Development and internal validation of models to predict postoperative troponin and myocardial injury after noncardiac surgery by incorporating hypotension as a predictor: single-centre retrospective cohort study.

BACKGROUND

Myocardial injury after noncardiac surgery (MINS) is associated with poor outcomes. Current prediction models have limitations and rarely incorporate postoperative hypotension. Our primary objective was to develop and validate a prediction model incorporating postoperative hypotension to predict subsequent postoperative high-sensitivity cardiac troponin T concentrations and MINS.

METHODS

We conducted a single-centre retrospective cohort study of patients aged ≥45 yr who underwent inpatient noncardiac surgery between 2020 and 2021 at St. Paul's Hospital, Canada. We excluded patients with missing vital signs or postoperative troponin concentrations, those who underwent repeat surgery <72 h after surgery, and those who were readmitted <72h after surgery. We developed multivariable linear mixed-effects models using percentage of time with postoperative hypotension (mean arterial pressure <75 mm Hg) on postoperative days (PODs) 0-2, age, sex, emergency surgery, and revised cardiac risk index to predict maximum troponin concentration on PODs 1-3 (primary model) and MINS (secondary model). Internal validation was performed using repeated patient-based five-fold cross-validation with 100 repetitions, using the root mean square error (RMSE), marginal R-squared (R2), and conditional R2 as metrics.

RESULTS

We included 846 patients (mean age 72 yr [standard deviation 11]; 57.4% male). Notably, 35% of patients (299 of 846) developed MINS between PODs 0 and 3. The primary linear mixed-effects model for log-transformed troponin showed a conditional R2 of 0.94, a marginal R2 of 0.20, and a cross-validated RMSE of 0.90 during internal validation. The internally validated MINS model area under the receiver operating characteristic curve was 0.69 (95% CI 0.63-0.75). The inclusion of intraoperative hypotension did not improve the prediction of troponin or MINS.

CONCLUSIONS

A model including postoperative hypotension to predict postoperative troponin concentrations fits our data well, but did not show sufficient accuracy and is not ready for clinical practice. Better-quality data and additional clinical variables may enhance model development.Trial registration. Clinicaltrials.gov NCT05188001, registered 7 January 2022.