Skip to main content
. Author manuscript; available in PMC: 2011 Jun 23.
Published in final edited form as: Arthritis Care Res (Hoboken). 2010 Aug;62(8):1120–1127. doi: 10.1002/acr.20184

Table 2.

Variables selected for the complete algorithm (narrative and codified EMR data) from the logistic regression in order of predictive value.

Variable Standardized regression coefficient Standard error
Positive predictors
NLP RA 1.11 0.48
NLP seropositive 0.74 0.26
ICD9 RA normalized* 0.71 0.23
ICD9 RA 0.66 0.44
NLP erosion 0.46 0.29
Codified RF negative 0.36 0.36
NLP methotrexate 0.3 0.34
codified anti-TNF** 0.29 0.3
NLP anti-CCP positive 0.27 0.25
NLP anti-TNF*** 0.2 0.36
NLP other DMARDs 0.13 0.34
Negative predictors
ICD9 JRA -0.98 0.9
ICD9 SLE -0.57 1.09
NLP PsA -0.51 0.74
*

ICD9 RA normalized= ln (# ICD9 RA codes per subject at least one week apart)

**

codified anti-TNF= etanercept, infliximab (adalimumab was not available in our EMR)

***

NLP anti-TNF= adalimumab, etanercept, infliximab