Persistent oro-facial pain (POFP) is disabling, and patients' treatment outcomes are difficult to predict; psychosocial factors play a role. The West Haven-Yale Multidimensional Pain Inventory (MPI) is a self-report measure, which to our knowledge, has not been studied across primary and secondary care in heterogeneous POFP. Assess the MPI's ability to predict clinical outcome in POFP patients across primary and secondary care settings receiving usual care. About 146 patients receiving usual care for POFP were recruited from primary and secondary care medical and dental practices in north-east England. Participants completed the MPI (v3) and Graded Chronic Pain Scale (GCPS) at recruitment, and after 6, 12, 18 and 24months. The Patient Health Questionnaire-4 (PHQ-4) was completed at recruitment, 12, and 24months. 'Good' and 'poor' outcome status was assigned to participants based on their mode dichotomised GCPS score across timepoints. Logistic regression was used with overall GCPS outcome (good/poor) as the dependent variable and MPI subscale scores, demographic variables, and PHQ-4 scores as predictors. 110 participants had a 'good', and 36 had a 'poor' outcome. In the 'poor' outcome group, age, mean income, and life control scores were lower; deprivation, months in pain, PHQ-4, pain severity, interference, and affective distress scores were higher. In the 'good' group, MPI scores improved over time. Interference was the only consistent predictor of 'poor' outcome in the logistic regression model (OR: 1.14-1.98, p<0.05). The MPI interference subscale may help to identify patients with POFP who are likely to have consistent pain-related disability over time; it may therefore be useful clinically to identify patients likely to need early intervention.