Self-reported symptoms can be used to predict whether a person will develop Covid-19, according to a study led by researchers at Kings College London.
Previous studies have tried to predict disease using symptoms from the peak of infection, which isn’t useful for early detection.
The new research, published in journal The Lancet Digital Health, collected data from the ZOE COVID Symptom Study smartphone app.
Researchers assessed three models to predict Covid from self-reported symptoms, which involved using one dataset to train a given model and testing performance on another. The training set included almost 183,000 people who reported early symptoms from 16 October to 30 November 2020, while the test dataset consisted of over 15,000 participants.
The three models were: 1) a statistics method called logical regression; 2) the UK National Health Service (NHS) algorithm, and; 3) an Artificial Intelligence (AI) approach known as a ‘hierarchical Gaussian process’.
Of the three models, the AI approach performed the best.