Do multiple community-based interventions on health promotion tackle health inequalities?

We conducted a systematic review according to the PRISMA-Equity 2012 extension guidelines for systematic reviews with a focus on health equity [13]. An additional checklist shows this in more detail [see Additional File 1].

The search was limited to articles published in English and German during the period January 1, 1999, to August 31, 2019. The reason for choosing this period of time is that earlier reviews mentioned above [9, 10] are focussed on specific topics and do not contain the more recent studies and newer health promotion strategies; in addition, the present review should complement our prior review on the effectiveness of community-based health promotion interventions in the last 20 years [14]. We searched PubMed and PsycINFO databases (advanced search: title/abstract) using the string “(health promot* OR disease prevention OR intervention*) AND (neighbo$rhood OR communit* OR area* OR district* OR ward* OR urban OR rural) AND (social determinant* OR occupation* OR education* OR socio* status OR income OR SES OR SEP OR social status OR equalit* OR inequalit* OR equit* OR inequit* OR disparit*) AND (effect* OR benefit* OR health outcome* OR impact* OR influence*) AND (randomi$ed OR trial OR quasi-experiment* OR pretest OR posttest OR pre-post OR time series OR controlled stud* OR before and after OR trend OR longitudinal) NOT (clinical OR review OR study protocol)”. PubMed search resulted in 2563 hits, while PsycINFO came to 554 entries; after duplicates were removed 2876 records remained (see Fig. 1). In addition, all included primary studies from our previous review on the effectiveness of community-based interventions were searched [14].

We included primary studies which evaluated the effectiveness of complex community-based health promotion interventions in high, middle and low income countries on any health outcome, and which reported differences in intervention effects between SES groups and/or socially deprived populations overall. Table 1 shows all inclusion and exclusion criteria to select eligible studies.

Studies included in the review should refer to the term “community” as a geographically defined area (e.g. neighbourhood, city, village), but not as a whole state or country. Further inclusion criteria were: peer-reviewed articles, intervention studies, assessment of one or more SES factors as well as multi-component strategies of intervention. The latter means that at least two different intervention types were provided, which also reflect social-ecological approaches of health promotion and prevention [7]. This includes changes based on the following six broad types of intervention: 1) social marketing (e.g. campaigns using mass media), 2) individual and group education (e.g. classroom instruction), 3) networking/partnership (e.g. advocacy groups), 4) environmental changes (e.g. greenspaces, availability of healthy food), 5) regulatory interventions (e.g. smoking policies), and 6) improving “sense of community” (e.g. neighborhood parties) [9, 15]. In the literature, these interventions are understood to be interacting components that address different dimensions of complexity and mediating effects, e.g. regarding the outcome, target groups or stakeholders [6].

Screening followed a two-step process with articles filtered by title/abstract, and full text. First author of this review (SN) initially screened the potentially relevant studies. The second author (OK) independently reviewed articles retained for inclusion in the preliminary phase. Full texts of the remaining 87 articles were read and checked by both authors for eligibility, of which 67 were excluded (see Fig. 1). If there were different views, a third colleague was asked to review the article in question (please see Acknowledgement), and a consensus was reached between the authors. Main reasons for exclusion were a missing consideration of SES, single intervention strategies, non-interventional study, and the focus on a non-community setting (e.g. school, workplace). Finally, we extracted data on characteristics of 23 included studies: author and publication date, study design, risk of bias, intervention types, outcome types, and the main findings on health inequalities.

Due to the heterogeneity of the included studies regarding methodological aspects (e.g. target populations, measurement instruments, statistical methods) as well as of the interventions and outcomes no meta-analysis was conducted. However, to identify trends and provide summary statements on intervention-generated inequalities, simple assessments were made for three possible findings by SES: increased inequalities, reduced inequalities, and no difference by SES. Ambivalent results include studies where there was an overall benefit for a disadvantaged population but no effect on health-related outcomes for any SES sub-group within population [8].

The twenty-three studies included in this review were conducted in high or upper-middle income countries (using the World Bank classification). Of these, 14 were set in Europe [16‐29], five in North America [30‐34], three in Australia/Oceania [35‐37], and one study took place in a country in Central America [38] (see Table 2). The studies used various designs, including RCTs (n = 1) [38], cluster randomized trials (n = 3) [25, 30, 34], quasi-experimental designs (n = 11) [16, 18, 22, 23, 27‐29, 33, 35‐37], pre-post-tests without control group (n = 3) [17, 19, 31], and secondary analyses (n = 5) [20, 21, 23, 26, 32]. The sample sizes at baseline varied from 200 to nearly 250,000 in 60 intervention areas, with study populations having a variety of socio-demographic and socio-economic backgrounds. Few studies took place in what can be considered rural areas, while the majority of studies was located in urban neighborhoods, districts or cities.

With regard to the six types of interventions mentioned above, we found notable differences in the number and combination of these strategies. Sixteen articles explicitly emphasized the component of networking and partnership with local organizations (e.g. sport clubs) and volunteers. Other strategies were rather traditionally shaped, including some types of individual or group education (18 studies), social marketing (11 studies) and/or strategies to promote the “sense of community” (10 studies). Many studies used environmental change or regulatory strategies in specific settings (20 and 8 studies, respectively). Only one of the programs contained elements of all six strategies [16]. Six programs comprised five strategies [22, 23, 25‐27, 32], five consisted of four [21, 23, 25, 28, 37], six of three [17, 19, 29, 33, 34, 36] and five of two strategies [18, 20, 28, 30, 36].

Twelve studies aimed at the improvement of health behavior (e.g. physical activity, F&V intake, and smoking), eleven studies examined self-reported mental and physical health, and five studies additionally examined anthropometric outcomes.

Included studies were assessed for risk of bias using the “Quality Assessment Tool for Quantitative Studies”, developed by the Effective Public Health Practice Project (EPHPP) [39]. Studies were scored against six criteria (selection bias, study design, confounders, blinding, data collection method, withdrawals and drop-outs), and the number of weak ratings was summed up to give a global quality score. Of the 23 studies reviewed, ten studies were found to be of strong quality (43.8%) [23, 25, 26, 27, 28, 30, 31, 34, 36 39]. Nine studies (39.1%) were moderate in quality [16, 17, 19‐21, 23, 32, 34, 37], and four studies (17.4%) were weak in quality [18, 28, 31, 36]. 52.2% of all studies showed poor ratings or could not be evaluated regarding withdrawals and drop-outs. An additional file shows this in more detail [see Additional File 2].

In the data extraction, we explicitly aimed to identify studies which were carried out in socially disadvantaged communities and/or conducted analyses of outcome measures by subgroups of SES such as income, education, and/or occupation. Out of the 23 studies that met our inclusion criteria, 13 (56.5%) reported reductions of health inequalities in the entire (deprived) neighborhood or community [16, 17, 19‐21, 25, 28‐30, 32, 34, 35, 38]. However, this included one study in which there was little impact on health equity, but even more negative effects [16]. In two further studies (8.7%), despite the lack of evidence at the level of the entire community, beneficial effects could be found in persons who belonged to the most disadvantaged subgroup within the area [27] or who were particularly exposed to the intervention [37].

Among the remaining eight (34.8%) studies, no differences were found between intervention and control areas [18, 23‐26, 32, 35]. Only in one differential sub-analysis a small intervention effect was found among those with a low educational level for energy intake and walking/bicycling [40]. No other studies except the two mentioned had analyzed the outcomes by socio-economic subgroups within an area to search for a specific social gradient in health. These results seem to be rather sobering and inconsistent, involving both large-scale (e.g. [25]) or smaller (e.g. [23]) programs with a very diverse mix of intervention strategies and outcome parameters. An important point in this context, however, is that there was, with one exception [16], no evidence that community-based programs included in our review resulted in any widening of health disparities (“intervention-generated inequalities” or IGIs) at community-wide level.

There are a number of limitations concerning the present review. Initial searches of databases identified several thousand references, but the small number of eligible studies suggests that few health outcome evaluations of complex community-based interventions have been published in peer-reviewed journals in the last 20 years. Thus, the review described here is possibly not exhaustive and does not cover studies from low income countries. Our search strategy may not have revealed a complete list of all studies describing intervention effects by SES due to limitations of the Pubmed and PsycINFO databases. Single interventions and/or sub-settings (e.g. school, kindergarten) were excluded. Finally, the vast majority of studies identified targeted on effects at a low-SES population level, and did not explore differential effects on inequalities by SES sub-groups.