Handwork vs machine: a comparison of rheumatoid arthritis patient populations as identified from EHR free-text by diagnosis extraction through machine-learning or traditional criteria-based chart review

Table 2 Comparison of baseline characteristics between the machine learning defined case selection (cutoff=0.83) and the two criteria based selections

	Patients from the cohort with EHR data and classification data
	Predicted case based on machine learning (cutoff=0.83)	1987 criteria Based cases	2010 criteria Based cases
N^☨	373	357	426
Proportion women	0.65	0.63	0.66
Proportion anti-CCP2-positive	0.52	0.49	0.49
Proportion RF-positive	0.56	0.57	0.58
Median DAS44 at baseline	2.8	2.9	2.9
Median BMI	26.0	25.6	25.6
Median ESR	25	29	27
Median CRP	9.5	10.2	9.0
Median age at inclusion	57.2	58.6	57.2
Median symptom duration at diagnosis (days)	92.0	90.0	91.0
Median number of swollen joints	5	6	6

P values were calculated with the Pearson chi-squared for proportions, Mann-Whitney U for medians: *p<0.05; **p<0.01, ***p<0.001; ^☨Not statistically tested

ISSN: 1478-6362