Comparing theory and non-theory based implementation approaches to improving referral practices in cancer genetics: a cluster randomised trial protocol

CRC surgery, pathology, and familial cancer clinic (FCC) clinical data will be retrospectively collected across all hospital networks for patients admitted during a 24-month period prior to study commencement, to demonstrate LS referral pathway clinical practice. The data will be cross-referenced with LS referral maps to be constructed for each hospital network. The proportion of patients who transition to each step, in adherence to the pathway map will be calculated, leading to the identification of bottlenecks. The referral maps and bottlenecks will be compared across hospital networks, with particular attention to potential differences in processes between geographical regions. A validated TDF-based questionnaire (Influences on Patient Safety Behaviours Questionnaire) [37] will be used to quantitatively evaluate psychosocial barriers to risk-appropriate completion of the LS referral pathway. Qualitative focus group data will be used to understand key barriers in context.

The primary post-implementation quantitative outcome will be the change in the proportion of patients with risk-appropriate completion of the site-specific LS tumour testing and referral pathway within 2 months of CRC resection. Secondary outcomes will be (1) proportion of high risk-patients who were referred to the FCCs (sensitivity of referral), (2) proportion of referred patients who attended FCC and (3) proportion of patients with missing testing and referral data.

The research team approached ten Australian hospital networks to participate. Due to the difficulty of obtaining the number of CRC resections performed by hospital per year, the eligibility criteria used to identify networks was defined as “hospitals where CRC resections are routinely performed”. A hospital network may consist of a single hospital or several hospitals (including co-located or off-site pathology laboratories and FCCs) in which staff and resources are shared. In each cluster, at least one hospital of > 500 beds is included. The Chief Investigator steering committee (see Additional file 2 for study personnel roles and responsibilities) was consulted to identify hospitals to approach to participate. Of the ten hospital networks approached, two were excluded. Research governance applications for eight networks (each consisting of 1–3 individual hospital sites) are being prepared for or reviewed by hospital research governance offices, and formal enrolment will be concluded when research governance approval is granted for each hospital network. No exact numbers of patients with colorectal cancer who undergo resection were available for the individual hospitals; rough estimates are 100–250 per hospital network per annum.

At each hospital network, a locally employed healthcare professional will be appointed as a paid Implementation Lead for one day per week over a 24-month period to coordinate the implementation approach and to oversee data collection. They will have a professional understanding of the LS patient pathway, and the ability to motivate other LS stakeholders to engage with implementation processes. These criteria will be used to select the most appropriate candidate for the Implementation Lead role at each hospital network.

Using snowball recruitment methods, stakeholders in the LS patient referral pathway (e.g. surgeons, pathologists, oncologists, registrars, genetic counsellors) will be invited by the Implementation Lead to participate in one or more study activities relevant to their role at timepoints throughout the implementation approach. These include process map meetings, questionnaires, focus groups and intervention implementation planning meetings.

The validated IPSBQ [37] will be used to assess barriers to the relevant target behaviour for change in relation to LS identification and referral, incorporating additional items to reflect the 14-domain TDF developed by Cane et al. (2012) [19]. Participants in the theory-based implementation arm will be provided with results from the IPSBQ to facilitate barrier identification and development of targeted intervention strategies as part of the implementation process. Participants in the non-theory trial arm will also complete the IPSBQ, but only for the purpose of measuring changes in perceived barriers pre-post intervention.

The content of the theory-based focus group schedules and additional materials (see Additional file 2) was developed based on the TDFI to encourage discussion of barriers explicitly in the context of the TDF [16, 38].

The content of the non-theory based focus groups and additional materials (see Additional file 2) was developed to promote generic discussion of barriers based on intuition and tacit knowledge, and does not include IPSBQ results or TDF-guided questions about barriers and interventions [38].

The standard training content received by both groups was developed based on common components of comprehensive implementation initiatives utilised by health services for the purposes of improvement [16, 39]. Differences in training materials used in the trial arms will be distinguished only by the inclusion of additional TDF-guided content in the theory-based implementation arm. The training will comprise an introduction to implementation science and specific instructions, practical activities and tools for moving through the 7 phases of the implementation approaches described below (see Additional file 2 for description of training contents). Implementation Leads will be informed of planned teleconference calls with the research team over the study period to support the delivery of each phase, and provided with additional tools for recording study-related activities for process evaluation purposes. These teleconferences will be used to identify and address challenges to participant recruitment and to record and manage any unintended effects of the interventions. Site-specific data management plans, including measures to protect confidentiality of individual participants, will be communicated to Implementation Leads during teleconferences.

The Implementation Leads at each site will undertake data extraction using the clinical practice data extraction tool to establish baseline evidence on existing LS referral practices (see Additional file 2 for data variables). These data will be de-identified and securely transported to the research team. The research team will carry out data checks and undertake site specific analyses for the purpose of presenting local results back to each site.

At each site, key LS patient pathway stakeholders will be identified by the Implementation Lead and invited to form an Implementation Team of 8–10 multidisciplinary change agents [16, 21, 40]. This will consist of healthcare professionals and administrators working in departments involved in the LS identification and referral pathway (including, but not limited to genetic counsellors, colorectal surgeons, pathologists, oncologists, nurses, administrative staff, multidisciplinary team (MDT) meeting coordinators). These individuals will bring expertise from various roles and departments, and therefore increase the likelihood of implementation success [41]. The Implementation Team (led by the Implementation Lead and with support from the research team) will work together through each phase and will also assist in recruiting other LS patient pathway stakeholders to relevant study activities.

At each site, the Implementation Lead and Implementation Team will participate in two meetings to identify site-specific target behaviours for change. Meeting one will involve a discussion of LS referral pathways, where participants will discuss what they believe current processes and practice to be. Based on these discussions, the Implementation Lead will generate a LS referral pathway process map, which will be later refined in collaboration with the research team. These site-specific process maps will allow the research team to identify key aspects of the process to improve and any key hospital differences.

Prior to meeting two, baseline clinical data (extracted in phase 1 and analysed by the research team) will be mapped to relevant components of the referral process. During the meeting, Implementation Team members will reflect on their experiences of and beliefs about the referral process in light of the baseline clinical data, and explore possible system and/or behavioural change targets. Identifying the target for behaviour change will involve consideration of the impact of changing the behaviour on the desired outcome; whether the behaviour change is achievable; the likely impact (positive or negative) on other, related behaviours and the degree to which the behaviour can be measured [42]. The Implementation Lead and research team will use this information to confirm the target behaviour(s) for change.

LS patient pathway stakeholders in both arms will complete the IPSBQ [37], and be invited to participate in a focus group. Focus groups will be facilitated (by the Implementation Lead) according to a theory-based or non-theory based format, depending on trial arm allocation (see Additional file 2 for focus group schedules). A member of the research team (NT) will attend meetings via video link to support discussions and elicit more in-depth responses.

For both arms, the Implementation Lead will invite LS patient pathway stakeholders to participate in a second focus group to identify and confirm intervention strategies. As in phase 4, focus groups will be facilitated differently according to trial arm allocation (see Additional file 2 for focus group schedules). Detailed descriptions of the intervention strategies used to address key barriers will be reported in accordance with Standards for Reporting Implementation Studies (StaRI) guidelines [43].

Intervention strategies generated in phase 5 will be implemented using general implementation science principles (i.e. the need for management support and commitment among target group members, use of boundary spanners, flexibility driven by local context, incorporation into established structures) [16]. With a template provided by the research team, Implementation Leads will create a report that will outline findings from phase 1–5, the intervention package and a plan for implementation. This will be submitted to senior management from departments affected by any proposed intervention strategies (e.g. CRC surgery, pathology and FCCs), and may require revision based on management feedback. Once management approves the proposed implementation plan, intervention strategies will then be presented to relevant forums (e.g. MDT meetings, pathology team meetings) to gain feedback that will allow the development of a broader structure to disseminate findings. The Implementation Lead will then work collaboratively with the Implementation Team to implement the intervention package over a 6-month period.

Following the 6-month intervention implementation period, LS patient pathway stakeholders from both arms will be invited to complete a second IPSBQ to assess any changes in perceived barriers to improving LS referral and diagnosis, as part of the simultaneous process evaluation. In addition, post-intervention clinical data will be extracted and compared to baseline clinical data to assess the impact of the intervention on clinical practice change, via assessment of the proportion of patients with CRC with risk-appropriate completion of the LS referral pathway.

The qualitative and descriptive outcomes (including the LS referral process maps, the descriptive baseline clinical data, the IPSBQ questionnaire results) are not amenable to sample size calculations. While it would be ideal to carry out a power calculation for the proportion of patients with risk-appropriate completion of the LS referral pathway, it is not possible in this situation due to the lack of exploratory trials in this area, including paucity of information on baseline outcomes, the number of patients per hospital and the intra-cluster coefficient (ICC) for outcomes in this trial. The power calculation will therefore be carried out once the baseline information is obtained. We also acknowledge that despite random allocation of the eight clusters, the characteristics may not be balanced between the two arms of the trial, representing a limitation of this trial.

For each hospital network, the baseline clinical data extracted in phase 1 will be analysed for concordance with the hospital-specific LS referral pathway. Patients who transition through each step - in adherence to the pathway map - will be leading to the identification of bottlenecks. In the event that 100% adherence is demonstrated at baseline, investigators from the relevant site will be offered the option of continuing or withdrawing from the trial.

Results from the IPSBQ will be analysed using descriptive statistics. Inter-item correlation will be tested to assess the internal consistency of each IPSBQ subscale (optimal range = 0.15–0.50). Pre-post questionnaire results from both trial arms will be used to assess changes in barriers as part of the aforementioned process evaluation.

Stage 1: in line with previous approaches [16], focus group data will be thematically analysed using a deductive approach, with coding according to the TDF domains. The key barrier domains emerging from the focus groups will be cross referenced with those identified by the IPSBQ, and discrepancies noted.

Stage 2: analysis of the top 3–4 barrier domains will identify overlap with existing theories of behaviour change. These theories will be reviewed to identify experimental data that has demonstrated the mechanistic effect of specific behaviour change strategies for eliciting behaviour change. This review will be used to identify categories of BCTs that may be effective [44, 45]. These techniques will be operationalised into context-specific intervention strategies to address the key barriers to generate an initial intervention package, which will be reviewed and refined in collaboration with each Implementation Lead.

Stage 1: inductive analysis will be applied to identify key intuitively derived barriers reported in the focus groups.

Stage 2: context-specific intervention strategies will be identified to address the key barriers to generate an initial intervention package, which will be reviewed and refined in collaboration with each Implementation Lead.

Deductive analysis on the second set of focus group data will be applied in both trial arms, based on the APEASE [38] criteria, to assess the perceptions of the feasibility and impact of each proposed intervention strategy. From this, the core intervention components will be refined.

The analysis of each dichotomous study outcome will be conducted at the individual patient level, using generalised estimating equation (GEE) adjustment for the clustering of patients within hospital networks. We will estimate the adjusted relative proportion (RR) of patients with the outcome using a generalised linear regression model with Poisson distribution, log link and robust standard errors. The exposure variable will be the implementation intervention status of the respective hospital network at the time of the patient’s CRC resection (pre or post implementation-intervention). We will carry out a two-sided test for each outcome, with the significance threshold defined using the Bonferroni correction for the number of outcomes tested. Missing testing and/or referral data will be treated as tests not carried out, or referral not completed, respectively. As covariates, we will consider time period (3-month intervals), time since implementation of intervention strategies (3-month intervals), patient age at operation, sex and cancer stage. The Quasi-Akaike information criterion will be used to select which covariates to include in the final model. To account for non-independence of within-hospital outcomes, all GEE analyses will be carried out with exchangeable working correlation structures. We will carry out sensitivity leave-one-out analyses by individually excluding hospital networks that accounted for more than 20% of the total patient numbers. If private patient status can be reliably obtained in at least six of eight trial sites, a supplementary analysis will be performed using private patient status as an additional covariate.

We note previous studies have found that a small number of clusters can lead to inflated type I error rates, and so-called small-sample corrections have been proposed to address this. However, these small-sample corrections were found to be too conservative for eight clusters [46]. Consequently, we will carry out a supplementary analysis using small-sample correction only for those outcomes that are robustly significantly associated with the implementation intervention status.