Evaluating a Web-Generated Headache History System

In a preliminary clinical trial an automated, web-based headache history system was tested in order to see if it could produce an accurate, expert report and clinical diagnostic impression. The system was developed by two board certified headache specialists who created a decision-tree based, interactive algorithm on which the web-tool is based. The system would ask the patients a question and then, based on their response to that question, would select an appropriate next question to ask, repeating this process until a diagnosis was reached. An average of 187 questions were given to each patient, but the amount ranged from as few as 56 questions to as many as 347 questions, reflecting the dynamic nature of the questionnaire system. To test the system six headache experts each sent six patients with primary headaches to the system’s website after they had completed consultations with them. The experts then compared their diagnostic impressions to those of the on-line system to evaluate the system’s accuracy. During the initial run the system correctly identified 44% of migraine without aura and 20% of migraine with aura, but after some adjustments to the program’s rules a second run yielded accuracies of 94% and 60% respectively. The system was also able to differentiate between migraine and non-migraine patients with 97% accuracy. This trial showed that an automated, web-based headache history system can serve as a reliable aide for the collection and interpretation of patient data. Although automated systems cannot take the place of an actual headache expert’s care, they have the potential to be powerful aides to researchers, physicians, and headache specialists.