Employing a user-centered cognitive walkthrough to evaluate a mHealth diabetes self-management application: A case study and beginning method validation

Introduction: Self-management of chronic diseases using mobile health (mHealth) systems and applications is becoming common. Current evaluation methods such as formal usability testing can be very costly and time-consuming; others may be more efficient but lack a user focus. We propose an enhanced cognitive walkthrough (CW) method, the user-centered CW (UC-CW), to address identified deficiencies in the original technique and perform a beginning validation with think aloud protocol (TA) to assess its effectiveness, efficiency and user acceptance in a case study with diabetes patient users on a mHealth self-management application.

Materials and methods: A total of 12 diabetes patients at University of Utah Health, USA, were divided into UC-CW and think aloud (TA) groups. The UC-CW method included: making the user the main evaluator for detecting usability problems, having a dual domain facilitator, and using three other improved processes: validated task development, higher level tasks and a streamlined evaluation process. Users interacted with the same mHealth application for both methods. Post-evaluation assessments included the NASA RTLX instrument and a set of brief interview questions.

Results: Participants had similar demographic characteristics. A total of 26 usability problems were identified with the UC-CW and 20 with TA. Both methods produced similar ratings: severity across all views (UC-CW = 2.7 and TA = 2.6), numbers of problems in the same views (Main View [UC-CW = 11, TA = 10], Carbohydrate Entry View [UC-CW = 4, TA = 3] and List View [UC-CW = 3, TA = 3]) with similar heuristic violations (Match Between the System and Real World [UC-CW = 19, TA = 16], Consistency and Standards [UC-CW = 17, TA = 15], and Recognition Rather than Recall [UC-CW = 13, TA = 10]). Both methods converged on eight usability problems, but the UC-CW group detected five critical issues while the TA group identified two. The UC-CW group identified needed personalized features for patients' disease needs not identified with TA. UC-CW was more efficient on average time per identified usability problem and on the total evaluation process with patients. NASA RTLX scores indicated that participants experienced the UC-CW half as cognitively demanding. Common themes from interviews indicated the UC-CW as enjoyable and easy to perform while TA was considered somewhat awkward and more cognitively challenging.

Conclusions: UC-CW was effective for finding severe, recurring usability problems and it highlighted the need for personalized user features. The method was also efficient and had high user acceptance. These results indicate UC-CW's utility and user acceptance in evaluating a mHealth self-management application. It provides an additional usability evaluation technique for researchers.