| dc.description.abstract | Abstract Objectives The purpose of this retrospective cross sectional study is to identify the minimal amount of checklist variables you must consider to diagnose an ECG as either normal or abnormal, and warrant further investigation. The aim is to combat the wide variability of checklists used by clinicians and health care professionals when interpreting ECG’s by creating an evidence based checklist that both increase productivity and decrease the amount of error during ECG interpretation. Method We compiled a comprehensive set of ECG checklist items from well known sources such as UpToDate and American Heart Association that covers the most common ECG phenomenon. We then exported 308 adult ECGs with various diagnosis and normal ECG’s from the PTB-XL database on physionet.org. Further, we analyzed each of the 308 ECG’s using the comprehensive checklist and categorized each variable as true or false according to diagnostic criteria, as well as each ECG as a whole with their status as normal or abnormal. From this comprehensive checklist we generated all possible sub-combinations of checklist items and selected the lists with the least amount of items necessary for diagnosing an ECG as abnormal in at least 95 % of the cases. Following this we verified the selected checklists against the same 308 ECG’s, only this time using the clinical remarks from the cardiologists annotating the database instead of our own interpretation. Using the cardiologists interpretation as gold standard, we want to find the specificity and sensitivity of the final checklists in the specific dataset, resulting in the most optimal checklist with regards to specificity and sensitivity. Results By evaluating every combination of checklist variables, excluding QTc, P amplitude, and P duration due to substantial variance between the interpretations, we derived six checklists, each comprised of seven variables. All lists exhibit similarity in the percentage of correctly identified true abnormalities, ranging from 95.3 % to 95.7 % of cases. Further analyzing these lists against the clinical remarks we see that the list; «1) Rhythm, 2) frequency, 3) axis, 4) T inversion, 5) ST depression, 6) ST elevation and 7) Sokolow-Lyon’s criteria for left ventricular hypertrophy» was slightly superior with a sensitivity of 89.0% and specificity of 77.8% Conclusion Recognizing the degree of bias in our study, and acknowledging many aspects that could be improved for future studies, we believe we can screen ECG’s faster with a more condensed checklist. Thus using more time on the ECG’s we mark as abnormal, of which a more thorough assessment is warranted. To finally evaluate this list, it should be tested out in a prospective clinical setting by health care profession with little ECG training, utilizing cardiologist as the study Gold Standard. Especially verifying the sensitivity and specificity of those lists, and determining an acceptable rate of false normal ECGs. | eng |