Cluster identification, selection, and description in cluster randomized crossover trials: the PREP-IT trials

In clinical research, large high-quality randomized controlled trials (RCTs) are considered the highest level of evidence to determine the effectiveness of an intervention [1]. While most RCTs follow a parallel group design and randomize individual participants to one or more intervention and control groups, there are certain situations in which randomization at this level is not feasible or practical [2]. Cluster randomized trials overcome these challenges by randomizing predetermined groups of participants (i.e., clusters) to interventions [3]. In cluster randomized crossover (CRXO) trials, each cluster serves as their own control group by participating in both the treatment and control arm at least once at various periods of time throughout the trial. An advantage to this design [CRXO] over the parallel group cluster randomized trial is that fewer clusters are required to achieve statistical power.

Researchers undertaking CRXO trials must ensure that they take into consideration multiple differences in the conduct and reporting between RCTs and CRXO trials [4, 5]. One of these fundamental differences is the requirement to provide information on the characteristics and flow of clusters throughout the conduct of a CRXO trial (i.e., enrolment, allocation, follow-up, analysis), as described in the CONSORT extension for cluster randomized trials [4]. This is accomplished by including a priori cluster selection criteria (similar to participant eligibility criteria) and a cluster flow diagram that shows the cluster selection and flow over the duration of the trial (similar to the patient flow diagram). This information enables knowledge users to assess the generalizability of the trial and determine if the results are applicable to their setting and practice.

To disseminate the early knowledge gained from our experience initiating two large infection prevention CRXO trials in orthopedic fracture patients, we describe our cluster identification and selection process. Additionally, we present data on the key characteristics of the clusters to illustrate the level of data collection that is required for appropriate reporting in CRXO trials.

The Aqueous-PREP and PREPARE trials represent two ongoing CRXO trials within the field of orthopedics. The successes we have experienced with cluster identification and selection provide support for the feasibility of using this trial design in orthopedic trauma trials. As part of the PREP-IT program’s commitment to transparency and upholding rigorous reporting standards, this paper clearly defines the cluster unit, provides detailed information about the cluster identification and selection process, and describes key characteristics of the included clusters (Fig. 4). These are some of the key reporting requirements described in the CONSORT extension for CRTs [4].

Given that the Aqueous-PREP and PREPARE trials use a pragmatic design, when selecting orthopedic practices (i.e., clusters), we aimed to include a sample that was representative of the orthopedic practices that treat the patient population in “real world” usual care settings. This is an essential component of pragmatic trials which aim to determine how well interventions work in the “real world” as opposed to explanatory trials which aim to determine how well interventions can work in ideal settings [8]. To accomplish this goal, we designed our cluster eligibility criteria to be as broad as possible while still ensuring that participating orthopedic practices would have the infrastructure and patient volume to successfully follow the protocol and meet study timelines. Through our cluster selection process, we identified 58 potentially eligible orthopedic practices and ultimately selected 5 sites to participate in both the Aqueous-PREP and PREPARE trials, 10 in the Aqueous-PREP trial only, and 15 in the PREPARE trial only. This represents an inclusion rate of 26% and 34%, respectively. Using the Pragmatic-Explanatory Continuum Indicator Summary (PRECIS-2) toolkit, a tool designed to measure where a trial falls on the pragmatic versus explanatory continuum on several domains, the setting of the Aqueous-PREP and PREPARE trials received a score of 4 out of a possible 5 indicating that the trials are “mostly pragmatic.” This score is reflective of the fact that most of our included orthopedic practices are affiliated with teaching hospitals that are level I trauma centers and regional referral centers versus a mix of trauma centers and local community hospitals. This is unsurprising as large academic hospitals tend to have both the infrastructure and patient volume to make participation in the PREP-IT trials feasible, whereas smaller community hospitals may have insufficient research infrastructure.

Despite our rigorous cluster selection process, we withdrew one cluster as they were unable to comply with the Aqueous-PREP protocol [6]. Specifically, this cluster was unable to meet the a priori threshold 90% of patients receiving the allocated antiseptic preparation solution. Additionally, the clinical site was unable to complete the case report forms, which led to incomplete data submission. These issues were the result of changes in research infrastructure from the time of site selection to site initiation. The Executive Committee, with support from the Methodology Core, made the decision to withdraw this site. As the criteria for cluster withdrawal were established prior to the initiation of the trial and a small proportion of data are affected, bias should not be introduced as a result of the decision to withdraw this cluster. Additionally, withdrawing this cluster will improve the overall integrity of the trial, as allowing them to continue in Aqueous-PREP would have led to increased protocol violations, treatment contaminations, and missing data.

As the primary outcome for both the Aqueous-PREP and PREPARE trials is SSI, we documented infection control processes. To remain pragmatic with data collection, we documented the infection control processes at three different levels: cluster level, surgeon level, and patient level. The most important SSI prevention strategies and infection strategies that are likely to vary from patient-to-patient, including antibiotic prophylaxis and blood sugar control, are documented at the patient level on the case report forms. Other infection control details (e.g., hair removal methods, type of surgical gown) are collected at the surgeon level, as these typically vary by surgeon but remain consistent across patients. Infection control measures that are implemented at the hospital level, including normothermia and decolonization protocols, are documented at the cluster level.

We aimed to select orthopedic practices that follow infection prevention protocols that are representative of the usual care that fracture patients would receive in the real world. We have found some variation in the infection control processes across the clusters in both the Aqueous-PREP trial and the PREPARE trial. Specifically, we found that 64% of all orthopedic practices in the Aqueous-PPREP trial perform active surveillance cultures for MRSA (64%). In the PREPARE trial, we found that only 20% of orthopedic practices perform active surveillance for MRSA in open fracture patients and 15% for closed fracture patients. In both trials, there was substantial variation in the use of decolonization protocols for MRSA (protocols in use at 64% of Aqueous-PREP practices and 60% of PREPARE practices).

Providing the knowledge user with a high level of detail regarding the hospital characteristics, orthopedic surgery details, and infection prevention protocols and procedures allows the knowledge user to determine the generalizability of the trial and the applicability to their hospital and surgical practice.

When undertaking CRXO trials, it is imperative to follow rigorous reporting standards, as outlined in the CONSORT extensions for CRTs and CRXO trials, including providing knowledge users with a clear definition of the cluster unit, a thorough description of the cluster identification and selection process, and sufficient description of key cluster characteristics. The PREP-IT trials illustrate how this can be accomplished using two large infection prevention CRXO trials in orthopedic fracture patients as an example.