An Artificial Neural Network Approach for Predicting Functional Outcome in Fibromyalgia Syndrome after Multidisciplinary Pain Program

Objective The objective of this study was to evaluate the ability of artificial neural networks (ANNs) to predict, on the basis of clinical variables, the response of persons with fibromyalgia syndrome (FMS) to a standard, 4‐week interdisciplinary pain program. Design The design of this study is retrospective longitudinal. Setting Fibromyalgia outpatient clinic in a tertiary‐care general hospital. Subjects The subjects of this study include outpatients with FMS. Intervention Multidisciplinary pain program including pain pharmacotherapy, cognitive‐behavioral therapy, physical therapy, and occupational therapy. Outcome Measures Reliable change (RC) of scores on the Stanford Health Assessment Questionnaire (HAQ), and accuracy of ANNs in predicting RC at discharge or at 6‐month follow‐up as compared to Logistic Regression. Results ANN‐based models using the sensory‐discriminative and affective‐motivational subscales of the McGill Pain Questionnaire, the HAQ disability index, and the anxiety subscale of Hospital Anxiety and Depression Scale at baseline as input variables correctly classified 81.81% of responders at discharge and 83.33% of responders at 6‐month follow‐up, as well as 100% of nonresponders at either evaluation time‐point. Logistic regression analysis, which was used for comparison, could predict treatment outcome with accuracies of 86.11% and 61.11% at discharge and follow‐up, respectively, based on baseline scores on the HAQ and the mental summary component of the Medical Outcomes Study—Short Form 36. Conclusions Properly trained ANNs can be a useful tool for optimal treatment selection at an early stage after diagnosis, thus contributing to minimize the lag until symptom amelioration and improving tertiary prevention in patients with FMS.