Optimising personal continuity for older patients in general practice: a study protocol for a cluster randomised stepped wedge pragmatic trial

Surveys will be used to investigate patients’ and GPs’ views on personal continuity to determine relevant components of the intervention. The surveys will contain questions regarding GPs’ and patients’ views on personal continuity, factors influencing personal continuity and their suggestions on how to improve personal continuity in general practice. Both questionnaires will be based on previous research by Schers et al. [5, 6]. and the results of the literature study performed in phase 0.

For the patient survey, we will recruit 35 GPs from general practices in two regions in the Netherlands and instruct them to take a random sample of 30 patients aged 65 years and older from the electronic medical record (EMR) (n = 35*30 = 1050). The selected patients will receive a postal survey with an accompanying letter from their GP. For the GP survey, we will randomly select 100 GPs for participation through various local GP networks. The GPs will receive a digital survey through Survalyzer: an online program for distributing, collecting and analysing surveys [59].

In order to prioritise the components of the intervention, focus groups will be organised to discuss the results of the literature study and surveys. GPs, patients, practice nurses and practice assistants will discuss the perceived threats and opportunities for personal continuity in daily practice, share opinions on how to improve personal continuity and discuss perceived barriers and facilitators for introducing these improvements. We aim to include 6–12 participants per focus group and to conduct 4 focus groups of 45–90 min. Two focus groups will consist of only GPs, one will consist of only patients aged 65 years or older, and one will include both practice nurses and practice assistants. Practice assistants are employed in all Dutch general practices [60], perform receptionist duties and have their own consultation hours for supportive medical tasks making them important stakeholders for assessing personal continuity in Dutch general practice. Participants will be purposefully selected in order to maximise the variance in participant characteristics such as gender, age and practice style. An external moderator with experience in moderating focus group discussions and affiliation with the research topic will conduct the focus groups. An observer/recorder will be present at each focus group session. Each focus group will be audio-recorded and transcribed. The transcripts will be checked for accuracy and corrected if necessary.

At the end of phase I, the results from the surveys and focus groups will be used to perform a reactive Delphi study among members of our research group (HS, JB, MS, AU, HH and OM) to determine the definite components of the intervention. Here, participants will be asked to give their opinion on predetermined topics in multiple rounds until group consensus is reached [61]. In round one, the members of the research group will receive an overview of draft components from the surveys and focus groups. They will be instructed to rate these draft components on four items: effectiveness and financial, organisational and time feasibility on a five-point Likert scale. Additionally, the members of the research group will be invited to elaborate on their decision and suggest changes. The Likert-scale scores will be trichotomised into negative, neutral and positive scores for purposes of analysis, according to previous research [62, 63]. Consensus will be defined as at least 75% agreement on either a positive, neutral or negative judgement [56]. Next, two members of our research group will independently review the scores of round one and determine which components will be included or excluded from the intervention based on group consensus on the four items.

In round two, the members of the research group will receive an overview of the results of round one, including a list of components on which no consensus has been reached during the review process. For each component on this list, we will ask the members of the research group a single yes/no question (“would you include this component in the intervention”) and invite them to elaborate on their decision. The inclusion or exclusion of components will be determined by majority decision (> 50%). An overview of all components, including scores and comments, will be distributed to the members of our research group inviting them to react to the results.

The literature study, surveys, focus groups and Delphi study will result in a list of definite components which will be conjoined in a complex intervention for optimising personal continuity for older patients: the TOOL-kit.

30 General practices from the Amsterdam (n = 15) and Nijmegen (n = 15) areas (Fig. 2) will be included. Practices will be included purposefully to comprise a broad spectrum of practice characteristics based on level of urbanisation, practice size and social-economic status of the practice area. Van Stippend and Schers observed that group practices tend to be more at risk for discontinuity (unpublished results, see Additional file 2). Therefore, only general practices employing three or more GPs will be included.

From the participating general practices, we will recruit GPs and practice assistants for the evaluation of the intervention. All GPs and practice assistants working in permanent employment of the practice are eligible for inclusion. Per practice, a minimum of three GPs and one practice assistant will be included resulting in the enrolment of at least 90 GPs and 30 assistants for the evaluation. This threshold was chosen to comprise a purposeful sample of the study population with sufficient variation in population characteristics while maintaining trial feasibility in regard to recruitment. Participating practices will be asked to bring forward eligible GPs and practice assistants for recruitment at the moment of inclusion.

Practices will be randomised to a wedge using computer-generated random numbers in Microsoft Excel 2016 by a researcher from our group. General practices randomised to wedge 1 will start with the intervention immediately after baseline measurement. Practices of wedge 2 will start in the control condition, providing 3 months of usual care before receiving the intervention. Practices of wedge 3 will start in the control condition, providing usual care for 6 months before start of intervention.

Researchers will be blinded to group assignment during data analysis where possible. Due to the study design and nature of the intervention, researchers, GPs and practice assistants cannot be blinded to group assignment.

Individual GPs and practice assistants will receive digital information about the study and a consent form. They can give consent for participation by returning the consent form to a designated researcher from our group.

The TOOL-kit will be introduced in all participating practices during the course of the study. After receiving the TOOL-kit, practices will be instructed to use it to optimise personal continuity for their older patients. Because the TOOL-kit is introduced on a practice level, all employees and all patients of a practice will be exposed to the intervention. Its introduction is the complex intervention of the study.

The TOOL-kit will consist of multiple components defined in phase 0, I and II. As complex interventions may work best when adapted to local circumstances [48], the TOOL-kit will contain both components which are obligatory and identical for all practices and optional components tailored to the local circumstances of a practice. The components will include professional-oriented strategies (e.g. educational materials, audits and feedback), patient-oriented strategies (e.g. decision aids, arrangements for (direct) contact with healthcare providers) and organisation-oriented strategies (e.g. arrangements for follow-up of patients, revision of professional roles) [58].

During the trial, all practices will ensure GP care will continue unrestricted. All participating general practices will receive an information leaflet instructing them to continue using the Dutch College of GPs (NHG) guidelines for clinical decision-making during the trial.

The primary outcome measure will be continuity of care from the patients’ perspective using the Nijmegen Continuity Questionnaire (NCQ) (see Table 1). The NCQ consists of 28 items within three subscales: ‘personal continuity: care provider knows me’, ‘personal continuity: care provider shows commitment’ and ‘team/cross-boundary continuity’. Items are scored on a five-point Likert scale, with an additional option to choose ‘?’ (‘I don’t know’). The NCQ has been developed for measuring continuity of care in Dutch general practice and has been validated for measuring and evaluating continuity of care interventions in healthcare systems [65, 66]. As the NCQ is not specifically adapted to the constructed intervention and may not detect all direct or indirect effects, we will perform an additional process evaluation to detect intervention-specific effects [48].

The secondary outcome measurements during the trial will constitute the process evaluation. This will include measuring implementation, acceptability and feasibility of the TOOL-kit and performing an analysis of barriers and facilitators for successful implementation (Table 1).

Implementation will be evaluated using the guidance proposed by Saunders et al. [67]: the extent to which the components of the TOOL-kit are performed as planned (fidelity), the degree to which the components were provided to GPs and practice assistants (dose delivered) and the extent to which these participants have adopted the components in daily practice and are performing the desired behaviour (dose received).

The acceptability and feasibility of the TOOL-kit will be determined by evaluating the experiences of the GPs and practice assistants with the intervention. This will include the satisfaction with the intervention, its perceived appropriateness to daily practice, work related stress, job satisfaction and support among participants for continued use of the TOOL-kit.

Barriers and facilitators as experienced by the GPs and practice assistants will be collected using a framework proposed by Fleuren et al. (see Additional file 3) [68]. This framework includes 50 factors that can influence the implementation difficulty of a health care innovation.

The sample size calculation was based on our primary outcome measure (the NCQ). For both NCQ distribution cycles in month 0 and month 12, we will approach 100 patients per practice. We expect at least 35 patients per practice to participate, resulting in a total of 1050 patients in each cycle. When the mean NCQ score is compared between cycles using a mixed linear model with a random effect for practice and two-sided testing at a significance level of 5%, the expected sample size allows detection of all standardized effect sizes of 0.12 and larger with at least 80% power irrespective of the intra-cluster correlation coefficient. The standardized effect size is defined as the difference in means divided by the total standard deviation incorporating both the between- and within-practice variation.

The NCQ will be distributed to patients in month 0 and 12 in all practices (Table 1). The participating general practices will be provided with a selection algorithm for identification of 100 eligible patients from the EMR. Eligibility criteria are: aged 65 or older, registered at the participating general practice, able to speak Dutch or English, community living, no severe cognitive impairment, and at least 1 GP consultation in the past 12 months. Eligible patients will receive a postal invitation for participation together with the questionnaire.

Digital surveys will be distributed to all GPs and practice assistants using Survalyzer at baseline (month 0) and during follow-up at 3, 6, 9, 12, 15 and 18 months. The baseline survey will address barriers and facilitators for implementation related to the practice organisation, structure and working methods. This survey will contain the same questions for all practices. The other surveys will prospectively measure the implementation, acceptability and feasibility, barriers and facilitators of the intervention during the trial. Because the intervention will be tailor-made per practice, each practice will receive surveys containing questions specifically formulated for that practice to ensure optimal data collection and ability to register unexpected events [48].

Two semi-structured telephone surveys will be held with coordinating GPs of all practices within one month and three months after the introduction of the TOOL-kit in that practice. In interview one, the fidelity, acceptability and other barriers for implementation of the TOOL-kit application will be adressed. Interview two will assess the practice improvement plan for personal continuity on the same parameters as interview one. In month 18, a concluding group interview will be held by videoconference per practice with all GPs and practice assistants to conclude the process evaluation.

The interviews will be conducted by a trained interviewer, voice recorded and transcribed verbatim afterwards.

Analysis of the transcripts will be conducted by two researchers using the framework method as it allows for a structured approach of the data and easier recognition of patterns [69]. The researchers will first familiarise themselves with the contents of the transcripts and subsequently code parts of the transcript using ATLAS.ti software (v.8.4). The researchers will identify themes using the codes and will review the themes for accuracy in order to identify the relevant theoretical domains.

Descriptive statistics will be used to summarize population characteristics. For the primary analysis, the mean NCQ score at baseline and at 12 months follow-up will be compared using a linear mixed model with a random effect for practice and a fixed effect for time point. To analyse if differences depend on longer exposure to the TOOL-kit, the main effect of wedge number and the two-way interaction between wedge number and time point will be determined.

Secondary outcomes measures from the surveys will be compared before and after the intervention and between measurements on different time points using mixed linear models or generalized estimating equations depending on the type of outcome measure. This accounts for clustering of participants within the practices. For outcomes that are collected in more than two periods, the length of exposure to the TOOL-kit main determinant will be considered using the control condition as a reference. A two-sided significance level of 5% will be used for all analyses.