A Bayesian sensitivity analysis of out-of-hospital 12-lead electrocardiograms: implications for regionalization of cardiac care.
Academic Emergency Medicine · 2007-11-30 · Method & validation
Abstract
BACKGROUND: The effectiveness of out-of-hospital regionalization of ST-elevation myocardial infarction (STEMI) patients to hospitals providing primary percutaneous coronary intervention depends on the accuracy of the out-of-hospital 12-lead electrocardiogram (PHTL). Although estimates of sensitivity and specificity of PHTL for STEMI have been reported, the impact of out-of-hospital STEMI prevalence on positive predictive value (PPV) has not been evaluated. OBJECTIVES: To describe the relationship between varying population STEMI prevalences and PHTL predictive values, using ranges of PHTL sensitivity and specificity. METHODS: The authors performed a Bayesian analysis using PHTL, where values for sensitivities (60%-70%), specificities (98%), and two prevalence ranges (0.5%-5% and 5%-20%) were derived from a literature review. PPV prediction intervals were compared with three months of prospective data from the Los Angeles County Emergency Medical Services Agency STEMI regionalization program. RESULTS: When the estimated prevalence of STEMI in the out-of-hospital population is 5%-20%, the median PPV of the PHTL is 83% (95% credible interval [CrI] = 53% to 97%). However, if the population prevalence of STEMI is between 0.5% and 5%, the median PPV is 43% (95% CrI = 12% to 86%). When the PPV prediction intervals were incorporated with the Los Angeles County Emergency Medical Services Agency data, the PPV was 66%. CONCLUSIONS: Even when assuming high specificity for PHTL, the false-positive rate will be considerable if applied to a population at low risk for STEMI. Before broadening application of PHTL to low-risk patients, the implications of a high false-positive rate should be considered.