Skip Ribbon Commands
Skip to main content
SharePoint

PublicComment

 Feedback

If you have any comments or questions on this page, please email us at:
 

Computer assisted STEMI ECG interpretation

Question Type:
Diagnostic
Full Question:
Among adult patients with suspected STEMI outside of a hospital (P), does the use of computer-assisted ECG interpretation  (P), does does the use of computer-assisted ECG interpretation   (I), compared with compared with physician ECG interpretation and/or clinical diagnosis of STEMI  (C), change identify STEMI on an ECG with acceptable rates of false negatives to allow earlier identification and false positives minimizing unnecessary angiography (O)?
Consensus on Science:
For the important outcomes of FP and FN, we have identified very-low-quality evidence (downgraded for risk of bias, inconsistency, and imprecision) from 2 cohort studies(Clark 2010, 1696-1702; Bhalla 2013, 211-216) enrolling 1112 patients/ECGs of FP for STEMI recognition ranging from 0% to 8.7% (assuming STEMI prevalence of 5% [highest expected FP results]) and FN ranging from 4.4% to 8.4% (assuming STEMI prevalence of 20% [highest expected FN results]). Note that sensitivity ranged from 0.58 to 0.78, and specificity ranged from 0.91 to 1. For the important outcome of FP/all positive results, we identified very-low-quality evidence (downgraded for risk of bias, inconsistency, and imprecision) from 6 observational studies(van't Hof 2006, 1255; Brown 2008, 158-161; Youngquist 2008, 784-787; de Champlain 2014, 94-105; Squire 2014, 1-8) enrolling 1949 ECGs of FP/all positive results for STEMI recognition ranging from 0% to 42.9%.
Treatment Recommendation:
We suggest computer-assisted ECG interpretation can be used as an adjunct* to recognize STEMI, given the high specificity of the computer algorithms evaluated (weak recommendation, very-low-quality evidence). We suggest computer-assisted ECG interpretation not be used alone to rule out STEMI, because of the poor sensitivity and thus the considerable risk for FN results of the computer algorithms evaluated (weak recommendation, very-low-quality evidence). *The computer-assisted ECG interpretation can be used as an adjunct or in conjunction with the interpretation of a physician or other trained professional. In this way, recognition of STEMI by the computer interpretation can be verified by individual interpretation, and lack of recognition by the computer would not be used solely to rule out STEMI. Values, Preferences, and Task Force Insights In making this recommendation, we put a higher value on minimizing treatment delays of patients with STEMI over possible wasted resources resulting from FP system activation. Recognition of STEMI on ECG may achieve highest accuracy if computer-assisted interpretation is implemented as an adjunct to on-site healthcare provider interpretation in the context of strong initial education programs, quality assurance programs, and ongoing oversight. As was pointed out in the public comments, it is difficult to perform head-to-head comparisons or combine data from these studies, because they have used different proprietary computer interpretation algorithms and different gold standards. It is likely that different algorithms perform differently. Computer interpretation algorithms can be updated periodically, which may change their effectiveness, making previous studies less relevant unless the algorithm and version are the same as is used in your setting. Last, some of the algorithms can now be adjusted to favor either lower FP results or lower FN results, depending on the needs or how it is used. Therefore, in choosing to use such a computer algorithm as an adjunct, careful consideration of the individual algorithm’s reported performance and evaluation of this in your own setting are key. The use of computer ECG interpretation did not yield equally effective performances across the various systems of care where it has been used with observed sensitivities ranging from 0.58 to 0.78 and specificity ranging from 0.91 to 1. This may be due to the algorithm performance (different performance with different types of STEMI), but it may also be related to the quality of obtained ECG and the level of training and individual expertise in acquiring the ECG. It is possible that the performance characteristics of a computer algorithm are different in controlled, in-hospital settings in stable patients compared with prehospital settings. Therefore, each system of care has to evaluate performance of any specific algorithm in the particular context where the algorithm is used. Diagnostic performance should always be considered in conjunction with local STEMI prevalence, because very high or low prevalence rates may lead to unacceptable FP and/or FN rates despite sensitivity and specificity rates that may seem satisfactory as stand-alone values. This approach may give important clues as to whether this method fits best in comparison with other existing options of ECG interpretation such as transmission of ECG for interpretation by an experienced provider.
CoSTR Attachments:
ACS 559 GRADE_CoSTR_n.docx    
Data bias computer int CK1_n.xlsx    

 Contact Us

 
If you have any comments or questions on this page, please email us at: