Susceptibility of AutoML mortality prediction algorithms to model drift caused by the COVID pandemic.

BACKGROUND: Concept drift and covariate shift lead to a degradation of machine learning (ML) models. The objective of our study was to characterize sudden data drift as caused by the COVID pandemic. Furthermore, we investigated the suitability of certain methods in model training to prevent model degradation caused by data drift. METHODS: We trained different ML models with the H2O AutoML method on a dataset comprising 102,666 cases of surgical patients collected in the years 2014-2019 to predict postoperative mortality using preoperatively available data. Models applied were Generalized Linear Model with regularization, Default Random Forest, Gradient Boosting Machine, eXtreme Gradient Boosting, Deep Learning and Stacked Ensembles comprising all base models. Further, we modified the original models by applying three different methods when training on the original pre-pandemic dataset: (Rahmani K, et al, Int J Med Inform 173:104930, 2023) we weighted older data weaker, (Morger A, et al, Sci Rep 12:7244, 2022) used only the most recent data for model training and (Dilmegani C, 2023) performed a z-transformation of the numerical input parameters. Afterwards, we tested model performance on a pre-pandemic and an in-pandemic data set not used in the training process, and analysed common features. RESULTS: The models produced showed excellent areas under receiver-operating characteristic and acceptable precision-recall curves when tested on a dataset from January-March 2020, but significant degradation when tested on a dataset collected in the first wave of the COVID pandemic from April-May 2020. When comparing the probability distributions of the input parameters, significant differences between pre-pandemic and in-pandemic data were found. The endpoint of our models, in-hospital mortality after surgery, did not differ significantly between pre- and in-pandemic data and was about 1% in each case. However, the models varied considerably in the composition of their input parameters. None of our applied modifications prevented a loss of performance, although very different models emerged from it, using a large variety of parameters. CONCLUSIONS: Our results show that none of our tested easy-to-implement measures in model training can prevent deterioration in the case of sudden external events. Therefore, we conclude that, in the presence of concept drift and covariate shift, close monitoring and critical review of model predictions are necessary.

Anaesthesiologists' guideline adherence in pre-operative evaluation: a retrospective observational study.

BACKGROUND: Surveys suggest a low level of implementation of clinical guidelines, although they are intended to improve the quality of treatment and patient safety. Which guideline recommendations are not followed and why has yet to be analysed. In this study, we investigate the proportion of European and national guidelines followed in the area of pre-operative anaesthetic evaluation prior to non-cardiac surgery. METHODS: We conducted this monocentric retrospective observational study at a German university hospital with the help of software that logically links guidelines in such a way that individualised recommendations can be derived from a patient's data. We included routine logs of 2003 patients who visited our pre-anaesthesia outpatient clinic between June 2018 and June 2020 and compared the actual conducted pre-operative examinations with the recommendations issued by the software. We descriptively analysed the data for examinations not performed that would have been recommended by the guidelines and examinations that were performed even though they were not covered by a guideline recommendation. The guidelines examined in this study are the 2018 ESAIC guidelines for pre-operative evaluation of adults undergoing elective non-cardiac surgery, the 2014 ESC/ESA guidelines on non-cardiac surgery and the German recommendations on pre-operative evaluation on non-cardiothoracic surgery from the year 2017. RESULTS: Performed ECG (78.1%) and cardiac stress imaging tests (86.1%) indicated the highest guideline adherence. Greater adherence rates were associated with a higher ASA score (ASA I: 23.7%, ASA II: 41.1%, ASA III: 51.8%, ASA IV: 65.8%, P < 0.001), lower BMI and age > 65 years. Adherence rates in high-risk surgery (60.5%) were greater than in intermediate (46.5%) or low-risk (44.6%) surgery (P < 0.001). 67.2% of technical and laboratory tests performed preoperatively were not covered by a guideline recommendation. CONCLUSIONS: Guideline adherence in pre-operative evaluation leaves room for improvement. Many performed pre-operative examinations, especially laboratory tests, are not recommended by the guidelines and may cause unnecessary costs. The reasons for guidelines not being followed may be the complexity of guidelines and organisational issues. A software-based decision support tool may be helpful. TRIAL REGISTRATION: ClinicalTrials.gov ID NCT04843202.