Search

BackgroundDetecting bacteremia among surgical in-patients is more obscure than other patients due to the inflammatory condition caused by the surgery. The previous criteria such as systemic inflammatory response syndrome or Sepsis-3 are not available for use in general wards, and thus, many clinicians usually rely on practical senses to diagnose postoperative infection. ObjectiveThis study aims to evaluate the performance of continuous monitoring with a deep learning model for early detection of bacteremia for surgical in-patients in the general ward and the intensive care unit (ICU). MethodsIn this retrospective cohort study, we included 36,023 consecutive patients who underwent general surgery between October and December 2017 at a tertiary referral hospital in South Korea. The primary outcome was the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC) for detecting bacteremia by the deep learning model, and the secondary outcome was the feature explainability of the model by occlusion analysis. ResultsOut of the 36,023 patients in the data set, 720 cases of bacteremia were included. Our deep learning–based model showed an AUROC of 0.97 (95% CI 0.974-0.981) and an AUPRC of 0.17 (95% CI 0.147-0.203) for detecting bacteremia in surgical in-patients. For predicting bacteremia within the previous 24-hour period, the AUROC and AUPRC values were 0.93 and 0.15, respectively. Occlusion analysis showed that vital signs and laboratory measurements (eg, kidney function test and white blood cell group) were the most important variables for detecting bacteremia. ConclusionsA deep learning model based on time series electronic health records data had a high detective ability for bacteremia for surgical in-patients in the general ward and the ICU. The model may be able to assist clinicians in evaluating infection among in-patients, ordering blood cultures, and prescribing antibiotics with real-time monitoring.

Background: Pace mapping (PM) is used to identify the origin of ventricular arrhythmias (VAs). For intramural VAs, the site of origin often cannot be reached and therefore PM is less accurate.Objective: The purpose of this study was to assess the value of single- and dual-site pace maps to differentiate intramural from nonintramural VAs.Methods: In 18 consecutive patients with idiopathic intramural VAs, pace mapping was performed at 2 breakthrough sites in adjacent anatomic structures. Twelve-lead electrocardiograms of the 2 pace maps were averaged in MATLAB and compared (correlation coefficient [CC]) with the targeted VA. Dual-site pace mapping was performed in a control group of 18 patients with nonintramural VAs at the sites of earliest electrical activation and a breakthrough site in an adjacent anatomic location.Results: Dual-site pace maps had a higher CC than did best single-site pace maps (0.87 ± 0.1 vs 0.81 ± 0.16; P = .02) in patients with intramural VAs. At the site of origin, single-site pace maps had a higher CC than did dual-site pace maps obtained from adjacent anatomic locations (0.93 ± 0.04 vs 0.89 ± 0.05; P = .0004) in patients with nonintramural VAs. Sensitivity, specificity, positive predictive value, and negative predictive value of dual-site pace maps for identifying an intramural VA were 89%, 82%, 84%, 88%, and 86%, respectively. Furthermore, the receiver operating characteristic curve analysis revealed that a CC cutoff value of ≤0.86 for a single-site pace map best differentiated intramural from nonintramural VAs.Conclusion: A higher CC value for a dual-site pace map obtained from the earliest breakthrough site as well as a CC cutoff value of ≤0.86 for a single-site pace map obtained from the site of earliest electrical activation can best differentiate intramural from nonintramural VAs. [ABSTRACT FROM AUTHOR]