A Systematic Review of Interventions to Improve Diabetes Care in Socially Disadvantaged Populations

  1. Richard H. Glazier, MD1234,
  2. Jana Bajcar, EDD25,
  3. Natalie R. Kennie, PHARMD25 and
  4. Kristie Willson, MHSC4
  1. 1Centre for Research on Inner City Health, St. Michael’s Hospital, Toronto, Ontario, Canada
  2. 2Department of Family and Community Medicine, St. Michael’s Hospital, Toronto, Ontario, Canada
  3. 3Department of Family and Community Medicine, University of Toronto, Toronto, Ontario, Canada
  4. 4Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada
  5. 5Faculty of Pharmacy, University of Toronto, Toronto, Ontario, Canada
  1. Address correspondence and reprint requests to Richard H. Glazier, Centre for Research on Inner City Health, St. Michael’s Hospital, 30 Bond St., Toronto, Ontario, Canada. E-mail: richard.glazier{at}utoronto.ca

Abstract

OBJECTIVE—To identify and synthesize evidence about the effectiveness of patient, provider, and health system interventions to improve diabetes care among socially disadvantaged populations.

RESEARCH DESIGN AND METHODS—Studies that were included targeted interventions toward socially disadvantaged adults with type 1 or type 2 diabetes; were conducted in industrialized countries; were measured outcomes of self-management, provider management, or clinical outcomes; and were randomized controlled trials, controlled trials, or before-and-after studies with a contemporaneous control group. Seven databases were searched for articles published in any language between January 1986 and December 2004. Twenty-six intervention features were identified and analyzed in terms of their association with successful or unsuccessful interventions.

RESULTS—Eleven of 17 studies that met inclusion criteria had positive results. Features that appeared to have the most consistent positive effects included cultural tailoring of the intervention, community educators or lay people leading the intervention, one-on-one interventions with individualized assessment and reassessment, incorporating treatment algorithms, focusing on behavior-related tasks, providing feedback, and high-intensity interventions (>10 contact times) delivered over a long duration (≥6 months). Interventions that were consistently associated with the largest negative outcomes included those that used mainly didactic teaching or that focused only on diabetes knowledge.

CONCLUSIONS—This systematic review provides evidence for the effectiveness of interventions to improve diabetes care among socially disadvantaged populations and identifies key intervention features that may predict success. These types of interventions would require additional resources for needs assessment, leader training, community and family outreach, and follow-up.

Diabetes affects at least 171 million people worldwide and is dramatically increasing in many countries as a consequence of increased obesity, sedentary lifestyle, and aging populations (1). Certain population subgroups are at higher risk of acquiring diabetes including nonwhite ethno-racial groups and those with low socioeconomic status (SES) (2). These same groups are often at higher risk of late diagnosis, poor diabetes control and self-management, the development of diabetes-associated complications, and more frequent emergency room visits and hospitalizations (38).

Interventions to improve diabetes outcomes can be directed at individuals with diabetes, health providers, or the health system. Patient-level interventions include those directed at improved self-management, including medication taking, diet, exercise, self-monitoring, and appropriate use of health care services (9). Continuing professional education and knowledge translation activities are examples of strategies intended to improve health provider care (10,11,12,13). Health system changes with the potential to improve access to care and quality of care include expanded hours of service, language translation, case management, reducing financial barriers to health providers and medications, and changes in health care provider roles (14,15).

When interventions are designed for general populations, however, disadvantaged groups may not be reached or may not be able to take full advantage of the opportunities provided to them. Socially disadvantaged groups may experience difficulties with language, cultural beliefs, transportation, getting time off work, child care, and low health literacy in addition to financial barriers to care. Given that these factors can influence uptake, interventions designed to reduce health inequalities may need to be targeted to specific audiences and place special emphasis on the way programs are designed and delivered to promote access to and use of resources and services (16).

In this systematic review, we define socially disadvantaged groups as those that have low SES or belong to an ethno-racial minority. We conceptualize social disadvantage as related to patient, provider, and health system factors that can affect self-management and provider management and ultimately manifest as clinical outcomes. The objective of this systematic review is to identify and synthesize evidence about the effectiveness of patient, provider, and health system interventions to improve diabetes care among socially disadvantaged populations.

RESEARCH DESIGN AND METHODS

Inclusion criteria

We included studies that targeted interventions toward socially disadvantaged adults with type 1 or type 2 diabetes and that were conducted in industrialized countries and measured self-management, provider management, or clinical outcomes. Self-management includes home glucose monitoring, regulation of diet and exercise, adherence to and self-adjustment of medications, and attendance at office appointments when needed or as scheduled by the provider. Provider management includes diagnostic testing, prescribing, educational and behavioral counseling, and referrals. Clinical outcomes include measures of fasting blood glucose, glycosylated hemoglobin (A1C), blood pressure, BMI and lipids, and patient-oriented outcomes of hospitalization, mortality, diabetes complications, and quality of life. To be included, studies had to be randomized controlled trials, controlled trials, or before-and-after studies with a contemporaneous control group. Articles written in all languages were included. Studies were excluded if they focused on specific age-groups such as youth, targeted only gestational diabetes, included only hospital process-of-care measures, or did not clearly specify the socially disadvantaged group being studied.

Search strategy

In consultation with an information specialist, electronic databases, including Medline, Embase, Cinahl, HealthStar, Cochrane Library, Sociological Abstracts, Social Science Citation Index, and International Pharmaceutical Abstracts, were searched for relevant studies from articles published in any language between January 1986 and December 2004. The references of review articles and of included original publications were also screened for potentially relevant studies.

Methods of the review

Two reviewers (N.R.K. and K.W.) independently screened citations and abstracts to identify articles potentially meeting the inclusion criteria. For those articles, full text versions were retrieved and independently screened by two reviewers to determine whether they met inclusion criteria. Disagreements about whether the inclusion criteria were met were resolved through discussion with a third reviewer (J.B. or R.H.G.).

Data extraction

Data extraction of relevant study information for articles meeting inclusion criteria was performed independently by two reviewers (N.R.K. and K.W.). Disagreements were resolved through discussion. A data extraction form was used to collect information on type of social disadvantage (low SES population or specific ethno-racial group), study objectives, study design, setting of intervention, study population (including demographics, SES, and health insurance), measures and scales used, and main findings.

To systematically describe the scope and components of the interventions, specific features were extracted from each intervention article using main components that have been previously described for the taxonomy of diabetes interventions by Elasy et al. (17). We extracted the main target of the intervention (patient, provider, or health system); type of needs assessment conducted, tailoring to the population; components of the intervention; diabetes-related content; timing/intensity; and the provider of the intervention. For articles that referenced previous publications (e.g., describing needs assessments or interventions), the original publication was retrieved and relevant data summarized.

Validity assessment

To assess the methodological quality of the intervention studies, we adapted the criteria suggested by Jadad et al. (18) if the study was described as randomized, if the sequence of randomization was described and appropriate, and if there was a description of withdrawals and dropouts. We did not assess blinding, as the types of interventions under study could not be blinded. We also assessed whether there was an adequate description of the intervention.

Data analysis

Due to heterogeneity of populations, interventions, and outcomes, it was not possible to combine outcomes in a meta-analysis. As an alternative, we used a recently described method to identify specific features of interventions that are likely to be associated with successful or unsuccessful outcomes (19). A total of 26 features were identified based on five categories: 1) target of the intervention, 2) how the intervention was designed, 3) setting of the intervention, 4) how the intervention was delivered, and 5) intensity and duration of the intervention (listed in Table 3). For each intervention, two reviewers (J.B. and N.R.K.) independently determined the absence or presence of these specific features. Disagreements were resolved through discussion with a third reviewer (R.H.G.).

Next, two reviewers (R.H.G. and N.R.K.) independently assessed whether the intervention resulted in a positive outcome that was both statistically and clinically significant. Statistical significance was set at P < 0.05, but due to the large number of possible outcomes, no predefined criteria for clinical significance were established. Instead, each reviewer provided a global judgment about clinical importance. In circumstances where studies reported multiple types of outcomes, judgments were primarily based on clinical measures. For studies that did not report clinical measures, provider management and self-management outcomes were used to make judgments. For each of the 26 features, we calculated the percentage success rate with and without the feature (percent of articles with the feature having positive outcomes minus percent of articles without the feature having positive outcomes). Positive and negative rate differences indicated that interventions with and without the feature, respectively, were more often associated with positive study results. Due to the small number of articles included, CIs were not calculated and we limited judgments about features to those that appeared in at least five included studies. Our use of this analytic approach was therefore more descriptive than quantitative. To assess whether our findings were related to the methodological quality of the included studies, two sensitivity analyses were performed as follows. 1) Studies were limited to those with a randomized controlled design, and 2) studies with randomized controlled designs were further limited to those in which randomization was described and appropriate and which had a description of withdrawals.

RESULTS

Description of studies

The initial search identified a total of 16,750 citations. After scanning titles of the citations, 615 were accepted for further screening and complete abstracts of these studies were reviewed. Of these, 163 citations were identified as potentially meeting the inclusion criteria. In addition, five articles were identified when scanning bibliographies of relevant published articles in the field. After examination of full text articles, 17 intervention studies (2036) were included in the review. Common reasons for exclusion were that the population under study did not meet the criteria for a socially disadvantaged population and/or there was no comparison group involved.

The characteristics of included studies appear in Table 1. Among the 17 studies, 7 (21,22,24,27,28,31,36) targeted low SES populations and 10 (20,23,25,26,29,30,3235) focused on specific ethno-racial groups, including African Americans (23,25,26,30,32,33), Mexican Americans (29,34), and Native Americans (33). Ten studies (21,22,25,28,29,30,32–34,36) were randomized controlled trials, two (26,27) were prospective controlled trials, and five (20,23,24,31,35) were comparative studies. Four studies (29,32,34,35) were community based, 10 (20,2228,30,31) took place in primary care settings, and 3 (21,33,36) were hospital based.

Eleven (2023,27,2933,35) of the 17 studies were judged by two reviewers to have both statistically and clinically significant positive outcomes. Due to the modest sample size of these studies, none were found to have statistically significant results that were not of clinical importance. Interventions were associated with significant improvement in A1C in 8 (2023,27,29,31,33) of 13 studies, weight/BMI in 2 (33,35) of 9, lipids in 2 (20,23) of 7, blood pressure in 2 (21,25) of 4, dietary intake in 0 of 4, eye examination in 3 (20,23,32) of 3, American Diabetes Association care indicators in 2 (23,28) of 3, physical activity in 1 (30) of 3, diabetes knowledge in 2 (29,30) of 2, physician trust in 1 (28) of 1, and mental and social well-being in 0 of 1 study.

Methodological quality

Ten (21,22,25,2830,3234) of 17 studies were described as randomized trials. The randomization process was described and appropriate in 6 (22,25,28,30,32,33) of 10 studies, all of which also had a description of withdrawals and dropouts and all had a well-described intervention.

Description of features of the interventions

Table 2 describes the features of the interventions. Some of the studies targeted more than one level; specifically, 12 (21,22,25,2730,3236) of 17 studies targeted the intervention at the patient level, 1 intervention (24) targeted the provider level, and 7 interventions (2023,25,26,31) targeted the system level. Three (27,29,30) of 17 studies described or referenced previous work (37,38) related to a needs assessment that had guided the development or adaptation of the intervention to their specific socially disadvantaged population (27,29,30). This generally included conducting meetings with staff involved with the patient group, conducting a needs assessment using community focus groups with the target population, testing of educational materials to ensure cultural appropriateness, or adapting the educational materials to low health literacy standards.

Ten (21,22,24,2730,33,34,36) of 17 studies described how specific aspects of the intervention were tailored to the socially disadvantaged population and described how the intervention was designed to meet the needs of the specific study population. Eight (21,22,24,29,30,33,34,36) of 17 studies considered literacy and language of the populations. Culturally specific references were used and described in five studies (22,29,30,33,34). Common materials that were used to adapt interventions included video and written materials; incorporated appropriate languages, cultural beliefs, and social values; targeted for low health literacy; and included records or logs that focused on recording behavioral modifications.

Six (28,29,30,33,34,36) of 17 studies provided diabetes education as the main focus of their intervention. Educational interventions used a variety of delivery methods including didactic lectures, skill building, problem solving, behavioral strategies (e.g., exercise), feedback (e.g., diaries), family member participation, and/or individualized assessment.

Six (2023,25,31) of 17 studies delivered an intervention that focused on innovative health care provider roles. These included individualized assessment using treatment algorithms by nurse case managers and clinical pharmacists and coordination of care by community health workers.

Other studies described interventions that were more focused on specific aspects of diabetes-related care. These included physical activity sessions (35), a focused education and support intervention to improve ophthalmic examinations (32), implementation of a visual communication tool (27), reminder cards for diabetes care indicators (24), and rapid A1C testing (26).

Analysis of features

We identified 26 features of the diabetes interventions (Table 3). Eight of these features appeared in fewer than five studies each, limiting any judgment about their effectiveness. Among the remaining 18 features, 13 were associated with positive rate differences and thus were more likely to be associated with successful interventions in socially disadvantaged populations. Features found in multiple studies with more consistent effects included cultural tailoring of the intervention, use of community educators or lay people leading the intervention, one-on-one interventions with individualized assessment and reassessment, use of treatment algorithms by various health care providers, focus on behavior-related tasks in the intervention, use of feedback about patient’s control of disease or performance, and high-intensity interventions (>10 contact times) delivered over a long duration (≥6 months). Interventions in multiple studies that were associated with the largest negative rate differences included those that used mainly didactic teaching or that focused only on diabetes knowledge.

Sensitivity analyses limiting results to only higher methodological quality studies indicated that most of the features had similar rate differences (remained positive or negative in the analysis), except that use of feedback to patients and treatment algorithms had an inconsistent pattern.

CONCLUSIONS

The results of this systematic review provide reason for optimism that interventions for socially disadvantaged populations can be effective and have the potential to reduce health disparities in diabetes care and outcomes. They also have practice and policy implications. The results indicate that short-term group-based didactic teaching aimed at improving diabetes knowledge may be of limited value for disadvantaged populations. To be successful, interventions appear to require organizational inputs and resources that go well beyond traditional diabetes education programs. The time and effort required to adapt interventions to the local community is likely to yield important benefits but must be done in each setting to fit local circumstances. The key aspects of tailoring in the reviewed studies were for culture and health literacy, both of which require substantial expertise and knowledge of local communities. Achieving a high degree of responsiveness to individual needs and making extensive use of behavioral methods requires one-on-one attention by highly trained personnel, again, with significant resource implications. If those personnel are community members, additional time and effort is required for their training, supervision, and support. Finally, successful interventions may require multiple contacts over a prolonged period of time.

Are such interventions affordable? Many of the resources needed for such community-oriented behaviorally mediated and intensive interventions are “up front,” including intervention design and recruiting and training community-based educators. Ongoing costs will mostly relate to multiple one-on-one sessions. The disadvantaged populations included in this study have significantly higher rates of diabetes complications, health services use, and poor control. A 1% drop in A1C has been associated with a 10% reduction in diabetes-related deaths and a 25% reduction in microvascular end points (39). Interventions targeting glucose, blood pressure, and lipid control simultaneously have the potential for up to 50% reductions in cardiovascular events and risk of microvascular complications (40). Reductions of 1% in A1C were achieved in several of the successful interventions included in this review, as were substantial improvements in blood pressure and lipids, so it is reasonable to expect major impacts on morbidity, mortality, and health services use from interventions such as these.

Our findings were less clear in several areas. Due to the small number of studies included and inconclusive results, it was not possible to clearly identify whether patient, provider, system-level interventions, or combinations of these are most effective. The results also did not point to hospital, primary care, or community as the best setting for intervention. Other features appeared to have positive effects, but definitive conclusions could not be made due to low numbers of studies identified. These features require further study and included needs assessment to design the intervention, delivery of comprehensive self-management education, problem solving tasks as part of the intervention, and the incorporation of family support.

The features of successful interventions identified in our review are consistent with those described for multiple health conditions as leading to positive outcomes for disadvantaged groups (16). Arblaster et al. (16) found that to reduce health inequalities among disadvantaged groups, intensive approaches, community involvement, face-to-face interventions, development of skills to promote behavior change, attention to the agent delivering the intervention, provision of prompts and reminders, and the use of prior needs assessment to inform intervention design were needed. They also found that multifaceted interventions and multidisciplinary approaches were desirable.

A review that was limited to diabetes self-management interventions in disadvantaged populations included five of the studies in this systematic review (41). That study found that interventions in community settings and those that explicitly address social-contextual issues tended to be more effective. They also identified the need to intervene at multiple levels with multiple strategies. Changing behavior and diabetes control in the long term may be much more difficult than achieving increases in knowledge. Active, hands-on, participatory, and behavioral approaches in self-management education appear to be more effective than didactic interventions, but long-term glycemic control tends to decline once interventions are over (9,42).

This study has several limitations. Social disadvantage is defined differently across studies, which limits their comparability (16,41). The studies included in this review primarily focused on black, Hispanic, and low-income people living in the U.S., and it is not known how these results relate to other socially disadvantaged populations. Our findings are based on a small number of heterogeneous studies that could not be combined in meta-analysis due to varying populations, interventions, and outcomes. The small number of studies and large heterogeneity also means that our conclusions about the effectiveness of various intervention strategies should be interpreted with caution. Few studies included in this review reported outcomes beyond 1 year, so the long-term effectiveness of interventions for disadvantaged populations is not known. Many of the included interventions had multiple components, and it was rarely possible to isolate the key elements that were responsible for beneficial outcomes. Additionally, no studies measured relevant patient-oriented outcomes. These issues represent areas for future research.

In conclusion, this systematic review highlights important features of interventions to improve diabetes care among socially disadvantaged populations. Key strategies include cultural tailoring of the intervention, community educators or lay people leading the intervention, one-on-one interventions with individualized assessment and reassessment, incorporating treatment algorithms, focusing on behavior-related tasks, providing feedback, and high-intensity interventions (>10 contact times) delivered over a long duration (≥6 months). These types of interventions would require additional resources for needs assessment, leader training, community outreach, and follow-up. Active and intensive strategies are justified, however, by the large burden of illness born by disadvantaged groups with diabetes, their greater need for health care services, and the demonstrated efficacy of diabetes control in reducing adverse outcomes. Further research is needed, in particular on which strategies are most effective for specific groups, on which interventions produce long-term improvement in clinical and patient-oriented diabetes outcomes, on interventions aimed at providers and the health system, and that incorporate a formal economic analysis.

Table 1—

Intervention studies

Table 2—

Features of interventions

Table 3—

Analysis of intervention components

Acknowledgments

This work is supported by the Ontario Ministry of Health and Long-Term Care and an unrestricted grant from BMO Financial Group.

The opinions, results, and conclusions are those of the authors, and no endorsement by the Ministry is intended or should be inferred. Rebecca Zakoor, Manager, St. Michael’s Hospital Health Sciences Library, generously assisted with development and implementation of electronic search strategy.

Footnotes

  • A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.

    DOI: 10.2337/dc05-1942

    • Accepted March 19, 2006.
    • Received October 11, 2005.

References

| Table of Contents