Prediction models for treatment success after an interdisciplinary multimodal pain treatment program

Abstract

Chronic musculoskeletal pain (CMP) poses a widespread health and socioeconomic problem, being the most prevalent chronic pain condition. Interdisciplinary multimodal pain treatment (IMPT) is considered the gold standard, offering cost-effective long-term care. Unfortunately, only a subset of patients experiences clinically relevant improvements in pain, fatigue, and disability post-IMPT. Establishing a prediction model encompassing various outcome measures could enhance rehabilitation and personalized healthcare. Thus, the aim was to develop and validate a prediction model for IMPT success in patients with CMP. A prospective cohort study within routine care was performed, including patients with CMP undergoing a 10-week IMPT. Success across four outcome measures was determined: patients' recovery perspective, quality of life (physical and mental), and disability. Sixty-five demographic and candidate predictors (mainly patient reported outcome measures) were examined. Finally, 2309 patients participated, with IMPT success rates ranging from 30% to 57%. Four models incorporating 33 predictors were developed, with treatment control being the sole consistent predictor across all models. Additionally, predictors effects varied in direction in the models. All models demonstrated strong calibration, fair to good discrimination, and were internally validated (optimism-corrected AUC range 0.69–0.80). Our findings show that treatment success can be predicted using standardized patient-reported measures, exhibiting strong discriminatory power. However, predictors vary depending on the outcome, underscoring the importance of selecting the appropriate measure upfront. Clinically, these results suggest potential for patient-centered care and may contribute to the development of a scientifically sound decision tool.