A real-world stepped wedge cluster randomized trial of practice facilitation to improve cardiovascular care

Practice facilitation is an approach to implementing evidence-based best practices in primary care [1]. It involves bringing an external healthcare professional (often a trained nurse with management experience) into a practice, in order to help primary care providers address the challenges associated with implementing evidence-based guidelines. Practice facilitators (also known as outreach facilitators, practice enhancement assistants, and practice coaches) are trained experts in initiating change in practices. They work with practices to identify areas for improvement, set care improvement goals, and provide tools and approaches to reach these goals. Facilitation has been found to lead to improvements in prevention, diabetes care, smoking cessation, and cancer care [2‐4]. A recent meta-analysis demonstrated that primary care providers are more likely to adopt evidence-based guidelines when supported by a facilitator [5]. Furthermore, a cost consequence analysis of a facilitation study that focused on improving preventive care in primary care practices demonstrated a 40 % return on intervention investment [6].

Facilitation programs are being broadly implemented by practice-based research networks, health departments, professional associations, and health plans. As such, ongoing evaluation is needed to determine the effectiveness of practice facilitation in real-world settings.

The adoption, optimal intensity, and duration of practice facilitation remain uncertain, as does the effectiveness beyond single-disease-focused facilitation programs [1, 2, 5, 7‐9].

We initiated the Improved Delivery of Cardiovascular Care (IDOCC) through an outreach facilitation project in 2007 to improve adherence to guidelines for the secondary prevention of heart disease, stroke, peripheral vascular disease, renal disease, and diabetes in primary care practices in a health region [10]. We hypothesized that facilitation would enable practices to improve overall adherence to cardiovascular care guidelines. This paper reports on our primary outcome of adherence to guidelines as measured by mean adherence to care delivery in the practice, based on patient-level data.

The IDOCC study protocol has been published elsewhere [10], so here, we provide a brief overview of the methods as per CONSORT [11].

We approached all 533 primary care practices in the Champlain region to participate in the IDOCC program. Ninety-nine were ineligible to participate because the practice was no longer in operation, the clinic was an exclusive walk-in practice, or the physician running the practice was planning to retire within the 2-year intervention time frame. Of the 434 eligible practices, 93 practices (21 %) (194 physicians) agreed to participate. Nine practices withdrew from the study prior to the initiation of the facilitation intervention for various reasons, including not having the office space to accommodate a chart abstractor and insufficient number of eligible high-risk patient charts, leaving 84 participating practices (182 physicians): 27, 30, and 27 in steps I, II, and III, respectively. The withdrawal of ten practices from the study resulted in a loss of only 12 physicians, suggesting that only small or single physician practices chose to withdraw. The first facilitation visit was on April 14, 2008, and the final site visit was March 27, 2012. The trial concluded as scheduled at the end of the intervention phase for step III. No practices withdrew after the start of the intervention.

Baseline characteristics of participating practices, providers, and patients are presented in Table 1. The majority of participating practices were fee-for-service (53.5 %) and located in urban areas (82.1 %).

We collected baseline data from 5292 patient medical charts, of which 627 (11.8 %) were lost at follow-up, primarily due to patients dying or moving within the study period. Chart abstraction data quality assessments via re-abstractions demonstrated an overall inter-rater reliability kappa value of 0.91 and a percent agreement of 93.9 %.

No practices received the intended intensity of facilitator visits. Only 32 practices had eight or more visits over 2 years. On average, facilitators had 6.6 (min: 2 max: 11) face-to-face visits with practices in year 1 and 2.5 (min: 0 max: 10) visits in year 2. In addition to face-to-face visits, facilitators communicated with practices an average of 1.7 times using phone or email in year 1 and 1.6 times in year 2.

Figure 2 presents the observed mean adherence scores for each study step during the baseline, intensive, and sustainability years. There is little change in the mean adherence score pre- and post-intervention for all practices in all three steps. Additionally, the three baseline time points for step III appear to show a (modest) upward secular trend in mean adherence, highlighting the importance of controlling for the time trend in the subsequent regression analysis. Fig. 3 presents the estimated effect of the IDOCC intervention presented as adjusted mean adherence score averaged over all practices in control, intensive, and sustainability conditions.

Using data from all three steps, the unadjusted (model A) least square mean differences showed an absolute decrease from baseline of 2.1 % in the intensive phase (95 % CI: −3.1 to −1.1 %) and a decrease of 4.7 % in the sustainability phase (95 % CI: −6.2 to −3.2 %) (see Table 2). After adjustment for patient characteristics (model B), there was an absolute decrease in mean adherence from baseline of 1.9 % in the intensive phase (95 % CI: −2.9 to −0.9 %) and a decrease of 4.2 % in the sustainability phase (95 % CI: −5.7 to −2.6 %). Adjusting for provider factors had little impact on these estimates.

In our pragmatic trial of practice facilitation, the intervention did not improve adherence to evidence-based guidelines for cardiovascular disease in primary care practices. These results contrast with findings from previously reported facilitation trials [5, 26‐28] and highlight the complexities and challenges of scaling up research studies into sustainable programs. Suboptimal intensity of the intervention, a broad focus on multiple chronic conditions, and measurement challenges are all factors that may have contributed to the null results.

A meta-analysis conducted of facilitation programs demonstrated that intensity as measured by number of facilitation visits was associated with greater effect size [5]. The IDOCC intervention was designed with the intent of having the facilitators meet face-to-face with practices 13 times (i.e., one visit every 4 weeks) in year 1 (intensive phase) and 4–6 times in year 2 during the sustainability phase [29, 30]. We were unable to attain this “dose” with our practices, who despite having the best intentions of engaging and meeting with the facilitators, were all unable to find the time for this frequency of meetings. Other facilitation studies have faced similar issues. For instance, facilitators in a group-randomized practice facilitation trial conducted in South Texas initially intended to make monthly facilitation visits over 12 months but due to competing demands within participating practices (e.g., EMR implementation, staff turnover) were only able to make six to seven visits during the 1-year study period [31]. The difficulties in achieving the requisite intensity of visits necessary to facilitate changes suggest that practice facilitation may not be as effective in complex, real-world settings. In our program, no incentives were provided apart from the opportunity to claim continuing professional development credits. Supporting practice change may require incentives to compensate providers for their time and/or policies designed to encourage a culture of quality improvement such as mandatory reporting of quality metrics and accreditation of primary care practices. Additionally, financial incentives have been identified as having a modest positive impact on primary care processes associated with quality management of chronic disease [32].

Improving quality of care delivery for people with multimorbidity is an area that requires more research. Although our study was grounded in the chronic care model approach with targeted practice level improvement activities across several dimensions within the model, it failed to improve care delivery. Previous facilitation studies primarily focused on single diseases such as diabetes and asthma [5, 33, 34], whereas our intervention attempted to improve adherence to guidelines for patients with a broad number of cardiovascular-related conditions (coronary artery disease, peripheral vascular disease, stroke/TIA, diabetes, and chronic kidney disease) and risk factors (hypertension, dyslipidemia, weight, and smoking). Implementing guidelines for people with multiple chronic conditions is complex. While physicians may be able to successfully focus on single diseases such as diabetes or asthma, improving adherence across multiple conditions is more challenging. A review of interventions for managing patients with multimorbidity found that programs targeting specific areas of concern for the patient (e.g., functional difficulties) were more effective than broader disease-oriented programs [35]. Likewise, recent studies (including ours) examining the implementation of practice facilitation programs targeting multiple diseases have found only modest or insignificant improvements in patient outcomes [31, 33].

Our analyses identified statistically significant decreases in our mean adherence score; however, these changes are likely the result of our study being overpowered rather than the impact of the program on clinical behavior. The decreases we detected were smaller than our pre-specified minimal clinical difference and hence unlikely to have any clinical significance. We chose to measure adherence to all indicators using a mean adherence score as we felt it would reflect the reality of managing multiple chronic conditions simultaneously. We hypothesized that practice facilitation’s focus on practice-level transformation would lead to changes, which would positively influence multiple processes of care and could be captured with a mean score. However, we were unable to demonstrate this. Measuring impact across multiple areas is challenging. Our adherence score was an unweighted mean in which all indicators were considered equally important. As a result, it is more difficult to interpret the clinical importance of observed changes: whereas a decrease of 2–4 % could represent deterioration in one to two individual indicators if practices improved in some areas but deteriorated in others, such changes would not be detectable. For these reasons, future analyses are being planned of the individual indicators of adherence.

Our findings highlight the complexities of expanding research evidence into effective, sustainable programs. Whether practices can commit the time required without policy or incentives is unknown. Further research is needed to better understand what impact practice facilitation can have on eliciting change at this level once implemented more broadly.