Integrating Laboratory Data and Continuous Telemonitoring to Improve Early Warning Scores

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Background: Early Warning Scores (EWS) are widely implemented to identify clinical deterioration in hospitalized patients. However, their accuracy remains limited by intermittent vital sign measurements and high false-positive rates. Continuous monitoring with wearable sensors offers more timely detection, and integrating laboratory parameters may further enhance predictive precision. This study evaluated whether combining continuous vital sign monitoring with routine laboratory tests improves the prediction of postoperative complications compared with standard EWS approaches. Methods: All adult patients admitted in 2023 to the surgical oncology ward of Catharina Hospital Eindhoven who underwent monitoring with a wireless accelerometer patch were included. Continuous heart and respiratory rate data were combined with laboratory parameters, including C-reactive protein (CRP) and leukocyte count. Postoperative complications (Clavien–Dindo ≥ II) were identified by chart review. Logistic regression was used to identify significant predictors and to construct the CLUE model (Combining Labdata Upgrading Early warning scores). Model performance was compared with the Continuous Remote Early Warning Score (CREWS) using the area under the receiver operating characteristic curve (AUC). Results: The final dataset included 198 complete observations from 155 patients. Heart rate, CRP, and leukocyte count were independently associated with postoperative complications. The CLUE model achieved an AUC of 0.71 (0.60-0.80) outperforming the CREWS model (AUC 0.57 (0.50-0.63)). Conclusion: Integrating laboratory data with continuous vital sign monitoring improved the discrimination of postoperative complications compared with vital sign based EWS alone. The CLUE model represents a clinically interpretable step toward more precise, multimodal early warning systems.

This publication has 16 references indexed in Scilit: