Prediction of Multisite Pain Incidence in Adolescence Using a Machine Learning Approach: A 2-Year Longitudinal Study

Multisite pain is a prevalent and significant issue among adolescents, often associated with adverse physical, psychological, and social outcomes. We aimed to (1) predict multisite pain incidence in the whole body and in the musculoskeletal sites in adolescents, and (2) explore the sex-specific predictors of multisite pain incidence using a novel machine learning (ML) approach (random forest, AdaBoost, and support vector classifier).

A 2-year longitudinal observational study (2013-2015) was conducted in a population-based sample of Finnish adolescents (N = 410, 57% girls, 12.5 years (SD = 1.2) at baseline). Three different data sets were used. First data included 48 pre-selected variables relevant for adolescents' health and wellbeing. The second data included nine physical fitness variables related to the Finnish national 'Move!' monitoring system for health-related fitness. The third data set included all available baseline data (392 variables). Multisite pain was self-reported weekly pain during the past 3 months manifesting in at least three sites and not related to any known disease or injury. Musculoskeletal pain sites included the neck/shoulder, upper extremities, chest, upper back, low back, buttocks, and lower extremities. Whole body pain sites also included the head and abdominal areas.

Overall, 16% of boys and 28% of girls developed multisite pain in the whole body and 10% and 15% in the musculoskeletal area during the 2-year follow-up. The prediction ability of ML reached area under the receiver operating characteristic curve 0.78 at highest but remained mainly < 0.7 for the majority of the methods. With ML, a broad variety of predictors were identified, with up to 33 variables showing predictive power in girls and 13 in boys.

The results highlight that rather than any isolated variable, a variety of factors contribute to future multisite pain.