Inequalities in health care among patients with type 2 diabetes by individual socio-economic status (SES) and regional deprivation: a systematic literature review

Type 2 diabetes mellitus (Type 2 DM) is one of the leading causes of death in the world with constantly increasing prevalence: in 2008, the prevalence was estimated to be 10% among adults aged 25+ years [1]. The rising burden of diabetes is associated with a constant increase in its complications, causing rising disability and booming health care costs ranging from 2.5% to 15% of annual health care budgets [2]. Good diabetes care is crucial to delay these complications. Good glycaemic control results in a reduction of complications and better patient outcomes. The data from the United Kingdom Prospective Diabetes Study (UKPDS) concerning patients with Type 2 DM suggested, for example, a decreased risk of 25% in retinopathy and nephropathy for every reduction of 1% in HbA1c [3].

Different patient groups are affected unequally by Type 2 DM. There are large differences, for example, concerning gender, age and race. Socio-economic status (SES) may influence access to and quality of care, social support and availability of community resources. It may also influence diabetes-related knowledge, communication with providers, treatment choices and the ability to adhere to recommended medication, exercise and dietary regimens [4, 5]. Thus, low SES could be associated with multiple risks. First, epidemiological studies have repeatedly confirmed the inverse association between Type 2 DM and SES [6‐9]. Also, regional deprivation (often used as a proxy for individual SES) was shown to have an independent influence on the incidence and prevalence of diabetes mellitus [10]. Second, some analyses focusing on patients with Type 2 DM indicate that health care could be worse for low SES groups, but this has rarely been studied in a systematic way.

To date, there is just one systematic review regarding inequalities in health care among patients with Type 2 diabetes by individual socio-economic status (SES) and regional deprivation. Ricci-Cabello et al. (2010) [11] focused on countries from the Organisation for Economic Cooperation and Development (OECD). They looked at differences by gender, ethnic group and SES, and for SES they reported results concerning diagnosis, control of diabetes and access to health services (four, seven and five studies respectively). They conclude that the results point towards health inequalities favouring higher SES groups, and that there is a great need for more empirical research. Their review covers publications up to 2007, and a number of papers have been published since. Thus, it was our objective to synthesize the current evidence, focusing on process and intermediate outcome indicators of health care for patients with Type 2 DM and on differences by individual SES and residential area deprivation.

In November and December 2012, we conducted a systematic literature search in PubMed for original studies published in English, German, French or Spanish. A combination of the following key words was used: diabetes, diabetes Type 2, quality of care, management, care, control, soci*, socio*, inequ*, differ*, dispar*, regio*, depriv* (and their translations) in title and abstract. The American Diabetes Association (ADA) guidelines for good diabetes control and quality of care were changed in 2002 (with subsequent improvements in 2005 and 2011) [12, 13] and, in 2005, the first Global Guideline for Type 2 Diabetes was published by the International Diabetes Federation (IDF) [14], complicating the comparison between previous and later studies. In order to improve the compatibility between studies, the time frame for the search was restricted to papers published between 1 January 2002 and 13 December 2012.

Altogether, 886 publications were eligible for further screening (see Figure 1). Titles and abstracts were screened, using the inclusion criteria outlined below and the PRISMA statement for reporting the results [15]. The references from selected articles were screened for further potentially relevant studies.

An article was included in the review: (a) if the study population comprises Type 2 DM patients or a mixed group of Type 1 and Type 2 DM patients; (b) if quantitative analyses are presented concerning process and intermediate outcome indicators of health care; (c) if individual SES and/or residential area deprivation is included as an independent variable. We focused the review on process indicators such as assessment of HbA1c (glycolised haemoglobin), BP (blood pressure), BMI (body mass index), and on intermediate outcome indicators such as quantitative measurement of HbA1c, BP and/or lipids in the blood. All these indicators relate directly to health care provision, they are usually well documented and show relatively short-term results. They are influenced by other factors as well, of course, including individual health behavior and social support. Health care plays an important role, though, and most guidelines for good diabetes control include targets such as reduction of high BP, smoking cessation and reduction of high body weight.

Quality of care could also be assessed by outcome indicators less closely associated with health care, such as complications, hospital admissions or cardiovascular disease (CVD) mortality [16]. Results concerning these ‘other outcome indicators’ are mentioned below just briefly. The exclusion criteria are: studies restricted to Type 1 DM and studies just focusing on social disparities by race, ethnicity and nationality or migration status. These social disparities could be closely linked to individual SES and residential area deprivation, of course, but we still believe that they touch on a different topic.

Different information was extracted from the studies finally selected, based on the data extraction sheet proposed by Aveyard [17]; the results are presented in Tables 1 and 2. We also tried to assess and compare the methodological quality of the studies; this task proved to be quite challenging. The studies included in the review were conducted in very different settings (e.g. hospital, primary care), using a wide variety of methodological approaches (e.g. national or local registry data, practice-level data, patient survey). It is hardly possible to compare them in a common scheme that would account for all these differences. Finally, we used an adopted version of the EPHPP (Effective Public Health Practice Project) quality assessment tool [18]. It includes different components (evaluation of selection bias, study design, confounders, blinding, withdrawals and drop-outs, intervention integrity). Each component must be rated as providing ‘strong’, ‘moderate’ or ‘weak’ evidence; the final scale summarizes these ratings. We applied relatively strict criteria: to be rated as ‘strong’ evidence, a study must have no weak component and at least three out of six strong ones. The criteria for ‘moderate’ evidence are: one ‘weak’ component, or no ‘weak’ and less than three ‘strong’ ratings. All other studies were rated as ‘weak’. Screening, information extraction and assessment was first performed by OG and WM, in case of disagreement a joint decision was reached together with AM.

Based on the inclusion and exclusion criteria listed above, 32 articles were identified for full-text review (see Figure 1).

Eleven of them were excluded for different reasons (no data on individual SES or regional deprivation, no data on quality of health care, or no analysis of the association between these two topics).Most of the remaining 21 papers are from Western European countries, i.e. from the UK (n = 8), Germany (n = 3) and one each from France, Italy and Spain. There are also three papers from the USA, two from Israel and one each from Taiwan and China. They comprise cross-sectional (n = 12) and cohort studies (n = 9), and most used the retrospective data (n = 17). Some (n = 8) are limited to Type 2 DM patients; the others use a mixed sample of both Type 1 and Type 2 DM patients. Sample size varies between 135 and 1,941,517 patients; mean (or median) age varies between 52.6 and 68.1 years. The process and intermediate outcome indicators (including cut-off points of good control) are presented in Table 1. The main characteristics and results of the studies are summarized in Table 2.

Studies included here come from a variety of countries, and they vary widely in study design and sample size. About one third use data from registries at the national [19], regional [24, 26, 27, 29‐31, 35, 37, 38] or practice level [20, 28, 33, 34]. These registries usually provide high quality data with few missing values, but they mostly lack information on individual SES. Other studies were conducted in a hospital setting [21, 23, 39], and again data quality should be high. Owing to potential selection bias, it could be difficult, though, to apply the results to the general population. A third group of studies is based on surveys [22, 23, 32, 36]. They might be more representative of the whole population, but they often lack information regarding process indicators. Each study design has its specific advantages and drawbacks, and this also applies to the studies included in this review.

About half of the studies include information on individual level SES, for example by educational level or per capita household income (see Table 2). The other studies include information on regional deprivation, probably often because data on individual SES were not available and regional deprivation was taken as a proxy. Regional deprivation is assessed in different ways, e.g. by the Townsend score [22], the IMD [24] or by a deprivation score developed in Spain [28]. Individual level SES and regional deprivation could both reveal important information, as it could be important to improve health care for specific social groups and/or for specific regions.

The studies reviewed here reflect the problem that there is no unified, internationally accepted definition concerning the quality of health care for people with Type 2 DM. Different approaches to defining and measuring health care quality are used, and it is hardly possible to combine the results in a quantitative meta-analysis. Also, different studies use different sets of indicators. Microalbuminaemia is assessed in just five of these 21 studies, for example, and six studies do not include any process indicator. All but three studies include an intermediate outcome indicator (e.g. comparing HbA1c values with a cut-off value for good glycaemic control), but different cut-off values are used in different studies, making it even more difficult to directly compare the empirical results.

Despite these multifaceted differences between the 21 studies reviewed here, it is still possible to identify some main results: (a) Process indicators of health care quality are often significantly associated with individual SES, i.e. patients with low SES often receive these measurements and assessments less frequently than patients with high SES (e.g. concerning HbA1c, lipids, retinopathy, BP, BMI, smoking status, microalbuminaemia). (b) Individual SES is often significantly associated with poor glycaemic control, i.e. HbA1c values above the recommended clinical target. (c) Concerning BP and lipids above the proposed level, there seems to be no clear association with individual SES. (d) Diabetes-specific complications such as retinopathy and nephropathy seem to be more common among patients with low (as compared with high) individual SES, but these associations have rarely been assessed. (e) The associations with health care quality are mostly less pronounced and consistent if social differences are assessed by regional deprivation instead of individual level SES. (f) Concerning process indicators, some studies indicate that health care quality is worse in more deprived regions, but these associations are statistically significant only in some of them. (g) A similar, but more consistent picture is seen for glycaemic control. HbA1c values indicating poor glycaemic control are found more often in more deprived regions, and in most studies this difference is statistically significant. (h) Some studies indicate that the targets concerning BP and lipids are achieved less often in more deprived regions, but the level of statistical significance is reached only in some of them. (i) Few studies assess the association between regional deprivation and diabetes-specific complications such as retinopathy and nephropathy; they indicate that it is positive and significant.

Some limitations of the review need to be considered. First, we tried to find all relevant studies published up to December 2012, by following a well established, standardized scheme [15], by including all papers irrespective of the statistical significance of their empirical results, by including papers written in different languages (i.e. English, German, French and Spanish) and by performing a counter-checked evaluation of the papers. Our search was confined to publications that could be found in the PubMed database. Following the recommendation from a reviewer we later repeated the search in the EMBASE database (using the identical strategy as outlined above; see ‘Methods’), and we found that all publications identified in EMBASE had also been included in the first group of papers (n = 886) identified in PUBMED (see Figure 1). But of course we cannot rule out the possibility that we have missed relevant publications.

Second, just one third of these studies are clearly restricted to Type 2 DM patients [22, 29, 31‐34, 36, 39], i.e. the others also include Type 1 DM patients. This should not be a major problem, though, as Type 2 DM accounts for about 90% of all diabetes patients [1]. Of course, there could also be important inequalities concerning health care for type 1 DM [10], but this review focuses as much as possible on type 2 DM. Third, social inequalities can be assessed by a number of indicators besides SES and regional deprivation (e.g. migrant status and ethnicity). These other indicators are not included here, although they might well interact with SES and regional deprivation [5].

The review could be an important contribution to the discussion on health care inequalities among patients with Type 2 DM. Most would agree that these inequalities should be reduced, but to date there seems to be just one review that summarizes the results and stresses the need for further research. The previous review from Ricci-Cabello et al. [11] covered publications up to 2007 and included 25 papers for final assessment; just two of these are included in our review as well [34, 36]. This little overlap is also due to the fact that the two reviews differ in a number of ways: The review from Ricci-Cabello et al. [11] also includes studies that define social inequalities by gender and ethnic group, that study inequalities in diagnosis of DM, and it includes four studies that are restricted to type 1 DM.

Adding some recommendations for further research, we would like to highlight eight topics: First, despite the great variety of proposals regarding appropriate assessment methods [16, 41, 42], quality of diabetes care should be measured in a way that enables benchmarking and comparison. A standard set of indicators for assessing good quality health care should be provided. Process indicators may not be a good predictor of patient outcome; thus, it would be better to assess patient outcomes directly [43]. For example, concerning intermediate outcome indicators, the same cut-off value should be used in all studies, preferably based on current international guidelines. This is easier said than done, of course; there are still different guidelines. Concerning cut-off points for ‘poor glycaemic control’, for example, it has been proposed to use HbA1c values above 6.5% [14] or 7% [13]. The National Diabetes Quality Improvement Alliance [44] suggests HbA1c levels above 9%. In addition, there are different opinions regarding cut-off points, especially for the elderly [12]. The standard set should also include outcome indicators.

Second, it is important to distinguish clearly between the individual and the regional level. Assessing SES at the individual level highlights the following problem: health care for people with Type 2 DM could be influenced by a number of individual risks and resources such as poor health behaviour or good social support. It would be important to know which of these risks (or resources) are especially strong (or weak) among patients with low SES and how their risks could be reduced and their resources improved. Assessing SES by regional deprivation points to another problem: resources for people with Type 2 DM could differ between regions. There could be differences concerning health care (e.g. availability of specialized physicians) and other factors such as facilities for physical activities. Important questions would be, for example: What kind of resources are especially weak in more deprived regions, and how could they be improved? What role could be played by improving health care? Also, it is possible that the proportion of patients with low individual SES is especially high in the most deprived regions, that the differences between regions could largely be explained by these individual level differences. Therefore it is necessary to differentiate clearly between effects at the individual and at the regional level, i.e. to conduct multilevel analyses. These analyses are well known in public health research [45], but we are not aware of a study on social and regional inequalities in health care for patients with Type 2 DM that is based on this kind of analysis.

Third, some associations with low individual SES seem to be fairly well established, concerning process indicators (e.g. less frequent measurement of HbA1c and lipids), intermediate outcome indicators (e.g. higher prevalence of HbA1c values above the proposed level) and diabetes-specific complications (e.g. higher prevalence of retinopathy). Some associations with regional deprivation seem to be fairly well established as well, concerning intermediate outcome indicators (e.g. HbA1c values) and diabetes-specific complications (e.g. retinopathy). The potential reasons behind these associations are poorly understood however. They may relate to social differences in training, awareness of early symptoms and compliance [40, 46, 47], for example, but the ‘web of causation’ has not yet been described in any detail. Health outcomes are influenced by many factors, and health behaviour plays an important role in explaining social differences [48]. The influence of health services may be limited [41], but it would be important to study the interplay between health care, on the one hand, and health behaviour, diabetes self-management and the achievement of clinical targets, on the other, especially among low SES patients.

Fourth, meta-analyses could be important especially for assessing effects that are relatively small. It is rather difficult to merge the empirical information from different studies (as mentioned above), but the chances to conduct meta-analyses should be explored in more detail. Fifth, registries provide an excellent database for these studies, but they often lack information on individual SES such as educational level or income. It is recommended that this information is added wherever possible. Sixth, different countries have different systems of health promotion and health care (e.g. concerning co-payments), and the problems and causes of social inequalities in health care may differ as well. One implication is that policy recommendations for a specific country will have to be based on empirical studies conducted in this country. Another implication is that cross-country comparisons are difficult to conduct, as all these differences will have to be taken into account, but these comparisons offer an important chance for studying the associations between system characteristics and health care inequalities. Seventh, health care for patients with type 2 DM is similar to health care for patients with type 1 DM, but the two groups of DM patients are quite different. It would be important to add a review focused on patients with type 1 DM. Last but not least, there is a great need for intervention studies, i.e. studies that compare different interventions aimed at reducing social differences in health care for people with Type 2 DM.

The available evidence clearly points to the fact that socio-economic inequalities in diabetes care do exist. Low individual SES and regional deprivation are often associated with worse process indicators of care and worse intermediate outcomes (e.g. poor glycaemic, lipid and BP control), resulting in higher risks of microvascular (e.g. retinopathy) and macrovascular complications (e.g. myocardial infarction, stroke). These differences exist even in countries with a health care system that is clearly based on solidarity between all socio-economic groups (e.g. Germany). Much is known about good quality diabetes care; it is important to study in more detail why patients with low SES profit less from this knowledge.

This work was partially supported by the Kompetenznetz Diabetes mellitus (Competence Network for Diabetes mellitus) funded by the Federal Ministry of Education and Research (FKZ 01GI1104A).