Electronic Medical Record Specialty Group Comparison by Multinomial Logistic Regression

In this study we investigate the application of logistic regression to audit log data to identify novel differences in the electronic medical record (EMR) experience between primary care (PC), surgical (S), and non-surgical medical (NSM) specialty groups using logistic regression. While 12 of the 23 candidate features identified through subject matter expert interview were significant in the univariate model, many dropped below the level of significance in the multinomial model. The final model of best fit included only eight features. Total overall time in EMR and teamwork for orders were predictive for PC, while a higher percentage of documentation, patient discharge, and chart review time, in particular review of clinical notes, were more predictive for S and NSM specialty groups. This study provides single site evidence of significant EMR workflow differences between specialty groups as measured through audit log data.