Study design and protocol of a stepped wedge cluster randomized trial using a practical implementation strategy as a model for hypertension-HIV integration — the MAP-IT trial

Two key components define our protocol: a hypertension management program and an external support system for nurses. A nurse-led (TASSH) program will be implemented at all PHCs. TASSH includes identification of patients with hypertension, treatment of patients with uncontrolled hypertension using antihypertensive medication based on a standard treatment and drug titration protocol (see Supplemental file A), lifestyle counseling, and referral to physicians for complicated cases of hypertension (i.e., nonresponsive to first-line treatment after maximum doses, BP ≥ 180/110 mm Hg, or with comorbid diabetes mellitus, heart attack, stroke, kidney disease, or heart failure, or pregnancy following study enrollment) [17]. With the exception of TASSH focusing on nurses providing hypertension care, these components of TASSH are similar to the Nigerian Hypertension Treatment Protocol. The experimental implementation component, PF, is an approach to service integration that will be introduced to all PHCs in a stepwise manner over the duration of the trial. PF involves provision of external expertise on practice redesign and a tailored approach to evidence-based care [18‐22]. It includes both a role (practice facilitator) and a process for supporting primary care practices to build motivation and capacity at the systems and individual levels to improve organizational performance [19, 21‐24]. Practice facilitators (typically nurses) are trained to work with primary care practices to develop the necessary skills to adapt evidence-based strategies, e.g., task shifting/sharing, to their practice environment [25]. In this study, the practice facilitators are nurses with an average of 35 years in clinical practice.

This study occurs in multiple phases [a pre-implementation (UG3), implementation (UH3), and post-implementation], applies several implementation science frameworks, and uses a stepped wedge cluster randomized trial (CRT) study design. Figure 1 overviews the two implementation science frameworks — the integrated Promoting Action on Research Implementation in Health Services (i-PARiHS) [26] and the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) [27] — used in the pre- and post-implementation phases. Below, we provide details on each phase.

The pre-implementation phase used the i-PARiHS in inquiring about the existing health system and PHC context to inform the development of a context-specific PF strategy. We assessed barriers and facilitators of integrating TASSH into PHCs for hypertension management through structured facility assessments and in-depth interviews with stakeholders within the PHCs and primary health system in Akwa Ibom. A pilot study was conducted during the pre-implementation phase wherein the feasibility of TASSH plus PF intervention was assessed in two PHCs. The facilities included in the pilot are not participating in the implementation phase.

In the implementation phase, we are using a SW-CRT design to introduce PF at the PHCs. The 30 PHCs were randomly allocated into five cohorts of six PHCs, and all begin in the TASSH only condition, which serves as the usual care control. Figure 2 overviews the stepped wedge design for this study. Given the SW-CRT, the length of time in TASSH alone (without PF) will vary, with a minimum duration of 3 months for the first cohort. Three months prior to the TASSH plus PF experimental condition starting, each cohort will have a training period. During this period, nurses and other staff at the PHCs receive training (or retraining) in delivering TASSH. Once the 3-month training period is completed, the PHCs will participate in a 12-month implementation period during which PF will occur.

Nurses at these PHCs will implement all aspects of TASSH in their clinical care for PLWH with uncontrolled hypertension. PFs will support nurses during the 12-month period by helping with hypertension treatment issues and addressing barriers to TASSH implementation. The RE-AIM framework defines measures to evaluate the effect of PF on TASSH adoption after 12 months of PF compared to TASSH adoption without PF (primary outcome). RE-AIM facilitates evaluation of the effect of PF on BP control (secondary outcome) and the level of TASSH implementation fidelity (secondary outcome).

Following each cohort’s 12-month implementation phase, each PHC transitions to the post-implementation phase (follow-up period). Nurses will continue to implement TASSH. However, PF will end. Six months into this post-implementation period (18 months following the start of PF), we will evaluate the sustainability of TASSH within the PHCs (secondary outcome). Figure 3 provides an overview of the study design including the pre-implementation, implementation, and post-implementation phases.

This study is approved by the Institutional Review Boards (IRBs) at the University of Abuja (Abuja, Nigeria), Ibom Multi-Specialty Hospital (Uyo, Akwa Ibom), and New York University (NYU) Grossman School of Medicine. The study is registered at Clinicaltrials.gov (NCT05031819).

In a pragmatic study design which stems from the fields of service delivery and policy evaluation, the SW-CRT allows for robust evaluation while factoring in logistical (i.e., rolling out of facilities) or political (i.e., relating to the policymakers and service managers) constraints common with a standard parallel randomized controlled trial (RCT) design [28]. In this study, the sequential crossover of PHCs from TASSH-only to TASSH-plus PF permits rigorous evaluation of the implementation strategy effects while enabling all study facilities and participants the benefit of TASSH with PF training.

Along with the robust nature of the stepped wedge design, we will evaluate the PF experimental condition using the RE-AIM framework. Widely used in both health research and health intervention programming and evaluation, RE-AIM offers a systematic guide for collecting and analyzing data to examine all aspects of planning, implementation, and potential sustainability [29]. For this study, RE-AIM is most appropriate to evaluate and identify the program elements critical for implementation and sustainability.

Thirty-two PHCs from a network of 59 PHCs across Akwa Ibom were recruited into the study — 2 for the pre-implementation phase and 30 for the implementation phase. These PHCs support a sizeable number of PLWH. Adoption will be measured on all eligible persons to be screened for hypertension. Eligible patients are identified from the roster of PLWH during routine visits and through linkages from community screening and proximal HIV clinics. In the 30 implementation study sites, we will recruit and enroll an average of 28–32 patients, for a maximum of 960 PLWH participants with uncontrolled hypertension to be followed longitudinally for BP outcomes. Eligibility criteria include the following: (1) HIV-positive persons aged 18 years or older; (2) either currently enrolled in HIV treatment services at one of the 30 study PHCs, receiving hypertension treatment services at one of the 30 study PHCs, or can be enrolled at one of the 30 study PHCs to receive hypertension services for the entire study duration; (3) (for those being followed for BP control) have elevated BP between 140–179 mm Hg systolic and/or 90–109 mm Hg diastolic, as determined by the average of the latter two of three separate BP readings during one clinic visit; and (4) be able to provide consent. The inclusion criteria are based on WHO’s CVD treatment guidelines [30‐32].

Those who refuse, who are unable to provide informed consent, or with any of the following criteria at baseline will be ineligible for inclusion in the longitudinal BP cohort: (1) BP of ≥ 180/110 mm Hg; (2) known history of kidney disease, heart disease, diabetes mellitus, transient ischemic attacks, stroke, or heart failure; and (3) are pregnant or breastfeeding at the time of enrollment.

As previously noted, our Lagos study [16] provides thorough details of TASSH and the approach for developing the PF implementation strategy. Here, we briefly describe both and specify any potential differences for this study.

TASSH uses a four-step approach for identifying, counseling, treating, and referring (ICTR) and the 5As counseling strategy (ask, assess, advice, assist, & arrange). The four steps are defined as follows: (1) identifying PLWH patients with uncontrolled hypertension (BP > 140/90); (2) initiating counseling by giving clear advice on how PLWH patients can modify their behaviors; (3) treating PLWH with uncontrolled hypertension by prescribing medication treatment according to the Nigerian Simplified Hypertension Protocol; and (4) referring patients with severely elevated BP and/or a chronic comorbid condition, including established CVD, to physicians for further care. The 5 As have been implemented in previous TASSH studies [11, 16, 17, 25] and applied to other health outcomes [33, 34].

As described, we will test the effect of PF on the adoption of TASSH and systolic BP reduction. We used a context-specific approach to develop and guide the PF implementation strategy based on three main components: (1) formation of an advisory board to guide the PF intervention, (2) development of the PF strategy, and (3) training of community nurses on TASSH protocol implementation.

The advisory board constitutes members from the National Agency for the Control of AIDS (NACA), the Federal Ministry of Health (FMOH), the Directorate of Nursing, Akwa Ibom State Ministry of Health, Akwa Ibom State Agency for Control of AIDS (AKSACA), two physicians, two community nurses, and three patient advocates. Development of the PF strategy included the following: (1) training the practice facilitators using a train-the-trainer model [35] guided by the study-developed three Es, i.e., engaging, enhancing, and evaluating the nurses on performing their TASSH duties and delivering the hypertension protocol, (2) identifying site champions and coordinators, (3) building consensus for quality improvement targets, (4) implementing practice changes to support TASSH implementation, and (5) peer-to-peer collaboration. Additional details on the training approach for TASSH and the PF implementation strategy are detailed in prior work [16].

PF is designed to stimulate specific, actionable steps that PHCs can use to build a robust and context-appropriate foundation for integrating TASSH components. Senior or retired healthcare nurses are trained as coaches and help develop the skills needed in order to adopt evidence-based interventions. The PF strategy combines one-on-one onsite facilitation with shared learning opportunities across practices via peer-to-peer phone calls and text messaging. Practice facilitators are expected as follows: (i) coordinate meetings with PHC nurses, (ii) assist PHC nurses in setting performance goals, (iii) strategize on how to best implement TASSH, and (iv) test workflow changes within PHCs.

During the implementation phase, practice facilitators will support PHCs and nurses implementing TASSH. Facilitators will coordinate monthly telephone calls and onsite meetings (twice during the first month and monthly thereafter). Key tasks for the onsite meetings include reviewing baseline data and assisting nurses to set targets for their weekly workload at the first meeting and troubleshooting workflow issues in subsequent visits. Monthly check-in phone calls will provide opportunities for nurses to raise challenges that may arise with implementing TASSH.

In the pilot phase, two PHCs were recruited to refine and understand data collection strategies, instruments, training timing and methods, and study recruitment methods. For the SW-CRT, 30 PHCs are randomized in five cohorts of six PHCs. Within each PHC, we anticipate enrolling approximately 28–32 PLWH on average. The enrollment targets for each PHC are based on its smaller, mid-sized, or large caseload. We do not anticipate any attrition in study sites. In the event that a site does drop out for any reason, we have adequate numbers of sites with similar characteristics to recruit from PHCs that the Agency oversees.

Study biostatisticians developed the randomization sequence to allocate each PHC to a stepped wedge cohorts. These allocations are kept in a secure electronic format inaccessible to study sites, in accordance with CONSORT guidelines. Study coordinators will be informed of which PHCs are allocated to each cohort just before each training period begins. Because of the nature of the intervention, it is impossible to mask individuals, nurses, practice facilitators, and study coordinators to the group assignment when implementation begins. One study biostatistician will remain blinded to treatment assignment until all data have been collected and the database is locked.

Adoption (i.e., uptake) is the utilization of components of TASSH. TASSH adoption will be defined as nurses showing utilization of all of the following components: (1) identifying PLWH with uncontrolled hypertension, (2) measuring PLWH’s anthropometrics and BP with a valid automated device following standard procedures, (3) initiating lifestyle counseling and medication treatment every 1–3 months, and (4) referring PLWH with severely elevated BP and complicated hypertension to physicians for further care. Adoption will be measured at the clinic level, in aggregate, each month from the baseline period of the first cohort through the end of follow-up, for all those PLWH with BP screens by case managers or those previously diagnosed with hypertension at a prior visit by the nurses. Table 1 expands on the specific measures of adoption.

Secondary outcomes include BP control, implementation fidelity, and TASSH sustainability. BP control is defined as a BP less than 140/90 mm Hg. For those consented and enrolled, at baseline visit, three BP readings will be taken by nurses using an automated BP monitor with the participant seated comfortably for 5 min prior to the measurements, following standard guidelines. The average of the latter two BP readings will be used as the measure for each visit. The same procedure will be followed for each participant visit, including the 12-month visit. Uncontrolled BP is defined by an average clinic systolic BP ≥ 140 mm Hg or diastolic BP ≥ 90 mm Hg following the guidelines set forth by WHO for CVD treatment [31, 32] and in accordance with the Nigeria Hypertension Treatment Protocol.

Fidelity to TASSH will be measured throughout the study and defined by five dimensions: (1) adherence to the program protocol, (2) dose of the program delivered, (3) quality of program delivery, (4) participant responsiveness, and (5) program differentiation. Tools to assess fidelity were developed for this study and adapted from the current literature [39‐41]. Fidelity measures will evaluate adherence from the perspective of the PF, nurses, and the participants.

Sustainability is defined as the maintenance of TASSH adoption at HIV clinics at 18 months (6 months after the end of each cohort’s TASSH + PF phase). Sustainability will be assessed similarly to adoption (as defined above) and qualitatively, based on interviews with nurses and clinic leadership at 18 months. The research coordinator will conduct the interviews with two nurses and one key leadership personnel at each clinic. The interviews will be guided by i-PARiHS and inquire about attitudes regarding TASSH implementation, barriers, facilitators, and implications for scalability. All interviews will be recorded, transcribed, and analyzed with NVivo Version 11.

As the level of randomization for this SW-CRT is at the practice level, there are several important study design features: K, the number of primary care practices, and m, the average number of PLWH recruited per practice, as well as the relatedness of individuals within sites (when individual level data are utilized for analysis). Specifically, the m participants will be nested (grouped) within the K practices. This nesting will lead to additional dependency in outcomes of participants within a practice; the degree of this dependency, the intraclass correlation coefficient (ICC), will contribute to the necessary increase in sample size to maintain the same level of power as would result if the unit of randomization was independent individual participants rather than participants within practices. Prior studies examining the effects of TASSH on BP control have demonstrated improvements in BP control level ranging from 20 to 25% over usual care. We expect a somewhat smaller effect, as we are testing the effect of PF plus TASSH on BP control.

Trials targeting BP control have found ICCs for BP control ranging between 0.02 and 0.06. In addition, preliminary evidence from our UG3 pilot phase suggests that ICC estimates for BP control under UC align with these estimates. Adoption will be measured as outlined above through aggregate data at the site level each month. It is expected that TASSH adoption will be lower than 50% before PF. We expect the estimates in Table 2 to be conservative if the baseline proportion for adoption is less than 40%. We will test our primary and secondary hypotheses at the 5% two-sided type 1 error level. For secondary measures that will utilize individual level data, we expect to enroll 28–32 individuals on average, expecting approximately 10% attrition over the course of the study period, resulting in at least 25 on average per site.

We will classify four intervals of the PF intervention, each 3 months in duration (start of PF + TASSH to 3 months; 4 to 6 months; 7 to 9 months; 10 to 12 months) to explore the change in adoption over the intervention period. Adoption at the cluster/site level will be described and compared across intervals. To examine whether the rate of adoption increases over time, a GLMM will be used to assess the effect of time on adoption rates. Specifying a main effect of time (categorized as defined above), we will interpret a positive main effect to indicate that rates of adoption increase over time. Covariates strongly related to adoption as well as study design factors which may influence levels of adoption will be identified and may be included in adjusted analyses. Proposed factors include site size and patient characteristics such as sex and age (as a proportion of the total). These will be a limited set of covariates, given the limited degrees of freedom available for these site/cluster-level analyses. Adjustment variables will be identified prior to data exploration and PF initiation. The denominator degrees of freedom for the F-test for the intervention effect contrast will be adjusted for the number of clusters and will consider any group level covariates utilized in adjustment [42]. The test of PF intervention to usual care will be evaluated at the two-sided 5% level.

Maximum likelihood estimation of mixed-effects models will be used to account for missing data. In the context of a mixed-effects model, this is equivalent to an assumption of data missing at random. We will conduct sensitivity analyses assessing this assumption including comparison of participants with and without missing values with respect to baseline and practice characteristics. If differential patterns emerge, we will consider the use of multiple imputation by chained equations and/or inverse probability weighting to adjust for missing data.

While primary analyses will focus on estimation through GLMM and take all information gathered during the usual care and PF periods into account, we will also consider the evaluation of PF during the “active stepped wedge” phase, the period in which at least one block differs in terms of treatment group (those periods when all blocks are in usual care or PF are excluded). These evaluations will compare estimates of the GLMM method and those which estimate the treatment effect through robust nonparametric permutation methods. GLMMs require assumptions on the random effects, which if violated result in biased treatment effect estimates [46, 47]. The proposed permutation tests do not require these assumptions on the correlation structure but require that all clusters are not in the same treatment condition. We will assess the robustness of GLMM assumptions through the comparison of results from methods in this reduced set of observations.