Application Value of a Machine Learning Model in Predicting Mild Depression Associated with Migraine without Aura

Abstract

Aims/Background To investigate the application value of a machine learning model in predicting mild depression associated with migraine without aura (MwoA). Methods 178 patients with MwoA admitted to the Department of Neurology of the First Affiliated Hospital of Anhui University of Traditional Chinese Medicine from March 2022 to March 2024 were selected as subjects. According to their inpatient medical records, 38 patients were selected as the validation group by random number method, and the remaining 140 patients were included in the modelling group. According to the diagnosis results, the patients in the modelling group and validation group were further divided into a MwoA with mild depression group and a MwoA without mild depression group. Results The results of univariate analysis and Multivariate logistic regression analysis showed that gender, course of disease, attack frequency, headache duration, Migraine Disability Assessment Questionnaire (MIDAS), and Headache Impact Test-6 (HIT-6) score were independent influencing factors for mild depression in MwoA patients (p < 0.05). The receiver operating characteristic (ROC) analysis results showed that the area under the curve of the established prediction model for MwoA patients with mild depression in the modelling group and the validation group was 0.982 and 0.901, respectively, the sensitivity was 0.978 and 0.857, respectively, and the specificity was 0.892 and 0.929, respectively. Conclusion Gender, course of disease, seizure frequency, headache duration, MIDAS score, and HIT-6 score are independent influencing factors for mild depression in patients with MwoA. The model displays good performance for the prediction of mild depression in patients with MwoA.