Determination of the effectiveness of two methods for usability evaluation using a CPOE medication ordering system

To assess the effectiveness of two usability evaluation methods, cognitive walkthrough (CW) and think aloud (TA), for identifying usability problems and to compare the performance of CW and TA in identifying different types of usability problems.

A CW was performed by two usability evaluators and 10 physicians were recruited to perform a TA usability testing of a CPOE system (Medicator). The severity of identified usability problems was determined and the usability problems were categorized based on the User Action Framework (UAF). The potential of usability problems to cause medication errors was also determined. The thoroughness, validity and effectiveness of the two methods were compared.

Fifty seven unique usability problems of different severity, spread over the four phases of interaction as defined by the UAF, were identified. The effectiveness of the TA method for identifying usability problems was 0.08 higher than that of the CW (0.70 vs. 0.62). The thoroughness (the extent to which a method can identify existing usability problems) of the TA was higher for the "Planning" and "Assessment" phases and lower for the "Translation" phase (as defined by UAF). The thoroughness of TA for identifying problems that may potentially result in medication errors was higher than that of CW (0.81 vs. 0.68). The number of usability problems identified by each of the methods was significantly less than the total number of detected real usability problems in Medicator (p<0.001). The observed differences between the number of real usability problems identified by CW and TA (38 vs. 41), the difference between the average severity of the detected problems by CW and TA (2.37 vs. 2.41), and the difference for identifying problems potentially resulting in medication errors (15 vs. 18) were not statistically significant (p>0.4).

This study shows that although TA showed a slightly better effectiveness, there is no significant difference between the performance of the CW and the TA methods in terms of number of usability problems identified and the mean severity of these problems. Since no single evaluation method will uncover all of the usability problems a combination of methods is advised as the most appropriate approach, especially if usability problems can lead to potentially fatal outcomes.