Introduction
Clinical trials are vital for informing pharmaceutical companies about drug action and efficacy, and for demonstrating the safety of new drugs. However, they can be notoriously expensive, time-consuming to conduct, and the commercial implications of failure are great. It is paramount that the trial data that is collected is valid, reliable and is useful in meeting the aims of the trial. Consequently, the healthcare industry is turning to the power and increasing capabilities of digital technology to improve this process, more specifically by collecting trial data from participants using smartphone apps.
Clinical trials: how technology is driving digitisation
One particularly challenging variable to measure via an app is a trial participant’s self-report stress levels. Historically, this has often been measured in clinical trials through a series of face-to-face meetings between a trial participant and a clinician, during which the participant answers a range of questions regarding their stress levels over a certain period of time. These measurements may be taken daily, weekly or less often, and the clinician may use one of several standardised and widely accepted measurement scales, such as the Perceived Stress Scale [1]. Collecting this self-report data remotely via an app has many benefits including: participants not needing to rely on their memory regarding their stress levels over the previous weeks providing more accurate, honest and up-to-date data; more data points for clinicians; quicker and cheaper data collection; and data that may be more in-line with the speed in which mental health problems occur.
As part of a wider project at Cambridge Consultants, we have developed an app (Verum Trial App) to facilitate the self-reporting of stress measurements for use in situations such as clinical trials. The design challenges of this are numerous, thus requiring very careful consideration at every stage of the development. It is not as simple as just transferring the questions from the existing paper-based measurement scales into an app, and it is not as simple as just adhering to established UX design guidelines such as Neilsen’s Heuristics for User Interface Design [2] and the W3C Guidelines for Web Design and Applications [3]. Some of the unique challenges that Cambridge Consultants faced when developing Verum Trial App are discussed in more detail below.
Question and response format
The question and response format used in the Verum Trial App had to be thoughtfully designed. While it may be possible to simply transfer questions from existing measurement methods to an app without changing the wording, the format of the question and response method may have to change. For example, the question may be “in the last month, how often have you been upset because of something that happened unexpectedly?” and the response may be a scale consisting of five points. However, a five-point scale, including an explanation of what each point means, may not fit horizontally within the width of a typical smartphone screen. An alternative to this could be to use a horizontal slider-style response option, with the description of the response (e.g. ‘almost never’) only displayed when the slider is at the appropriate point on the scale. This limits the amount of information on the screen but could add potential for bias in that the slider has to start somewhere on the scale to allow the user to move it and could infer a ‘correct’ response to the participant. The scale therefore may be best placed vertically, but again this could introduce subtle biases if a participant interprets this as a form of hierarchy in response options [4]. There isn’t necessarily a ‘correct’ approach to question and response format, and Figure 1 shows how both of these options have been used in the Verum Trial App, but the development of the app emphasised the importance of giving it careful consideration.
Amount and complexity of information
There is reduced opportunity for providing additional information or clarification when the participant is providing data remotely via an app (compared to that provided by a clinician in person). Therefore, not only should fonts be of adequate size, contrast, appropriately formatted for the size of device screen etc., questions also should be worded in the least complex way to avoid confusion, with an appropriate reading level for the user group. There are several methods for determining whether the text is of an appropriate reading level (such as a Flesch–Kincaid Readability Test), however, avoiding introducing complexity in other ways may be less straight-forward. For example, by phrasing some questions in the positive and others in the negative it can encourage participants to think more carefully about their answers and not just provide the same answer to each question. However, this could add complexity and room for confusion and error which could result in participants either not filling in their answers, or unknowingly giving the wrong answers which would risk invalidating the data.
Information access
When a paper-based measurement scale is presented to a participant, all the information can be presented on one sheet of paper. Given the limited size of a smartphone interface this information may have to be broken into individual chunks, making easy navigation between screens and access to the relevant information at the right time challenging. All the information that the participant requires should be easily visible to them but must also be done without overloading the participant with information. Pop-ups are generally not a suitable solution for apps due to the small screen space available, and they don’t make sense on touchscreens. Our solution was to precede important screens with introduction screens containing additional information. In this way on-screen content could be minimised for readability, especially for those steps which require greater user focus. An example of this is the ‘welcome’ screen shown in Figure 2, which was provided before the first set of questions.
Influencing stress levels
It is also paramount that the interaction that the participant has with the app does not cause additional stress to the participant. The frequency of participant interaction with the app should provide sufficient data for the clinical trial, whilst minimising user burden. In some situations, it is possible that simply asking a participant how stressed they are feeling, particularly if they are not used to thinking about their stress levels on a regular basis, could make them more aware of their feelings of stress, which may cause them to feel even more stressed. This could be amplified if the app relies upon regular notifications, warnings and reminders for participants to fill in their data. Therefore, the design of the graphical user interface (GUI), the frequency of data collection and the mechanism of reminding participants should be carefully considered. The app should also be enjoyable and simple to use so as not to cause additional frustration and stress, especially for participants who may not be particularly confident in using a smartphone.
Participant engagement
Maintaining participant engagement is another challenge in clinical trials, potentially amplified by asking participants to fill in data more frequently. Therefore, the amount of time required for data input should be kept to a minimum. With this in mind, Cambridge Consultants developed the Verum Trial App to automatically import data from other sources such as FitBit or MyFitnessPal, and then simply asking the participant to confirm that the data are correct (see Figure 3). Cambridge Consultants also limited the number of questions asked, and then introduced additional more in-depth questions at less frequent intervals so as not to overwhelm participants but also to refresh their interest and keep them engaged with the questions. It is also important that the GUI is designed to discourage participants from ‘skipping through’ the data entry if they are bored or short of time. In the Verum Trial App, Cambridge Consultants limited participants progress through the app until the participant interacted with the response input (e.g. by partially moving a slider) before they were allowed to click the ‘next’ button. Finally, it was a conscious decision in the Verum Trial App to keep the use of language relatively direct, through formal and instructional language. The aim of this was to come a cross as authoritative and encourage participants to take the trial seriously and therefore fill in the data in a timely manner.
Aesthetic considerations
The overall aesthetic portrayed to the participants also requires special consideration in the case of an app designed to measure stress in clinical trials. For example, the use of images in the app was carefully considered as it has been found that images can impact on participant responses. For example, including images of unhealthy people or of healthy people has been shown to impact on a participant’s rating of their own health [5]. Cambridge Consultants also implemented a visual language around the idea of calm, modern, clean, medical and professional, to encourage participants to take the data entry seriously, avoid adding stress, and give them a sense of trust in the app and clinical trial. Example screens portraying this aesthetic can be seen in Figure 4.
Take-away messages
Whilst there are many benefits to using apps to collect data in clinical trials, there are challenges when collecting stress data remotely, without negatively impacting on the validity or reliability of the data collected. Furthermore, modern-day app users have high expectations regarding usability, aesthetics, navigability and consistency with other apps, so are easily perturbed by apps that do not compete with market leaders in this regard. In addition to the design challenges discussed in this blog, there are of course other challenges which need to be overcome, such as the cost of the app development, the depth of data that can be collected, and disrupting an industry which may not be that familiar with using digital technology in this way. Whilst there may not be a right or wrong way of addressing some of these design challenges, we can be sure that this is a particularly interesting field, merging expertise from the fields of Human Factors, UX Design and data collection, and we hope to see it result in some exciting developments in the clinical trial process.
References
[1] Cohen, S., Kamarck, T. and Mermelstein, R., 1994. Perceived stress scale. Measuring stress: A guide for health and social scientists, pp.235-283.
[2] Neilsen’s Heuristics https://www.nngroup.com/articles/ten-usability-heuristics/
[3] W3C Guidelines https://www.w3.org/standards/webdesign/
[4] Galesic, M., Tourangeau, R., Couper, M.P. and Conrad, F.G., 2008. Eye-tracking data: New insights on response order effects and other cognitive shortcuts in survey responding. Public Opinion Quarterly, 72(5), pp.892-913.
[5] Couper, M.P., Conrad, F.G. and Tourangeau, R., 2007. Visual context effects in web surveys. Public Opinion Quarterly, 71(4), pp.623-634.
