Natural Language Processing in the Electronic Medical Record: Assessing Clinician Adherence to Tobacco Treatment Guidelines

摘要: Background Comprehensively assessing care quality with electronic medical records (EMRs) is not currently possible because much data reside in clinicians’ free-text notes. Methods We evaluated the accuracy of MediClass, an automated, rule-based classifier EMR that incorporates natural language processing, whether clinicians: (1) asked if patient smoked; (2) advised them to stop; (3) assessed their readiness quit; (4) assisted quitting by providing information or medications; and (5) arranged for appropriate follow-up (i.e., 5A’s smoking-cessation care). Design analyzed 125 known smokers at each four HMOs 2003 2004. One trained abstractor HMO manually coded all 500 according smoking cessation was addressed during routine outpatient visits. Measurements For patient’s record, we compared presence absence as human coder MediClass. measured chance-corrected agreement between raters MediClass using kappa statistic. Results “ask” “assist,” among coders indistinguishable from humans (p>0.05). “assess” “advise,” agreed more other than they did (p Conclusions performance appears adequate replace care, allowing automated assessment clinician adherence one most important, evidence-based guidelines preventive health care.