Predicting abatacept retention using machine learning

Table 3 Key baseline data comparisons of the training/validation and test cohorts for 12- and 6-month^a models

Variable	Training/validation (n = 4218)	Test (n = 1055)	P value
Age, years	57.7 (12.7), n = 4214	57.7 (12.4), n = 1055	0.829
Female, %	79.1, n = 3337	78.7, n = 830	0.723
BMI, kg/m²	27.1 (5.7), n = 3973	27.1 (5.7), n = 1006	0.989
RA duration, years	11.1 (9.3), n = 4184	11.2 (9.8), n = 1053	0.590
TJC28	9.4 (6.8), n = 4039	9.3 (7.1), n = 998	0.720
SJC28	6.5 (5.4), n = 4052	6.3 (5.2), n = 1003	0.253
DAS28 (CRP)	4.8 (1.1), n = 3264	4.8 (1.1), n = 815	0.678
DAS28 (ESR)	5.2 (1.3), n = 3194	5.3 (1.3), n = 785	0.282
Physician Global Assessment, VAS	58.0 (19.4), n = 3600	58.8 (19.2), n = 908	0.254
Patient Global Assessment, VAS	61.5 (21.6), n = 3548	62.4 (21.3), n = 895	0.234
Patient pain, VAS	61.4 (22.0), n = 3544	62.8 (21.6), n = 891	0.085
HAQ-DI	1.4 (0.7), n = 3101	1.4 (0.7), n = 768	0.492

All data are mean (SD) unless specified. P values for continuous variables are from a t-test and for categorical variables from a chi-squared test
^a6-month retention from baseline
BMI body mass index, CRP C-reactive protein, DAS28 Disease Activity Score in 28 joints, ESR erythrocyte sedimentation rate; HAQ-DI Health Assessment Questionnaire-disability index, RA rheumatoid arthritis, SD standard deviation, SJC28 swollen joint count in 28 joints, TJC28 tender joint count in 28 joints, VAS visual analog scale

ISSN: 1478-6362