Facilitation of team-based care to improve HTN management and outcomes: a protocol for a randomized stepped wedge trial

During the usual care period, patients at the sites receive standard HTN care delivered by their primary care providers.

Facilitators collaborate with practice leadership and staff to implement TBC principles (i.e., shared goals, clear roles and responsibilities, effective communication, mutual trust and measurable processes/outcomes (i.e., monitoring and evaluation)) to facilitate the adoption of five evidence-based HTN management strategies [11, 42, 43]. These include: 1) population health (e.g., collecting patient race and ethnicity data to address disparities), 2) panel management (e.g., identifying and reaching out to patients who have uncontrolled HTN), 3) measuring BP accurately, 4) improving medication adherence, and 5) self-management support. Facilitators use a toolkit, developed by BEHS and the research team, that describes the activities to be completed before, during, and after each visit. Facilitators are expected to complete at least 16 in-person practice site visits during the 12-month intervention (Table 2). The purpose of the first two visits is to orient practices to the goals of the project and to conduct a needs assessment. This includes evaluating current roles and responsibilities for each member of the practice and obtaining a baseline assessment of care processes. For example, assessing the current application of TBC principles, in relation to HTN management (e.g., whose role is it to assess BP, what are their modes of communication).

In visit 1, facilitators use an adapted version of the Primary Care Team Guide Assessment (PCTGA) [44] to establish current workflows across key domains related to TBC. Specifically, the tool was adapted to assess at baseline the extent to which practices have adopted principles of TBC in the context of using specific HTN management strategies. For example, the tool assesses to what extent the practice defines team roles and practice goals in relation to providing self-management support. For each domain, practices are scored along a continuum from level D (just beginning to make changes) to level A (has achieved the most important changes in that domain). Practice facilitators are also trained to use the tool as a method for engaging practices in a discussion about expected challenges to integrating a TBC approach into current workflows, and specifically, to implement evidence-based HTN strategies. Using findings from the PCTGA and the roles and responsibilities worksheet, the facilitators complete a HTN management checklist after the first visit. These three tools provide the necessary information to support practice facilitators in tailoring the intervention to practice context. During the second visit, facilitators use their findings to further engage the practice in exploring potential challenges to implementing TBC and HTN strategies, review the practice’s baseline BP quality measures, and begin to expand on core QI concepts (e.g., plan do study act cycles and performance reporting). In the following visits, facilitators work with practices to integrate a TBC approach for each of the five HTN management strategies. Table 3 outlines the PF activities for those visits. These include conducting workflow analyses, using root cause analyses to identify barriers and suggest tailored solutions, and supporting the practice to redefine roles and responsibilities based on those analyses and updating electronic health record (EHR) templates. For the final visit, facilitators reassess practice progress using the PCTGA and reinforce action steps for sustaining gains.

Practice facilitators attend 15 h of didactic and experiential training over three days. BEHS facilitators are all certified facilitators and have received additional training on motivational interviewing, and implementing quality improvement strategies and systems (e.g., adapting the practice EHR to generate population-based blood pressure reports). Therefore, the training focuses on increasing capacity to support practices to apply principles of TBC in the context of improving the implementation of the five HTN management strategies. Specifically, the training reviews TBC principles, HTN management strategies, the activities that they are expected to complete during each visit, including conducting workflow and RCA, and uses case examples to engage facilitators in role-playing exercises. Practice facilitators are also trained to complete the assessment tools. Prior to attending the trainings, facilitators are expected to review the PF toolkit, complete required readings, and view required videos. The required readings focus on teams, team functioning, and the principles of TBC, while the videos focus on leadership, communication, and fostering mutual trust and support in primary care practices.

Facilitators participate in weekly one-on-one meetings with a PF supervisor. Supervision meetings review site visits and facilitators/barriers to implementing TBC and HTN management strategies. Supervisors use the PCTGA and other tools that the PFs completed, as well as PF visit data from Salesforce, a cloud-based customer relations management system [45], (see fidelity measures, below) to guide the conversation. PF supervisors use a standardized form to guide the discussion and document challenges and the next steps to address those challenges. Additionally, facilitators meet monthly to share field experiences. Based on supervision meetings, additional PF training is provided as necessary.

A secure, online, password-protected database built on a REDCap platform and linked to a secure web-based platform is used for survey collection and tracking. All EHR data is aggregated count data only. These data after extraction from the EHR are exchanged via secure file transfer protocol between BHES and NYU. The transfer of deidentified salesforce data follows the same data security protocol. Interviews are audiotaped, transcribed and stored in a password protected database. An independent data monitoring safety board is convened annually to review study progress, and monitor study participant safety.

To measure our primary outcome (Table 4), adoption of TBC for improving adoption of HTN strategies, we adapted two instruments: the 29-item Primary Team Dynamics Survey Measure and a subscale from the Team Process Survey Measure. The Primary Team Dynamics Survey Measure is a validated instrument designed to measure team dynamics in primary care settings [46]. The 29-item instrument examines seven factors that map to four out of five elements of TBC: clear roles, shared goals, mutual trust, and effective communication. Respondents will be asked to rate their level of agreement on a scale from 1 (Strongly disagree) to 5 (Strongly agree) to statements related to TBC. Participants also complete the 5-item monitoring progress toward goals subscale of the team process measure. Respondents rate the extent to which their team participates in activities related to goal monitoring on a scale of 1 (Not at all) to 5 (To a very great extent). Practice staff in each wave complete the TBC Assessment Tool at baseline and 12- and 18-month period (end of intervention period). Both surveys were adapted to include language specifically related to HTN management. This allows us to measure the degree to which practice teams adopt TBC strategies for HTN management.

We assess the percentage of patients with BP < 140/90 [48]. Data on BP is obtained through the EHR. Using the National Quality Forum-endorsed BP control measure [48], the denominator is defined as the count of all individuals, ages 18–85, with a diagnosis of HTN, who were seen at least once during the measurement period. The numerator represents the count of individuals who met the denominator criteria whose latest measured BP in the measurement period was less than 140/90 [49]. We obtain reports on BP control at baseline, and every six months thereafter, coinciding with the stepped-wedge design, using the most recent prior 6-month measurement period. BP data is extracted from the EHR systems through existing reporting and registry tools. We also collect aggregated data on race/ethnicity and insurance status among patients with HTN.

The primary outcome is the overall average score on the TBC assessment tool. The secondary outcome is patient BP control. The analysis of the effect of PF on the TBC adoption outcome measure will be based on a linear mixed model. Specifically, to assess the PF intervention effect compared to usual care, we will use a model with a random site effect, and fixed effects for time and study condition. Condition will be dummy coded with the control condition as a reference category. The fixed-effects coefficient for the contrast of the PF condition with the control condition will indicate the direction and magnitude of the difference on the overall average TBC score. Inclusion of a fixed effect for time accounts for any general time trends. We also will explore whether the time coefficient should have a random effect to capture differences in time trends across practice sites. Confidence intervals (95%) will be reported to convey precision. In addition to the effects of time and condition, covariates (e.g., organizational readiness) will be considered for inclusion in the mixed-effects model to reduce the within-group variance, if the potential covariate is predictive of the outcome.

Models will be fit using the glmmTMB package using R software [51, 52]. Each of the PF intervention effects will be tested using a two-sided level of significance α = 0.05. For the blood pressure secondary outcome, we will also conduct sub-analyses to assess outcomes by race/ethnicity.

Potential moderator effects will be examined by including separately each organizational characteristic as a main effect and in interaction with the treatment condition indicator (Control vs. PF vs. Sustainability). A significant interaction effect would indicate the effect of the condition depends on the moderator, and contrasts of the control and PF periods at different levels of the moderator would be undertaken to understand the pattern of the interaction. We will conduct further analysis to explore the possibility that changes in organizational readiness mediate the effect of PF on primary and secondary outcomes. We will conduct preliminary analyses to assess whether PF (versus Control) is associated with changes at the site level. If there is a detectable effect of the PF intervention on these measures, we will assess whether the changes in these measures are associated with changes in overall average TBC and HTN management scores and changes in the rate of BP control using the Baron and Kenny “product method” as a starting point of the analysis [53]. If necessary, we will explore more complex structural equation models (SEM) adapted for Poisson outcomes [54] as well as more recently developed causal inference methods [55, 56] to evaluate the mediation.

With the pre-specified number of sites (90) and design (Table 1; stepped-wedge design with 3 waves and 6 time periods; n = 432 TBC overall average scores), we investigated the PF intervention effect that is detectable with at least 80% power of a 2-sided significance test with α = 0.05. Power was calculated by simulating ten thousand datasets with modest correlation of TBC overall average scores over time (r = 0.35; ICC = 0.32) [57], and a standardized mean difference between control and PF conditions of d = 0.5 (i.e., a difference of half a standard deviation). Power was 0.92 to detect this effect with the planned number of sites. To allow for withdrawal of a small number of sites after baseline, we plan to enroll 90 practice sites (30 in each wave). Power for the secondary blood pressure outcome was also calculated by simulating ten thousand datasets. In all conditions, we assume an average of 150 patients with HTN per site and time period (i.e., 64,800 observations of patient BP over all sites and time periods). The intraclass correlation for BP control was specified as ICC = 0.85 based on an estimate from our previous EvidenceNOW study [25]. For the control condition, we specified a rate of BP control of 70%. For the PF condition, we specified an increase in the rate of BP control to 75%, and for the follow-up condition, we specified a decrease to 72%. The difference between the control and PF conditions corresponds to IRR = 1.07. Power to detect this IRR was > 0.99.

We will calculate adherence to the PF protocol as the number of components fully or partially implemented divided by the total number of possible components [58]. We will assess visit dose (i.e., number of visits completed, type of visit conducted (in person vs. remote), and length of visit). We will compare pre and post-intervention HTN checklist responses to changes in the number of evidenced-based HTN strategies implemented at each practice.

Interviews are transcribed verbatim and imported into Atlas.ti for analysis. The study uses an abductive content analysis that systematically integrates practices from inductive analysis (e.g., grounded theory) [59] and deductive analysis (i.e., applying codes associated with the guiding framework [CFIR]), while remaining open to informative deviations from those frameworks. Coding begins with an independent reading of the transcripts to identify preliminary themes, relevant patterns, and generative questions, followed by focused coding to identify clustered concepts, organize ideas, identify major emergent themes, and link them to relevant theoretical constructs. Throughout, coders will meet to review their coding, conduct team debriefing meetings, and reach a consensus on code names and meanings [60]. Once all transcripts are collaboratively coded, analytic domains are identified, and major and minor thematic areas described. Two coders independently code at least 10 interviews to establish the inter-rater reliability (IRR) and if it is inadequate (Krippendorff’s alpha < 0.80) [61], the lead investigators will work collaboratively to refine and/or clarify the coding scheme and provide additional coder training. The research team refines and/or clarifies the coding scheme and provides additional coder training. Double coding continues until adequate IRR is achieved.

The analysis integrates the qualitative and quantitative data in a nested, parallel mixed-methods design [62‐64]. Each interview is linked to descriptors of the originating study site’s characteristics, including potential moderators, mediators, and outcomes. These descriptors are used during analysis to categorize and interpret qualitative data in Atlas.ti using group functions. We triangulate findings from the baseline assessment, 12, and 18, -month adoption and sustainability data waves by developing a thematic matrix that includes practice and team-level characteristics and compare side-by-side those factors that were identified as facilitating or hindering the introduction, adoption, and sustainability of TBC for HTN strategies. This analysis examines how themes emerging from the qualitative data converge with, contextualize, explain, or expand the quantitative findings.