Diagnostic performance of manual and automated Washington code vs. the Louvain program

A newly implemented computerized version of the Washington code (WC) has been tested against both the manual coding and the Louvain vectorcardiogram (VCG) program on a clinical data set of 250 Frank records validated independently from the electrocardiogram (ECG). Total accuracy was 52.6% for the automated WC, as compared with 59.9% for the manual WC and 72.8% for the Louvain program. The low performance of the automated WC in the diagnosis of myocardial infarction was due to a high proportion of concomitant false diagnoses of ventricular hypertrophy. The agreement with the manual coding was low because of the differential diagnosis applied in the automated version. Current criteria of the WC cannot be recommended for clinical application