Multilevel Resilience Resources and Cardiovascular Disease in the United States: A Systematic Review and Meta-Analysis

Abstract

Objective:

This systematic review and meta-analysis aimed to quantify the relationship between resilience resources at the individual (e.g., optimism), interpersonal (e.g., social support), and neighborhood (e.g., social environment) levels, and cardiovascular outcomes among adults in the United States.

Methods:

On 9/25/2020, electronic databases (PubMed, Embase, CINAHL, PsycINFO) were systematically searched for randomized controlled trials, non-randomized intervention studies, and prospective cohort studies that examined the relationship between resilience resources at the individual, interpersonal, or neighborhood level and cardiovascular outcomes. Studies that met the eligibility criteria were summarized narratively and quantitatively. Because relevant search results yielded only observational studies, risk of bias was assessed using an adapted version of the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool.

Results:

From 4,103 unique records, 13 prospective cohort studies with a total of 310,906 participants met the eligibility criteria, and 6 of these studies were included in the meta-analyses. Most relevant studies found that higher levels of individual-level resilience resources were associated with lower incidence of adverse cardiovascular outcomes, with point estimates ranging from 0.46 to 1.18. Interpersonal-level resilience resources (i.e., social network) were associated with a lower coronary heart disease risk (risk ratio, 0.76; 95% CI, 0.56–1.02). Neighborhood-level resilience resources (i.e., perceived social cohesion and residential stability) were associated with a lower odds of stroke (odds ratio, 0.92; 95% CI, 0.84–1.01).

Conclusions:

Evidence suggests that higher levels of resilience resources are associated with better cardiovascular outcomes. However, more prospective studies with diverse populations are needed to strengthen the evidence.

Historically, cardiovascular disease (CVD) has been the leading cause of death in most racial/ethnic groups in the United States (U.S.) (ACC, 2019; CDC, 2017). Given the historical and current contexts of racism and inequality in the U.S., the burden of CVD is unequally distributed, where African American adults have a higher prevalence and incidence of CVD outcomes compared to White adults (Barr, 2016; Havranek et al., 2015). Therefore, the U.S. government has identified cardiovascular health as a high priority with the goal of improving cardiovascular health in all groups and reducing CVD disparities (ODPHP).

A resilience-based approach as one strategy to improve cardiovascular health may be warranted. Resilience has been defined in various ways; however, the general consensus in the literature is that resilience involves a capacity to positively adapt after experiencing an adversity (Fletcher & Sarkar, 2013; Luthar et al., 2000; Masten et al., 1990). This capacity includes utilizing resilience resources to overcome adversities and attenuate the negative impacts of serious threats to health and development (Fletcher & Sarkar, 2013; Luthar et al., 2000; Masten & Obradovic, 2006). In the resilience literature, recent scholarship has shifted towards conceptualizing resilience at multiple levels (Distelberg et al., 2015; Martin et al., 2015). Several common protective factors (e.g., self-efficacy, social support, neighborhood safety) have emerged that may contribute to resilience at the individual, interpersonal, or neighborhood level. In general, these protective factors represent the concept of resilience (Martin et al., 2015) in the sense that individuals who possess higher levels of these factors, henceforth referred to as resilience resources, are thought to be more resilient.

There are several potential mechanisms through which resilience resources may influence cardiovascular health. Resilience resources may have protective effects on CVD through indirect behavioral pathways (e.g., physical activity) that positively influence health (Gallo et al., 2004). Further, the Reserve Capacity Model postulates that reserve capacity (i.e., resilience resources), comprised of individual resources, such as optimism, and interpersonal resources, such as social support, mediates the effects of adversities on cardiovascular health by reducing physiological stress responses, and promoting adaptive coping which, facilitates positive cardiovascular outcomes (Gallo, de Los Monteros, et al., 2009; Gallo & Matthews, 2003). Hence, resilience resources may be a prime target for interventions to improve cardiovascular health.

While resilience research has focused historically on resources at the individual level, resilience also operates at other levels such as the interpersonal and neighborhood levels (Schetter & Dolbier, 2011; Shaw et al., 2016). Studies on individual-level resilience resources, such as positive psychological well-being, have shown these resilience resources to be associated with better cardiovascular outcomes (Boehm & Kubzansky, 2012; DuBois et al., 2015; Sin, 2016). For example, one study in the U.S. reported that optimism was associated with a lower odds of heart failure (odds ratio (OR), 0.52; 95% confidence interval (CI), 0.33–0.81) (Kim, Smith, et al., 2014). Likewise, interpersonal resilience resources, such as social support, have been shown to be associated with better cardiovascular health outcomes, including 50% higher survival from CVD-specific mortality among those with stronger social relationships (Holt-Lunstad et al., 2010). Furthermore, evidence suggests that neighborhood-level resilience resources have been shown to be associated with favorable CVD outcomes, as shown in large observational cohort studies in the U.S. (Barber et al., 2016; Kaiser et al., 2016; Unger et al., 2014; Wang et al., 2017). For example, people with the highest perceived neighborhood social cohesion experienced a lower odds of myocardial infarction (MI) compared to those with the lowest perceived neighborhood social cohesion; but the results were imprecise (OR, 0.69; 95% CI, 0.40–1.20) (Kim, Hawes, et al., 2014). However, evidence has also been reported in studies evaluating the relationship between multiple levels of resilience resources and CVD outcomes that conflicts with the previously mentioned studies. For example, when considering coronary heart disease (CHD) incidence, religiosity (an individual-level resilience resource) has been shown to have a higher association (hazard ratio (HR), 1.18; 95% CI, 0.97–1.43) (Schnall et al., 2010), and social network (an interpersonal-level resilience resource) has been shown to have no association (risk ratio (RR), 0.99; 95% CI, 0.81–1.20) (Eng et al., 2002). Despite some mixed findings, most evidence suggests that resilience resources exist at multiple levels, and may be associated with better cardiovascular outcomes.

In light of promising evidence supporting the association between greater resilience resources at multiple levels (as the exposure) and better cardiovascular health outcomes, there is a need for a summary of the published literature on prospective studies on this topic. Such a summary may help inform public health interventions and practices targeted at building resilience and improving cardiovascular health. Existing systematic reviews on resilience and CVD have investigated only one resilience resource (e.g., optimism) (Cohen et al., 2016; Rasmussen et al., 2009; Rozanski et al., 2019), investigated resilience resources at only one level (individual-level or interpersonal-level) (DuBois et al., 2015; Holt-Lunstad et al., 2010), and/or did not specifically investigate CVD outcomes (Cal et al., 2015; Ghanei Gheshlagh et al., 2016; Kim et al., 2019). This systematic review and meta-analysis will more comprehensively examine the relationship between resilience resources at the individual, interpersonal, and neighborhood levels and CVD outcomes.

Thus the objectives are to: 1) conduct a systematic review of the published literature regarding the relationships between resilience resources at various levels, as the exposures, and CVD outcomes in U.S. adults; 2) conduct a meta-analysis that quantifies these relationships; 3) identify relevant gaps in the literature; and 4) provide recommendations for future research. The focus on U.S. adults is based on the American Heart Association’s (AHA) goal to reduce CVD and improve CVH. In addition, the U.S. government has continued to set national objectives in preventing CVD and improving overall CVH (Healthy People 2030) (ODPHP). Further, the relationship between resilience and CVD may vary across countries perhaps due to differences in the types or extent of adversities faced and cultural meanings regarding resilience resources (Canvin et al., 2009; Chen & Miller, 2012; Gallo, Penedo, et al., 2009; Rosengren et al., 2019; Ruiz et al., 2018; Ungar, 2011).

Methods

Eligibility criteria

English-language, published, peer-reviewed studies that: 1) enrolled participants 18 years of age or older without a history of CVD at baseline who resided in the U.S.; 2) included, as exposure(s), resilience resources at the individual (e.g., coping, optimism), interpersonal (e.g., social support), and/or neighborhood (e.g., social environment) levels; 3) reported, as outcome(s), (i) incidence of any CVD event as a single outcome (e.g., stroke, CHD, fatal/non-fatal MI, congestive heart failure, coronary artery disease) or as a composite outcome (e.g., CVD-specific incidence or mortality, including both stroke and CHD), or (ii) a Life’s Simple 7 (LS7) score for ideal cardiovascular health (measured by physical activity, diet, weight, smoking, blood pressure, cholesterol, and blood glucose); and 4) were randomized controlled trials (RCTs), non-randomized intervention studies, or prospective observational cohort studies were included. Supplemental Table 1 includes a list of resilience resources of interest that was used to develop the search terms and informed the screening process. This list was based on prior research on resilience constructs (Distelberg et al., 2015; Martin et al., 2015) that are consistent with the conceptualization of multilevel resilience in this review.

Search strategy

Four electronic databases (PubMed, Embase, CINAHL, and PsycINFO) were searched from inception to September 25, 2020, as outlined in Supplemental Tables 2–5. To identify RCTs, the Cochrane Highly Sensitive Search Strategy study design filter for PubMed and PsycINFO was used, which was also adapted for Embase. For CINAHL, the RCT study design filter developed by Glanville et al. (2019) was used. To identify prospective observational cohort studies, a cohort study filter for each database was developed that was similar to the RCT filter. The search was limited to English language and human subjects. A health sciences librarian at Brown University with expertise in systematic review searches helped develop the search strategy.

Study selection and data collection

One investigator (JP) performed the initial search and deduplication of records. Two investigators (JP and RM) independently screened the title and abstracts of all records based on the eligibility criteria using Covidence (https://www.covidence.org). Full text of the abstracts that met eligibility criteria were reviewed independently for eligibility by the same two investigators. Disagreements between JP and RM regarding eligibility were resolved by a third investigator (CJH). Articles that met the eligibility criteria based on their full text were narratively summarized.

A data extraction form (in Covidence) was pilot tested using five randomly selected articles. The form included details pertaining to study design, population, exposures, outcomes, and results. Two investigators (JP and RM) performed data extraction independently and resolved disagreements by discussion.

Unit of analysis

The study (and not the article) was considered as the unit of analysis, and, as such, data were extracted separately for each study. If two or more articles had the same or very similar participants, the articles were considered as representing the same study. When the same exposure-outcome relationship (from the same study) was reported in multiple articles, results were extracted for the exposure-outcome relationship from the article with the largest sample size.

Risk of bias

Assessment of risk of bias in individual studies was adapted from the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool (Sterne et al., 2016). The ROBINS-I tool was chosen to assess the risk of bias because the search results yielded only observational studies. Had studies of other designs been found, the appropriate tools to assess bias would have been used, such as the Risk of Bias 2.0 (RoB 2) tool for RCTs (Sterne et al., 2019). The following domains for risk of bias were assessed: (1) bias due to confounding, (2) bias in selection of participants into the study, (3) bias in measurement of the exposure, (4) bias due to missing data, (5) bias in measurement of the outcome, and (6) bias due to selective reporting of results for a particular exposure or outcome among multiple measures of exposure or outcome, or selective reporting of results for a particular subgroup of study participants. In accordance with ROBINS-I, for each study, each domain was rated as: “low risk”, “moderate risk”, “serious risk”, or “critical risk”. For example, a low risk rating for a domain implied that the study was comparable to a well-performed randomized trial with respect to that domain, while a moderate risk implied that the study was comparable to a sound randomized trial, but not a well-performed trial (Sterne et al., 2016). The overall risk of bias for a study was rated based on the individual domains (Sterne et al., 2016). Using the data extracted, two investigators (JP and RM) independently assessed the risk of bias for each domain. There were no disagreements.

Synthesis of Results

The results were summarized both narratively and, when possible, quantitatively. Quantitative findings were grouped first by resilience level and then by outcome. Groupings were determined a priori. Random-effects meta-analyses of the quantitative results from each relevant study were performed using OpenMeta[Analyst] (http://www.cebm.brown.edu/openmeta/). As measures of association, RRs, HRs, or ORs and corresponding 95% CIs for each dichotomous outcome were used. To estimate an overall measure of the relationship between resilience and cardiovascular outcomes, data were pooled across all relevant studies included in the meta-analysis. Statistical heterogeneity among pooled studies was evaluated using the chi-squared test for heterogeneity and the I² statistic. Because this chi-squared test has low power to detect heterogeneity in the context of a small number of studies, a P-value threshold of 0.1 was used to determine statistical significance. An I² threshold of 50% was used to indicate substantial heterogeneity, but only used this threshold in the context of considering the clinical and methodological heterogeneity across the studies (Higgins et al., 2003). A sensitivity analysis was conducted by excluding studies with an overall serious or critical risk of bias from the meta-analysis.

In accordance with the recent statistical and other literature regarding hypothesis testing, study findings based on point estimates and confidence intervals were interpreted in terms of compatibility with the data rather than statistical significance (Amrhein et al., 2019; Wasserstein et al., 2019).

Results

Study selection

Figure 1 depicts the identification of studies included in the narrative summary and meta-analysis. The electronic database search identified 4,103 unique records. After the screening of titles and abstracts, 74 articles were deemed eligible for full-text review, of which, 17 articles met the full inclusion criteria. The 13 studies reported in these 17 articles were included in the narrative summary, of which 6 studies, reported in 8 articles, were included in the meta-analyses.

Figure 1.

PRISMA Flow diagram of the study selection process

Study characteristics

Table 1 summarizes the characteristics of the 13 studies included in the narrative review, all of which were prospective observational cohort studies. A total of 310,906 participants were included across these 13 studies. The sample sizes of participants in these studies ranged from 1,122 (Kubzansky et al., 2011) to 92,395 (Schnall et al., 2010). The ages of participants ranged from 18 (Vogt et al., 1992) to 90 years (Kubzansky et al., 2011); however, several studies enrolled only older adults (ages 55 years or older) (Colantonio et al., 1992; Freedman et al., 2011; Ostir et al., 2001). Most study participants were female; however, some studies enrolled only males (Eng et al., 2002; Kawachi et al., 1996; Kubzansky et al., 2011; Reed et al., 1983). With the exception of 4 studies (Felix et al., 2019; Hussein et al., 2018; Ostir et al., 2001; Reed et al., 1983), participants were overwhelmingly White (≥79%). Participants in most studies were highly educated, with the exception of the Colantonio et al. (1992) study, in which 55% of participants had less than a high school education. Studies followed-up participants for 2 (Freedman et al., 2011) to 16 years (Li et al., 2016), and study baseline data were collected from 1970 (Vogt et al., 1992) to 2006 (Kim, Hawes, et al., 2014; Kim, Park, et al., 2013; Kim, Smith, et al., 2014).

Table 1.

Characteristics of the studies that met the inclusion criteria for the narrative review

Author(s) (Year)	Cohort	Baseline Year(s)	Follow-up (years)	Number of Participants at Baseline	Age of Participants at Baseline	Study Inclusion Criteria	Resilience Resource	Outcomes
Colantonio et al. (1992)	Community-setting in New Haven, Connecticut	1982	7 years	2,812	≥65	Non-institutionalized men and women, 65 years of age and older	Individual: religiosity, religion as source of support. Interpersonal: social network, no. of support, religious service attendance	Stroke
Kawachi et al. (1996) (Eng et al., 2002; Kawachi et al., 1996)	Health Professionals Follow-up Study	1988	4 years	32,624	40–75	Health professionals (men) in 1986 and responded to questionnaires from 1988 to 1992	Interpersonal: social network	CVD-specific mortality, stroke, total CHD
Freedman et al. (2011)	Health and Retirement Study (HRS)	2002	2 years	12,777	≥55	HRS participants aged 55 years or older in 2002 who responded to the 2004 wave	Neighborhood: residential stability	Stroke, heart problems
Vogt et al. (1992) (Hibbard & Pope, 1993; Vogt et al., 1992)	HMO	1970–71	15 years	2,573	≥18	Northwest Kaiser Permanente (Portland, OR) random 5% sample in 1970 who were enrolled for at least 2 years	Interpersonal: social network, marriage satisfaction	IHD, stroke
Hussein et al. (2018)	Multi-Ethnic Study of Atherosclero sis (MESA)	2000–02	12.2 years (median)	5,608	45–84	Adults aged 45–84 years enrolled in the MESA who selfidentified as white, black, Chinese, or Latino and were free of clinically overt CVD at baseline	Neighborhood: social environment. Interpersonal: social support	CVD incidence
Kim et al. (2014) (Kim, Hawes, et al., 2014; Kim, Park, et al., 2013; Kim, Smith, et al., 2014)	HRS	2006	4 years	6,808	≥55	50% of participants who were interviewed face-to-face at study baseline (2006) and followed-up in 2008 and 2010	Individual: optimism. Neighborhood: perceived neighborhood social cohesion	Stroke, MI, heart failure
Kim et al. (2017)	Nurses’ Health Study	2006	6 years	70,021	30–55 in 1976	Female registered nurses who were 30–55 years of age in 1976 and followed-up from 2006 to 2012	Individual: optimism	Heart-disease and stroke-specific mortality
Kubzansky et al. (2011)	Normative Aging Study	1986	12.7 years (mean)	1,122	40–90	Established by the Veterans Administration in 1961; community-dwelling men aged 21 to 80 years in Boston area	Individual: self-regulation	Non-fatal MI, total CHD
Li et al. (2016)	Nurses’ Health Study	1996	16 years	74,534	30–55	Nurses aged 30 to 55 years from across the United States at study baseline (1996) through 2012	Interpersonal: religious service attendance	CVD
Ostir et al. (2001)	North Carolina EPESE study	1986	6 years	2,478	≥65	Non-institutionalized, community-dwelling adults in 1986 and followed-up through 1992	Individual: positive affect	Stroke
Reed et al. (1983)	Honolulu Heart Program	1971	8 years	4,389	45–65	Men of Japanese ancestry born between 1900 and 1919, and living in Oahu in 1965	Interpersonal: social network	Total CHD, MI, angina
Schnall et al. (2010)	WHI Observationa l Study (WHI-OS)	1994–98	7.7 years	92,395	45–65	Women aged 50–79 years at 40 clinical centers across the U.S.	Individual: religion as strength/comfort. Interpersonal: religious affiliation, religious service attendance	CHD
Felix et al. (2019)	WHI-OS and Clinical Trials (WHI-CT)	1994–98	12.5 years (mean)	92,395	50–79	Black women aged 50–79 years at 40 clinical centers across the U.S. and responded to the Extension 2 study (2010–2015)	Individual: modified Brief Resilience Scale	CVD incidence

CHD coronary heart disease; CVD cardiovascular disease; EPESE Established Populations for Epidemiologic Studies of the Elderly; HMO health maintenance organization; HRS Health and Retirement Survey; IHD ischemic heart disease; MI myocardial infarction; MIDUS National Survey of Midlife Development; NSHAP National Social Life, Health, and Aging Project; WHI Women’s Health Initiative

Examined exposures

The included studies examined resilience resources at the individual, interpersonal, or neighborhood level alone, or at two levels. Seven studies assessed individual-level resilience resources, which included religiosity, dispositional optimism, self-regulation, positive affect, or resilience using the modified Brief Resilience Scale. Seven studies assessed interpersonal-level resilience resources, which included social network, social support, religious service attendance, or companionship in marriage (Eng et al., 2002; Hibbard & Pope, 1993; Kawachi et al., 1996; Li et al., 2016; Reed et al., 1983; Vogt et al., 1992). Three studies assessed neighborhood-level resilience resources, which included residential stability or social environment (social cohesion and neighborhood safety) (Freedman et al., 2011; Hussein et al., 2018; Kim, Hawes, et al., 2014; Kim, Park, et al., 2013). Four studies assessed resilience resources at two levels; 2 studies assessed individual and interpersonal-level resilience resources (Colantonio et al., 1992; Schnall et al., 2010), 1 study assessed interpersonal and neighborhood-level resilience resources (Hussein et al., 2018 ), while 1 study assessed individual and neighborhood-level resilience resources (Kim, Hawes, et al., 2014; Kim, Park, et al., 2013; Kim, Smith, et al., 2014).

Reported outcomes

Included studies reported stroke, CHD, ischemic heart disease, MI, incident CVD, CVD-specific mortality, heart failure, and angina. No studies reported an LS7 score for ideal cardiovascular health as an outcome. The most frequently reported outcomes were CHD (8 studies), stroke (7 studies), and composite outcomes (i.e., CVD incidence or mortality; 4 studies). Across studies, there was a high level of consistency with regard to how CVD incidence and mortality were measured. The incidence of CVD outcomes was ascertained most often via self-reports from participants or close contacts and/or medical record reviews (International Classification of Diseases-9 [ICD-9] code classifications). For medical record reviews, CVD was classified via the following ICD-9 codes that indicated a CVD diagnosis or outcome of interest: 410–414 (CHD), 798 (CHD), and 430–438 (stroke).

Risk of bias

Table 2 summarizes the risk of bias assessment results for all 13 studies. In terms of bias in the measurement of outcomes and bias due to selective reporting of results, all 13 studies were rated at low risk. In terms of bias due to missing data, 3 studies were rated at moderate risk (Eng et al., 2002; Freedman et al., 2011; Kawachi et al., 1996; Ostir et al., 2001; Reed et al., 1983), and all others at low risk. In terms of bias in the measurement of exposure, only 1 study was rated at moderate risk (Felix et al., 2019), while others were rated low risk. In terms of selection bias, 3 studies were rated at low risk (Hibbard & Pope, 1993; Kim, Hawes, et al., 2014; Kim, Park, et al., 2013; Kim, Smith, et al., 2014; Schnall et al., 2010; Vogt et al., 1992), 9 at moderate risk (Colantonio et al., 1992; Eng et al., 2002; Felix et al., 2019; Freedman et al., 2011; Hussein et al., 2018; Kawachi et al., 1996; Kim et al., 2017; Kubzansky et al., 2011; Li et al., 2016; Reed et al., 1983), and 1 at serious risk, which was due to the study sample representing less than two-thirds of the total eligible study population (Ostir et al., 2001). Most studies (11 of 13) were rated at moderate risk of bias due to confounding; among the non-moderate risk studies, 1 had low risk (Li et al., 2016) and 1 had critical risk, which was due to only unadjusted estimates being reported in the study (Colantonio et al., 1992). As guided by the ROBINS-I tool, most included studies were categorized as having at least a moderate risk of bias due to confounding based on the criteria that all studies were observational cohort studies and were expected to have residual confounding, even after accounting for known confounders.

Table 2.

Risk of bias assessments

Author (Year)	Bias due to confounding	Bias in selection of participants	Bias in measurement of exposure	Bias due to missing data	Bias in measurement of outcomes	Bias due to selective reporting	Overall risk of bias
Colantonio et al. (1992)	Critical	Moderate	Low	Low	Low	Low	Critical
Kawachi et al. (1996) (Eng et al., 2002; Kawachi et al., 1996)	Moderate	Moderate	Low	Moderate	Low	Low	Moderate
Freedman et al. (2011)	Moderate	Moderate	Low	Low	Low	Low	Moderate
Vogt et al. (1992) (Hibbard & Pope, 1993; Vogt et al., 1992)	Moderate	Low	Low	Low	Low	Low	Moderate
Hussein et al. (2018)	Moderate	Moderate	Low	Low	Low	Low	Moderate
Kim et al. (2014) (Kim, Hawes, et al., 2014; Kim, Park, et al., 2013; Kim, Smith, et al., 2014)	Moderate	Low	Low	Low	Low	Low	Moderate
Kim et al. (2017)	Moderate	Moderate	Low	Low	Low	Low	Moderate
Kubzansky et al. (2011)	Moderate	Moderate	Low	Low	Low	Low	Moderate
Li et al. (2016)	Low	Moderate	Low	Low	Low	Low	Moderate
Ostir et al. (2001)	Moderate	Serious	Low	Moderate	Low	Low	Serious
Reed et al. (1983)	Moderate	Moderate	Low	Moderate	Low	Low	Moderate
Schnall et al. (2010)	Moderate	Low	Low	Low	Low	Low	Moderate
Felix et al. (2019)	Moderate	Moderate	Moderate	Low	Low	Low	Moderate

Table 2 also provides the overall risk of bias assigned to each study. Most studies (11 of 13) were rated at moderate overall risk of bias, 1 at serious overall risk of bias, and 1 at critical overall risk of bias, which was mainly due to confounding (i.e., unadjusted estimates) and selection bias.

Narrative summary of results

Of the 7 studies that examined individual-level resilience resources (Colantonio et al., 1992; Felix et al., 2019; Kim et al., 2017; Kim, Smith, et al., 2014; Kubzansky et al., 2011; Ostir et al., 2001; Schnall et al., 2010), 4 studies found that higher levels of individual resilience resources were associated with lower incidences of adverse cardiovascular outcomes (i.e., stroke and CHD). The remaining 3 studies were inconsistent; 2 reported no association with stroke (RR, 0.99; 95% CI, 0.82–1.20) or incident CVD (RR, 1.05; 95% CI, 0.70–1.59), while the other reported a higher hazard of CHD (HR, 1.18; 95% CI, 0.97–1.43) (Colantonio et al., 1992; Felix et al., 2019; Schnall et al., 2010).

Seven studies examined resilience resources at the interpersonal-level (Colantonio et al., 1992; Eng et al., 2002; Hibbard & Pope, 1993; Hussein et al., 2018; Kawachi et al., 1996; Li et al., 2016; Reed et al., 1983; Schnall et al., 2010; Vogt et al., 1992). One study reported lower risk of CVD-specific mortality in participants with a higher interpersonal-level resilience resource (i.e., religious service attendance) (Li et al., 2016). Two studies provided evidence that higher interpersonal resilience resources (i.e., social support and religious service attendance), was compatible with either lower or higher risk of CHD (Hussein et al., 2018; Schnall et al., 2010). Four studies examined social networks as an interpersonal resilience resource, where 2 studies showed a lower incidence of adverse cardiovascular health outcomes (i.e., CHD and stroke incidence) among participants with larger social networks (Colantonio et al., 1992; Eng et al., 2002; Kawachi et al., 1996). One study found no association (Reed et al., 1983) while another study indicated lower incidence of CHD with larger social networks (Hibbard & Pope, 1993; Vogt et al., 1992). However, this last study (Hibbard & Pope, 1993; Vogt et al., 1992) also reported findings that were compatible with higher incidence of stroke, captured by the wide confidence interval.

Among the 3 studies that assessed the relationship between resilience resources at the neighborhood level and CVD outcomes (Freedman et al., 2011; Hussein et al., 2018; Kim, Hawes, et al., 2014; Kim, Park, et al., 2013), 1 study found that a higher neighborhood-level resilience resource was associated with a lower likelihood of CHD and stroke (Kim, Hawes, et al., 2014; Kim, Park, et al., 2013), while 2 studies reported that the data were compatible with a small reduced incidence of CVD (Freedman et al., 2011; Hussein et al., 2018).

Quantitative summary of results

Individual-level resilience.

Studies that assessed individual-level resilience resources, comparing participants with the highest level of resilience resources to the lowest reported point estimates (i.e., RR, HR, or OR) that ranged from 0.46 (95% CI, 0.26–0.82) (Kubzansky et al., 2011) to 1.18 (95% CI, 0.97–1.43) (Schnall et al., 2010) for CHD outcomes, and 0.61 (95% CI, 0.43–0.85) (Kim et al., 2017) to 0.99 (95% CI, 0.82–1.20) (Colantonio et al., 1992) for stroke outcomes. Because of substantial heterogeneity among the studies assessing the relationships between individual-level resilience resources and stroke (I² = 59.13%, p = 0.09; 3 studies) and CHD (I² = 85.03, p < 0.01; 3 studies (Kim, Smith, et al., 2014; Kubzansky et al., 2011; Schnall et al., 2010)), an overall measure of the relationship between individual-level resilience resources and any CVD outcome was not reported.

Interpersonal-level resilience.

Among the studies that assessed the relationship between interpersonal-level resilience resources and stroke, point estimates comparing participants with the highest level of resilience resources to those with the lowest level ranged from 0.50 (95% CI, 0.25–1.00) (Kawachi et al., 1996) to 1.00 (95% CI, 0.59–1.67) (Vogt et al., 1992); with CHD as an outcome, point estimates ranged from 0.67 (95% CI, 0.43–1.00) (Vogt et al., 1992) to 1.08 (95% CI, 0.91–1.29) (Schnall et al., 2010). Note that these ranges include all estimates reported in the studies, not just the estimates that were chosen for the meta-analysis.

In the meta-analyses, having a larger social network as an interpersonal-level resilience resource was associated with lower incidences of stroke (RR, 0.77; 95% CI, 0.57–1.04; 3 studies (Colantonio et al., 1992; Kawachi et al., 1996; Vogt et al., 1992); Figure 2) and CHD (RR, 0.76; 95% CI, 0.56–1.02; 2 studies (Kawachi et al., 1996; Vogt et al., 1992); Figure 2). The studies included in the above meta-analyses of interpersonal-level resilience resources and CVD outcomes reported findings in HRs and RRs. Further, higher interpersonal-level resilience resources were associated with a lower risk of the composite outcome of CVD-specific mortality (defined as stroke and CHD-related deaths) (RR, 0.73; 95% CI, 0.63–0.84; 2 studies (Eng et al., 2002; Li et al., 2016); Figure 3).

Figure 2.

Forest plots of associations between social network (interpersonal-level resilience resource) and incidence of stroke (a) and coronary heart disease (b) from random-effects meta-analyses. Abbreviations: NR, not reported; CI, confidence interval.

Figure 3.

Forest plot of interpersonal-level resilience and cardiovascular disease mortality (composite outcome defined as stroke and coronary heart disease-related deaths), from a random-effects meta-analysis. Abbreviation: CI, confidence interval.

Neighborhood-level resilience.

In studies of the relationship between neighborhood-level resilience resources and stroke, point estimates ranged from 0.87 (95% CI, 0.77–0.99) to 1.02 (95% CI, 0.82–1.27) (Freedman et al., 2011), while point estimates for CHD ranged from 0.82 (95% CI, 0.66–1.02) (Kim, Hawes, et al., 2014) to 0.99 (95% CI, 0.88–1.12) (Freedman et al., 2011).

In the meta-analyses, increased neighborhood-level resilience resources were associated with lower odds of stroke (OR, 0.92; 95% CI, 0.84–1.01; 3 studies (Freedman et al., 2011; Kim, Park, et al., 2013); Figure 4) and CHD (OR, 0.95; 95% CI, 0.87–1.04; 3 studies (Freedman et al., 2011; Kim, Hawes, et al., 2014); Figure 4).

Figure 4.

Forest plots of associations between neighborhood-level resilience resources and incidence of stroke (a) and coronary heart disease (b) from random-effects meta-analyses. Abbreviations: NR, not reported; CI, confidence interval.

Sensitivity analysis

A sensitivity analysis that excluded 1 study (Colantonio et al., 1992) with an overall serious or critical risk of bias was conducted for the interpersonal-level resilience resources (i.e., larger social networks) and lower incidence of stroke (RR, 0.77; 95% CI, 0.60–0.99; 3 studies (Colantonio et al., 1992; Kawachi et al., 1996; Vogt et al., 1992); Figure 2). Upon exclusion of this study the point estimate changed slightly, but the confidence interval was wider (RR, 0.73; 95% CI, 0.37–1.46; I² = 60.2%, p = 0.11; Figure not shown).

Discussion

Overall summary of evidence

This study systematically reviewed and provided quantitative summaries of the published literature examining the association between multilevel resilience resources and CVD outcomes among adults in the U.S. The findings suggest an inverse relationship between resilience resources and adverse cardiovascular health outcomes, especially CHD and stroke. These results were generally consistent across individual, interpersonal, and neighborhood resilience resource levels. The studies identified were generally at a moderate overall risk of bias. The findings from this study support a component of the Reserve Capacity Model where resilience resources have potential protective effects on physical health outcomes such as CVD whereby higher levels of resources (i.e., reserve capacity) may contribute to reduced adverse CVD outcomes (Gallo & Matthews, 2003).

Individual-level resilience.

Although an overall summary measure for the relationship between individual-level resilience and CVD was not reported due to the substantial heterogeneity observed across the included studies, the narrative summary and quantitative findings generally suggested an inverse association between higher levels of resilience resources at the individual level and adverse cardiovascular health outcomes. This finding supports past reviews of studies in both the U.S. and other countries (Cohen et al., 2016; DuBois et al., 2015; Rozanski et al., 2019) that have concluded that individual-level resilience resources are associated with improved CVD outcomes. However, this review builds on this prior work by providing a summary of the relationships between various individual-level resilience resources and CVH with a focus on prospective studies among U.S. adults.

Interpersonal-level resilience.

The narrative summary and quantitative findings for interpersonal-level resilience may indicate a protective effect of resilience on CVD outcomes. Evidence from the meta-analysis suggested that having a larger social network (measured by social network index) was found to be associated with a reduced risk of stroke, CHD, and CVD-specific mortality. However, the association between interpersonal-level resilience and stroke was stronger but less precise after the exclusion of a study with an overall serious or critical risk of bias. The findings from this review are similar to past systematic reviews that reported reduced CHD mortality (Holt-Lunstad et al., 2010) and incidence (Barth et al., 2010) among participants with higher interpersonal-level resilience resources, such as social support.

Neighborhood-level resilience.

There was some evidence for an inverse relationship between neighborhood-level resilience resources and adverse cardiovascular health outcomes. Evidence from the meta-analyses suggested a small protective effect of higher neighborhood-level resilience resources (measured by perceived neighborhood social cohesion and residential stability) on stroke and CHD. To our knowledge, no past systematic reviews have assessed neighborhood-level resilience resources and cardiovascular health outcomes.

As mentioned earlier, the findings in this study are consistent with one component of the Reserve Capacity Model proposed by Gallo and Matthews (2003). However, the model also posits that resilience resources may attenuate the negative impacts of adversities, such as stressors, on physical health. Although this systematic review and meta-analysis did not focus on a specific adversity, the resilience resources that met the eligibility criteria did include several protective factors that are believed to reflect the multilevel conceptualization of resilience (e.g., self-efficacy, social support, and neighborhood cohesion). Additional research should explore adversities in relation to these resilience resources to be consistent with the Reserve Capacity Model framework, and to further support the evidence reported in this review that in the face of negative life events, resilience resources have protective effects on adverse CVD outcomes.

Limitations and strengths

There are several limitations to the current systematic review and meta-analysis. First, it is possible that some eligible studies were missed during the search process. The search terms may not have included all relevant resilience resources that reduce the risk of adverse cardiovascular outcomes. However, the search strategy employed in this study was based on a summary of common resilience resource concepts (Distelberg et al., 2015; Martin et al., 2015). Second, it is possible that some eligible studies were missed during the screening process. During the abstract screening, abstracts were required to have mentioned cardiovascular outcomes. It is possible that studies that reported the association between resilience resources and cardiovascular outcomes as part of a larger set of outcomes in the full text, but did not mention those outcomes in the abstract, were missed. Third, the results from this systematic review and meta-analysis are limited by the quality of the included studies. The findings from the studies included in this review were from secondary data analyses of observational studies thus resulting in potential threats to internal validity such as residual confounding and selection bias.

Despite these limitations, this systematic review and meta-analysis has several strengths. First, the systematic review methodology adheres to contemporary standards for literature retrieval, screening, data extraction, risk of bias assessment, and qualitative and quantitative syntheses (AHRQ; Higgins et al., 2019; Higgins et al., October, 2019; IOM, 2011). Second, the reporting of this systematic review and meta-analysis adheres to the PRISMA Reporting Guidelines (Liberati et al., 2009; Moher et al., 2009). Third, this review includes only prospective studies to reduce various challenges that may arise from cross-sectional or case-control studies, such as reverse causation and recall bias. Finally, to our knowledge, this paper provides the most up-to-date and comprehensive review of the published scientific evidence regarding the relationship between multilevel resilience resources and CVD outcomes among U.S. adults.

Gaps in the literature and suggestions for future research

Based on this systematic review and meta-analysis, additional research on neighborhood-level and multilevel resilience resources are needed. Generally, more prospective studies including observational cohorts and RCTs conducted at all levels of resilience are needed, as shown by the limited number of studies that met the eligibility criteria for this systematic review (i.e., 13 in total). More specifically, our review only identified 3 studies that examined neighborhood-level resilience and 4 that assessed two levels of resilience, compared with the 7 studies that assessed individual-level resilience and the 7 that assessed interpersonal-level resilience.

Although there were 4 studies that assessed two levels of resilience, these studies did not assess the joint effect of multiple sources of resilience. Instead, all studies included in this review that examined multiple levels assessed one resilience resource at a time, which precluded reporting findings of simultaneously possessing resources at multiple levels or multiple resources within a level. Once stronger evidence is provided that suggests that possessing resilience resources is inversely associated with the occurrence of CV outcomes, future research should estimate the joint effects of resilience resources that were identified as being protective on relevant CV outcomes to further inform intervention development.

In terms of CVD outcomes, no prospective studies that examined the relationship between resilience resources and LS7 scores for cardiovascular health were identified. Existing evidence concerning resilience resources and LS7 scores are from cross-sectional studies in well-established cardiovascular health cohorts. These cross-sectional studies have shown a positive relationship between higher resilience resources and ideal cardiovascular health (Hernandez et al., 2018; Hernandez et al., 2015; Sims et al., 2019). However, further research on resilience resources and LS7 scores for ideal cardiovascular health in prospective studies is needed. As suggested by the Reserve Capacity Model, resilience resources may have positive effects on CVD morbidity and mortality through the biobehavioral pathway that includes behavioral mechanisms, such as physical activity and diet, and physiological/metabolic mechanisms, such as obesity (Gallo & Matthews, 2003). Thus, LS7 components are intermediates to CVD that could be amenable to potential resilience-based interventions.

To allow for better comparisons across different studies that examine the same resilience resources, future research should use standardized measures of resilience, or clinically meaningful thresholds for scales that measure resilience constructs, such as optimism and social networks (Kawachi et al., 1996; Kim, Smith, et al., 2014). Further, standardized measures should minimize the use of quartile cut-points given that quartiles present challenges for comparison across studies because resilience score distributions for each quartile will most likely differ across studies. These differences across studies are likely due to differences in participant characteristics. Moreover, standardization of resilience resource measures should be done across measures that have the same cultural meaning. Lastly, standardizing resilience resource measures could better facilitate harmonizing data from multiple cohort studies and conducting meta-analyses.

Finally, participants in the summarized studies were overwhelmingly White, and most had more than a high school-level education. Prior work (Canvin et al., 2009; Chen & Miller, 2012; Gallo, Penedo, et al., 2009; Ruiz et al., 2018; Ungar, 2011) suggests that the relationship between resilience and CVD outcomes may differ by race/ethnicity and socioeconomic position (SEP) due to differences in experiences of adversities (e.g., stressors) (Gallo, 2009; Gallo & Matthews, 2003; Williams et al., 2016). Specifically, some racial/ethnic minority subgroups and individuals living in areas of lower SEP experience more stressors (Clark et al., 2013; Kershaw et al., 2015; Sims et al., 2012). If resilience resources are activated in the face of adversity or more beneficial with greater adversity experienced, we would expect resilience resources to have a greater impact among these subgroups with greater experiences of stressors. Therefore, the findings from this systematic review and meta-analysis may not be generalizable to some racial/ethnic minorities or lower SEP groups. Additionally, while a few studies included participants comprised entirely of males, most participants included in the studies were female.

To be able to explore the impact of SEP, race/ethnicity, and gender, future research should include socioeconomically and racially/ethnically diverse populations as well as more diverse gender groups, and report data for each of these subgroups. The inclusion of more diverse populations should provide additional evidence regarding whether resilience resources are potential intervention targets for improving CVH and eliminating CVD-related health disparities.

Conclusion

This systematic review and meta-analysis summarizes and contributes to the growing evidence suggesting the possible protective effect of higher multilevel resilience resources on cardiovascular health outcomes. The results from this narrative and quantitative syntheses are promising. However, more prospective studies focusing on multiple levels of resilience are needed to inform public health interventions that might target modifiable resilience resources so as to reduce the occurrence of adverse cardiovascular outcomes and promote cardiovascular health.