Machine learning approach to dynamic risk modeling of mortality in COVID-19: a UK Biobank cohort study

Dabbah, M. A.; Reed, A. B.; Booth, A. T. C.; Yassaee, A.; Despotovic, A.; Klasner, B.; Binning, E.; Aral, M.; Plans, D.; Labrique, A. B.; Mohan, D.

The COVID19 pandemic has resulted in over two million deaths globally. There is an urgent need for robust, scalable monitoring tools supporting resource allocation and stratification of high-risk patients. This research aims to develop and validate prediction models, using the UK Biobank to estimate COVID19 mortality risk in confirmed cases. We developed a random forest classification model using baseline characteristics, pre-existing conditions, symptoms, and vital signs, such that the score could dynamically assess risk of mortality with disease deterioration (AUC: 0.92). The design and feature selection of the framework lends itself to deployment in remote settings. Possible applications include supporting individual-level risk profiling and monitoring disease progression across high volumes of patients with COVID19, especially in hospital-at-home settings. The COVID19 pandemic has precipitated over 100 million confirmed cases and 2.3 million deaths globally. The impact of the pandemic has not been limited to healthcare systems: a ripple effect has resulted in wide-ranging economic and social disruption. Interventions to reduce transmission, such as lockdowns, travel restrictions, and re-allocation of health resources, are critical to limiting the impact. Although large-scale vaccination programmes have begun, many countries globally will not have widespread access to vaccines until 2023, meaning that non-pharmaceutical interventions are likely to remain indispensable national strategies for some time. COVID19 shows highly varied clinical presentation. A significant proportion (17 to 45%) of cases are asymptomatic and require no specific care. Conversely, reviews of severe complications have found that up to 32% of hospitalized COVID19 patients are admitted to ICU7. Between these two extremes, typical symptoms include fever, continuous cough, anosmia, and dyspnoea, which may range from requiring only self-management at home to inpatient care. Understanding which individuals are most vulnerable to severe disease, and thereby in most need of resources, is critical to limit the impact of the virus. Decision-making at all levels requires an understanding of individuals risk of severe disease. Various patient characteristics, comorbidities, and lifestyle factors have been linked to greater risk of death and/or severe illness following infection. Furthermore, socioeconomic factors have also been linked as risk factors for COVID19 mortality. Once patients are infected with SARS CoV 2, additional physiological parameters, such as symptoms and vital signs, can inform real-time prognostication13. Laboratory testing and imaging can also inform risk stratification for early, aggressive intervention, though this data is only accessible to hospital inpatients, who are likely to be already severely affected. Robust, predictive models for acquisition and prognosis of COVID 1916 18 and resource management have been developed to support risk stratification and population management at scale, offering important insights for organizational decision-making. However, the individual is currently overlooked, and granular, patient-specific risk-scoring could potentially unify decision-making at all levels. Existing individualized risk scores, however, often conflate risk of COVID19 acquisition with risk of mortality following infection, which can limit their utility in patient management. For prediction models to achieve impact at scale, assessment of risk factors should be inexpensive and accessible to the general population, ideally without the need for specialized testing or hospital visits. Such risk prediction tools, enabling improved patient triage, could be used to further increase the efficiency of, and confidence in, hospital at home solutions, which have shown promise in reducing hospital burden throughout the pandemic. Risk scores in these circumstances need to be dynamic and contemporaneous, ideally incorporating symptoms and vital sign data to maximise utility to clinical and research teams. Therefore, the primary aim of this study is to develop and validate a population-based prediction model, using a large, rich dataset and a selective, clinically informed approach, which dynamically estimates the COVID19 mortality risk in confirmed diagnoses.