Fig. 3

Feature importance plots of characteristics for A LASSO and B random forests. Feature importance plots for A the DAS28-ESR model with LASSO algorithm, B the components of DAS28-ESR model with LASSO algorithm, C the DAS28-ESR model with random forests methods, and D the components of DAS28-ESR model with random forests methods are provided below. Feature importance was determined based on standardized LASSO coefficients and the Gini score for random forests. The most important feature was set to 100, and the rest is relative to that feature. DAS28ESR, Disease Activity Score with 28-joint count with erythrocyte sedimentation rate; RA, rheumatoid arthritis; SJC66, 66 Swollen Joint Count; ESR, erythrocyte sedimentation rate; ACPA, anti-citrullinated protein antibodies; TJC68, 68 Tender Joint Count; CRP, C-reactive protein; HAQ, Health Assessment Questionnaire Score; PhGA, Physician’s Global Assessment of Disease Activity; PtGA, Patient’s Global Assessment of Disease Activity; MI, missing indicator