Logistic discriminant analysis to clinical practice: the development of clinical score to predict failure of reperfusion following thrombolytic therapy for acute myocardial infarction
by
Wanzhen Gao
Green Lane Hospital
Coauthors: John K. French, Krishnan Ramanathan, James T. Stewart, Pierre Théroux, Harvey D. White

To develop an early, rapid, accurate and noninvasive means of predicting failure of reperfusion which is measured by TIMI flow in order to help identify patients who might benefit from further reperfusion strategies that may limit myocyte necrosis and preserve ventricular function, Logistic regression was used to perform a discriminant analysis including ST recovery, myocardial protein levels and the clinical variables. The "goodness of fit" of the model was evaluated by calculating the Hosmer-Lemeshow statistic. The score of each factor was derived from the parameter estimate of the model. The ability of the model to identify failure to achieve TIMI-3 flow was evaluated using sensitivity, specificity and predictive values.

Hosmer-Lemeshow statistic for the model was Chi-square=5.173, df=8, p=0.74. The maximum score was 8 points. A score of less than or equal to 2 was 96% accurate for excluding failure to achieve TIMI-3 flow, whereas a score of greater than or equal to 7 was 90% accurate for excluding TIMI-3 flow. A score of greater than or equal to 5 predicted failure to achieve TIMI-3 flow with 77% sensitivity, 79% specificity, 74% positive predictive value and 82% negative predictive value. The use of this score may facilitate triage of patients to additional reperfusion therapies.

Date received: August 29, 2001