SEARCH

SEARCH BY CITATION

We read Dr. Smith’s letter with keen interest in the experiences of those working out the details of the reorganization of emergency medical services to better care for patients with ST-elevation myocardial infarction (STEMI). Out-of-hospital recognition of STEMI can lead to decreased activation and door-to-balloon times, as has been shown previously.1,2 However, the limitations of out-of-hospital 12-lead electrocardiogram (ECG) in predicting the need for emergent catheterization are an important consideration in the context of STEMI regionalization efforts. The discrepancy observed between Smith’s favorable experience and our own may result from any one or combination of the following issues:

  • 1
    Random variation: Differences between the correspondent’s positive predictive value (PPV) and our PPV in the Los Angeles County Emergency Medical Services Agency (LACEMSA) may be due to random variation in either or both data sets. Our prior credible interval (CrI) for expected PPV in a moderate-risk population was 53% to 97%, however, which while less likely, includes Smith’s value.3
  • 2
    Selection/prevalence: The prevalence of STEMI in the tested population is in part determined by the criteria used for patient selection—and the interpretation of those criteria—by the paramedics in the field. Using our published probability distributions for sensitivity and specificity, a Markov-Chain Monte Carlo back-calculation of the prevalence of STEMI in Smith’s tested population produced a median value of 24% (95% CrI = 4% to 64%). This prevalence is considerably higher than that observed among the LACEMSA patients, possibly suggesting a different selection process for testing.
  • 3
    Test characteristics: The sensitivity and specificity of the computer algorithm for STEMI in Smith’s tested population may be higher than the values considered in our analysis, assuming that they are using the same algorithm. Our probability distributions representing sensitivity and specificity were literature-based and included tolerances for both reasonably higher and lower values. Nonetheless, differences either in the test itself (i.e., the machine/algorithm) or in its characteristics in a particular population could explain the disparity. While blinded human interpretation of the ECG underperformed the computer algorithm in our model, recent work has shown that transmission of the out-of-hospital 12-lead ECG to an emergency physician, presumably in association with brief historical and demographic information, can improve the PPV.4
  • 4
    Criterion standard: There is no agreed upon standard for the definition of a true-positive, and, by extension, a false-positive STEMI on ECG.5 False-positives may be defined in many ways, from normalization of the ECG upon arrival to the emergency department to absence of significant coronary artery disease on cardiac catheterization. We proscribed no specific definition for a false-positive and chose to use the data as self-reported by receiving hospitals; as such, our definition is likely an aggregate of several working definitions. Thus, another reason for the differences in our experience may be differences in the definition of a false-positive.

We feel it is also important to emphasize our Bayesian approach. Instead of allowing the data by themselves to define our beliefs about the PPV, we combine in coherent manner prior information with the data thus far accumulated to describe our new state of knowledge. Let us assume that our previously described priors apply to Smith’s data (which may not be the case given the above considerations). If we use the correspondent’s data to revise our beliefs about the PPV as we did with the LACEMSA data, we should now believe that the median PPV is 90% (95% CrI = 80% to 97%) in a tested population at lower risk for STEMI and 92% (95% CrI = 83% to 97%) in a tested population at moderate risk (the definitions of lower and moderate risk are given in the original article).

Finally, we are all aware that in medicine, variation among individual experiences is the rule rather than the exception, and we have enumerated some possible sources of such variability. While we are appropriately impressed with the correspondent’s experience, it remains of paramount importance that agencies considering regionalization of cardiac care explore the likely impact in their own areas before implementation. Informed public health policies require an understanding of the benefits and limitations inherent in any particular strategy to improve and extend care and should make use of all available information in attempting to best anticipate future consequences. A sensitivity analysis remains an effective means of helping to accomplish this goal.

References

  1. Top of page
  2. References
  • 1
    Brown JP, Mahmud E, Dunford JV, Ben-Yehuda O. Effect of prehospital 12-lead electrocardiogram on activation of the cardiac catheterization laboratory and door-to-balloon time in ST-segment elevation acute myocardial infarction. Am J Cardiol. 2008; 101:15861.
  • 2
    Afolabi BA, Novaro GM, Pinski SL, Fromkin KR, Bush HS. Use of the prehospital ECG improves door-to-balloon times in ST segment elevation myocardial infarction irrespective of time of day or day of week. Emerg Med J. 2007; 24:58891.
  • 3
    Youngquist ST, Kaji AH, Lipsky AM, Koenig WJ, Niemann JT. A Bayesian sensitivity analysis of out-of-hospital 12-lead electrocardiograms: implications for regionalization of cardiac care. Acad Emerg Med. 2007; 14:116571.
  • 4
    Davis DP, Graydon C, Stein R, et al. The positive predictive value of paramedic versus emergency physician interpretation of the prehospital 12-lead electrocardiogram. Prehosp Emerg Care. 2007; 11:399402.
  • 5
    Larson DM, Menssen KM, Sharkey SW, et al. “False-positive” cardiac catheterization laboratory activation among patients with suspected ST-segment elevation myocardial infarction. JAMA. 2007; 298:275460.