Table 4 Practical Guidance for Internet-Based Treatment Prediction Research

1.Keep Assessments Brief

Retention was high for brief, self-report assessments and in particular weekly check-ins were well-received by patients wishing to track their progress through treatment. Cognitive tests were by far the most disliked component of our study. Considerable work is needed to make these more tolerable for participants

2.Ensure Incentives are Aligned

The key to quality data in an online environment is to keep incentives aligned. Participants in our study resonated with the mission of the study and/or enjoyed the opportunity for self-reflection. Future research should be sensitive to these motivations and (i) communicate the mission of the study clearly, early, and often, (ii) supply participants with information about study outcomes at the time of publication, (iii) solicit feedback from participants and (iv) consider a graphical display where service users can visualise their progress throughout treatment

3.Make Participation Easy

The ease of participation is imperative to achieving successful online recruitment, for example, allowing participants to complete assessments remotely and at a time convenient to them. In addition to a PC/laptop, smartphone and smartwatch may be incorporated in future for increased convenience in online data collection. They can further facilitate the collection of different sorts of data, such as mobility data, sleep, and experience sampling data

4.Issue Regular Reminders, be Flexible and Pragmatic

To encourage retention, a timely reminder for each assessment should be delivered a day prior to due date, and a small window for completion may be provided to increase flexibility for participants to complete each assessment. Sensitivity analyses can be used to ensure late or early assessments do not confound results

5.Data Quality is not a Given

Data quality indicators (e.g., catch questions, distraction probes, and stable variables for high test–retest reliability analysis) should continue to be included for assessing the quality of self-report online data. The online research environment changes and is potentially vulnerable to bots or dishonest respondents. To reduce the threat this poses to valid research, recruitment should be targeted to those initiating treatment and include a validation check (prescription photo, iCBT registration)