Back

What is the probability that this patient, who presents to a UK hospital, will be diagnosed with Covid-19? Prospective validation of the open-source CovidCalculatorUK resource.

Chapman, G. A.; Mundell, L.; Harrison, C. H.; Cargill, T.; Keating, O.; Johnson, M.; Smith, A.

2020-12-08 infectious diseases
10.1101/2020.12.06.20243691 medRxiv
Show abstract

IntroductionThe novel coronavirus SARS-CoV2 and the associated disease, Covid-19, continue to pose a global health threat. The CovidCalculatorUK is an open-source online tool (covidcalculatoruk.org) that estimates the probability that an individual patient, who presents to a UK hospital, will later test positive for SARS-CoV2. The objective is to aid cohorting decisions and minimise nosocomial transmission of SARS-CoV2. MethodsThis n = 500 prospective, observational, multicentre, validation study compared the CovidCalculatorUKs estimated probability of Covid-19 with the first SARS-CoV2 oropharyngeal/nasopharyngeal swab result for individual patients admitted to hospital during the study period (01.04.20 - 18.05.20). A comparison with senior clinicians estimates of the probability of Covid-19 was also made. ResultsPatients who were prospectively grouped, by the CovidCalculatorUK, into 0-30% estimated probability, 30-60% and 60-100% estimated probability went on to have first swab SARS-CoV2 positive results in: 15.7%, 30.5% and 61.9% of cases, respectively. CovidCalculatorUK performance demonstrated an area under the curve of 0.76 (95% CI 0.71 - 0.81) (p < 0.001). Senior clinician stratification of the estimated probability of Covid-19 performed similarly to the CovidCalculatorUK. ConclusionThe CovidCalculatorUK provides a reasonably accurate estimate of the probability of an individual testing positive on their first SARS-CoV2 nasopharyngeal/oropharyngeal swab. The CovidCalculatorUK output performs similarly to a senior clinicians estimate. Further evolution of the calculator may improve performance.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.