Applying Computerized Adaptive Testing to the FACE-Q Skin Cancer Module: Individualizing Patient-Reported Outcome Measures in Facial Surgery.
Ottenhof MJ., Geerards D., Harrison C., Klassen AF., Hoogbergen MM., van der Hulst RRWJ., Lee EH., Pusic AL., Sidey-Gibbons CJ.
BACKGROUND: Skin cancer is among the most frequently occurring malignancies worldwide, which creates a great need for an effective patient-reported outcome measure. Providing shorter questionnaires reduces patient burden and increases patients' willingness to complete forms. The authors set out to use computerized adaptive testing to reduce the number of items needed to predict results for scales of the FACE-Q Skin Cancer Module, a validated patient-reported outcome measure that measures health-related quality of life and patient satisfaction in facial surgery. METHODS: Computerized adaptive testing generates tailored questionnaires for patients in real time based on their responses to previous questions. The authors used an open-source computerized adaptive testing simulation software to run item responses for the five scales from the FACE-Q Skin Cancer Module (i.e., scar appraisal, satisfaction with facial appearance, appearance-related psychosocial distress, cancer worry, and satisfaction with information about appearance). Each simulation continued to administer items until prespecified levels of precision were met, estimated by standard error. Mean and maximum item reduction between the original fixed-length short forms and the simulated versions were evaluated. RESULTS: The number of questions that patients needed to answer to complete the FACE-Q Skin Oncology Module was reduced from 41 items in the original form to a mean of 23 ± 0.55 items (range, 15 to 29) using the computerized adaptive testing version. Simulated computerized adaptive testing scores maintained a high correlation (0.98 to 0.99) with the score from the fixed-length short forms. CONCLUSIONS: Applying computerized adaptive testing to the FACE-Q Skin Cancer Module can reduce the length of assessment by more than 50 percent, with virtually no loss in precision. It is likely to play a critical role in the implementation in clinical practice.