Study design

We conducted a prospective cohort study in the context of a programmatic roll-out of a CHW-led intervention. Based on positive findings from a pilot programme in two neighbouring communities, CES made a programmatic decision to scale the intervention to four additional communities in a stepwise fashion. To ensure optimal training, implementation and supervision, they implemented the intervention in one community at a time at 3-month intervals, and to respect equity between communities, they determined the order of the roll-out at random. Because the planned roll-out of the intervention mirrored that of a stepped-wedge trial,18 we emulated this study design by collecting outcomes data in all four communities at two occasions at baseline (ie, prior to implementation in any of the four communities) and prior to each expansion to a new community thereafter. We continued data collection until 12 months after the last community received the intervention. Figure 1 shows the timing of data collection and the CHW-led intervention roll-out.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Schematic of stepped-wedge CHW-led intervention roll-out and data collection. CHW, community health workers.

This study was reviewed and approved by institutional review boards of the Brigham and Women’s Hospital and the Instituto Tecnológico de Monterrey and conforms to the principles embodied in the Declaration of Helsinki. This study is registered at ClinicalTrials.gov (NCT02549495). Participants provided verbal informed consent which was documented in writing by study staff.

Study population

We conducted this study in four remote, rural communities with populations of 1500–2500 people in Chiapas, Mexico. Participants were Spanish-speaking. Each community had one clinic. Clinics maintain registries of patients with NCDs that served as the basis for eligibility determination and recruitment. Eligible patients were those who had a diagnosis of diabetes and/or hypertension, resided in a study community, were aged 18 years or older at time of enrolment and were prescribed daily medications by the clinic physician for control of their diabetes and/or hypertension. We excluded patients with secondary hypertension, type 1 diabetes, pregnancy and chronic use of glucocorticoids. We also excluded patients who, after enrolment but prior to the implementation of the intervention in the first community, were removed from diabetes or hypertension treatment by their physician, moved outside the study community, transferred care to another health facility or who no longer carried a diagnosis of diabetes or hypertension. The study size was determined by the number of patients meeting the eligibility criteria in the communities where the intervention was to be rolled-out.

NCD care in the study setting

CES is an affiliate of Partners in Health, a multinational non-governmental organisation with experience with CHW-led interventions to deliver care for a range of medical conditions, including NCDs.19 Since February 2012, CES partnered with the Mexican Ministry of Health to rehabilitate, staff and operate existing government primary care clinics. CES provides comprehensive primary care and manages patients with NCDs in accordance with national guidelines. The clinic physician directs care of patients with NCDs, which includes once-monthly clinic visits, therapy with common oral medications for diabetes (biguanides and sulfonylureas) and hypertension (ACE inhibitors, beta-blockers, diuretics, angiotensin II receptor blockers and calcium channel blockers) and community education. Study patients had Mexico’s national health insurance and there are no direct charges for clinic visits or medications.20 CES screens all adults in their catchment area for diabetes and hypertension using door-to-door active case finding, works with the Ministry of Health and other suppliers to eliminate medication stock-outs and provides clinical oversight and continuing education to physicians.21

Description of CHW activities

The CHW-led intervention designed by CES, titled ‘Acompañantes’, follows a ‘community-based accompaniment’ approach, which has been previously shown to be effective in improving medication adherence and disease outcomes among patients with HIV.22–24 In this approach, CHWs serve as a bridge between clinic and patient, promoting medication adherence, reinforcing basic disease education, providing psychosocial support and promoting active case retention. The CHWs in this intervention are women who were nominated at community meetings (either by self or community nomination), wherein CES staff presented the proposal, and then selected on the basis of a formal interview process focusing on leadership potential, motivation as well as basic literacy and education. They were trained in four-times-weekly group sessions for 1 month, covering basic pathophysiology, diagnosis and treatment of chronic diseases such as diabetes and hypertension, as well as practical training on the elements of a home visit and the logistical requirements of the role. They also participated in monthly refresher training sessions, covering themes such as brief motivational interviewing, recognising emergencies and complications and navigating interactions with challenging patients. CHWs work longitudinally with four to eight patients, conducting home visits which begin weekly then change in frequency based on a collaborative assessment of the patient’s needs by the CHW and clinic physician. CHWs escort patients to clinic visits and meet regularly with clinic physicians to discuss patient management. CHWs are compensated with household food and consumable items, worth a dollar amount approximately equivalent to the monthly stipend given to participants in Prospera (formerly Oportunidades), a Mexican national conditional cash-transfer programme.

Data collection

Data collection took place from March 2014 to January 2016 (see figure 1). We measured systolic and diastolic blood pressure (blood pressure, in mm Hg) among hypertensive patients, and haemoglobin A1c (A1c, in per cent haemoglobin A1c) among patients with diabetes. We measured blood pressure using three measurements using the Omron HEM 7080IT E automated blood pressure cuff. We measured A1c using the Bayer A1c NOW point-of-care device.

We assessed three aspects of self-reported adherence using scales adapted from prior studies of adherence to long-term therapy.25 26 ,27 The number of days during the last 5 and 30 days in which medication was not taken was assessed using 5-day recall and 30-day recall, respectively,25 26 and adherence behaviour was assessed using three questions assessed on a Likert scale, similar to the CASE adherence index and adapted to the local context: "How often do you take your medications in the way your doctor instructed?" "How many days a week do you forget to take your pills?” and "Some people choose to stop taking their medications. When was the last time you chose to stop taking your medications?"

Study staff visited patients at home and collected outcomes data in all communities at baseline and at 3-month intervals thereafter, timed just prior to the roll-out of the intervention in a new community. We also administered a questionnaire with basic demographic and income information at the time of enrolment. We classified patients as exposed to the intervention once it had been implemented in their community; however, because data collection occurred just prior to implementation, the first exposed data point for any patient was collected 3 months after implementation in their community. Local clinicians and programme personnel felt the effect of the intervention could be expected within 3 months of implementation.

Outcome definitions

We defined disease control as A1c <7 for patients with diabetes, and blood pressure <140/90 for hypertensive patients (<150/90 for patients aged at least 80 years). We classified a patient as optimally adherent by 5-day recall if s/he took medications on all of the preceding 5 days, and by 30-day recall if s/he took medications on at least 28 of the preceding 30 days. Adherence behaviour was classified as optimal if s/he reported perfect adherence (five of five possible points on a Likert scale) for two of three questions, and at least near-perfect adherence (four or more of five possible points on a Likert scale) on the third. For patients with diagnoses of both diabetes and hypertension, patients were required to meet the definition of optimally adherent for both sets of medications or controlled disease for both diseases.

Statistical methods and data analysis

We conducted individual-level mixed effects analyses, which included random intercepts for each individual and community to adjust variances for clustering. We modelled binary outcomes (optimally adherent vs not, controlled vs not) using a logit link. Independent variables included a binary variable to indicate whether the person lived in a community that was exposed to the intervention at each time point and an indicator variable for time point to flexibly model time and minimise residual confounding by time. We conducted stratified analyses to examine whether any effect of the intervention depended on disease (diabetes or hypertension) or baseline disease control (adherence outcomes only) and calculated P values for differences in ORs using Cochran’s Q-test for heterogeneity. Outcome assessments for individuals who withdrew from the study for any reason were included until the time at which they withdrew. We conducted two sensitivity analyses; in the first we adjusted for community, and in the second we excluded 10 individuals who were removed from diabetes and/or hypertension treatment by their providers during the study. Analyses were conducted using SAS V.9.3 (Cary Institute, Cary, North Carolina, USA). Statistical significance was defined at a threshold of P<0.05.