Introduction

The health status of a population is inextricably linked to social and economic conditions. In the United States (U.S.), we face unique challenges fueled by decades of inequity that require culturally appropriate health and social interventions. For example, African Americans are more than twice as likely to die from diabetes [1]. Diabetes, which is both an antecedent and moderating factor for cardiovascular disease, is the leading cause of death in the U.S.

Community health centers (CHCs)—previously referred to as neighborhood health centers—are community-based clinics that provide primary healthcare services to people with limited access to health care [2]. Currently, there are more than 1300 CHCs in the nation, that serve more than 27 million patients, 92% of whom are low-income and 62% of whom are racial/ethnic minorities [3]. Serving socioeconomically disadvantaged populations with heightened disease burden makes CHCs an ideal setting for implementing chronic care programs for vulnerable populations suffering from diabetes.

A number of systematic reviews were published that addressed diabetes interventions in primary care and community settings, but not CHCs. Previous systematic reviews found that theory-based lifestyle interventions (e.g., health belief model, transtheoretical model, Precede–Proceed model, theory of empowerment, social cognitive theory) [4], social network interventions [5], or interventions using community health workers [6, 7] or peers [8], and mHealth [9] were effective in improving hemoglobin A1C (HbA1C) among individuals with diabetes. Additionally, motivational interviewing by general practitioners [10] and nurse-led self-management support interventions [11] resulted in a significant improvement in HbA1C in patients with diabetes. Community pharmacist-led interventions did not report statistically significant effects of interventions on diabetes [12].

There is growing attention to the social determinants of health in order to better understand and act on the upstream socioeconomic drivers of poor health outcomes and higher economic costs for diabetes [13]. Given that CHCs serve as primary care homes for the nation’s most vulnerable populations [3], a comprehensive systematic review on CHC interventions to control diabetes among vulnerable populations is needed. To this end, the purpose of this study is to synthesize the evidence on CHC interventions. Specifically, we examined the characteristics of CHC interventions and the outcomes in people with diabetes. Our review systematically extends the previous efforts by providing an understanding of: (1) what constitutes CHC interventions (type and contents); (2) who delivers CHC intervention; and (3) how CHC interventions achieve desired effects.

Methods

Search Strategy

Following consultation with a health science librarian, four databases were searched. PubMed, EMBASE, CINAHL, and Scopus were searched using MESH and Boolean search techniques. Search terms included: "Community Health Centers” OR "community health center" OR "satellite center" OR "satellite centers" "neighborhood health centers" AND "Chronic Disease” OR "Diabetes Mellitus" OR "chronic disease" OR "chronic diseases" OR "chronic illness" OR "chronically ill" OR "chronic illnesses". Truncations of these words were also used in the electronic searches. See Appendix for specific terms used for each database.

Selection of Studies

The search was conducted in January 2018. A total of 1657 references were retrieved from the electronic searches and imported into the Covidence software. Of these, 765 duplicates were removed, and 892 studies were forwarded to title and abstract screening. Two reviewers independently conducted an initial screening of titles and abstracts for relevance to diabetes. After title and abstract review, 671 articles were excluded because they were irrelevant. Also, conference abstracts had limited information on study characteristics and thus were also excluded. 221 abstracts were identified for full review. Two reviewers independently evaluated full-text articles to determine eligibility. Articles were included in this review if the study was: about diabetes mellitus, published in the English language and involved participants who were 18 years and older. Additionally, only intervention studies that were conducted within a CHC setting in the U.S., and studies that reported patient outcomes were included. Subsequently, 192 articles were excluded for the following reasons: wrong study design (n = 68), abstract only (n = 33), wrong intervention (n = 25), wrong study population (n = 24), supplement (n = 21), wrong outcomes (n = 12), wrong setting (n = 4), wrong patient population (n = 3), wrong language (n = 2). All references were screened at every stage by two independent reviewers. Disagreements were resolved through consensus. Twenty-nine articles met the inclusion criteria. Figure 1 provides details of the selection process.

Fig. 1
figure 1

PRISMA diagram

Full size image

Data Extraction

Relevant data were extracted by four trained research assistants using a standardized data extraction form developed by the authors. The following data were extracted from the included studies: first author, year, study design, sample, recruitment methods, retention rate, setting, study outcomes and measurement, main findings, intervention type, modality and dose, control or comparison condition, and interventionist. An independent research assistant reviewed extracted data to check accuracy. Any discrepancies were resolved through discussions among all research assistants and authors.

Quality Appraisal

Each study was evaluated for its quality, based on published quality rating scales. Specifically, we used the quality rating scales published by the Joanna Briggs Institute for experimental and quasi-experimental studies [14]. The total quality rating scale score ranged from 0 to 13 for experimental studies and from 0 to 9 for quasi-experimental studies. Zero is the lowest quality and 13 (or nine) indicates the highest quality. Based on the possible range of scores, studies with quality ratings of 0–4, 5–8, and 9 + were categorized as low, medium, and high quality for experimental studies and 0–3, 4–6, and 7 + for quasi-experimental studies, respectively. Two authors rated each study for its quality independently. Inter-rater agreement statistics using percent agreement ranged from 44 to 100% (average 73%). Any discrepancies were resolved through team discussions.

Results

Overview of Studies

Table 1 summarizes the main characteristics of the 29 articles included [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. The review included 28 unique studies [15, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43] with one companion article [16]. The companion article [16] analyzed the cost-effectiveness of the parent study [15]. Two articles [36, 37] addressed the same intervention but with different outcomes and samples. Seventeen of the 28 studies were randomized controlled trials [15, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32] and the remaining studies (n = 10) were quasi-experimental [33,34,35,36,37,38,39,40,41,42,43]. Twenty-one studies focused solely on people with type 2 diabetes [15, 17, 19,20,21,22,23,24,25,26,27,28,29,30,31,32, 34, 39, 40, 42,43,44], while four studies focused on both type 1 and type 2 diabetes [18, 35, 36, 38]. One study included people with various chronic diseases including type 2 diabetes [33] and one study did not specify the type of diabetes [41].

Table 1 Study characteristics
Full size table

The sample sizes ranged from 14 [18] to 10,000 [41] and the enrolled participant age ranged from 18 [37] to 89 years [33]. The majority of participants were female who consisted of 52% [24] to 89% [19] of the study samples except for a few in which female participants made up 50% or less of the sample [38, 41, 42]. Ten studies focused solely on interventions targeting Hispanic and Latino patients [20, 26,27,28, 31,32,33,34, 40, 42] and several studies included populations where African Americans [15, 19, 25, 35, 43] or Hispanics [37, 38] comprised more than 50% of the study sample. Other minority ethnicities included were Native Americans and Asian Pacific Islanders [19], and Native Hawaiians [30].

Studies’ recruitment methods included reviewing medical records or billing claims [15, 20, 22, 26, 32, 39], household mailings [17, 28, 34, 40], and primary care provider referrals [17, 21, 24, 25, 27, 28, 30, 31, 34, 35, 40, 41, 43]. Retention rates were reported in 20 studies [17, 19, 20, 23, 25,26,27,28,29,30, 32, 33, 37, 39,40,41,42,43]. Of the studies, retention rates ranged from 18% retention over 8 months for both the control and intervention together in one study [28] to 100% for the intervention group over 12 weeks in another [31].

Characteristics of CHC Interventions

Table 2 describes detailed characteristics of CHC interventions included in the review. The CHC interventions varied greatly in terms of type and modality of the interventions. Twelve studies used one-on-one education sessions [15, 24, 25, 27,28,29, 32, 33, 37,38,39, 43] with three of the studies incorporating follow-up telephone calls to answer patients’ questions [15, 29, 39]. One CHC intervention [33] incorporated individual education sessions into the patient’s routine clinic visits, while the remaining individual education sessions were independent of routine visits to their primary care provider. Four interventions used group education sessions [19, 28, 29, 31] with the education sessions ranging from 1 to 12 sessions [28]. Five CHC interventions [17, 18, 21, 22, 44] used the telephone as the main method of communication with one intervention sending daily text messages following a one-day workshop [21]. In addition, three studies included telephone calls as a method of follow-up with participants [15, 24, 29]. Other intervention methods included one diabetic complication screenings [35], and a 1-day workshop that focused on mindfulness as a way to improve one’s management of diabetes [23].

Table 2 Characteristics of CHC interventions
Full size table

The main focus of the CHC interventions also varied. Clinical variables were measured as outcomes in numerous studies with 22 studies using HbA1C as a primary outcome [15, 17,18,19,20,21, 23, 24, 26, 27, 29,30,31,32, 34, 36,37,38,39,40, 42, 43] and eleven studies including cholesterol as an outcome [15, 17, 19, 20, 24, 26, 28, 29, 38,39,40]. Thirteen studies used diabetes knowledge as an outcome by educating patients about diabetes self-management topics (diet, exercise smoking cessation, and stress) [17, 19, 23,24,25, 31, 32, 34, 36, 38, 39, 42, 43]. Other studies included medication management [15, 19, 22, 29], goal setting and achievement [20, 33, 34], and depression screenings [34] as part of their interventions.

Additionally, two studies [18, 43] focused on increasing physical activity levels and one study [23] taught mindfulness techniques as a way to improve one’s self-management of diabetes. Adjusting education to the appropriate health literacy levels [17, 25, 34, 42] and bilingual and bicultural education [27, 28, 33, 34, 36, 40, 45] were incorporated into several studies. Furthermore, two studies [27, 28] culturally tailored group education sessions to include making traditional foods healthier and created Latin American styled soap opera dramas to convey diabetes-specific education in a familiar format. In addition, one study utilized physical activity [18] and three studies set behavior change goals [20, 33, 34] as the main methods of diabetes management. One of the goal setting studies [20] had patients create goals after completing a computer-based assessment of their motivational readiness. Then patients planned ways to achieve these goals with their healthcare provider and reviewed their goals at 3, 6, and 9-month follow-up visits.

The included studies used a variety of control or comparison conditions. For example, thirteen studies [17, 21, 22, 24,25,26, 29,30,31,32,33, 38, 39] provided routine care for their control group, whereas five studies enhanced their patient’s usual care with diabetes education packets [15, 18,19,20, 27]. One of the studies using enhanced usual care [27] also gave their control group their lab results. Seven studies had no control group [35,36,37, 40,41,42,43].

Throughout the studies, different interventionists were used to deliver the intervention. Providers used to deliver the diabetes education interventions included one or more of the following: physicians [19, 20, 35], nurse practitioners [15, 19, 42], registered nurses [17, 21, 24, 29, 32, 33, 35, 39], pharmacists [22, 29, 38], and dietitians [19, 24, 29, 32, 35, 39] with providers often trained as certified diabetes educators. Studies also used community health workers [15, 24, 36, 43], medical assistants [24, 32, 39,40,41,42], and peer educators [26, 39] also known as “promotoras.” Promotoras in these studies [26, 39] were former intervention recipients who were selected to complete a training and mentoring program and then deliver supervised education sessions while following a detailed curriculum.

Intervention fidelity was monitored in several interventions, but it cannot be detailed extensively due to the lack of reporting among all of the studies in our systematic review. Of available studies in which fidelity was discussed, one study used direct observation [40] as a method to ensure that the content for group education sessions was delivered correctly. Group education sessions also reported using protocol adherence check [28] to maintain intervention fidelity. Peer educators in two different studies [30, 39] followed scripted manuals in order for the same curriculum to be delivered to all participants. One study audio recorded classes and reviewed that all appropriate content was covered in order to ensure fidelity [26]. During individual education sessions, Allen et al. [15] confirmed fidelity by recording the number and length of patient-provider encounters, documenting protocol adherence, and conducting quality assurance assessments.

Effects of CHC Interventions

Twenty-two CHC interventions used HbA1C as a primary outcome [15, 17,18,19,20,21, 23, 24, 26, 27, 29,30,31,32, 34, 36,37,38,39,40, 42, 43]. Of the twenty-two studies that included HbA1c as an outcome, 14 studies [15, 19, 26, 27, 29, 30, 32, 34, 36,37,38,39,40, 43] saw statistically significant decreases in HbA1c, eight studies [17, 18, 20, 21, 23, 24, 31, 42] did not see statistically significant decreases, and one study [28] saw significant decreases at 4 months, but not at 12 months. Additionally, Sinclair et al. [30] reported significant increases in diabetes understanding (p < 0.001) and self-management (p < 0.001). Five studies that implemented both individual and group education sessions also reported significant decreases in HbA1c [19, 26, 29, 39, 40], triglycerides [19], and total cholesterol [26, 39].

The five CHC interventions that were delivered primarily over the phone [17, 18, 21, 22, 41] showed mixed effects on people with diabetes. In all four studies where clinical outcomes (i.e. HbA1c, BMI, cholesterol, weight) were measured [17, 18, 21, 22], telephone counseling exhibited no significant improvements on these outcomes. One study [41] that used the telephone to remind patients of scheduled appointments and tests seven days prior reported an increased number of patients seen in clinic and an increase in eye, foot, dental, and urine exams.

Other outcomes included diabetes goal attainment and self-efficacy. Anderson et al. [34] found that participants who achieved their goals successfully maintained or decreased their HbA1c of < 7.0% with an average decrease of 0.9 percentage points over the 3-year study period. Hargraves et al. [24] saw a significant difference (p < .01) in goal-setting between patients with or without a community health worker, but no statistically significant clinical outcomes were associated with goal-setting. Another study [20] that set goals after assessing a participant’s motivational readiness saw reductions in HbA1c levels. Self-efficacy promotion was also mixed. Two studies (routine clinic visit and group education session) did not see significant differences in self-efficacy [31, 33], while one study [21] that used a 1-day workshop with text message reminders found significance (p = 0.008) in self-efficacy scores after 3-month compared to baseline.

Quality of Included Studies

Results of the quality assessments are detailed in Table 3. The average quality scores for the seventeen randomized controlled trials [15, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32] was 8.5 of a maximum possible score of 13 (range 3 to 10). Eight studies were rated high-quality (9 or higher) [15, 17, 20, 23, 26,27,28, 31], eight studies fell under the medium-quality category (scores of 5 to 8) [19, 21, 22, 24, 25, 29, 30, 32], and one had a low rating scale of 3 [18]. Five of the RCTs either did not apply true randomization for assignment of participants to treatment groups [8, 24, 29] or it was unclear if true randomization was used [21, 22]. There were eleven quasi-experimental studies scored [33,34,35,36,37,38,39,40,41,42,43] with an average quality rating of 6.9 (range 6 to 8; maximum possible score = 9). Nine studies met the criterion of high-quality (7 or greater) [33, 34, 36, 37, 39,40,41,42,43], while the other two were of medium-quality (both with scores of 6) [35, 38]. Only three of quasi-experimental studies had control groups [33, 38, 39].

Table 3 Study quality ratings
Full size table

Discussion

To the best of our knowledge, this is the first systematic review that provides a critical appraisal of CHC interventions targeting persons with diabetes. The included studies evaluated a variety of clinical outcomes encompassing HbA1c, blood pressure, and lipids, as well as psychosocial and behavioral outcomes such as diabetes knowledge, self-management, adherence to treatment regimen (e.g., screening for complications), goal setting, and self-efficacy. CHC interventions were generally effective in HbA1c reduction either via individual education [15, 27, 32, 36, 37], group education [30, 43] or both [19, 26, 29, 39, 40], although insignificant HbA1c reductions were also noted in nine studies [17, 18, 20, 21, 23, 24, 28, 31, 42]. CHC interventions were also effective in improving lipids [15, 19, 26, 39] and systolic blood pressure [15, 39] but only a small number of studies addressed them as study outcomes. Similarly, there was limited evidence to show that CHC interventions were effective in improving diabetes knowledge [30], self-management [30], screening for complications [41], goal attainment [20, 24, 34], and self-efficacy [21]; two studies failed to find significant changes in self-efficacy as a result of CHC intervention [31, 33].

The different type and modality of CHC interventions used in the included studies might explain why some interventions were effective in achieving study outcomes while others were not. For example, the CHC interventions primarily delivered over the phone failed to yield any improvement in clinical outcomes including HbA1c [17, 22], in comparison to other forms of educational interventions using face-to-face interactions which most often resulted in significant reductions in HBA1c [15, 19, 26, 27, 29, 30, 32, 36, 37, 39, 40, 43]. The findings suggest that CHC interventions would benefit from incorporating face-to-face education as an essential component of the intervention and that phone-based delivery be considered only as a means to supplement and not to replace the face-to-face interaction.

The majority of study participants included in the studies were racial/ethnic minorities and female. Given the demographics of patients served at CHCs, it is not surprising that the CHC interventions involved racial/ethnic minorities as study participants, most often of Hispanic [20, 26,27,28, 31,32,33,34, 37, 38, 40, 42] or African American ethnicities [15, 19, 25, 35, 43]. Female patients were also overrepresented in the study samples included in the review. Recent epidemiological studies point to higher prevalence rates of type 2 diabetes in subgroups of Asian American and Pacific Islanders (e.g., Pacific Islanders, South Asians, and Filipinos) than any other racial/ethnic groups, including minorities traditionally considered high risk such as African Americans, Latinos, and Native Americans [46, 47]. Future CHC intervention research should consider targeted recruit strategies to identify and enroll more diverse and understudied groups such as Asian American and Pacific Islanders and male participants in the study sample.

Only one article [16] included in the review examined the cost-effectiveness of the CHC intervention reported in the primary study [15]. In the era of healthcare expenditure ever so rising, a cost-efficient model of diabetes care is an important avenue for future research. For example, prior reviews of community health worker intervention studies [48,49,50,51] concluded that working with CHWs is a reimbursable, alternative model of care for promoting cardiovascular risk reduction including diabetes management in low-income, racial and ethnic minority populations. One of the main aims of the U.S. Affordable Care Act (ACA) is to reduce the costs of healthcare for individuals and the government [52]. The ACA acknowledges community health workers as an important part of healthcare teams for the delivery of care, particularly among medically underserved populations and communities. Clearly, the ACA presents opportunities to include community health workers as a core component of effective CHC intervention teams.

There are methodological issues to be taken into consideration when interpreting the findings in this review. While 17 of 28 studies (61%) included in the review [15, 17, 20, 23, 26,27,28, 31, 33, 34, 36, 37, 39,40,41,42,43] were of high quality, many lacked methodological rigor, which might have led to negative findings. For example, 11 of 28 studies included in the review [33,34,35,36,37,38,39,40,41,42,43] used a quasi-experimental study design and seven of them [35,36,37, 40,41,42,43] lacked a control group, hence subject to substantial threats to internal validity. In addition, majority of the studies using an RCT design lacked any discussion regarding blinding [15, 18, 19, 21,22,23,24,25,26,27, 29,30,31,32], hence it was unclear whether interventionists took on multiple roles (e.g., recruitment and/or data collection responsibilities) as well as the delivery of the intervention. Study members with multiple roles would be expected to lead to disclosure of group allocation and therefore could result in bias in how the intervention is delivered and also measurement bias, hence, threatening the internal validity of the results. Finally, intervention fidelity was monitored in less than one third of the studies [15, 26, 28, 30, 39, 40] included in the review. Variability in fidelity of CHC intervention implementation could be possible explanations for non-significant effects found in some studies. Previous systematic reviews [50, 53] underscored the importance of required training and competency levels for community interventionists in relation to assigned responsibilities. There is a strong need for studies to clearly elaborate the contents and processes of interventionists training such as competency evaluation and supervision to optimize the use of this approach.

A few limitations of this review should be noted. First, it is possible that we did not find all relevant articles in the literature. To avoid this, we conducted an extensive systematic electronic search using a compressive list of MeSH terms, after a consultation with an experienced health science librarian, in addition to hand searches of references of the identified studies. We did not include grey literature such as reports from organizations and conference abstracts; this could result in publication bias for this review, but the included studies are also final (not preliminary, as conference results may be) and peer-reviewed. The CHC predominantly serves low-income, racial/ethnic minority populations. Thus, our findings may not be applicable to other populations such as mid- or high-income populations. Finally, we included only articles written in English; therefore, the findings cannot be generalized to populations for which studies are published primarily in non-English languages.

Conclusions

CHC-based health promotion is a growing trend for the care of low-income patients. CHCs have a unique role in spearheading community health promotion and may be the mechanism by which to establish close ties between healthcare providers and community members who often lack sufficient resources and support for diabetes control. CHC intervention models also support the movement from a healthcare system that focuses only on “sickness care” to one that is also “prevention-focused.” Our review of 29 articles shows that CHC interventions are in general effective in promoting glucose control when using face-to-face interactions in low-income, underserved, and racial and ethnic minority patients with diabetes. There was positive but limited evidence to show that CHC interventions were effective in improving diabetes knowledge, self-management, screening for complications, goal attainment, and self-efficacy. Our findings support the use of CHC intervention for glucose control and suggest the need for more rigorous and continued evaluations of this intervention approach.