As described earlier, OPEM was originally conceived as a two-by-two factorial design. This design was modified for the second and third iteration, transforming it into an incomplete two-by-three factorial randomised trial, for reasons documented in the OPEM trial protocol [
3]. In the second iteration, the additional two groups had a reminder note added to the short directive message, formatted as a pad of patient-aimed reminder slips (short directive and pad, short directive and pad plus long discursive message). In the third iteration, the additional two groups had an outsert message developed based on the TPB, in comparison with the "standard" short messages similar to those developed for the first two iterations. The development of the psychologically informed outsert message is described in the "TRY-ME" Study Protocol [
8]. Table
1 describes the groups in each iteration.
Table 1
Description of the intervention groups within the two replicates of the OPEM Trial
| |
Insert
|
No insert
|
OUTSERT
|
Patient Reminder Note
| 1. Insert & Outsert & Patient Reminder | 2. Outsert & Patient Reminder Note |
|
No Patient Reminder Note
| 3. Insert & Outsert | 4. Outsert only |
NO OUTSERT
| | 5. Insert Only | 6. No PEM |
REPLICATE 3: Diuretics for first-line treatment of hypertension
|
| |
Insert
|
No insert
|
OUTSERT
|
Theory-based Outsert
| 1. Insert & Theory-based Outsert | 2. Theory-based Outsert Only |
|
Non-theory-based outsert
| 3. Insert & Non-theory-based Outsert | 4. Non-theory-based Outsert only |
NO OUTSERT
| | 5. Insert Only | 6. No PEM |
Theory-based process evaluations collect data on theoretical constructs alongside randomized trials to explore potential causal mechanisms. We hypothesize that the OPEM intervention causes changes in physicians' intentions due to improved attitudes or subjective norms with little or no change in perceived behavioural control. We will test this hypothesis using the TPB model that incorporates these constructs [
9]. We will develop theory-based surveys using standard methods [
10,
11] based upon the TPB for the second and third replications, and survey a subsample of recipients from each arm of the trial two months before and six months after the dissemination of the index edition of
informed ( given the timing of the funding application and decision, we were unable to conduct a theory-based replication for the first replication of the OPEM trials). We will use Dillman's total design method to maximise response rates [
12]. Analysis initially will assess the internal reliability of the measures, and use regression to explore the relationships between predictor and dependent variable (intention to undertake the recommended practice). We will then compare groups using methods appropriate for comparing independent samples (t-tests to compare two groups, analysis of covariance to compare groups adjusting for differences in baseline performance) to determine whether there have been changes in the predicted constructs (attitudes, subjective norms or intentions) across the study groups as hypothesised. We will use the Cox-Wermuth method (described below) for exploring dependencies and associations within systems to explore whether there is convergence between the theory-based process evaluation results and the main trial results.
Phase 1. Development of survey instruments
We will develop the survey instrument using standard methods [
10]. TPB instruments can be developed based upon direct measures of the TPB constructs, or based on belief measures of the TPB constructs. The direct measures are relatively straightforward to develop and are relatively short and easy to complete (three to five items per construct, i.e., a total of 15–20 items). In contrast, belief-based measures are more complex to develop, and are considerably longer and more complex to complete. Belief-based measures are likely to be most beneficial if the aim is content-focused, that is if the goal is to identify specific beliefs that could be effectively targeted by an educational intervention. In the present study, the aim is to identify the causal mechanisms through which the OPEM interventions do or do not work; direct measure surveys are generally sufficient for this purpose and are more likely to be acceptable to physicians especially for repeated surveys.
We therefore plan to use a direct measure survey. Careful specification of the behavior is essential during the development of TPB surveys. We will decide on the specification of the behavior based on drafts of the short and long educational messages and the primary outcome for the OPEM trial. The specified behavior will be defined in terms of the TACT (target, action, context and time) principle (for example, prescribing diuretics as the first line treatment in newly diagnosed elderly hypertensive patients in the next six months). We will measure generalized intention via respondents' responses to three items measured on a seven point response format ("I will <behaviour>", " I plan to <behaviour>", and "I intend to <behaviour>". For example, "I plan to prescribe thiazide diuretics in newly diagnosed elderly hypertensive patients in the next six months"). Our direct measure of attitude will use a common stem (for example, "For me, prescribing thiazide diuretics in newly diagnosed elderly hypertensive patients in the next six months would be: ...") and four items using evaluative bipolar adjectives with a seven point response format (for example, "good practice...bad practice"). We will use both instrumental items (reflecting whether the behavior achieves something, for example, "<behaviour> is necessary..... unnecessary") and experiential items (reflecting how the respondents feel when performing the behaviour, for example, "satisfying..... not satisfying"). The specification of the bipolar adjectives will be considered carefully during both the development and pilot testing of the interview. Our direct measure of subjective norms will involve three items with a seven point response format anchored by "strongly agree" to "strongly disagree" (for example, "Most people who are important to me think that <behaviour>", "It is expected of me that I <behaviour>", and "I feel under social pressure to <behaviour>", for example, "I think most general practitioners/family physicians would approve of me prescribing thiazide diuretics in newly diagnosed elderly hypertensive patients in the next six months"). Our direct measure of perceived behavioral control will involve four items with a seven point response format. We will use items relating to both difficulty (whether the respondent thinks that she can actually do the behavior, e.g., "Doing the <behavior> is difficult for me","I am confident that I could <behavior>"), and controllability (whether the respondent believes that she is in control of the behavior, e.g., "There are factors outside of my control that would prevent me from prescribing thiazide diuretics in newly diagnosed elderly hypertensive patients in the next six months"). We will distribute items throughout the questionnaire so that questions used to assess different measures are interspersed to avoid a response set bias. We will also measure habit (past behaviour) by asking the respondents: "Thinking about your last ten elderly patients newly diagnosed with uncomplicated hypertension, for how many of them did you prescribe thiazide diuretics as a first-line drug treatment?" The survey will also include demographic questions to provide information about the sample.
We anticipate that each survey will have 15–20 items and could be completed by practitioners in 5 – 7.5 minutes. Initial drafts of each survey will be circulated within the OPEM, and OPEM theory-based process evaluation project teams to ensure face and content validity. We will pilot each survey with six family physicians using a semi-structured interview format.