RESUMO
Background: The field of machine learning has been evolving and applied in medical applications. We utilised a public dataset, MIMIC-III, to develop compact models that can accurately predict the outcome of mechanically ventilated patients in the first 24 h of first-time hospital admission. Methods: 67 predictive features, grouped into 6 categories, were selected for the classification and prediction task. 4 tree-based algorithms (Decision Tree, Bagging, eXtreme Gradient Boosting and Random Forest), and 5 non-tree-based algorithms (Logistic Regression, K-Nearest Neighbour, Linear Discriminant Analysis, Support Vector Machine and Naïve Bayes), were employed to predict the outcome of 18,883 mechanically ventilated patients. 5 scenarios were crafted to mirror the target population as per existing literature. S1.1 reflected an imbalanced situation, with significantly fewer mortality cases than survival ones, and both the training and test sets played similar target class distributions. S1.2 and S2.2 featured balanced classes; however, instances from the majority class were removed from the test set and/or the training set. S1.3 and S 2.3 generated additional instances of the minority class via the Synthetic Minority Over-sampling Technique. Standard evaluation metrics were used to determine the best-performing models for each scenario. With the best performers, Autofeat, an automated feature engineering library, was used to eliminate less important features per scenario. Results: Tree-based models generally outperformed the non-tree-based ones. Moreover, XGB consistently yielded the highest AUC score (between 0.91 and 0.97), while exhibiting relatively high Sensitivity (between 0.58 and 0.88) on 4 scenarios (1.2, 2.2, 1.3, and 2.3). After reducing a significant number of predictors, the selected calibrated ML models were still able to achieve similar AUC and MCC scores across those scenarios. The calibration curves of the XGB and BG models, both prior to and post dimension reduction in Scenario 2.2, showed better alignment to the perfect calibration line than curves produced from other algorithms. Conclusion: This study demonstrated that dimension-reduced models can perform well and are able to retain the important features for the classification tasks. Deploying a compact machine learning model into production helps reduce costs in terms of computational resources and monitoring changes in input data over time.
RESUMO
INTRODUCTION: Emergency departments (EDs) at public hospitals in Vietnam typically face problems with overcrowding, as well as being populated by a wide variety of illnesses, resulting in increasing dissatisfaction from patients. To alleviate these problems, we used the increasingly popular value-stream mapping (VSM) and lean strategy approaches to (1) evaluate the current patient flow in EDs; (2) identify and eliminate the non-valued-added components; and (3) modify the existing process in order to improve waiting times. METHODS: Data from a total of 742 patients who presented at the ED of 108 Military Central Hospital in Hanoi, Vietnam, were collected. A VSM was developed where improvement possibilities were identified and attempts to eliminate non-value-added activities were made. A range of issues that were considered as a resource waste were highlighted, which led to a re-design process focusing on prioritizing blood tests and ultrasound procedures. On the administrative side, various measures were considered, including streamlining communication with medical departments, using QR codes for healthcare insurance payments, and efficient management of X-ray and CT scan online results. RESULTS: By implementing a lean approach, the following reductions in delay and waiting time were incurred: (1) pre-operative test results (for patients requiring medical procedures/operations) by 33.3% (from 134.4 to 89.4 min); (2) vascular interventions by 10.4% (from 54.6 to 48.9 min); and (3) admission to other hospital departments by 49.5% (from 118.3 to 59.8 min). Additionally, prior to the implementation of the lean strategy approach, only 22.9% of patients or their proxies (family members or friends), who responded to the survey, expressed satisfaction with the ED services. This percentage increased to 76.5% following the curtailment of non-value-added activities. Through statistical inferential test analyses, it can be confidently concluded that applying lean strategy and tools can improve patient flow in public/general hospital EDs and achieve better staff coordination within the various clinical and administrative hospital departments. To the authors' knowledge, such analysis in a Vietnamese hospital's ED context has not been previously undertaken.