Ethics approval was obtained from the Human Research and Ethics Committee of Peninsula Health (Reference number QA/69001/PH-2020-234264) and Monash Health (Reference number QA/71890/MonH-2020-241855). Informed consent was waived by ethics committees as data was already collected as part of routine quality assurance processes.

Postoperative patients admitted to ICU over a 4-year period from April 2016 to May 2020 which corresponded to the introduction of hs-cTn measurements at our study sites were screened. Patients were included in the study if they were admitted to ICU following elective or emergency surgery and had a hs-cTn assay on the day of admission to ICU. There were no strictly defined criteria for testing of hs-cTn in ICU, but it was generally performed in patients older than 60 years, had cardiovascular comorbidities or were requiring vasoactive agents. Patients were excluded if they had been diagnosed with an acute myocardial infarction perioperatively or during their current hospital admission. Patients included were stratified into two groups based on their serum hs-cTn: a low troponin group (< 15 ng/L in females and < 33 ng/L in males) and a high troponin group (> 15 ng/L in females and > 33 ng/L in males). These values were based on the normal reference range in our laboratories. The hs-cTn values were assayed with UniCel DxI 800 platform (Beckman Coulter). The upper reference limit was 15 ng/L in females and 33 ng/L in male patients. Data were collected from our ICU databases, hospital pathology databases and individual patient case records.

Data on physiological, laboratory variables and scores derived from scoring systems (American Society of Anaesthesiologists physical status classification system [ASA] and acute physiology age and chronic health evaluation III score [APACHE III]) during the first 24 h were collected, and the most abnormal values during the first 24 h were analysed. The biochemical variables analysed included high-sensitivity troponins, lactate, sodium, potassium, blood glucose, haemoglobin, white cell count, platelet count, bilirubin, albumin, creatinine and urea. Physiological variables included age, sex, heart rate, blood pressure (systolic, diastolic and mean), respiratory rate, temperature, Glascow Coma Scale, FiO2 requirements, PaO2, PaCO2, pH, HCO3–, whether or not invasive or non-invasive ventilation was used and whether or not inotropes or vasopressors were used during the first 24 h of ICU admission.

All patients admitted to ICU had fixed patient to nurse ratio depending on the monitoring and treatments required. Patients were nursed 1:1 if the patients required mechanical ventilation or had haemodynamic instability that required vasoactive medications and 1:2 otherwise. All patients had continuous monitoring of ECG, oxygen saturation, blood pressure and respiratory rate during their ICU stay. All patients had ECG on admission to ICU and at least once daily while in ICU. Over 90% of the patients had invasive haemodynamic monitoring using intra-arterial catheters.

The primary outcome was in-hospital mortality. The secondary outcomes included ICU mortality, ICU and hospital length of stay, development of acute renal failure and in patient cardiac arrest. ARF was defined as a 24 h urine output < 410 ml and serum creatinine ≥ 133 μmol/L and no chronic dialysis.

All analyses were performed with SAS software version 9.4 (SAS Institute, Cary, NC, USA). Baseline and outcome variables were compared between groups (high vs low hs-cTn) using chi-square tests or Fisher’s exact tests, as appropriate, for categorical variables; Student’s t tests for normally distributed continuous variables; and Wilcoxon rank sum tests otherwise, with results presented as frequency (proportion), mean (SD), and median (interquartile range [IQR]), respectively. Univariable and multivariable analyses for hospital mortality were performed using logistic regression modelling with results presented as odds ratios (OR) and 95% confidence intervals (95% CI). Variables with a p < 0.05 on univariable analysis or those deemed to be clinically relevant were considered for inclusion in the multivariable regression model. The variables included in the final model were highest heart rate, APACHE III score and highest hs-cTn on day of admission to ICU. The interaction between hs-cTn and type of surgery (elective vs emergency) was assessed by fitting main effects for hs-cTn, type of surgery and their two-way interactions. The prognostic value of hs-cTn in predicting hospital mortality was assessed by calculating area under the receiver operating characteristic curves (AUROC). The AUROC was interpreted as follows: 0.9–1, high accuracy; 0.7–0.9, moderate accuracy; 0.5–0.7, low accuracy and 0.5 a chance result [18]. The optimal cut-off point for hs-cTn to predict hospital mortality was determined using Youden’s index [19, 20]. Troponin was analysed as a continuous variable in all regression analyses. All calculated p values were two-tailed and p < 0.05 indicated statistical significance.