Accuracy of computerized electrocardiographic identification of left ventricular hypertrophy as determined by echocardiographic measurements of left ventricular mass: evaluation of a widely used computer program.

The Hewlett-Packard 1000 electrocardiographic management system employs a user-interactive computer with revisable software. The diagnostic accuracy of this system in predicting left ventricular hypertrophy has been evaluated by comparing computer-predicted with anatomic left ventricular hypertrophy. The latter was defined as a left ventricular mass greater than 247 g as determined by M-mode echocardiography within 1 week of the computerized electrocardiogram. In this study, the Hewlett-Packard system was evaluated in 134 consecutive patients having anatomic left ventricular hypertrophy and 157 similarly studied patients with a normal left ventricular mass. By means of various combinations of voltage criteria, ST-T wave changes, abnormal QRS duration or ventricular activation time, left atrial enlargement and left axis deviation, the computer correctly identified 43 of 134 patients with left ventricular hypertrophy (sensitivity 32%); left ventricular hypertrophy was suggested or identified in an additional 18 patients with a normal left ventricular mass (specificity 89%). Graded probability statements (that is, the "strength" of the left ventricular hypertrophy diagnosis) increased with ventricular mass not only in patients with anatomic left ventricular hypertrophy but also in the patients with a "false positive" computerized electrocardiogram. Overall diagnostic accuracy was improved only in men by adjusting the definition of left ventricular hypertrophy for sex (at least 300 g for men and 220 g for women). Diagnostic accuracy was significantly better in patients 65 years of age and younger than in older patients (p = 0.007). It is concluded that the overall performance of this computer program (version ECLSB6) requires modification to improve its accuracy in identifying left ventricular hypertrophy.