Ian CK Wong
Profile Url: ian-ck-wong
Researcher at University of Hong Kong
Background: The coronavirus disease 2019 (COVID-19) has become a pandemic, placing significant burdens on the healthcare systems. In this study, we tested the hypothesis that a machine learning approach incorporating hidden nonlinear interactions can improve prediction for Intensive care unit (ICU) admission. Methods: Consecutive patients admitted to public hospitals between 1st January and 24th May 2020 in Hong Kong with COVID-19 diagnosed by RT-PCR were included. The primary endpoint was ICU admission. Results: This study included 1043 patients (median age 35 (IQR: 32-37; 54% male). Nineteen patients were admitted to ICU (median hospital length of stay (LOS): 30 days, median ICU LOS: 16 days). ICU patients were more likely to be prescribed angiotensin converting enzyme inhibitors/angiotensin receptor blockers, anti-retroviral drugs lopinavir/ritonavir and remdesivir, ribavirin, steroids, interferon-beta and hydroxychloroquine. Significant predictors of ICU admission were older age, male sex, prior coronary artery disease, respiratory diseases, diabetes, hypertension and chronic kidney disease, and activated partial thromboplastin time, red cell count, white cell count, albumin and serum sodium. A tree-based machine learning model identified most informative characteristics and hidden interactions that can predict ICU admission. These were: low red cells with 1) male, 2) older age, 3) low albumin, 4) low sodium or 5) prolonged APTT. A five-fold cross validation confirms superior performance of this model over baseline models including XGBoost, LightGBM, random forests, and multivariate logistic regression. Conclusions: A machine learning model including baseline risk factors and their hidden interactions can accurately predict ICU admission in COVID-19.
Background: Pulmonary hypertension, a progressive lung disorder with symptoms such as breathlessness and loss of exercise capacity, is highly debilitating and has a negative impact on the quality of life. In this study, we examined whether a multi-parametric approach using machine learning can improve mortality prediction. Methods: A population-based territory-wide cohort of pulmonary hypertension patients from January 1, 2000 to December 31, 2017 were retrospectively analyzed. Significant predictors of all-cause mortality were identified. Easy-to-use frailty indexes predicting primary and secondary pulmonary hypertension were derived and stratification performances of the derived scores were compared. A factorization machine model was used for the development of an accurate predictive risk model and the results were compared to multivariate logistic regression, support vector machine, random forests, and multilayer perceptron. Results: The cohorts consist of 2562 patients with either primary (n=1009) or secondary (n=1553) pulmonary hypertension. Multivariate Cox regression showed that age, prior cardiovascular, respiratory and kidney diseases, hypertension, number of emergency readmissions within 28 days of discharge were all predictors of all-cause mortality. Easy-to-use frailty scores were developed from Cox regression. A factorization machine model demonstrates superior risk prediction improvements for both primary (precision: 0.90, recall: 0.89, F1-score: 0.91, AUC: 0.91) and secondary pulmonary hypertension (precision: 0.87, recall: 0.86, F1-score: 0.89, AUC: 0.88) patients. Conclusion: We derived easy-to-use frailty scores predicting mortality in primary and secondary pulmonary hypertension. A machine learning model incorporating multi-modality clinical data significantly improves risk stratification performance.