Validation of the myocardial-ischaemic-injury-index machine learning algorithm to guide the diagnosis of myocardial infarction in a heterogenous population: a prespecified exploratory analysis
Dimitrios Doudesis, Kuan Ken Lee, Jason Yang, Ryan Wereski, Anoop Shah, Athanasios Tsanas, Atul Anand, John W. Pickering, Martin Than, Nicholas L. Mills, Nicholas L. Mills, Fiona E. Strachan, Christopher Tuck, Anoop Shah, Atul Anand, Andrew R. Chapman, Amy V. Ferry, Kuan Ken Lee, Dimitrios Doudesis, Anda Bularga, Ryan Wereski, Caelan Taggart, Matthew T.H. Lowry, Filip Mendusic, Dorien M Kimenai, Dennis Sandeman, Philip D Adamson, Catherine L. Stables, Catalina A. Vallejos, Athanasios Tsanas, Lucy Marshall, Stacey Stewart, Takeshi Fujisawa, Mischa Hautvast, Jean McPherson, Lynn McKinlay, Ian Ford, David E. Newby, Keith A.A. Fox, Colin Berry, Simon Walker, Christopher J. Weir, Alasdair Gray, Paul Collinson, Fred S Apple, Alan Reid, Anne Cruikshank, Iain Findlay, Shannon Amoils, David McAllister, Donogh Maguire, Jennifer S. Stevens, John Norrie, Jack PM Andrews, Alastair J. Moss, Mohamed Anwar, John Hung, Jonathan Malo, Colin Fischbacher, Bernard Croal, Stephen J Leslie, Catriona Keerie, Richard Parker, Allan Walker, Ronnie Harkess, Tony Wackett, Roma A. Armstrong, Laura Stirling, Claire MacDonald, Imran Sadat, Frank Finlay, Heather Charles, Pamela Linksted, Stephen G. Young, Bill Alexander, Chris Duncan
Abstract
BACKGROUND: Diagnostic pathways for myocardial infarction rely on fixed troponin thresholds, which do not recognise that troponin varies by age, sex, and time within individuals. To overcome this limitation, we recently introduced a machine learning algorithm that predicts the likelihood of myocardial infarction. Our aim was to evaluate whether this algorithm performs well in routine clinical practice and predicts subsequent events. METHODS: threshold (49·7). The trial is registered with ClinicalTrials.gov, NCT01852123. FINDINGS: score ≥49·7; specificity 95·0% [94·6-95·3], positive predictive value 70·4% [68·7-72·0]). At 1 year, subsequent myocardial infarction or cardiovascular death occurred more often in high-probability patients than low-probability patients (520 [17·6%] of 2961 vs 197 [1·5%] of 12 983], p<0·0001). INTERPRETATION: algorithm could improve the diagnosis and assessment of risk in patients with suspected acute coronary syndrome. FUNDING: Medical Research Council, British Heart Foundation, National Institute for Health Research, and NHSX.