Our meta-analysis suggests that CHW interventions lasting at least 12 months result in a modest reduction in A1c, as compared to usual care. We found evidence in meta-regression that greater A1c reduction may be achieved in populations with higher A1c levels at baseline. It is also possible that studies with a more visit-intensive CHW protocol might have shown greater efficacy. Unfortunately, detailed data describing the exact number of CHW-participant encounters was not available for several studies, and this precluded a meta-regression analysis to assess whether efficacy varied significantly according the intensity of the intervention.
Given that studies used different laboratory methods to measure A1c, we estimated the pooled effect size as standardized mean difference (SMD), which measures the incremental A1c reduction by the intervention, above and beyond usual care, in standard deviation units. The pooled SMD we report is usually considered to reflect a small difference between treatment and control groups.27 However, as pointed out by Durlak,28 rigid categorizations of therapeutic efficacy based on SMD values can be misleading. In addition, although methodologically warranted, the assessment of efficacy of in standard deviation units may be difficult to understand for most readers. Examination of individual study results as unadjusted changes in mean A1c reduction may offer an additional perspective, provided that it is interpreted with due caution. In that sense, eight of the nine studies showed a larger unadjusted mean A1c reduction in the CHW arm than in the control arm—in relative terms, that reduction was 1.6 to 12 times greater in the CHW arm. Only the study by Tang et al. showed a lower relative mean A1c reduction by CHWs. However, as we already discussed, the design of that study probably handicapped the evaluation against CHWs. Their control arm was not limited to “usual care”, but had an intervention carried out by Peer Leaders, who were, by design, in more frequent contact with participants than the CHWs. It should be noted that, given its small sample size, that study was not influential in our meta-analysis, and removing it did not substantially modify our findings.
A comparison with the efficacy reported for other interventions in lowering A1c levels may also be informative. Previous meta-analyses have shown similar effect sizes for A1c reduction by other interventions. Polisena et al. estimated an SMD (95 % CI) of 0.21 (0.08 to 0.35) for home telemonitoring interventions.29 In a meta-analysis by Harkness et al., psychosocial interventions aimed at improving both the physical and mental health of people with diabetes achieved an SMD of 0.29 (0.21 to 0.37) for A1c reduction.30 Interventions that applied the Chronic Care Model to diabetes care resulted in a standardized A1c reduction of 0.19 (0.10, 0.29).31
The limitations of the available evidence used for our meta-analysis must be considered. First and foremost, the small number of eligible randomized controlled trials resulted in a wide confidence interval for the pooled estimate, limited our ability to rule out publication bias,25,26 and reduced the confidence in the meta-regression results.17 Second, there was considerable variability in trial design, including target population, intervention components, participation of other health care professionals, trial length, and baseline A1c values of the study participants. However, it is reassuring to note that the outcomes heterogeneity across studies, as measured by the I2 statistic, was moderate, suggesting that in spite of methodological differences, the efficacy estimates, that is, the observed effects of the CHW intervention, were not excessively heterogeneous. In addition, the studies we meta-analyzed reported substantial attrition, but had, in general, a low risk of bias. Several trials did not report allocation concealment during randomization, but this does not necessarily mean allocation was not concealed. For example, a study by Devereaux et al. reviewed 98 randomized trials, and then contacted investigators to identify under-reporting of concealment. They found that allocation concealment had not been reported in 54 (55 %) of the publications, but when contacted directly, investigators reported concealing allocation in 96 % of the trials.
Despite the above noted limitations, our meta-analysis provides useful information for health care organizations, and clinicians who are interested in the implementation of CHW programs for diabetes, and for researchers planning studies in this field. In addition, the completion of currently ongoing trials should add much needed information, and increase our confidence in meta-analytic estimates.32–36 The efficacy of CHWs may vary depending on whether they work from within the community, or integrated into multidisciplinary health care teams. The CHWs worked in a team that included other health care professionals, either a nurse practitioner or a diabetes nurse, in only four of the nine long-term studies. Only two of those nine studies assessed the CHW intervention as a “stand alone” protocol, in which the CHW intervention did not actively involve the participant’s PCP.12,21 At this point, there is insufficient data to determine whether the efficacy of a CHW intervention program is enhanced through the collaboration with other health care professionals. Of note, there is a strong movement, supported by public health experts, third party payers, and governmental institutions, towards the implementation of a patient-centered Medical Home.37 It is thus relevant to assess the participation of CHWs in Medical Homes created for disadvantaged populations. Another important question for future research is whether telephone-based interventions may be efficacious when in-person protocols are not feasible.
The great variability that we found in CHW intervention models in randomized trials is probably a reflection of the heterogeneity seen across CHW clinical programs being implemented across the United States.38 There is widespread consensus regarding the need for an evidence-based approach to standardize the training curriculum and certification of CHWs, and to validate the protocols and tools they implement in their work, while preserving the flexibility to adapt to the specific needs of the communities they serve. System-based initiatives that integrate CHWs into multidisciplinary teams to provide care to underserved populations have already been implemented across the United States, but with great variability in resources. There remains a great need for systematic support and development in most states. A comprehensive review in Massachusetts, published in 2010, identified four areas in which public health officials should act to achieve efficacious and sustainable CHW models: 1) development of appropriate infrastructure, 2) providing CHWs with a professional identity through clear definition of core competencies and roles in the healthcare system, 3) workforce development with training, certification, and continuing education, and 4) financing.6 One essential component in the development of sustainable CHW workforce is the guidance and financial support provided by State governments. A 2013 review by the CDC showed that only seven states had laws authorizing Medicaid to reimburse for CHW services.39 In addition, only five states had enacted legislation to create a CHW certification process or required CHWs to be certified. The work carried out in the state of Massachusetts could be used as a model by others, as it addressed to four core areas delineated above. More recently, the Affordable Care Act has provided new opportunities. A review by Katzen and Morgen, from the Center for Health Law and Policy Innovation at Harvard, identified three ways three ways the Affordable Care Act (ACA) has “opened doors for CHWs”.40 First, the ACA has increased access to preventive health services under Medicaid, and has clarified that states may designate non-licensed providers (i.e., CHWs) to provide preventive services. Second, the ACA offers state Medicaid programs the opportunity to create “Health Homes” for beneficiaries living with chronic illness, and those homes may include a role for CHWs. Third, the ACA created funding for State Innovation Models to help states improve health outcomes and quality of care while slowing growth in health costs; at the time of the report, four states had included CHWs in their innovation models.
In conclusion, our meta-analysis of CHW interventions has found a modest reduction in A1c, as compared to usual care. The estimated effect was larger in studies with higher mean baseline A1c, suggesting that people with poorer glycemic control may benefit more.