Using registry and questionnaire data from 47,235 chronic pain patients, we evaluated whether supervised learning outperforms clustering in predicting nine one-year outcomes. Supervised models showed small but consistent improvements (best RMSE 5.49 vs. 5.92; R2 0.34 vs. 0.19), with overall comparable performance.