Predictive Value of the Electrocardiogram in Acute Myocardial Infarction—Reply

In Reply: Although calculations such as sensitivity and positive predictive values could be applied to our data, these measures are most commonly used in diagnostic studies to compare a new test to a criterion standard. Likelihood ratios are also primarily used in diagnostic studies, but occasionally are applied to risk factors for cardiovascular disease and the prediction of death.¹ In our prognostic study we did not assess the ability of the ECG to diagnose death or AMI, but rather to examine its independent association with mortality and its ability to estimate risk. This prognostic information is best described by the odds ratios (adjusted by logistic regression), a technique that is familiar to most physicians.² Furthermore, likelihood ratios have the disadvantage of requiring an assessment of the probability of the condition in question and conversion to odds.³ The quantitative methods recommended by Dr Salmi are useful in diagnostic studies but the likelihood ratios are rarely used by physicians in everyday practice. A study of 300 physicians found that only 0.7% used likelihood ratios when evaluating the usefulness of diagnostic studies, and physicians preferred more "informal direct" methods.⁴