External validation of the myocardial-ischaemic-injury-index machine learning algorithm for the early diagnosis of myocardial infarction: a multicentre cohort study

Background The myocardial-ischaemic-injury-index (MI3 ) is a novel machine learning algorithm for the early diagnosis of type 1 non -ST -segment elevation myocardial infarction (NSTEMI). The performance of MI3 , both when using early serial blood draws (eg, at 1 h or 2 h) and in direct comparison with guideline-recommended algorithms, remains unknown. Our aim was to externally validate MI3 and compare its performance with that of the European Society of Cardiology (ESC) 0/1h-algorithm. Methods In this secondary analysis of a multicentre international diagnostic cohort study, adult patients (age >18 years) presenting to the emergency department with symptoms suggestive of myocardial infarction were prospectively enrolled from April 21, 2006, to Feb 27, 2019 in 12 centres from five European countries (Switzerland, Spain, Italy, Poland, and Czech Republic). Patients were excluded if they presented with ST -segment -elevation myocardial infarction, did not have at least two serial high-sensitivity cardiac troponin I (hs-cTnI) measurements, or if the final diagnosis remained unclear. The final diagnosis was centrally adjudicated by two independent cardiologists using all available medical records, including serial hs-cTnI measurements and cardiac imaging. The primary outcome was type 1 NSTEMI. The performance of MI3 was directly compared with that of the ESC 0/1h-algorithm. Findings Among 6487 patients, (median age 61<middle dot>0 years [IQR 49<middle dot>0-73<middle dot>0]; 2122 [33%] female and 4365 [67%] male), 882 (13<middle dot>6%) patients had type 1 NSTEMI. The median time difference between the first and second hs-cTnI measurement was 60<middle dot>0 mins (IQR 57<middle dot>0-70<middle dot>0). MI3 performance was very good, with an area under the receiver-operating-characteristic curve of 0<middle dot>961 (95% CI 0<middle dot>957 to 0<middle dot>965) and a good overall calibration (intercept -0<middle dot>09 [-0<middle dot>2 to 0<middle dot>02]; slope 1<middle dot>02 [0<middle dot>97 to 1<middle dot>08]). The originally defined MI3 score of less than 1<middle dot>6 identified 4186 (64<middle dot>5%) patients as low probability of having a type 1 NSTEMI (sensitivity 99<middle dot>1% [95% CI 98<middle dot>2 to 99<middle dot>5]; negative predictive value [NPV] 99<middle dot>8% [95% CI 99<middle dot>6 to 99<middle dot>9]) and an MI3 score of 49<middle dot>7 or more identified 915 (14<middle dot>1%) patients as high probability of having a type 1 NSTEMI (specificity 95<middle dot>0% [94<middle dot>3 to 95<middle dot>5]; positive predictive value [PPV] 69<middle dot>1% [66<middle dot>0-72<middle dot>0]). The sensitivity and NPV of the ESC 0/1h-algorithm were higher than that of MI3 (difference for sensitivity 0<middle dot>88% [0<middle dot>19 to 1<middle dot>60], p=0<middle dot>0082; difference for NPV 0<middle dot>18% [0<middle dot>05 to 0<middle dot>32], p=0<middle dot>016), and the rule-out efficacy was higher for MI3 (11% difference, p<0<middle dot>0001). Specificity and PPV for MI3 were superior (difference for specificity 3<middle dot>80% [3<middle dot>24 to 4<middle dot>36], p<0<middle dot>0001; difference for PPV 7<middle dot>84% [5<middle dot>86 to 9<middle dot>97], p<0<middle dot>0001), and the rule-in efficacy was higher for the ESC 0/1h-algorithm (5<middle dot>4% difference, p<0<middle dot>0001). Interpretation MI3 performs very well in diagnosing type 1 NSTEMI, demonstrating comparability to the ESC 0/1h-algorithm in an emergency department setting when using early serial blood draws. Copyright (c) 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.