Affordable Access

Prognostic modelling of COVID-19 using artificial intelligence in a UK population

Authors
  • Abdulaal, A
  • Patel, A
  • Charani, E
  • Denny, S
  • Mughal, N
  • Moore, L
Publication Date
Jul 24, 2020
Source
Spiral - Imperial College Digital Repository
Keywords
Language
English
License
Green
External links

Abstract

Background: The current severe acute respiratory syndrome-coronavirus disease (SARS-CoV-2) outbreak is a public health emergency which has had a significant case-fatality in the United Kingdom (UK). Whilst there appear to be several early predictors of outcome, there are no currently validated prognostic models or scoring systems applicable specifically to SARS-CoV-2 positive patients. Objective: To create a point-of-admission, mortality-risk scoring system utilising an artificial neural network (ANN). Methods: We present an ANN which can provide a patient-specific, point-of-admission mortality risk prediction to inform clinical management decisions at the earliest opportunity. The ANN analyses a set of patient features including demographics, comorbidities, smoking history and presenting symptoms and predicts patient-specific mortality risk during the current hospital admission. The model was trained and validated on data extracted from 398 patients admitted to hospital with a positive real-time reverse transcriptase polymerase chain reaction (rt-PCR) test for SARS-CoV-2. Results: Patient-specific mortality was predicted with 86.25% accuracy, with a sensitivity of 87.50% (95% CI: 61.65% to 98.45%) and specificity of 85.94% (95% CI: 74.98% to 93.36%). The positive predictive value was 60.87% (95% CI: 45.23% to 74.56%), and the negative predictive value was 96.49% (95% CI: 88.23% to 99.02%). The (AUROC) was 90.12%. Conclusions: This analysis demonstrates an adaptive ANN trained on data at a single site, which demonstrates the early utility of deep learning approaches in a rapidly evolving pandemic with no established or validated prognostic scoring systems.

Report this publication

Statistics

Seen <100 times