The web-based IME will deliver one of the interventions used in the paper-based IME  -- a persuasive communication -- and a second, new intervention developed for the web-based IME. The persuasive intervention aimed to reinforce the GP's beliefs about the positive consequences of managing sore throat without prescribing antibiotics. The materials for the persuasive communication can be downloaded at http://www.biomedcentral.com/content/supplementary/1472-6963-8-11-s2.doc. This intervention did change GPs' beliefs in the paper-based IME, and we aim to see if this is replicated in the web-based IME.
However, we will also deliver a new intervention that takes advantage of the possibilities of web delivery. This will allow a comparison between an intervention simulated using web and paper-based methods and, additionally, provide a head-to-head comparison between that intervention and an intervention developed without the constraints of paper delivery.
The new intervention, designed to take advantage of the possibilities offered by web-based delivery, will be designed using predictors identified from the first questionnaire and by using appropriate behaviour change techniques [16, 17]. There is no widely agreed taxonomy by which to specify methods of delivery of behaviour change techniques; we will use as a starting point the suggestions of the Evidence-based Behavioural Medicine Committee (EBMC) . Suggested dimensions are: provider, format, setting, recipient, duration, and frequency. We will use the EBMC suggestions to map the effectiveness of the interventions in the primary studies included in the systematic reviews populating it with the delivery methods used in the studies in the systematic reviews of strategies to decrease antibiotic prescribing [19, 20] and the results from the first questionnaire. We will send the results of the mapping exercise to a small number of experts for external validation.
We will use 'intervention mapping'  to systematically consider combinations of content and methods of delivery; there are likely to be logical combinations. By considering such combinations we will produce a range of potential new interventions.
We will pilot each of the potential new interventions (up to three) with groups of three to six GPs. Whilst delivering the intervention we will ask them to 'think aloud' about their reactions to the intervention, which allows us to identify problems and fine-tune content and delivery. We will select the most promising intervention for evaluation in the web-based IME. GPs involved in developing the intervention will not be included in the web-based IME.
For the web-based IME, the primary outcomes will be behavioural intention and behavioural simulation, which have been shown to be reliably related to behaviour [16, 22]. Three questions will assess GPs' strength of intention to manage URTIs without antibiotics (e.g., over the next 12 weeks, when a patient presents with a URTI, I have in mind to manage them without prescribing an antibiotic (rated on 7-point Likert scales 'Strongly Disagree' - 'Strongly Agree'). Responses will be summed with a low score corresponding to a low intention to prescribe antibiotics. Sixteen clinical scenarios from the paper-based IME work will provide the materials to measure simulated behaviour (eight in the first questionnaire, eight more in the second). Respondents will be asked how they will manage the patients in the scenarios and asked to rate, on a score out of 10, the difficulty of making their clinical decision.
Secondary outcomes are concerned with evaluating the relative utility of web-based and paper-based IMEs. For this, we will compare the following outcomes for each method.
1. Behaviour change techniques that can be operationalised
We will develop a matrix of behaviour change techniques that can be operationalised by each IME method. This work will be done with researchers (especially health psychologists) unconnected with the project in brainstorming workshops. Project team members will present work done using paper-based and web-based IMEs, and participants will work alone to place behaviour change techniques into the matrix. Following this group discussion will be used to reach consensus on which techniques could be supported by each method. We will do this with more than one group of researchers (e.g., at weekly departmental research meetings at collaborator sites). Here we are looking for the method that can be to deliver most behaviour change techniques.
2. Complexity of scenarios that can be delivered
We will develop a table of factors (e.g., people involved, context) that we are able to vary in clinical scenarios that can be delivered by each IME method. We will use brainstorming workshops involving researchers unconnected to the project to validate this list (i.e., do they agree with it) and make suggestions as to whether it can be extended (or reduced) for each IME method. We may use the same groups of researchers as for point one above although not at the same workshop. Here we are looking for the method that would allow more factors of a scenario to be varied.
3. Identification of predictors
The first part of the web-based IME will evaluate whether the delivery mechanism of the IME (paper or web) affects predictors of GP behaviour. We anticipate that the predictors coming from the two methods will be similar because only the delivery method will be different; the web-based IME will use materials from the earlier paper-based IME. For example, evidence of habitual behaviour was strongly correlated with behavioural intention in the earlier paper-based IME. When delivering the same scenarios to GPs in a web-based IME, we anticipate (but do not know) that we will also find a strong correlation between habit and intended behaviour. Therefore, here we are seeking confirmation that delivery method alone does not lead to widely different predictors being identified by the IME.
4. Recruitment (reaching target and time taken to do so)
Here, we will assess three things: time (in days) spent by the project team to recruit to target; time (in days) between sending an invitation to a potential participant and receiving a completed questionnaire; and the number of invitations necessary to receive one completed questionnaire
These will be measured in the first part of the web-based IME and compared to estimates of timing (admittedly, less accurate) from the earlier paper-based IME. We will, however, get an indication of how the two methods compare. The embedded trial of paper invitation to GPs to participate in our study versus email invitation will provide comparative data on these two recruitment methods, which will be of wider relevance than recruitment to IMEs. Here we are looking for the fastest and least labour-intense method of recruiting participants to an IME.
5. Ability to change targeted constructs
The theories used in paper- and web-based IMEs target particular constructs (e.g., GPs' beliefs about antibiotic prescribing) and the effect an intervention has on these targets is measured using a score. The web-based IME will involve a direct comparison of an intervention from the paper-based IME and a new intervention that makes use of the possibilities offered by web-based delivery. We will also compare the effect on targeted constructs of the same persuasive communication intervention delivered by both methods. By comparing each IME method's ability to change scores on the targeted constructs we will be able to make a statement as to which IME method is most effective.
6. Effect size for persuasive communication intervention compared to control
The persuasive communication intervention and control were delivered in an earlier paper-based IME. By delivering the same intervention in a web-based IME we can compare the effect on the primary outcome (intention to prescribe an antibiotic) of the two delivery methods. Here, we are looking for the greatest effect size.
7. Qualitative work: ease of administration and GP feedback
Running an IME is not trivial and a method that provides some administrative benefits (e.g., easier to put scenarios together, easier to pilot scenarios with some participants, easier to change the order of scenarios, easier to deliver to participants, easier to manage collected data) has efficiency advantages. We will compare, via interviews, experiences of the research team with running paper-based and web-based IMEs. This will, of course, be subjective (although item four will provide some quantitative information about effort linked to recruitment) but will nevertheless provide useful information about the effort required by researchers to run an IME via each approach. Here, we are looking for the method that is considered easiest to use and which uses fewest resources for an IME.
We will also consider qualitative feedback from GPs on the web-based delivery method and the utility of decision time data collected in the web-based IME, which is impossible to measure with paper-based IMEs. This work will also consider how the differing configurations of the interventions might explain their effects. Here, we are looking for feedback from GPs that identifies whether they think web-based IMEs have face validity, that the approach is feasible, and that they would participate in future web-based IMEs.
For the email versus postal invitation trial, the primary outcome is the number of GPs completing the first questionnaire. There are no secondary outcomes.