Incorporating Laboratory Data into Machine Learning Models Improves Predictions of In-hospital Morality After Rapid Response Team Call: A Follow-up Study
CCCF ePoster library. Reardon P. Nov 7, 2018; 233406; 6
Dr. Peter Reardon
Dr. Peter Reardon
Login now to access Regular content available to all registered users.

You may also access this content "anytime, anywhere" with the Free MULTILEARNING App for iOS and Android
Abstract
Rate & Comment (0)
Introduction: Clinical decision-making is challenging in the rapid response team (RRT) setting due to limited decision time and rapidly progressing illness.1 Machine learning models (MLM) are relatively new in acute care, but have shown promise as clinical decision aids.2,3 A recent novel MLM was developed to examine predictors of mortality among patients seen by the RRT.4 This model outperformed the National Early Warning Score with an AUC of 0.78.4 However, it is hypothesized that laboratory investigations and clinical history may improve the accuracy of the MLM.



Objectives: We developed a MLM for in-hospital mortality prediction among patients seen by the RRT with the goal of assessing the effect of augmenting the model by adding laboratory results and clinical history.



Methods: We performed a retrospective analysis on patients seen by the RRT between May 2012 and May 2016 at two tertiary care hospitals. We gathered information on patients using the Ottawa Hospital Data Warehouse, and supplemented with an additional chart review to obtain details regarding the RRT activation and the patient’s clinical history. We derived and internally validated a machine learning model using a Gradient Boosted Decision Tree model to predict a primary outcome of 30-day in-hospital mortality. We then augmented the model with laboratory investigations and clinical history to our model. We also used a subset of patients to compare our model’s accuracy to the one recently developed by Shappell et al.4



Results: A total of 6131 RRT activations occurred during the time period. A high quality subset of 256 patients with complete data including investigations and clinical history was selected for model development. The median age of 68 (IQR 57 – 79) and median days since admission of 3 (IQR 1 – 8). The 30-day mortality rate was 27.7%. Our base model performed with an AUC of 0.72 (95% CI 0.57 – 0.87). When the model was augmented with laboratory investigations, clinical history, or both, the AUC improved to 0.78 (95% CI 0.64 – 0.91), 0.73 (95% CI 0.58 – 0.88), and 0.77 (95% CI 0.63 – 0.91) respectively. Important base mortality predictors in the model were the temperature, length of service prior to RRT call, systolic blood pressure, and age. Important laboratory markers in the augmented model were anion gap, lactate, and bicarbonate. The highest weighted clinical history variable was advanced malignancy. A sample of 50 patients were run through the Shappell et al. model, which performed with an AUC of 0.72 (95% CI 0.57 – 0.87) in our cohort, which is similar to our base model.



Conclusion: Machine learning models can identify mortality predictors among RRT patients and potentially supplement clinical decision making. Models that incorporate laboratory investigations and clinical history may improve model accuracy.



 
    This eLearning portal is powered by:
    This eLearning portal is powered by MULTIEPORTAL
Anonymous User Privacy Preferences

Strictly Necessary Cookies (Always Active)

MULTILEARNING platforms and tools hereinafter referred as “MLG SOFTWARE” are provided to you as pure educational platforms/services requiring cookies to operate. In the case of the MLG SOFTWARE, cookies are essential for the Platform to function properly for the provision of education. If these cookies are disabled, a large subset of the functionality provided by the Platform will either be unavailable or cease to work as expected. The MLG SOFTWARE do not capture non-essential activities such as menu items and listings you click on or pages viewed.


Performance Cookies

Performance cookies are used to analyse how visitors use a website in order to provide a better user experience.


Save Settings