1414 Care Volume 41, July 2018

Michael S. Spencer,1 Edith C. Kieffer,1 Outcomes at 18 Months From Brandy Sinco,1 Gretchen Piatt,1 Gloria Palmisano,2 Jaclynn Hawkins,3 a Community Health Worker Alana Lebron,4 Nicolaus Espitia,1 Tricia Tang,5 Martha Funnell,1 and and Peer Leader Diabetes Michele Heisler1 Self-Management Program for Latino Adults Diabetes Care 2018;41:1414–1422 | https://doi.org/10.2337/dc17-0978 CLIN CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL

OBJECTIVE This study evaluated the effectiveness of a community health worker (CHW) diabetes self-management education (DSME) program, followed by two different approaches

to maintain improvements in HbA1c and other clinical and patient-centered outcomes over 18 months.

RESEARCH DESIGN AND METHODS The study randomized 222 Latino adults with type 2 diabetes and poor glycemic control from a federally qualified health center to 1) a CHW-led, 6-month DSME program or 2) enhanced usual care (EUC). After the 6-month program, participants randomized to the CHW-led DSME were further randomized to 1)12monthsofCHW- delivered monthly telephone outreach (CHW-only) or 2) 12 months of weekly group sessions delivered by peer leaders (PLs) with telephone outreach to those unable

to attend (CHW+PL). The primary outcome was HbA1c. Secondary outcomes were blood pressure, lipid levels, diabetes distress, depressive symptoms, understanding of diabetes self-management, and diabetes social support. Assessments were conducted at baseline and at 6, 12, and 18 months.

RESULTS

Participants in the CHW intervention at the 6-month follow-up had greater decreases 1University of Michigan, Ann Arbor, MI 2 in HbA1c (20.45% [95% CI 20.87, 20.03]; P < 0.05) and in diabetes distress (20.3 Community Health and Social Services (CHASS), points [95% CI 20.6, 20.03]; P < 0.05) compared with EUC. CHW+PL participants Detroit, MI 3 maintained HbA improvements at 12 and 18 months, and CHW-only participants Michigan State University, East Lansing, MI 1c 4University of California at Irvine, Irvine, CA maintained improvements in diabetes distress at 12 and 18 months. CHW+PL 5University of British Columbia, Vancouver, participants also had significantly fewer depressive symptoms at 18 months British Columbia, Canada compared with EUC (22.2 points [95% CI 24.1, 20.3]; P < 0.05). Participants in Corresponding author: Michael S. Spencer, CHW-led DSME had significant improvements in diabetes social support and in [email protected]. understanding of diabetes self-management at 6 months relative to EUC, but Received 23 May 2017 and accepted 7 April 2018. theseinterventioneffectswerenotsustainedat18months. Clinical trial reg. no. NCT00800410, clinicaltrials .gov. CONCLUSIONS © 2018 by the American Diabetes Association. This study demonstrates the effectiveness of a 6-month CHW intervention on key Readers may use this article as long as the work diabetes outcomes and of a volunteer PL program in sustaining key achieved gains. is properly cited, the use is educational and not for profit, and the work is not altered. More infor- These are scalable models for centers in low-resource settings for mation is available at http://www.diabetesjournals achieving and maintaining improvements in key diabetes outcomes. .org/content/license. care.diabetesjournals.org Spencer and Associates 1415

Type 2 diabetes (T2D) disproportionately intervention in improving clinical outcomes patient-centered outcomes compared affects low-income racial and ethnic mi- (HbA1c, blood pressure, and lipid levels), with the delayed control group (8). Both nority groups and is a growing public psychosocial outcomes (depressive symp- studies were limited to outcomes mea- health problem. Diabetes self-management toms and diabetes-related distress), di- sured immediately after the conclusion education (DSME) is necessary but often abetes self-management behaviors, and of the 6-month CHW-led program. The insufficient to sustain the substantial self- understanding of diabetes self-management current study was conducted with a third management effort needed during a life- compared with enhanced usual care (EUC) cohort of participants with poorly con- time with diabetes (1). Many patients need immediately after the intervention at trolled diabetes. on-going diabetes self-management sup- 6 months. We then examined two alter- port (DSMS) (2). DSMS is defined as native lower-intensity approaches from Study Procedures “activities that assist the individual with 6 to 18 months of follow-up. Specifically, The University of Michigan Institutional diabetes to implement and sustain the we examined whether an additional PL Review Board approved this study. From on-going behaviors needed to manage intervention, implemented after the con- October 2009 to February 2013, we re- their illness (3).” clusion of the 6-month CHW-led interven- viewed computer-generated lists of all As trusted members of their commu- tion, enhanced maintenance of improved potentially eligible CHASS patients with nities, trained community health workers outcomes compared with monthly CHW- -diagnosed diabetes who were fi (CHWs) are particularly effective in reach- only follow-up or EUC alone. at least 21 years old and self-identi ed ing and providing both DSME and DSMS to as Latina/o. Individuals with physical lim- members of communities of color who itations preventing participation, termi- face numerous barriers to diabetes self- RESEARCH DESIGN AND METHODS nal health conditions, serious psychiatric management (4–7). CHW interventions Study Population illness, and self-reported excessive alco- in diabetes have led to improved self- Founded in 1999, the Racial and Ethnic hol or illicit drug use were excluded. monitoring, self-care, lifestyle change, Approaches to Community Health (REACH) From the initial list of 1,049 patients, and blood glucose control outcomes com- Detroit Partnership is a community-based 295 were eligible based on intake screen- pared with control groups (3–5,7,8). Most participatory research (CBPR) coalition of ing, 565 did not meet inclusion criteria, studies, however, report postintervention community organizations, academic insti- 126 refused to participate, and 63 were outcomes only at 6 and 12 months. A key tutions, and health care systems that has unable to be contacted (see Fig. 1). There challenge is to sustain immediate gains used CBPR approaches to design, imple- were no differences in sex and ethnicity after a DSME intervention over longer ment, and evaluate interventions aimed at between eligible participants and non- periods of follow-up (4,8). improving diabetes care and outcomes in participants,butnonparticipantswerean One particularly promising approach east and southwest Detroit (15). All work average of 3.7 years older than partic- tosustaingainsfrommoreintensiveCHW has been conducted with the Community ipants (P = 0.012). Of the 295 eligible, interventions is to offer less intensive Health and Social Services Center (CHASS), 234 initially signed consents and 12 support services by peer leaders (PLs) afederallyqualifiedcommunityhealthcen- withdrew before baseline, leaving 222 (4,8). PLs share important characteristics ter serving the predominantly Latino com- to be randomized. After participants with participants, including having dia- munity in southwest Detroit, and guided provided informed consent, laboratory betes, and are volunteers or receive by a steering committee of partnership and anthropometric measurements were small stipends to reimburse any expenses members. collected, and participants completed (9–12). Some PL-led interventions have We developed a culturally tailored baseline questionnaires. The 222 partic- ipantswerefirstrandomizedintotheCHW resulted in improvements in HbA1c com- DSME curriculum, “Journey to Health/ pared with control groups (13). Other PL El Camino a la Salud,” grounded in the intervention arm (n = 149) or the EUC interventions did not result in greater im- empowerment approach that emphasizes (n = 73) arm using a computer-generated process with concealed allocation. At provements in HbA1c than control groups a collaborative approach to facilitate self- but found improved self-empowerment directed behavior change of patients 6 months (immediately after the CHW scores, self-care indicators (9), patient ac- (15,16). Empowerment-based approaches intervention), CHW intervention partic- tivation (12), and diabetes-related distress are effective in improving chronic disease ipants were further randomized into (10). One study found a PL program sus- self-management among racial and ethnic the CHW-only intervention (n =89)or tained gains in glycemic control and several minority patients (17). the CHW+PLintervention(n=60)groups. key patient-centered outcomes achieved We evaluated the 6-month CHW in- EUC through a CHW-led DSMS program, but tervention in two prior cohorts of adult The EUC group received a 2-h class con- there was no comparison with a usual care AfricanAmericanandLatinopatientswith ducted by a research assistant covering fi group (14). poor glycemic control. In the rst, we how to interpret their clinical and an- Given the effectiveness and potential compared the effectiveness of the CHW thropometric results. EUC participants low cost of PL-led DSMS, further research intervention with a concurrent observa- were contacted once each month to fi is needed to test whether PLs may be tional control group and found signi cant update contact information. effective in helping to maintain gains improvements in HbA1c and other clinical achieved through more intensive DSME outcomes (18). Our second cohort study, CHW-Led DSME programs led by CHWs compared with using a randomized, 6-month delayed Three trained CHWs conducted activi- usual care. Therefore, the current study control group design, also found signif- ties during the initial 6 months of the examined the effectiveness of a CHW icant improvements in HbA1c and other intervention period. The CHWs were all 1416 Outcomes From CHW and PL DSME Program Diabetes Care Volume 41, July 2018

Figure 1—Consolidated Standards of Reporting Trials flow diagram.

Spanish-fluent Latinas who had complet- behaviormodification strategies,cultural and help participants formulate their ed high school or had a GED and were competency, and CPBR principles. CHWs own action plans. recruited from the southwest Detroit were trained in empowerment-based During the 6-month intervention, community. The CHWs underwent .160 approaches to inform their approach to CHWs conducted 1) DSME classes, 2) h of CHW training, .80 h of diabetes ed- each component of the intervention (19), two 60-min home visits each month, ucation, including home visit experiences, including motivational interviewing (20), and 3) one clinic visit with the participant and training in human subjects protocols, which is used to elicit participants’ goals and his or her provider. The care.diabetesjournals.org Spencer and Associates 1417

diabetes self-management classes, Jour- and its care, and set self-management instrument that assesses emotional dis- ney to Health/El Camino a la Salud, were goals. The PLs helped participants set tressandfunctioningspecifictolivingwith culturally tailored group classes in both goals using the same five-step goal- diabetes, with higher scores indicating English and Spanish (8,16). Eleven 2-h setting model described above. PLs also higher levels of distress (25). We assessed group sessions of 8–10 participants were provided support to participants by dis- diabetes-specific social supportwithan held every 2 weeks at community loca- cussing psychosocial concerns, identify- adapted version of the Diabetes Support tions. The development, implementa- ing facilitators and barriers to behavior Scale, a six-item instrument that assesses tion, and evaluation of these curricula change, taking inventory of support sour- perceived social support as it relates to are described in depth elsewhere (16). ces, and developing strategies to navigate meeting emotional needs, seeking advice, CHWs used the empowerment approach the health care system. and obtaining information, with higher to diabetes education by eliciting par- To ensure regular contact with each scores indicating more support (26). De- ticipants’ experiences and requests for participant, PLs made a telephone sup- pressive symptoms were assessed with the information (17). CHWs also helped par- port call to any participant who had not Patient Health Questionnaire-9 (27). Un- ticipants improve communication skills attended a session over 3 consecutive derstanding of diabetes self-management with their providers and facilitated nec- weeks. During the telephone support calls, was computed from 16 questions from the essary referrals to other service systems. PLs facilitated a conversation that closely Diabetes Care Profile (28,29). During home visits, and phone calls every mirrored support activities conducted in 2 weeks, CHWs helped participants set the group setting. Statistical Methods goals using the five-step goal-setting All baseline characteristics were compared model, which included 1)exploringa CHW-Only between the intervention groups and EUC participant-identified problem, 2)discus- After the 6-month intervention, partic- group with the Fisher F test for one-way sing the emotional impact of the problem, ipants randomized to this group received ANOVA (30). The log-rank test was used 3) selecting a self-management goal, 4) monthly telephone calls from a CHW who to compare “diabetes duration (years)” developinganactionplan,and5)executing had led their DSME group to check in and between the two groups (31). Categorical and evaluating the action plan (21). assess continued progress in setting and variableswere comparedbetweengroups meeting diabetes care goals. with the Pearson x2 test or with the Fisher exact test for rare outcomes. CHW+PL All Participants The PLs were recruited by the CHWs Information about community activities that Participants were analyzed as part of among patients with diabetes who had were free and publicly available was pro- their original group assignments. The previously successfully completed the vided to all study participants, who all were outcomes were evaluated for interven- Journey to Health/Camino a la Salud pro- receiving ongoing health care at CHASS. tion effects by using linear mixed models gram. They completed 46 h of training led (LMMs)toaccountforrepeatedmeasures. by the CHWs over 12 weeks (11). Besides Study Measures Because education differed significantly the initial training, the CHWs led monthly The primary clinical outcome was HbA1c, between the treatment groups, all out- booster support sessions with the PLs and measured with a Bayer DCA 2000+ An- come analyses were adjusted with a binary provided them with ongoing supervision. alyzer (23). This assay has a test coefficient indicator for high school graduation. All participants in the PL intervention of variation ,5% as required by the The outcomes in the LMMs were meas- completed the CHW intervention from National Diabetes Data Group. Secondary urements at baseline and at 6, 12, and baseline to 6 months and then were clinical outcomes included a lipid panel 18 months, with covariates of indica- randomized to participate in the PL in- (total cholesterol, LDL cholesterol [LDLc], tor variables for time (baseline, 6, 12, tervention. Adapted from the Lifelong and HDL cholesterol [HDLc]), using the 18 months), group, group 3 time inter- Management program of Tang et al. Cholestech LDX (Cholestech, Hayward, action, and education. Medication inten- (22), the PL intervention was designed CA) point-of-care machine that meets sification was included for HbA1c and to provide patients with ongoing emo- National Cholesterol Education Program cholesterol but omitted for blood pressure tional and behavioral support through guidelines for measuring lipid levels (24). becauseofminimalchangeinmedications weekly drop-in group-based sessions and Systolic and diastolic blood pressures during the study period (32,33). To ensure follow-up telephone contacts from 6 to were taken with two readings on a Welch that medication treatment intensification 18 months. Participants were invited Allyn Speidel & Keller sphygmomanometer; did not confound the intervention effect, to attend weekly group diabetes self- the averaged readings were used in the HbA1c and lipids outcomes were analyzed managementsessionsasoftenasneeded. analysis. All participants were weighed on with and without medication intensifica- Based on patient-empowerment princi- an EverWeigh lithium digital scale. Height tion (i.e., change in number or dose of ples, discussion topics were driven by and waist circumference were measured medicine). Changes in medications were patients’ self-identified priorities, ques- bythe same technician at each time point. calculated from baseline to 6 months, 6 to tions, and concerns (19). The PLs sought to BMI was calculated as weight in kilograms 12 months, and 12 to 18 months. complete the following five tasks at each divided by the square of height in meters. Changes in physical and psychosocial session: discuss recent self-management Waist circumference was measured at the outcomes (and 95% CI) from baseline to challenges, share feelings about these umbilicalwaist using the Tech-Med model each follow-up time point were estimated challenges and other aspects of living with (cat. no. 4414) measuring tape. for the intervention and control groups diabetes,engageingroup-basedproblem We assessed diabetes-related distress by post hoc contrasts. From baseline to solving, address questions about diabetes using the Diabetes Distress Scale, a 17-item 6 months, there were two treatment 1418 Outcomes From CHW and PL DSME Program Diabetes Care Volume 41, July 2018

groups, EUC and CHW. After 6 months, was therefore included as a covariate in A different pattern was apparent for the CHW group split into the CHW-only outcome analyses, coded as a binary in- LDLc and blood pressure levels. There were and CHW+PL groups. Thus, estimates dicator for high school graduation. No no differences in LDLc levels among groups after 6 months compared the EUC, CHW- physical or psychosocial outcomes signif- from baseline to 6 and to 12 months. At only, and CHW+PL groups. To check for icantlydifferedbetweengroupsatbaseline. 18 months, significant decreases in LDLc multiple comparisons, all P values for Figure 2 shows change in HbA1c levels from baseline in the CHW+PL group the contrasts were double-checked by over time. Among participants receiving (212.3 mg/dL [95% CI 223.1, 21.6]; using Monte Carlo simulation (34). For the CHW intervention, mean HbA1c de- P , 0.05) were observed. However, the each outcome analysis, we included all creased by 20.51% (95% CI 20.75, 20.26; intervention effect for LDLc was not signif- available data, consistent with the intent- P , 0.001) from baseline to 6 months. A icant. Intervention effects were not found to-treat principle. No demographic or significant intervention effect was demon- for HDLc, total cholesterol, or BMI. Simi- baseline outcome measures differed stratedcomparedwithEUC(0.45%[95% larly, no intervention effects were found for significantlyby whether HbA1c was miss- CI 20.87, 20.03]; P , 0.05). From 6 to blood pressure; only within-group differ- ing at the 18-month follow-up. All signif- 12months,improvementsinHbA1cwere ences were observed over time. No inter- icance tests were two-tailed using a = sustained for participants randomized vention effects were observed for waist 0.05. SAS 9.4 software was used for all to the CHW+PL group (20.63% [95% circumference or the waist-to-hip ratio. analyses. CI21.06,20.19];P , 0.01)butnotthe Figure 3 shows the mean estimates To ensure that missing data were not CHW-only or the EUC groups. The inter- overtimefromLMMsforthepsychosocial biasing the results, all outcomes were an- vention effect for the CHW+PL group at outcomes. Depressive symptoms signif- alyzed with all available data and by using 12 months continued to be significant icantly decreased for participants receiv- multiple imputations with chained equa- (20.60% [95% CI 21.18, 20.01]; P , ing the CHW intervention from baseline tions (35). Twenty imputations were in- 0.05). From 12 to 18 months, the CHW to 6 months (21.1 points [95% CI 21.9, cluded to obtain .99% relative efficiency, +PL group maintainedreductionsinHbA1c 20.4]; P , 0.01), but no intervention with no changes in results (36). (20.56%[95%CI21.06, 20.05]; P , effect was observed. At 18 months, 0.05), and the intervention effect at 18 the CHW-only and the CHW+PL groups RESULTS months was also significant (20.76% both maintained the reductions in depres- Table 1 presents the baseline character- [95% CI 21.48, 20.05]; P , 0.05). HbA1c sion symptoms compared with baseline. isticsofthestudyparticipants.Educational results were similar after adjusting for At 18 months, only the CHW+PL grouphad status differed by treatment group and medication intensification. asignificant intervention effect compared

Table 1—Baseline characteristics of REACH Detroit, cohort 3 Group EUC CHW-only CHW+PL Total Characteristic (n = 73) (n = 89) (n = 60) (N = 222) P value for between groups Age (years) 48.5 (10.0) 48.2 (10.7) 50.2 (11.1) 48.9 (10.6) 0.488a Female, n (%) 49 (67.1) 54 (60.7) 32 (53.3) 135 (60.8) 0.269b High school graduate, n (%) 32 (43.8) 19 (21.3) 17 (28.3) 68 (30.6) 0.008b Employed full or part time, n (%) 32 (43.8) 37 (41.6) 26 (43.3) 95 (42.8) 0.954b Antihyperglycemic medication, n (%) No medication 3 (4.1) 4 (4.5) 3 (5.0) 10 (4.5) 0.839c Only oral diabetes medication 49 (67.1) 64 (71.9) 45 (75.0) 158 (71.2) Insulin, with or without medication 21 (28.8) 21 (23.6) 12 (20.0) 54 (24.3) Physiological measures a HbA1c (NGSP %) 7.7 (1.8) 7.7 (1.7) 8.2 (2.2) 7.8 (1.9) 0.136 LDLc (mg/dL) 95.4 (28.9) 92.2 (29.8) 102.1 (35.3) 95.8 (31.2) 0.219a HDLc (mg/dL) 37.6 (11.6) 40.7 (13.6) 40.5 (16.8) 39.6 (13.9) 0.347a BP (mmHg) Systolic 133.3 (15.8) 131.7 (17.5) 134.8 (17.8) 133.1 (17.0) 0.550a Diastolic 80.1 (10.9) 78.9 (10.0) 81.2 (10.1) 79.9 (10.3) 0.402a BMI (kg/m2) 32.3 (5.4) 33.9 (7.3) 33.1 (7.6) 33.2 (6.8) 0.299a Waist circumference (inches) 39.5 (5.4) 40.9 (5.6) 41.8 (6.3) 40.7 (5.8) 0.071a Psychological measures DDSd 2.0 (1.0) 2.2 (1.1) 2.0 (1.0) 2.1 (1.0) 0.657a DSSe 4.0 (1.1) 4.3 (1.0) 4.0 (1.2) 4.1 (1.1) 0.098a PHQf 4.8 (4.3) 5.8 (5.2) 6.5 (6.1) 5.7 (5.2) 0.187a DCPg 2.8 (0.9) 2.8 (0.8) 2.9 (0.8) 2.8 (0.8) 0.673a Data are mean (SD) unless otherwise indicated. BP, blood pressure; DCP, Diabetes Care Profile; DDS, Diabetes Distress Scale; DSS, Diabetes Support Scale; PHQ, Patient Health Questionnaire. Bold values are statistically significant (P , 0.05). aF test for equal means. bPearson x2 test. cFisher exact test. dDDS: ,2, little or no distress; 2 to ,3, moderate distress; $3, high distress. eDSS: 1 = strongly disagree to 6 = strongly agree. fInterpretation of depression level from the PHQ: 1–4, minimal; 5–9, mild; 10–14, moderate; 15–19, moderately severe; 20–27, severe. gDCP, which measures understanding of diabetes self-management: 1 = poor to 5 = excellent. care.diabetesjournals.org Spencer and Associates 1419

Figure2—TrajectoryofphysicaloutcomesovertimefromLMM.MeanestimatesfromLMM.Covariatesincludetimepoint,treatmentgroup,interaction between time and treatment group, and high school education. B, significant intervention effect relative to EUC (control) group and CHW-combined (CHW-only and CHW1PL groups) at 6 months; B1, significant intervention effect for CHW-only group at 12 or 18 months; B2, significant intervention effect for CHW+PL group at 12 or 18 months. *P , 0.05.

with EUC (22.2 [95% CI 24.1, 20.3]; P , 0.01,0.6];P,0.05)andtheCHWgroup(0.8 P,0.05)comparedwithEUCat12months 0.05). [95% CI 0.6, 1.0]; P , 0.001). However, the but not in the CHW-only group compared Diabetes-related distress significantly CHW group had a significantly higher with EUC(0.2 [95%CI 20.1,0.4]).However, decreased from baseline to 6 months increase than the EUC group, with an there was no significant intervention for participants receiving the CHW in- interventioneffectof0.5(95%CI0.2,0.8; effect at 18 months. None of the signi- tervention(20.4points[95%CI20.6,20.2]; P , 0.01). There were no significant ficant P values lost significance after P , 0.001), with an intervention effect differences across groups for the later being recomputed with Monte Carlo of 20.3 (95% CI 20.6, 20.03; P , 0.05). time periods. simulation. Improved distress scores were main- The EUC and CHW groups signifi- tained at 12 and 18 months for the cantly improved in their understanding CONCLUSIONS CHW-only and the CHW+PL groups. In of diabetes management from baseline Thisstudyfoundasignificantintervention addition, intervention effects were found to 6 months, with a significant interven- effect for HbA1c among low-income, ur- at12months(20.3[95%CI20.6,20.01]; tion effect for the CHW group relative banLatinoadultswithT2D,withdecreases P , 0.05) and 18 months (20.4 [95% to EUC (0.2 points [95% CI 0.01, 0.4]; P , at 6 months among participants receiving CI 20.7, 20.1]; P , 0.05) for the CHW- 0.05). The EUC, CHW-only, and the CHW the CHW DSME intervention. Participants only group but not for the CHW+PL +PL groups each maintained improve- in the CHW+PL group maintained these group. ment in the understanding of diabet- improvements at 12 and 18 months. Al- From baseline to 6 months, diabetes- es management at 12 and 18 months. though the CHW+PL group sustained sig- related support increased significantly An intervention effect was found in the nificant improvements in LDLc and in for both the EUC (0.3 points [95% CI CHW+PL group (0.3 [95% CI 0.1, 0.6]; systolic and diastolic blood pressure at 1420 Outcomes From CHW and PL DSME Program Diabetes Care Volume 41, July 2018

Figure 3—Trajectory of psychological outcomes over time from LMM. Mean estimates from LMM. Covariates include time point, treatment group, interaction between time and treatment group, and high school education. PHQ-9, Patient Health Questionnaire-9. B, significant intervention effect relative to EUC (control) group and CHW-combined (CHW-only and CHW1PL groups) at 6 months. B1, significant intervention effect for CHW-only group at 12 or 18 months; B2, significant intervention effect for CHW+PL group at 12 or 18 months. *P , 0.05; **P , 0.01.

18 months, differences between groups the REACH Detroit Partnership on the combination of CHW and PL services in these clinical measures were not effectiveness of the culturally tailored, after an empowerment-based DSME significant. The CHW+PL group had a CHW-led DSME program immediately af- provides an efficient and low-cost means significant intervention effect on reduc- ter the intervention at 6 months (8–18,37). for continued support for diabetes self- ing depressive symptoms at 18 months. A strength of this study is that it extends management. Diabetes-related distress decreased sig- those findings by demonstrating that clin- This study reinforces the importance of nificantly for CHW intervention partici- ical and psychosocial outcomes can be examining outcomes beyond the conclusion pants at 6 months, with intervention sustained to 18 months through the on- of short-term diabetes self-management effects at both 12 and 18 months for going support of PLs trained and super- education programs. Although HbA1c, the CHW-only group. Diabetes-related vised by CHWs compared with usual care. diabetes-related distress, diabetes-related support had a significant intervention Although CHW diabetes interventions support, and self-management knowledge effect at 6 months. Understanding of have demonstrated varying degrees of improved immediately after the 6-month diabetes management showed a signifi- short-term success (5), very few studies CHW intervention, LDLc and depressive cant intervention effect at 6 months, and haveexaminedoutcomesforan18-month symptom outcomes did not show a signif- the intervention effect was sustained to period. There is overwhelming evidence icant intervention effect until 18 months. It is 12 months in the CHW+PL group. that for many adults with diabetes, gains important to note that our intervention did These findings build on prior research achieved through short-term diabetes self- not target LDLc or depression as primary in several key ways. Most significantly, management programs are not sustained outcomes, although information on both the study confirms previous research by without ongoing support (38,39). The was provided through the curriculum. The care.diabetesjournals.org Spencer and Associates 1421

intervention also did not target weight loss, seeking to provide effective services to RB, Eds. Hamilton, Ontario, Canada, BC Decker, which may account for the nonsignificant low-incomeLatinocommunities.Ourfind- 2001, pp. 207–251 results for BMI and waist circumference. ings also demonstrate the feasibility of 2. Haas L, Maryniuk M, Beck J, et al.; 2012 Standards Revision Task Force. National standards Increasing education and support for these conducting rigorous research involving for diabetes self-management education and outcomes as part of our intervention could low-income communities of color using support. Diabetes Care 2012;35:2393–2401 improve these outcomes sooner and more CBPR principles and methods. Finally, our 3. Norris SL, Chowdhury FM, Van Le K, et al. effectively. Improvements at 18 months study contributes to the literature on the Effectivenessofcommunityhealthworkersinthe are encouraging for LDLc and depressive need for expanded and sustained CHW pro- care of persons with diabetes. Diabet Med 2006; 23:544–556 symptoms. grams and the hybrid CHW+PL model that 4. PalmasW,FindleySE,MejiaM, etal.Resultsof The current study also demonstrates we tested. The spread of these models has the northern Manhattan diabetes community that a linked CHW+PL intervention, with been slowed by inadequate and unstable outreach project: a randomized trial studying a CHWs providing initial and monthly funding. Policy changes supporting sustain- community health worker intervention to im- booster training and supervision for able funding and reimbursement models prove diabetes care in Hispanic adults. Diabetes Care 2014;37:963–969 PLs, can be a successful model. We did forCHWandPLprograms,aswellasongoing 5. Kim K, Choi JS, Choi E, et al. Effects of not include an arm that received only PL research on the cost effectiveness of these community-based health worker interventions to services. Although this could be considered programs, could lead to a greater reach in improve chronic disease management and care a limitation, the comparison of these two communities that seek to provide culturally among vulnerable populations: a systematic re- – nonprofessional interventions for DSME was appropriate care and achieve health equity. view. Am J Public Health 2016;106:e3 e28 6. Shah M, Kaselitz E, Heisler M. The role of nottheintentofthisstudy.Rather,wefound community health workers in diabetes: update on a cooperative model using CHWs and PLs current literature. Curr Diab Rep 2013;13:163– 171 was effective in sustaining gains achieved The authors thank the Acknowledgments. 7. American Public Health Association. Commu- through a more intensive, short-term DSME REACH Detroit Partnership and the dedicated nity health workers [article online]. Available from program led by CHWs. This is potentially staff at CHASS, including the outstanding CHWs https://www.apha.org/apha-communities/ and PLs. The authors thank the patients who a scalable and sustainable model for health member-sections/community-health-workers. Ac- participated in this intervention. care centers in low-resource settings and cessed 22 March 2017 Funding. ThisresearchwassupportedbyaPeers 8. Spencer MS, Rosland AM, Kieffer EC, et al. provides volunteer opportunities for pa- for Progress grantfrom the American Association Effectiveness of a community health worker tients who successfully complete CHW- of Family Foundation, by the National intervention among African American and Latino ledDSMEprogramsandwouldliketosupport Institute of Diabetes and Digestive and Kidney adults with type 2 diabetes: a randomized controlled Diseases (grant P30-DK092926 to the Michigan other patients grappling with diabetes. trial. Am J Public Health 2011;101:2253–2260 Center for Diabetes Translational Research and Several limitations are notable. First, 9. Baig AA, Benitez A, Locklin CA, et al.; Little grant R18-DK-0785501A1), and by the Centers Village Community Advisory Board. Picture good Latino participants were recruited from for Disease Control and Prevention (cooperative health: a church-based self-management inter- one federally qualified health center in agreement no. U50/CCU417409). The funding vention among Latino adults with diabetes. J Gen southwest Detroit. Thus, the general- sources had no role in the study design; data Intern Med 2015;30:1481–1490 collection; administration of the interventions; izability of this study is limited to this 10. de Vries L, van der Heijden AA, van ’t Riet E, analysis, interpretation, or reporting of data; or population. Second, psychological and et al. Peer support to decrease diabetes-related decision to submit the findings for publication. distress in patients with type 2 diabetes mellitus: behavioral measures were self-reported. Duality of Interest. No potential conflicts of Third, we experienced attrition in our interest relevant to this article were reported. design of a randomised controlled trial. BMC Endocr Disord 2014;14:21 Author Contributions. M.S.S., E.C.K., and M.H. sample through 18 months. Although 11. Tang TS, Nwankwo R, Whiten Y, Oney C. expected, the reduced sample size may contributed to study conception and design, data analysis and interpretation, and manuscript prepa- Outcomesofachurch-baseddiabetesprevention have affected our ability to detect some ration and were the principal investigators. B.S. program delivered by peers: a feasibility study. – statisticallysignificantresults.Futurestud- contributed to study conception and design, data Diabetes Educ 2014;40:223 230 ies should consider multisite or national collection, analysis, and interpretation, and manu- 12. Xue L. Effectiveness of a Peer Leader Sup- script preparation. G.Pi., J.H., A.L., and N.E. con- ported Diabetes Self-Management Support Pro- randomized controlled trials with larger gram on Patient Assessment of Care for Chronic sample sizes with representation from tributed to study conception and design, data analysis and interpretation, and manuscript prep- Conditions (PACIC). Pittsburgh, University of the diverse populations of Latinos in the aration. G.Pa. contributed to study conception and Pittsburgh, 2014 U.S. Fourth, as with other interventions, design, data collection, study implementation, and 13. Thom DH, Ghorob A, Hessler D, De Vore D, those willing to participate likely differ manuscript preparation. T.T. and M.F. contributed Chen E, Bodenheimer TA. Impact of peer health coaching on glycemic control in low-income significantly from nonparticipants, with to study conception and design of the PL in- tervention. M.S.S. is the guarantor of this work patients with diabetes: a randomized controlled nonparticipants often having worse – and, as such, had full access to all the data in the trial. Ann Fam Med 2013;11:137 144 clinical measures. We also excluded study and takes responsibility for the integrity of 14. Tang TS, Funnell M, Sinco B, et al. Compar- patients less likely to be able to partic- the data and the accuracy of the data analysis. ative effectiveness of peer leaders and commu- ipateintheintervention,furtherlimiting Prior Presentation. Parts of this study were nity health workers in diabetes self-management presentedinabstractformattheAmericanPublic support: results of a randomized controlled trial. generalizability. Diabetes Care 2014;37:1525–1534 Notwithstanding these limitations, Health Association Annual Meeting and Expo, Denver, CO, 29 October–2 November 2016. 15. Kieffer EC, Willis SK, Odoms-Young AM, et al. our findings provide further evidence for Reducing disparities in diabetes among African- culturally appropriate, theory-based CHW American and Latino residents of Detroit: the and PL interventions aimed at improving References essential role of community planning focus 1. Piette JD, Glasgow R. Strategies for improving groups. Ethn Dis 2004;14(Suppl. 1):S27–S37 diabetes self-management among Latinos behavioralhealthoutcomes among patients with 16. FeathersJT,KiefferEC,PalmisanoG,etal.The with T2D. The study offers encourage- diabetes: self-management, education. In Evi- development,implementation,andprocesseval- ment for underresourced health centers dence-Based Diabetes Care. Gerstein HC, Haynes uation of the REACH Detroit Partnership’s 1422 Outcomes From CHW and PL DSME Program Diabetes Care Volume 41, July 2018

diabetes lifestyle intervention. Diabetes Educ 24. Rogers EJ, Misner L, Ockene IS, Nicolosi RJ. 32. PiattGA,AndersonRM,BrooksMM,etal.3-year 2007;33:509–520 EvaluationofsevenCholestechL.D.Xanalyzersfor follow-up of clinical and behavioral improvements 17. Funnell MM, Anderson RM. Empowerment total cholesterol determinations. Clin Chem following a multifaceted diabetes care intervention: and self-management of diabetes. Clin Diabetes 1993;39:860–864 results of a randomized controlled trial. Diabetes 2004;22:123–127 25. Polonsky WH, Fisher L, Earles J, et al. Assess- Educ 2010;36:301–309 18. Two Feathers J, Kieffer EC, Palmisano G, et al. ing psychosocial distress in diabetes: develop- 33. Piatt GA, Orchard TJ, Emerson S, et al. Trans- Racial and Ethnic Approaches to Community mentofthediabetesdistressscale.DiabetesCare latingthechroniccaremodelintothecommunity: Health (REACH) Detroit partnership: improving 2005;28:626–631 results from a randomized controlled trial of a diabetes-related outcomes among African American 26. Barrera M Jr, Glasgow RE, McKay HG, Boles multifaceted diabetes care intervention. Diabe- and Latino adults. Am J Public Health 2005;95: SM, Feil EG. Do internet-based support inter- tes Care 2006;29:811–817 – 1552 1560 ventions change perceptions of social support?: 34. Edwards D, Berry JJ. The efficiency of sim- 19. AndersonRM,FunnellMM,ArnoldMS.Using an experimental trial of approaches for support- ulation-based multiple comparisons. Biometrics theempowermentapproachtohelppatientschange ing diabetes self-management. Am J Community 1987;43:913–928 behavior. In Practical Psychology for Diabetes Clini- Psychol 2002;30:637–654 35. van Buuren S. Multiple imputation of dis- cians. 2nd ed. Anderson B, Rubin RL, and Rubin RR, 27. KroenkeK,SpitzerRL,WilliamsJB.ThePHQ-9: crete and continuous data by fully conditional Eds. Alexandria, VA, American Diabetes Association, validity of a brief depression severity measure. J specification. Stat Methods Med Res 2007;16: 2003, pp. 3–12 Gen Intern Med 2001;16:606–613 219–242 20. Emmons KM, Rollnick S. Motivational inter- 28. Fitzgerald JT, Anderson RM, Gruppen LD, 36. Rubin DB. Multiple Imputation for Non- viewinginhealthcaresettings.Opportunitiesand fi limitations. Am J Prev Med 2001;20:68–74 et al. The reliability of the Diabetes Care Pro le response in Surveys. Hoboken, NJ, John Wiley & 21. Funnel MM, Anderson RM. Behavior change for African Americans. Eval Health Prof 1998;21: Sons, 2004 – strategies. In Medical Management of Type 2 52 65 37. Wagner EH. The role of patient care teams Diabetes. 5th ed. Burant CF, Ed. Alexandria, VA, 29. FitzgeraldJT,Davis WK, ConnellCM,Hess GE, in chronic disease management. BMJ 2000; – AmericanDiabetesAssociation,2004,pp.124–129 FunnellMM,HissRG.Developmentandvalidationof 320:569 572 22. Tang TS, Funnell MM, Noorulla S, Oh M, the Diabetes Care Profile. Eval Health Prof 1996;19: 38. NorrisSL,LauJ,SmithSJ,SchmidCH,Engelgau Brown MB. Sustaining short-term improvements 208–230 MM. Self-management education for adults with over the long-term: results from a 2-year diabetes 30. Fisher RA. The Design of Experiments. type 2 diabetes: a meta-analysis of the effect on self-management support (DSMS) intervention. Edinburgh, U.K., Oliver and Boyd, 1942 glycemic control. Diabetes Care 2002;25:1159– Diabetes Res Clin Pract 2012;95:85–92 31. Kalbfleisch JD, Prentice RL. The Statistical 1171 23. Arsie MP, Marchioro L, Lapolla A, et al. Evalu- Analysis of Failure Time Data. 2nd ed. [Internet], 39. Norris SL, Engelgau MM, Narayan KM. Effec- ation of diagnostic reliability of DCA 2000 for rapid 2011. Hoboken, NJ, John Wiley & Sons. Available tiveness of self-management training in type 2 and simple monitoring of HbA1c. Acta Diabetol from https://onlinelibrary.wiley.com/doi/book/ diabetes: a systematic review of randomized con- 2000;37:1–7 10.1002/9781118032985. Accessed 24 March 2017 trolled trials. Diabetes Care 2001;24:561–587