Summary
We successfully used the Precede/Proceed to structure a synthesis of the findings from a programme of study to inform a complex intervention. We concluded that interventions which reduce clinical uncertainty, reduce clinician/parent miscommunication, make clear delayed/no-antibiotic recommendations, and provide clinicians with alternate treatment actions have the best chance of reducing antibiotic prescriptions in primary care for childhood RTI.
Our synthesis provides a method for the development of future complex interventions in primary care using a theoretical framework combined with empirical findings. This publication makes explicit the steps we undertook in intervention design prior to piloting, and enables examination of any changes we might make to the intervention design between feasibility and effectiveness studies [
19].
Strengths and weaknesses of our approach
The strength of our approach was to allow strength of evidence to be considered within theoretically driven logic model development. We integrated quantitative and qualitative findings to develop an evidence-based model that reflected the experiences and views of parents and clinicians. We produced generalizable intervention recommendations, as well as the intervention developed for this context. This was possible because of the context of a National Institute for Health Research programme grant, enabling the same group of researchers to work together across phases of intervention development [
15], and working in an interdisciplinary team in a multi-method environment. We therefore see our funding context are a key element in our success.
The weakness of this approach was in moving from intervention recommendations, to decisions about how to implement these in practice. While we were confident about the principles guiding intervention development, we had to rely on expert opinion and traditional testing with users to inform the many practical decisions involved in designing the intervention. The effectiveness of web-based interventions varies greatly [
59] and very little is known at present about the mechanisms through which interventions work. Limited systematic research had been conducted to identify the design features most likely to result in behavioural change [
60,
61].
An advantage of the Precede/Proceed framework includes consideration of actions to change social norms, regulatory constraints, resource availability. However, our application of it was limited by the research evidence available concerning childhood RTIs. Within both the primary data we generated and the systematic reviews we undertook in our TARGET programme, there was a focus on drivers of prescribing within the patient consultation. The decision aids and clinician communication tools that act within consultations sit within a broader range of interventions which act to change the context for consultations. Recent evidence for population-level changes suggests that shifting social norms through mass communication [
62] and prescriber feedback [
63] can influence antibiotic prescribing behaviour.
We underestimated the barriers to use of a stand-alone website, and were not able to integrate this intervention into existing electronic medical notes systems.
Comparison to similar literature
Interventions informed by evidence
and theory have the best chance of altering health related behaviours [
15]. However, methods and advice for how to achieve this to develop a specific intervention are sparse [
17,
64,
65]. We believe our combination of strength of evidence applied to a theoretical structure, augmented with input from key stakeholders provides a useful model to achieve this.
We drew on a developing literature using logic models to inform mixed methods syntheses of research [
66,
67]. Our work complements the contemporaneous work of Yardley and colleagues in developing their ‘person-based approach to intervention development’ which similarly combines mixed methods primary research with theoretical models [
68]. Michie’s behaviour change wheel usefully synthesised behaviour change theories to identify key influences on individual behaviours [
69]. Our strength of evidence approach adds to this body of work a mechanism for identifying which actions were most likely to elicit change in this case.
Our approach is similar to those employed by O’Brien and colleagues to develop a web-based healthy lifestyles intervention for older adults [
65] and Salisbury and colleagues to develop a new telehealth intervention for chronic disease in adults using Precede/Proceed [
57]. Both of these studies used systematic reviews of the existing evidence, supplemented with additional primary research and stakeholder consultation [
57], or co-design workshops [
65]. Both study teams describe the success of this structured approached, and the tele-health intervention has since been tested, with modest benefits for patient health [
70]. However, like us, while the evidence review provided functional guidance, authors report little guidance about the design features of the website [
65].
Responding to the scale of the AMR challenge, there are very many interventions aimed at reducing antibiotic prescribing for RTIs in primary care, although fewer focus on children specifically. Many combine approaches drawing on previous research to suggest intervention elements. Four recent trials of family practitioner training using previous research to establish effective interventions and combine approaches in a new multi-component intervention with mixed success [
63,
71‐
73], in one case highlighting differential effects by age where a reduction in prescriptions was seen for adults and adolescents, but not children under 12 [
71]. Two studies are ongoing [
74,
75]. These approaches to changing the environment for prescribing and use of social norms including through peers are consistent with our model.
There is a growing literature testing methods to change professionals’ behaviour. Prior literature establishes that education alone is insufficient to change prescribing behaviour in primary care [
76], although we know that computerised reminders can be effective [
77]. Our research contributes to the gap identified in using computerised reminders as part of a more complex decision support system [
77], but our intervention would benefit from better integration into electronic health record systems.