Web usage data in clinical trials – how can we determine dose?

The use of web based interventions in clinical trials is on the increase (with a recent literature identifying in excess of 2800 journal articles relating to online intervention trials), as the internet provides an accessible mechanism for delivering intervention without the need for participants to travel to a clinic. However, it does not appear that the effectiveness of web interventions is being linked back to the actual usage of the intervention.
Collection of intervention use, or “dose”, data is becoming increasingly important, in order to inform “causal” analysis methods accounting for actual intervention use (rather than simply analysing according to randomisation, using “intention to treat”). In the context of an online intervention, however, it is not immediately obvious how best to define and measure “dose”.

Participants’ use of an online intervention can be recorded and monitored using various techniques, including Google Analytics (GA), server log data and customisable call backs to the projects servers; however, the reliability of these approaches is not guaranteed. In particular, it is likely that many users are unaware of the inaccuracies associated with GA data.

This project seeks to guide trialists on best practice of collection and use of online intervention usage data, to ensure consistent and reliable comparisons of web intervention “dose” to be evaluated across studies. This guidance material would ultimately be offered as an extension to CONSORT guidelines, relating specifically to the design, conduct and analysis of online intervention trials.
Scientific ObjectivesThe aim of this project would be to: 1) undertake a literature review to ascertain current practice among online intervention trials in terms of collecting, reporting and analysing web usage data; 2) use in-house generated web usage data and data collected for the REACT trial (an online trial evaluating the effectiveness of a peer supported self-management intervention (Relatives Education And Coping Toolkit) for relatives of people with psychosis or bipolar disorder) to compare GA data with in-house server log and video data, in order to demonstrate the extent of GA inaccuracies and determine best practice for capturing accurate representation of web usage (on features such as video and page downloads, module use and navigational patterns, allowing for time away from screen, multiple browsers, mouse hovering etc.); 3) demonstrate how to use web usage data (from REACT trial) to inform causal analyses (adjusting for patterns of navigational sequencing and module use in order to ascertain whether certain patterns of web use correspond to improved outcome) as an example of how to apply causal analysis in this field; 4) develop guidelines and toolkit for trialists on how best to collect, report and analyse web usage data as part of online intervention trials.
Person specificationThis studentship requires a degree/Masters in a relevant discipline demonstrating statistical/numerical skill. The ideal candidate would demonstrate an interest in the use of web based interventions for clinical research, with a statistical/numerical background of sufficient depth to competently apply current causal analysis methodology to this new field of online interventions.
Training and supportThe student would be provided with all relevant IT and statistical training by the supervisors and colleagues in Lancaster, Nottingham and UCL. Ian White (UCL) will provide statistical expertise on the application of causal methodology. Collaboration with the Spectrum Centre for Mental Health Research at Lancaster University (REACT trial) will be strengthened and links with research on User Interface design and usage analysis (Jason Alexander and Borja De Balle Pigem, Data Science Institute) undertaken at Lancaster will be utilised. Professor Chris Hollis (Director, NIHR MindTech HTC, University of Nottingham) will provide expertise on development and implementation of digital technologies to assess, treat and monitor mental health disorders.

Applicants should send a CV, academic transcripts, a letter of motivation and two names of referees who can send letters of recommendation to Nyree Collinson [Email Address Removed].