We appreciate the opportunity to respond to the comments of Dr. Darbar and colleagues and Dr. Zema on our clinical rule to identify patients with preserved LVEF after myocardial infarction.

Darbar and colleagues applied our clinical prediction rule to a subset of patients admitted to their coronary care unit. They had data on 55 of 75 randomly selected patients; only 13 of these patients were predicted to have an LVEF of 40% or greater. Measurements showed that only 9 of these 13 patients had preserved LVEF, yielding a positive predictive value of 69%; this is apparently much lower than the 98% positive predictive value yielded in our patient population. On the basis of their results, Darbar and colleagues question the generalizability of our prediction rule to patients after myocardial infarction. They suggest that their patients were unselected and heterogeneous and that our patients were, in part, selectively screened.

We have several concerns about their conclusions. First, their validation set was small, consisting of 75 randomly selected patients. We wonder in what way the selection was random. Data on LVEF were only available for 55 of these patients, and no data were provided for the 20 patients who did not have LVEF assessment. Second, Darbar and colleagues initially claim that their selection criteria included no history of congestive heart failure, but later in their letter they note that 19% of their patients had a history of congestive heart failure. We are unsure of the implications of this contradiction. Third, their quoted positive predictive value is based on a denominator of 13 patients, an extremely small group. Although the 95% CIs for their positive predictive value (a post-test probability) cannot be calculated on the basis of the data they provide, the upper limit of the CI would almost certainly overlap with the lower CI limit of 0.90 that we reported in our study. Finally, their patient population consisted solely of patients in the coronary care unit. These patients probably had lower LVEFs, and the likelihood of finding occasional false-negative results in these patients would be higher. We agree that the predictive value of the rule will differ depending on the population to which it is applied. Our population included patients who had smaller infarctions and who were treated in our intermediate-care units, not solely in our coronary care unit, which is reserved for sicker patients with larger infarctions. This factor would also explain why 40% of our patients were predicted to have preserved LVEFs, whereas only 24% of Darbar and colleagues' patients were predicted to have a LVEF of 40% or greater.

We appreciate Dr. Zema's suggestions and agree that, if validated, a combination of our prediction rule with other indicators of left ventricular systolic function might improve the negative predictive value of our rule. This combination could even more narrowly target those patients who have had a myocardial infarction and who require technology-based LVEF assessment, with further reductions in the cost of routine care after myocardial infarction.