Predictive modeling of perioperative patient deterioration: combining unanticipated ICU admissions and mortality for improved risk prediction.

OBJECTIVE: This paper presents a comprehensive analysis of perioperative patient deterioration by developing predictive models that evaluate unanticipated ICU admissions and in-hospital mortality both as distinct and combined outcomes. MATERIALS AND METHODS: With less than 1% of cases resulting in at least one of these outcomes, we investigated 98 features to identify their role in predicting patient deterioration, using univariate analyses. Additionally, multivariate analyses were performed by employing logistic regression (LR) with LASSO regularization. We also assessed classification models, including non-linear classifiers like Support Vector Machines, Random Forest, and XGBoost. RESULTS: During evaluation, careful attention was paid to the data imbalance therefore multiple evaluation metrics were used, which are less sensitive to imbalance. These metrics included the area under the receiver operating characteristics, precision-recall and kappa curves, and the precision, sensitivity, kappa, and F1-score. Combining unanticipated ICU admissions and mortality into a single outcome improved predictive performance overall. However, this led to reduced accuracy in predicting individual forms of deterioration, with LR showing the best performance for the combined prediction. DISCUSSION: The study underscores the significance of specific perioperative features in predicting patient deterioration, especially revealed by univariate analysis. Importantly, interpretable models like logistic regression outperformed complex classifiers, suggesting their practicality. Especially, when combined in an ensemble model for predicting multiple forms of deterioration. These findings were mostly limited by the large imbalance in data as post-operative deterioration is a rare occurrence. Future research should therefore focus on capturing more deterioration events and possibly extending validation to multi-center studies. CONCLUSIONS: This work demonstrates the potential for accurate prediction of perioperative patient deterioration, highlighting the importance of several perioperative features and the practicality of interpretable models like logistic regression, and ensemble models for the prediction of several outcome types. In future clinical practice these data-driven prediction models might form the basis for post-operative risk stratification by providing an evidence-based assessment of risk.

Postoperative circadian patterns in wearable sensor measured heart rate: a prospective observational study.

PURPOSE: This study aimed to describe the 24-hour cycle of wearable sensor-obtained heart rate in patients with deterioration-free recovery and to compare it with patients experiencing postoperative deterioration. METHODS: A prospective observational trial was performed in patients following bariatric or major abdominal cancer surgery. A wireless accelerometer patch (Healthdot) continuously measured postoperative heart rate, both in the hospital and after discharge, for a period of 14 days. The circadian pattern, or diurnal rhythm, in the wearable sensor-obtained heart rate was described using peak, nadir and peak-nadir excursions. RESULTS: The study population consisted of 137 bariatric and 100 major abdominal cancer surgery patients. In the latter group, 39 experienced postoperative deterioration. Both surgery types showed disrupted diurnal rhythm on the first postoperative days. Thereafter, the bariatric group had significantly lower peak heart rates (days 4, 7-12, 14), lower nadir heart rates (days 3-14) and larger peak-nadir excursions (days 2, 4-14). In cancer surgery patients, significantly higher nadir (days 2-5) and peak heart rates (days 2-3) were observed prior to deterioration. CONCLUSIONS: The postoperative diurnal rhythm of heart rate is disturbed by different types of surgery. Both groups showed recovery of diurnal rhythm but in patients following cancer surgery, both peak and nadir heart rates were higher than in the bariatric surgery group. Especially nadir heart rate was identified as a potential prognostic marker for deterioration after cancer surgery.

Prediction of postoperative patient deterioration and unanticipated intensive care unit admission using perioperative factors.

BACKGROUND AND OBJECTIVES: Currently, no evidence-based criteria exist for decision making in the post anesthesia care unit (PACU). This could be valuable for the allocation of postoperative patients to the appropriate level of care and beneficial for patient outcomes such as unanticipated intensive care unit (ICU) admissions. The aim is to assess whether the inclusion of intra- and postoperative factors improves the prediction of postoperative patient deterioration and unanticipated ICU admissions. METHODS: A retrospective observational cohort study was performed between January 2013 and December 2017 in a tertiary Dutch hospital. All patients undergoing surgery in the study period were selected. Cardiothoracic surgeries, obstetric surgeries, catheterization lab procedures, electroconvulsive therapy, day care procedures, intravenous line interventions and patients under the age of 18 years were excluded. The primary outcome was unanticipated ICU admission. RESULTS: An unanticipated ICU admission complicated the recovery of 223 (0.9%) patients. These patients had higher hospital mortality rates (13.9% versus 0.2%, p<0.001). Multivariable analysis resulted in predictors of unanticipated ICU admissions consisting of age, body mass index, general anesthesia in combination with epidural anesthesia, preoperative score, diabetes, administration of vasopressors, erythrocytes, duration of surgery and post anesthesia care unit stay, and vital parameters such as heart rate and oxygen saturation. The receiver operating characteristic curve of this model resulted in an area under the curve of 0.86 (95% CI 0.83-0.88). CONCLUSIONS: The prediction of unanticipated ICU admissions from electronic medical record data improved when the intra- and early postoperative factors were combined with preoperative patient factors. This emphasizes the need for clinical decision support tools in post anesthesia care units with regard to postoperative patient allocation.

Machine Learning for Postoperative Continuous Recovery Scores of Oncology Patients in Perioperative Care with Data from Wearables.

Assessing post-operative recovery is a significant component of perioperative care, since this assessment might facilitate detecting complications and determining an appropriate discharge date. However, recovery is difficult to assess and challenging to predict, as no universally accepted definition exists. Current solutions often contain a high level of subjectivity, measure recovery only at one moment in time, and only investigate recovery until the discharge moment. For these reasons, this research aims to create a model that predicts continuous recovery scores in perioperative care in the hospital and at home for objective decision making. This regression model utilized vital signs and activity metrics measured using wearable sensors and the XGBoost algorithm for training. The proposed model described continuous recovery profiles, obtained a high predictive performance, and provided outcomes that are interpretable due to the low number of features in the final model. Moreover, activity features, the circadian rhythm of the heart, and heart rate recovery showed the highest feature importance in the recovery model. Patients could be identified with fast and slow recovery trajectories by comparing patient-specific predicted profiles to the average fast- and slow-recovering populations. This identification may facilitate determining appropriate discharge dates, detecting complications, preventing readmission, and planning physical therapy. Hence, the model can provide an automatic and objective decision support tool.

The Accuracy of Wrist-Worn Photoplethysmogram-Measured Heart and Respiratory Rates in Abdominal Surgery Patients: Observational Prospective Clinical Validation Study.

BACKGROUND: Postoperative deterioration is often preceded by abnormal vital parameters. Therefore, vital parameters of postoperative patients are routinely measured by nursing staff. Wrist-worn sensors could potentially provide an alternative tool for the measurement of vital parameters in low-acuity settings. These devices would allow more frequent or even continuous measurements of vital parameters without relying on time-consuming manual measurements, provided their accuracy in this clinical population is established. OBJECTIVE: This study aimed to assess the accuracy of heart rate (HR) and respiratory rate (RR) measures obtained via a wearable photoplethysmography (PPG) wristband in a cohort of postoperative patients. METHODS: The accuracy of the wrist-worn PPG sensor was assessed in 62 post-abdominal surgery patients (mean age 55, SD 15 years; median BMI 34, IQR 25-40 kg/m2). The wearable obtained HR and RR measurements were compared to those of the reference monitor in the postanesthesia or intensive care unit. Bland-Altman and Clarke error grid analyses were performed to determine agreement and clinical accuracy. RESULTS: Data were collected for a median of 1.2 hours per patient. With a coverage of 94% for HR and 34% for RR, the device was able to provide accurate measurements for the large majority of the measurements as 98% and 93% of the measurements were within 5 bpm or 3 rpm of the reference signal. Additionally, 100% of the HR and 98% of the RR measurements were clinically acceptable on Clarke error grid analysis. CONCLUSIONS: The wrist-worn PPG device is able to provide measurements of HR and RR that can be seen as sufficiently accurate for clinical applications. Considering the coverage, the device was able to continuously monitor HR and report RR when measurements of sufficient quality were obtained. TRIAL REGISTRATION: ClinicalTrials.gov NCT03923127; https://www.clinicaltrials.gov/ct2/show/NCT03923127.

A wearable patch based remote early warning score (REWS) in major abdominal cancer surgery patients.

INTRODUCTION: The shift toward remote patient monitoring methods to detect clinical deterioration requires testing of wearable devices in real-life clinical settings. This study aimed to develop a remote early warning scoring (REWS) system based on continuous measurements using a wearable device, and compare its diagnostic performance for the detection of deterioration to the diagnostic performance of the conventional modified early warning score (MEWS). MATERIALS AND METHODS: The study population of this prospective, single center trial consisted of patients who underwent major abdominal cancer surgery and were monitored using routine in-hospital spotcheck measurements of the vital parameters. Heart and respiratory rates were measured continuously using a wireless accelerometer patch (HealthDot). The prediction by MEWS of deterioration toward a complication graded Clavien-Dindo of 2 or higher was compared to the REWS derived from continuous measurements by the wearable patch. MAIN RESULTS: A total of 103 patients and 1909 spot-check measurements were included in the analysis. Postoperative deterioration was observed in 29 patients. For both EWS systems, the sensitivity (MEWS: 0.20 95% CI: [0.13-0.29], REWS: 0.20 95% CI: [0.13-0.29]) and specificity (MEWS: 0.96 95% CI: [0.95-0.97], REWS: 0.96 95% CI: [0.95-0.97]) were assessed. CONCLUSIONS: The diagnostic value of the REWS method, based on continuous measurements of the heart and respiratory rates, is comparable to that of the MEWS in patients following major abdominal cancer surgery. The wearable patch could detect the same amount of deteriorations, without requiring manual spot check measurements.

Quality Improvement in the Preoperative Evaluation: Accuracy of an Automated Clinical Decision Support System to Calculate CHA2DS2-VASc Scores.

Background and Objectives: Clinical decision support systems are advocated to improve the quality and efficiency in healthcare. However, before implementation, validation of these systems needs to be performed. In this evaluation we tested our hypothesis that a computerized clinical decision support system can calculate the CHA2DS2-VASc score just as well compared to manual calculation, or even better and more efficiently than manual calculation in patients with atrial rhythm disturbances. Materials and Methods: In n = 224 patents, we calculated the total CHA2DS2-VASc score manually and by an automated clinical decision support system. We compared the automated clinical decision support system with manually calculation by physicians. Results: The interclass correlation between the automated clinical decision support system and manual calculation showed was 0.859 (0.611 and 0.931 95%-CI). Bland-Altman plot and linear regression analysis shows us a bias of -0.79 with limit of agreement (95%-CI) between 1.37 and -2.95 of the mean between our 2 measurements. The Cohen's kappa was 0.42. Retrospective analysis showed more human errors than algorithmic errors. Time it took to calculate the CHA2DS2-VASc score was 11 s per patient in the automated clinical decision support system compared to 48 s per patient with the physician. Conclusions: Our automated clinical decision support system is at least as good as manual calculation, may be more accurate and is more time efficient.

Advanced practice providers versus medical residents as leaders of rapid response teams: A 12-month retrospective analysis.

PURPOSE: In a time of worldwide physician shortages, the advanced practice providers (APPs) might be a good alternative for physicians as the leaders of a rapid response team. This retrospective analysis aimed to establish whether the performance of APP-led rapid response teams is comparable to the performance of rapid response teams led by a medical resident of the ICU. MATERIAL AND METHODS: In a retrospective single-center cohort study, the electronic medical record of a tertiary hospital was queried during a 12-months period to identify patients who had been visited by our rapid response team. Patient- and process-related outcomes of interventions of rapid response teams led by an APP were compared with those of teams led by a medical resident using various parameters, including the MAELOR tool, which measures the performance of a rapid response team. RESULTS: In total, 179 responses of the APP-led teams were analyzed, versus 275 responses of the teams led by a resident. Per APP, twice as many calls were handled than per resident. Interventions of teams led by APPs, and residents did not differ in number of admissions (p = 0.87), mortality (p = 0.8), early warning scores (p = 0.2) or MAELOR tool triggering (p = 0.19). Both groups scored equally on time to admission (p = 0.67) or time until any performed intervention. CONCLUSION: This retrospective analysis showed that the quality of APP-led rapid response teams was similar to the quality of teams led by a resident. These findings need to be confirmed by prospective studies with balanced outcome parameters.

Accuracy of vital parameters measured by a wearable patch following major abdominal cancer surgery.

INTRODUCTION: Recent advances in wearable technology allow for the development of wirelessly connected sensors to continuously measure vital parameters in the general ward or even at home. The present study assesses the accuracy of a wearable patch (Healthdot) for continuous monitoring of heartrate (HR) and respiration rate (RR). MATERIALS AND METHODS: The Healthdot measures HR and RR by means of chest accelerometry. The study population consisted of patients following major abdominal oncological surgery. The analysis focused on the agreement between HR and RR measured by the Healthdot and the gold standard patient monitor in the intensive and post-anesthesia care unit. RESULTS: For HR, a total of 112 h of measurements was collected in 26 patients. For RR, a total of 102 h of measurements was collected in 21 patients. On second to second analysis, 97% of the HR and 87% of the RR measurements were within 5 bpm and 3 rpm of the reference monitor. Assessment of 5-min averaged data resulted in 96% of the HR and 95% of the RR measurements within 5 bpm and 3 rpm of the reference monitor. A Clarke error grid analysis showed that 100% of the HR and 99.4% of the 5-min averaged data was clinically acceptable. CONCLUSION: The Healthdot accurately measured HR and RR in a cohort of patients recovering from major abdominal surgery, provided that good quality data was obtained. These results push the Healthdot forward as a clinically acceptable tool in low acuity settings for unobtrusive, automatic, wireless and continuous monitoring.

RegressionExplorer: Interactive Exploration of Logistic Regression Models with Subgroup Analysis.

We present RegressionExplorer, a Visual Analytics tool for the interactive exploration of logistic regression models. Our application domain is Clinical Biostatistics, where models are derived from patient data with the aim to obtain clinically meaningful insights and consequences. Development and interpretation of a proper model requires domain expertise and insight into model characteristics. Because of time constraints, often a limited number of candidate models is evaluated. RegressionExplorer enables experts to quickly generate, evaluate, and compare many different models, taking the workflow for model development as starting point. Global patterns in parameter values of candidate models can be explored effectively. In addition, experts are enabled to compare candidate models across multiple subpopulations. The insights obtained can be used to formulate new hypotheses or to steer model development. The effectiveness of the tool is demonstrated for two uses cases: prediction of a cardiac conduction disorder in patients after receiving a heart valve implant and prediction of hypernatremia in critically ill patients.