Abstract MP22: Feasibility of Electrocardiographic Identification and Classification of Myocardial Infarction Using Electronic Health Records. The Atherosclerosis Risk in Communities Study (ARIC)

Background: Standardized coding of 12-lead electrocardiograms (ECGs) enables an accurate classification of hospitalized myocardial infarction (MI). Although the costs of ECG coding are difficult to sustain in long-term community surveillance, automated protocols for identifying and classifying MI using electronic health records (EHR) have not incorporated ECG coding. We developed a process to retrieve and automatically code ECGs from EHR and compared its performance to ECGs coded at a specialized ECG reading center. Methods: Samples of EHR pertaining to hospitalizations previously included in hospital record abstraction and Minnesota coding of ECGs per the ARIC protocol were obtained. We assigned Minnesota codes (MCs) to the terms used by manufacturers of three ECG devices to interpret ECGs, extracted machine interpretation text from the EHR, and applied natural-language processing to match the interpretations to MCs. We defined MI by MCs as any major Q code (1.1.x, 1.2.x), or the co-occurrence of any min...