What impact do chronic disease self-management support interventions have on health inequity gaps related to socioeconomic status: a systematic review

We conducted a systematic review of the literature using the PRISMA reporting guidelines [29] to structure the report. We searched for full-text articles in English to December 2018 in the following databases: Cochrane database; PubMed; Cinahl; Embase; Proquest and Psychinfo. The search terms covered the following areas, using MeSH terms and synonyms: [1] Chronic condition, including diabetes, cardiovascular disease, musculoskeletal conditions and chronic pulmonary disease [2]; Self-management [3]; Socio-economic status, including associated terms such as inequity, disparity, ‘vulnerable groups’; and [4] Terms related to outcomes, efficacy, retention or participation. The PubMed search strategy is available in Additional file 1. No date filter was employed in order to obtain the widest possible search. In the course of the search thirteen related systematic reviews were located and their references were screened resulting in seven additional papers.

Inclusion and exclusion criteria are outlined in Table 1. We looked for four main chronic conditions: cardiovascular disease (CVD), musculoskeletal conditions (MSK), pulmonary disease (COPD) and diabetes. All these conditions contribute significantly to the burden of disease and share many common risk factors. We included studies of co/multimorbidity since this is representative of the CD population. A decision was made to focus only on socio-economic status (SES), which has well-documented and consistent effects on chronic disease, rather than on other WHO PROGRESS+ factors such as gender and ethnicity, which can vary between countries [19]. All studies needed to provide a comparison between a less and more advantaged group, based on income, education or socioeconomic area. Comparisons based on literacy or ethnicity were only included if there was a quantifiable relationship between these variables and other SES measures. As well as post-intervention outcomes such as behavioural or clinical changes, outcomes related to participation and dropout were included to fully capture potential areas of disparity. Study designs could include randomised controlled trials with subgroup analyses, pre-post designs, cross-sectional or longitudinal data analyses.

Title and abstract screening reduced the number of papers to 310. Articles were excluded according to the criteria outlined in Table 1. Common reasons for exclusion were no SMS intervention (e.g. studies of self-care or adherence behaviours); SES not quantified, and no measurement of SES disparity. A full list of reasons for exclusion of the 291 full-text articles is available in Additional file 2. Figure 1 illustrates the search process undertaken. One reviewer (RH) completed the initial search and a second reviewer (ES) independently assessed the final papers to ensure agreement on inclusion criteria. Nineteen studies were included in the review.

The data was summarised on the setting, study design, type of CD, sample size, description of intervention and control, outcomes or variables measured, follow-up time, results and study quality (Table 2 and Table 3). Table 4 and Table 5 summarises data related specifically to SES and disparity, including the theory behind the SM intervention (or study question for participation/attrition studies), intervention description, SES adaptations made, SES status of population, results in relation to SES, dropout rates and overall impact on SES disparity. Related papers were retrieved to provide additional data about the population or intervention as needed [31‐38].

Quality analysis was undertaken using the Joanna Briggs Institute checklists [30] for randomised controlled trials (RCTs) and observational studies, and the Sun/Oxman criteria [27, 28] for subgroup analyses.

No meta-analysis was possible due to the diversity of study designs, interventions and outcome variables.

Nineteen studies were identified, all published in English. Five studies looked at participation in SMS; five studied attrition from SMS programmes and nine assessed outcomes from SMS interventions. Interventions were very diverse, ranging from studies of the group-based Stanford Chronic Disease Self-Management Programme (CDSMP – 4 studies) to highly tailored 1–1 interventions. Table 2 and Table 3 details the main features of all studies.

Most studies were of moderate to good quality although two RCTs [39, 40] and three subgroup analyses [39‐41] rated poorly. A summary of quality ratings is included in Table 2 and Table 3 and a detailed table describing how each study was assessed is available in Additional file 3.

Responses to study questions

Four cross-sectional studies and one cohort study looked at initial participation in SMS programmes. All were large population surveys ranging from 2600 to 80,000 people. There were three reports on diabetes SMS education programmes [42‐44], one on the Stanford CDSMP [45] and the final study examined recruitment to an internet diabetes SMS programme [46]. In all studies, low SES (as measured by education, income or location) was significantly and consistently associated with lower levels of participation, suggesting that disparity in CDSM starts here. Some studies [43, 45] suggested that this imbalance was related to course availability, cost or marketing strategies. However, the studies which did match attendance to course availability and cost [42, 44] found that this did not influence participation in the low SES population. Glasgow [46] also compared participation rates in a self-selected (via media advertising) population to a referred population and found even greater disparity. As well as being of higher SES, the self-selected participants were those at lowest risk and least in need of the intervention.

There is consistent evidence that low SES is associated with lower levels of participation in SMS interventions, and some evidence that this is unrelated to access to SMS interventions.

Five studies examined attrition: two cross-sectional studies and three RCTs with subgroup analysis, with sample sizes from 100 to 300. Two RCTs [41, 47] were of more advantaged populations. Of these, one reported low (22.8%) completion rates of the Stanford CDSMP [41], but predictors were related to poor physical health rather than SES. Since this was a high-risk multimorbid rather than a low SES population, dropout likely reflects increased treatment burden, as noted in other multimorbid populations [48]. The second study [47], of a diverse urban population, reported no difference in use of a supported internet programme in terms of SES (education). This intervention had been carefully tailored to maximise engagement across population groups and included extensive community involvement in the design process. Three studies [49‐51] focussed on low SES populations. Two cross-sectional studies [49, 50] reported that dropout rates correlated to social stressors and lack of job flexibility, suggesting that attrition within a low SES population may be influenced by socioeconomic factors that are not captured by education or income alone. Finally, a small RCT [51] of a tailored group programme found that high levels of dropout were significantly associated with low income and education. By contrast, Horrell et al. [45] noted that although SES area predicted enrolment in the Stanford CDSMP, it did not affect rates of completion.

SES is not consistently associated with dropout from SMS interventions. SES may be one of a number of factors associated with programme attrition, as suggested by qualitative studies on this topic [52].

Nine studies looked at outcomes following SMS interventions, with four describing group interventions (including 2 of the Stanford CDSMP) and five individual [1] interventions. Only two of the RCTs [53, 54] were sufficiently powered for subgroup analysis and most had follow-up periods of 6 months or less.

Three of the nine studies featured outlier populations (in terms of age, sex and/or level of disadvantage), including the two lower-quality studies [39, 40] and the cohort study [55]. The findings from these studies may not be reliable or relevant to the wider low SES population.

The remaining six studies, of moderate to high quality, described broadly similar populations in terms of age, sex, education and income. Of these studies, one reported increased disparity following the intervention; two reported no change; and three studies reported a reduction in SES disparity.

Three of the studies, all individual interventions, described programmes specifically tailored for low SES groups, including extra supports and literacy adaptations. These included a 6-month peer support programme [56] and two 12-month phone support programmes [53, 57] (conducted by the same research group, but with different chronic diseases and interventions). All studies reported clinically and statistically significant changes in either hospitalisation rates [53] or HbA1c [56, 57] in favour of the intervention. Two of the studies also reported a reduction in SES disparity from the intervention, with low-literacy patients experiencing greater benefit from the intervention than their higher literacy counterparts. In an already low-SES population, this was found to be a stronger predictor than income or education. The third study (the peer support programme) reported no change in disparity, with benefits across all education levels and the greatest benefit experienced by those with poorer medication adherence and self-management ability.

The remaining studies – comprising one individual and two group interventions – did not provide specific tailoring for low SES participants. The individual intervention [58], a 6-week CBT programme designed to increase self-efficacy, found clinically significant improvements in depression only in the higher educated, with no change and higher rates of dropout in the lower educated. The group interventions, which were both for people with heart failure, included the 6-week CDSMP and a year-long SMS group programme. The CDSMP study did show short-term benefits as compared to usual care, but no overall gains at 6 or 12 months. The lower educated patients did better than their higher educated counterparts in terms of cardiac quality of life (QOL) (p = 0.018) over 12 months, suggesting a reduction in SES disparity, although it was not clear whether this was clinically significant. The second group programme [54] used an active education control and found no additional benefit from an SMS group. Low-income participants receiving the intervention did have a longer time to cardiac event (death or hospitalisation), but this was not statistically significant. Overall there was no change in SES disparity, nor any added benefit from the intervention.

There is limited evidence to suggest that SES does affect outcomes following SMS interventions. Interventions that were tailored for low SES participants reported significant improvements in clinical outcomes, which in some cases also included a reduction in SES disparity following the intervention.

This systematic review of disparities related to SMS interventions has reinforced observations [18‐20, 22, 25] that there is a lack of research in this area. Although many studies of low SES groups have been undertaken, very few have focused on whether the outcomes compare favourably to those in higher SES groups. There are practical and statistical challenges in comparing population subgroups. Many studies had SES groupings that were fairly homogenous, limiting the ability to compare outcomes within the analysis, and almost all subgroup analyses were insufficiently powered. Larger studies and co-operation between different study populations are needed so that there is a more distinct contrast between SES levels across groups.

Responses to study questions.

This review confirms that low SES groups are significantly less likely to participate in SMS interventions [42‐46]. Thus, healthcare disparity is increasing before an intervention even commences. In order to reach those who need the intervention, targeted recruitment and retention strategies will be needed. Self-selection runs the risk of spending limited resources on those who need them least [46].

The findings in relation to retention and dropout are less clear-cut, with few studies and small sample sizes. Social factors do appear to be important [49‐51], although a simple measure of SES may not capture the barriers to engagement.

With the limited number of high-quality studies available, there was some evidence that SES does affect outcomes following SMS interventions, depending on the type of intervention on offer. No trends were observed in terms of the SM components, which varied little between studies, or the type of service providers involved.

Programme structure (group or individual) did seem to affect both dropout rates and outcomes, with fewer benefits observed in the group interventions. In the few programmes that recorded dropout by SES, it appeared that attrition was also greater from group programmes (see Table 4 and Table 5). High rates of dropout from group programmes have been reported in several reviews of CD interventions in low SES and other vulnerable groups [21, 59], while other reviews [13, 60, 61] have noted that individually tailored interventions appear to reduce disparity. Other authors have noted that although group programmes provide beneficial social support and peer modelling [5], they can also present many barriers to a low SES population who may have less flexibility in terms of work, transport or caring demands [21, 59]. In the current review, interventions over longer time periods (6–12 months) also seemed to be more effective at reducing disparity [53, 56, 57], consistent with a CD review on similar populations [13].

This review suggests that SMS interventions may impact differently on low SES populations, and that more individualised treatment over longer time periods may be needed. Some writers have suggested that SES could be used as a ‘high risk’ predictor to identify those needing an earlier or more intensive intervention [23, 62], although this encompasses a large population group and has significant resource implications, emphasising the need for appropriate targeting of interventions.

Data from the current review indicates that low SES groups are heterogeneous, with additional factors such as literacy, social stressors and social capital influencing SM ability, engagement, health outcomes [49, 50, 53, 57] and thus disparity. Therefore, some low SES groups may benefit simply from better marketing of and access to generic SM courses [45] and lower-level interventions, while others will require a more intensive, tailored approach. The ability to accurately identify these groups, perhaps by using a triage instrument, could lead to more effective resource allocation, increased participation and better outcomes in terms of both efficacy and equity.

Few studies reviewed described the theory behind the proposed SMS intervention, as noted in other reviews of SMS [12, 63], although several referred to the role of self-efficacy [40, 54, 58, 64], as described in Bandura’s social-cognitive theory [4, 5]. The studies which targeted a low SES or otherwise diverse population did note particular challenges for disadvantaged groups in terms of knowledge or literacy [47, 53, 56, 57], and those which adapted to these challenges often had better outcomes. In contrast, ‘one size fits all’ programmes [45, 46, 54, 58] had fewer benefits, and in some cases increased disparity.

SMS approaches informed only by self-efficacy have been criticised as overly individualistic [10, 11, 15] and it has been observed that the relationship between self-efficacy and self-management ability is weaker in vulnerable groups [65], indicating that other barriers play an important part. Furthermore, since the development of self-efficacy depends both on one’s behaviour and on social/environmental feedback [66], several authors [11, 58] have suggested that increasing self-efficacy may be harder if environmental feedback (e.g. job or housing insecurity) negates a belief in control over one’s circumstances.

This suggests that for SMS interventions to be effective in low SES populations, attention should be paid to other factors that influence self-management ability. Health provider/system issues [67, 68]; resources (literacy, financial, job/carer demands) [67, 69‐71]; and condition demands (multimorbidity, treatment burden) [48, 71, 72] have been consistently identified in qualitative reviews as barriers to self-management. Each of these factors will impact disproportionately on a low SES population. Health providers/systems can be less accessible due to cost, literacy levels and a limited understanding of the social determinants of health by providers [67, 68]. Although few studies of SM in disadvantaged populations look at interventions at the health provider/system level [18, 21], it would seem a potentially effective way to reduce disparity without increasing the patient’s treatment burden.

Barriers related to resources and condition demands are far greater for the low SES population [73‐75], who have fewer financial and social resources; higher levels of overall social complexity (job/housing insecurity, family demands, trauma history [3]); and higher rates of multimorbidity at earlier ages [76]. They experience both more disease-related workload (treatment burden) and non-disease workload (life burden) [73, 77]. Unfortunately, many SMS interventions, especially those requiring regular attendances or homework, will increase workload. Approaches that reduce patient workload or increase access to resources are rarely tried, but are likely to be important in low SES groups [73]. Phone consultations, problem-solving of specific barriers, integrating healthcare with social services and directing interventions toward healthcare practitioners rather than individual patients can all reduce treatment burden and maximise resources. Coventry [76], in a qualitative study of SM and multimorbidity, identifies three factors required for engagement in SM: capacity (resources, knowledge and energy); responsibility (shared understanding between the patient and provider about how to manage the treatment workload) and motivation. All three are negatively impacted by low SES, yet many SMS interventions [10] aim to increase motivation without recognising responsibility or capacity, and thus may contribute to increasing disparity in low SES groups.

This review identifies important gaps in knowledge and potential directions for future research. It reveals the assumptions informing SMS approaches and the inadequacy of using ‘low SES’ to define a population group. The study limitations include the lack of published research on disparity in SM interventions. It was difficult to conduct a comprehensive literature search of this topic because many subgroup analyses were a relatively small part of the overall paper. It is possible that some studies were missed that may have provided useful data. Meta-analysis was not possible due to the variety of studies available; therefore, no strong conclusions can be formed. In addition, the methodology of many of the studies prohibited causal inference: several studies were cross-sectional and most subgroup analyses were underpowered or did not formulate a priori hypotheses.