A Digital Intervention for Respiratory Tract Infections (Internet Dr): Process Evaluation to Understand How to Support Self-care for Minor Ailments

Background Approximately 57 million physician appointments annually in the United Kingdom are for minor ailments. These illnesses could be self-cared for, which would potentially lower patients’ anxiety, increase their confidence, and be more convenient. In a randomized controlled trial of the Internet Dr digital intervention, patients with access to the intervention had fewer consultations for respiratory tract infections (RTIs). Having established intervention efficacy, further examination of trial data is required to understand how the intervention works. Objective This paper reports a process evaluation of Internet Dr usage by the intervention group. The evaluation aims to demonstrate how meaningful usage metrics (ie, interactions that are specific and relevant to the intervention) can be derived from the theoretical principles underlying the intervention, then applied to examine whether these interactions are effective in supporting self-care for RTIs, for whom, and at what time. Methods The Internet Dr trial recorded patients’ characteristics and usage data over 24 weeks. At follow-up, users reported whether their levels of enablement to cope with their illness changed over the trial period. The Medical Research Council process evaluation guidance and checklists from the framework for Analyzing and Measuring Usage and Engagement Data were applied to structure research questions examining associations between usage and enablement. Results Viewing pages containing advice on caring for RTIs were identified as a meaningful metric for measuring intervention usage. Almost half of the users (616/1491, 42.31%) viewed at least one advice page, with most people (478/616, 77.6%) accessing them when they initially enrolled in the study. Users who viewed an advice page reported increased enablement to cope with their illness as a result of having participated in the study compared with users who did not (mean 2.12, SD 2.92 vs mean 1.65, SD 3.10; mean difference 0.469, 95% CI 0.082-0.856). The target population was users who had visited their general practitioners for an RTI in the year before the trial, and analyses revealed that this group was more likely to access advice pages (odds ratio 1.35, 95% CI 1.159-1.571; P<.001). Conclusions The process evaluation identifies viewing advice pages as associated with increased enablement to self-care, even when accessed in the absence of a RTI, meaning that dissemination activities need not be restricted to targeting users who are ill. The intervention was effective at reaching the target population of users who had previously consulted their general practitioners. However, attrition before reaching advice pages was high, highlighting the necessity of prioritizing access during the design phase. These findings provide guidance on how the intervention may be improved and disseminated and have wider implications for minor ailment interventions.

Is the purpose of the session to collect data and/or use the intervention?
Questions are asked after login to ascertain user's purpose -either to use intervention for illness or to complete interim questionnaires. When is the intervention considered to be finished?
The study finishes after the follow-up questionnaire at 24 weeks. The intervention was finished when users no longer wished to access it. What prompts are used to encourage usage (e.g. emails, texts, notifications) and when are they sent?
Email prompts for 4 weekly data collection. Additional emails are sent after logging in during illness to prompt repeat login after 48 hours. Does the intervention contain 'tunnelled' (compulsory) sequences of pages which users have to view in order to move forward?
'Doctor's Questions' component is tunnelled as symptoms are assessed before providing illness management advice. Are users able to select linked sections they wish to view, and avoid others?
There are 3 linked menu components available from the home page. Follow-up: website satisfaction, problematic experiences of therapy scale 9 .

Log-data. Information automatically collected through engagement with an intervention.
What data is the software platform able to record? Time and date, pages viewed & order, time spent on pages, self-report measures. Are number, date and time of logins available by individual user?
Yes. Are individuals' total durations of usage accessible?
Need to be extracted. Are the number and time of usage prompts recorded?
Overall scheduled timings for emails are available, but not sent times by individual. Are there details for which pages were viewed, the sequential order and time spent viewing? Yes.

External data. Data collected independently but alongside intervention usage.
How and where is the data collected?
Hand collected by the research team from users' GP notes.
What data is collected?
Number of GP visits for RTI during trial, for year prior to trial, and co-morbid illnesses. Antibiotic prescriptions for RTI. Which of these measures relate to or may impact on the target behavior?
Number of GP visits for RTI during trial, for year prior to trial, and co-morbid illnesses. 3. Contextual data. Data indirectly related to the running of the intervention which may be influential over usage and analysis.

External factors. Structures and events which may influence participation in the intervention.
How are users recruited to the intervention?
Recruited via GP. Did any specific large-scale events, with the potential to impact on the intervention, occur during the period of the intervention? n/a

Previous theory and findings. Results of behavioral analyses carried out during intervention development (e.g. logic models), and analyses of clinical outcomes if available.
What are the hypothesised mechanisms of the intervention (e.g. as specified in the intervention's logic model)?
Viewing Doctor's Questions and Common Questions will raise users' understanding of their symptoms, leading to them being less likely to consult their GP. Viewing Treatment Options will raise users' self-efficacy in their ability to self-manage their illness, leading to them being less likely to consult their GP. Level of anxiety may be associated with usage. Which factors are identified as important in qualitative research, and can they be related to the variables collected in the trial (e.g. preferences for specific pages)?

N/a
Which variables are identified as relating to outcomes (e.g. behavioral determinants, theoretical constructs, health factors)?
Analysis of the data from the RCT focused on the outcomes of GP contacts and antibiotic use 10 . No analyses were carried out for usage, behavioral determinants or personal characteristics.