Introduction

Many governments currently embrace the notion that public health policy should address not only health itself but also its social determinants. For instance, educational attainment (or a lack thereof) is one important social determinant of health (SDH) and a useful indicator for socioeconomic status (SES), which also encompasses income, wealth, power, social class, etc. [1]. This may be explained by factors such as higher incomes among educated people, as well as higher levels of general literacy, which allow more educated individuals to make better-informed health-related decisions [2].

Positive general trends over time in major social determinants of health (e.g., increasing income and educational attainment) have been associated with increasing inequality in health between different socioeconomic groups [1]. This is also true for a welfare state like Sweden, where life expectancy has been increasing for people of all educational levels. However, this increase has lately been more prominent for people with high education, whereas women with only compulsory school (i.e., 9 years of education) have experienced the least improvements [3].

Usually, two strategies are often distinguished when discussing public health policy interventions that address single determinants of health in the context of reducing health inequalities: targeted interventions vs. population-wide interventions. Both strategies could aim at closing the health gap between worse-off groups and the rest of the population, but only population-wide interventions can address the health gradient across the whole population [1]. Whether a marginalized group or the whole population is targeted depends partly on the interpretation of the underlying mechanisms of social inequalities in health. The population-wide approach is usually preferred if the mechanisms are perceived to affect all social strata. However, if the mechanisms are primarily perceived to affect individuals at the low end of the socioeconomic spectrum, targeted interventions are preferred [4].

Population-wide interventions targeting individual behavior, however, may have no impact on health inequalities and could even risk exacerbating them, since their impact is often more prominent among those who have more resources, such as higher levels of education [5]. For example, many effective interventions for smoking cessation are more readily adopted by individuals with high educational levels, thus leading to greater health inequalities caused by smoking [6–7]. In contrast, targeted interventions by definition have an impact among less privileged or vulnerable groups exclusively [8–9], but targeting individuals with low educational levels might be counterproductive due to perceived stigmatizing effects or paternalistic connotations regarding this approach.

For that reason, it is not sufficient for policymakers who wish to reduce health inequalities to simply assume that they will accomplish this goal if they only have information that an intervention is effective overall, without considering the possible differential effects regarding education level. However, there are few published examples of meta-analyses or reviews of empirical studies regarding the differential effects of different intervention strategies to reduce health inequalities by addressing individual health behaviors [10]. Our study was performed to contribute to help fill in this knowledge gap.

In Sweden, there has been a public health debate about the most efficient way to reduce health inequalities between educational groups. This review was conducted for the Public Health Agency of Sweden to assess the magnitude of evidence regarding intervention evaluations with high-quality designs concerning health-related behavior, which have shown a higher impact among individuals with a low educational level, as well as the potential of reducing health inequality.

Methods

Rapid reviews use streamlined traditional methods for systematic review to help synthesize and communicate evidence within a shortened time frame [11]. The review was completed within four months and was conducted according to the PRISMA guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analysis [12]).

Study selection

The database searches yielded 1,929 articles. After duplicates were removed, 1,365 articles remained. One of the authors (AV) then screened all article titles and abstracts using inclusion criteria to determine which studies were relevant. Eligible for inclusion were: (1) studies with evaluations of non-healthcare-based interventions regarding health-related behavioral factors among different educational groups. Further, these should have been published in English in international peer-reviewed scientific journals between 1990 and 2015 from countries with developed welfare systems (i.e., from countries in Europe, North America, Australia, and New Zealand); (2) studies comparing those receiving the intervention with a control group and thereby applying one of the following study designs: randomized controlled trials or non-randomized trials with a cohort design (in the case of community interventions, a control area was required); (3) studies where the intervention included at least 100 individuals.

The main reason for exclusion was that the intervention did not measure educational status (n = 812). Studies were also excluded for not having original data (n = 169) or for having outcomes (individual health-related behavior) that did not conform to our study aim (n = 160). Some studies did not contain an intervention (n = 94), and there were studies included in the search terms that were unrelated to the study objective (n = 99).

Thus, of the 1,365 articles, 1,334 were excluded in the first screening process, which generated 31 articles. The full text of these articles were assessed independently by both authors (AV and P-OÖ) while applying the inclusion criteria with more detailed information, as well as quality criteria according to guidelines used by the Swedish Agency for Health and Technology Assessment and Assessment of Social Service, a public national agency in charge of providing impartial and scientifically reliable information to decision makers and health care providers [13]. The following aspects of quality were assessed: (i) confirmation of appropriate study design, (ii) possible selection bias due to sample recruitment procedures, (iii) use of validated measures regarding both exposure and outcome assessment, and (iv) reasonable control for confounding factors, such as age and gender imbalance between intervention and non-intervention groups. Differences in the primary judgment between the authors was resolved by discussion and consensus. Of the 31 articles left at this stage of the analysis, 22 were excluded for lacking an adequate measure of educational status (n = 9) or proper control group (n = 8). Five articles were excluded for lacking an appropriate study design, leaving nine articles for the final analysis (Fig 1 and S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart of the literature search and screening process.

https://doi.org/10.1371/journal.pone.0195774.g001

We concluded that we could not perform a formal meta-analysis because of different study designs, study settings, and quality of analysis. Furthermore, outcomes were not measured in ways that permitted comparison across the included studies. Therefore, we were limited to making an overall assessment of whether the evaluations of the interventions’ impacts yielded results indicating that inequalities in health could be reduced between groups and individuals with different educational backgrounds.

Results

Of the nine publications included, all applied prospective study designs. Five studies were randomized controlled trials [14–18], three were community interventions [19–21], and one was a quasi-experimental trial [22] (Table 1). Seven studies evaluated a targeted approach [14–16,17–19,22], and two evaluated a population-wide approach [20,21].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of studies included in the review.

https://doi.org/10.1371/journal.pone.0195774.t001

The sizes of the study samples varied between 161 and 3,371 participants included in interventions from five different countries: three from the Netherlands [14,16,20], three from the USA [17–19], one from Australia [15], one from Canada [22], and one from Norway [21] (Table 1). Two interventions occurred within school settings [21–22], two were web-based [14,19], two were implemented within a public health service setting [16,20], one involved a home-nurse visitation service [17], and one was in a maternal healthcare setting for first-time parent groups [15], and another one was in a mobile health communication research facility [18] (Table 1).

Discussion

This rapid review identified nine studies that evaluated interventions targeting health-related behavior, while taking educational inequalities into account. Five different health outcomes were analyzed: smoking, dietary intake, physical activity, mental health, and mammography.

Given the heterogeneity of the studies included, we could not perform a meta-analysis because of the different study designs, study settings, and quality of analysis. Furthermore, outcomes were not measured in ways that permitted comparison across the included studies. Therefore, we were limited to making an overall assessment of whether the evaluations of the intervention impacts yielded results indicating that inequalities in health could be reduced between groups and individuals with different educational backgrounds. Regarding interventions aiming at smoking cessation, we could not draw any decisive conclusions. Of the three studies found, only one [14] provided some support for benefiting those with lower education. We also found limited evidence for decreasing inequality through interventions regarding dietary intake [21], physical activity[22], and mental health [16–17]. The Norwegian school fruit intervention is particularly interesting since this program seems feasible to implement as a cost-effective health-promoting initiative. Since we only found one single study attempting to increase the use of breast cancer screening (mammography) [18], we concluded that there is not enough scientific evidence concerning the potential for increased health equity for this approach.

All in all, there are few intervention evaluations with high-quality designs regarding health-related behavior, which have shown a higher impact among individuals with a low educational level. Our results are in line with previous research that used less strict quality criteria for including studies. That research found limited evidence of effective community-based interventions, such as interventions aimed at diet and physical activities for socio-economically disadvantaged groups [23]. As previously mentioned, policies associated with positive developments in average public health have also been associated with increased inequality in health outcomes between different socioeconomic groups [1,3]. This may be due to high-SES individuals benefitting more from such interventions because of greater access to resources (such as education), which positively interact with the intervention [5,6–7]. In fact, two of the studies included in our review were interventions where the health-promoting initiative risked exacerbating inequalities in health instead of reducing them [14,19]. Nevertheless, we do not argue in favor of a nihilistic position regarding public health interventions concerning individual determinants of health. As suggested by the CSDH and many others, the underlying issue of health inequity is the systematic generation of a socially unjust distribution of individual level resources, such as money, power, and resources depending on elements of the social structure [24].

The likelihood of public-health interventions having a higher impact on health inequality could be increased by designing them to be embedded in the public welfare system, which has essentially been developed to reallocate money, power, and resources. A prominent example of this is the early interventions in terms of mother and child health services, which were developed in Scandinavian countries in the 1930s [25–26]. These services include an important element of resource transfer to individuals and families with low SES and were embedded in other elements of the welfare state, such as progressive taxation and gender equity policies [27]. A more recent example is the taxation of sugar-sweetened beverages or “soda tax” as it is sometimes referred to [28–29] which after its introduction in Mexico has resulted in a more pronounced decline of high-energy foods in low SES households.

Moreover, such embedded interventions should have greater potential to address the issue of the gradient of inequality across the spectrum of socioeconomic statuses rather than just the gap between the worst-off and the rest of the population [30]. Hence, the gradient in health inequalities could be efficiently addressed by the strategic principle of proportionate universalism [31], which involves a common integrated system of health and welfare services that serves the whole population according to need and is financed according to ability.

Conclusions

There are very few studies of high quality design that have examined the potential for interventions aimed at individual determinants of health to reduce health inequalities. This review could only identify nine intervention studies with limited comparability. Therefore, the main conclusion is that solid evidence is lacking for the mentioned interventions and that more research is needed to fill this gap in knowledge. However, some potentially promising interventions were identified, which merit further research.

Supporting information

Acknowledgments

This manuscript is based on a report commissioned by the Public Health Agency of Sweden in 2014–2015 that was part of a governmental assignment from the Ministry of Health and Social Affairs. The authors wish to thank Marlene Makenzius and Mikael Nordberg at the Public Health Agency of Sweden for valuable comments on that report.

Ethical approval was not required as no primary data were collected.