonly evaluated outcome in 18 studies;^{21,39,40,43–45,48,49,52,55,56,60,64,66,69,72,73,75} but was also the outcome with the highest proportion of uncontrolled studies. Notably, at least half of all studies in each intervention category measured subjective health. Combining the p values from the controlled studies suggest we reject the null hypothesis of no benefit in any study (p<0.001, 7 studies). The Albatross plot is depicted in Figure 3C and the estimated average effect size ranges from a small to medium improvement in subjective health.

Hospital Utilization

ED visits or hospitalizations were assessed in 15 studies; 80% reported reduced hospital use.^{39,50,52,54–56,63–66,68,72} Combining p values implies strong evidence (p<0.001, 8 studies) of positive effect on reducing hospital use but Figure 3D suggests any size of effect would be small.

Other Healthcare Utilization

Three studies reported increased primary care provider visits^21,64,70 and two studies reported decreased visits^55,66 although the latter were both uncontrolled before after (UBA) studies of low quality. Within studies, this outcome varied as well; for example, in one study the intervention decreased physician visits but increased nursing visits.³⁹ The calculated combined p value of 4 controlled studies was 0.98 (accepting the null of no benefit of social interventions collectively on primary care provider visits), reflected in Figure 3E.

Discussion

Summary of Results

By counting direction of effect in all studies, the authors show there is a signal for effectiveness on function, subjective health and hospital utilization outcomes greater than chance alone for complex interventions with a main component of the intervention targeting social vulnerability. Then, combining p values of studies with a control group provided strong evidence (p<0.001) that at least one of the interventions improved function, cognition, subjective health and hospital utilization in a positive direction, although the effect sizes estimated with the Albatross plots were small (possible exception of a medium effect size for subjective health). There was an inadequate number of studies evaluating death as an outcome, but the sole study showed a statistically significant reduction in mortality. There is no evidence to suggest the interventions in this review had a consistent impact on primary care utilization. Although a positive effect was coded as a reduction in healthcare utilization, increasing physician and nursing visits can be seen as a good outcome in certain circumstances (ie for older adults in need but with reduced access) and should be considered in the interpretation of results. The complex interventions also demonstrated substantial variations in the program components, settings within the community and targeted older adult population. This review aligns with a recent systematic review and meta-analysis also addressing the gap between social interventions and measuring objective health outcomes, which concluded governmental social policies are an overall promising way to improve population health but noted that studies are often underpowered to detect a health effect, which offers an explanation for the small or null associations found by this review.⁷⁶

Strengths & Limitations

This review is novel in that it has attempted to describe, and synthesize, complex interventions where the common thread is a strong focus on social interventions. This review looks at effectiveness on objective health outcomes, as opposed to the more common health outcomes of self-perceived improvements, lifestyle changes or surrogate markers (ie increased physical activity);¹⁵ albeit, the measured outcomes may not reflect those that are most valued by older adults. By including all types of studies, we are able to determine, in principle, that complex interventions focusing on reducing social vulnerability as a key intervention component does improve health outcomes. However, the choice to lump interventions instead of splitting the complex interventions means that the question of which type of intervention, for which populations, or which components of the intervention are most effective remains unanswerable. For example, Cappelli et al's recent review on social vulnerability suggested the quality of social relationship was a key factor in protecting older adults from functional decline.⁷⁷ Future research could answer this latter question using component network meta-analytic methods as recommended by the Cochrane Collaboration.

The biggest challenge of this review was managing the heterogeneity of the targeted populations, the SDOH addressed, and the measurement of health outcomes; hence synthesis via a meta-analysis was not feasible nor was answering questions of effect size and relative weights of each study according to quality. Generalizing across studies is also problematic given that 12 different countries and 34 distinct complex interventions, each with multiple components, are included in the findings. How much an intervention can impact is influenced by national or regional health and social policies. For example, the North American navigator interventions may not be as effective in Japan where older adults are automatically assessed for community resources under the national long-term care insurance scheme. Since the same intervention will have varied effects in different populations when the distributions of SDOH supports differ, there is a practical benefit of including all study types in this review, not only the ones with the best evidence or the most funding for a robust study design. Even if only RCTs were examined, the gold standard’s emphasis on “average effects” would not solve the problem of generalizability, nor would it help policymakers implement an intervention in their specific context. Said another way, the authors recognize that complexities create less definitive causal inferences which is a limitation of the review, but continue to search for methods to answer these types of questions without artificially fragmenting the duality of medical and social frailty.

The authors also recognize that this is a synthesis of social interventions that targeted reductions in social vulnerabilities through the social determinants. However, this is not the same as assuming the participants of the included studies were at the same level of baseline social vulnerability, which is another study limitation. A review of this nature, looking at interventions for a defined population with known social disadvantages would be a worthy endeavor.

While this review followed rigorous PRISMA-CI methodology, a single reviewer conducted the screening, data extraction and analysis as part of her doctoral studies. Ideally, a double screening and extraction approach would have been carried out to increase internal validity and reliability, and reduce the chance of random error or personal bias. However, since there was significant consultation between the authors to ensure adherence to the pre-established review protocol and to ensure consensus for any difficult screening questions, it is unlikely that double screening would change the overall findings of this paper.

Clinical, Research and Policy Relevance

Investing in social interventions to improve health has strong intuitive appeal, fundamentally based on the assumption that better population health requires interventions for socially vulnerable older adults. The question of how effective SV interventions may be remains difficult to answer with this review providing a small, but certain positive signal of effect among 34 complex interventions with a strong social focus. Furthermore, the findings of this review challenge researchers in this area to develop or revitalise research methodologies that can make cohesive sense of the heterogeneity of these studies, of the pragmatic nature of social interventions and of the wide variety populations. By providing evidence for the collective effectiveness and practicability of such interventions, there are implications for clinicians, policymakers and older adults themselves to implement and advocate for social interventions that will benefit their communities with objective health improvements as a motivator.

Conclusion

Attention to reducing an older adult’s social vulnerability may be helpful in accomplishing improvements to health. Complex interventions with a main component of the intervention targeting social vulnerability showed evidence of positive effects on function, cognition, subjective health and reduced hospital utilization. Moreover, this review demonstrates heterogeneity in intervention type and quality of studies and supports the need for more research in this area.