Back

A simple laboratory parameter facilitates early identification of COVID-19 patients

Li, Q.; Ding, X.; Xia, G.; Geng, Z.; Chen, F.; Wang, L.; Wang, Z.

2020-02-17 infectious diseases
10.1101/2020.02.13.20022830 medRxiv
Show abstract

The total number of COVID-19 patients since the outbreak of this infection in Wuhan, China has reached 40000 and are still growing. To facilitate triage or identification of the large number of COVID-19 patients from other patients with similar symptoms in designated fever clinics, we set to identify a practical marker that could be conveniently utilized by first-line health-care workers in clinics. To do so, we performed a case-control study by analyzing clinical and laboratory findings between PCR-confirmed SARS-CoV-2 positive patients (n=52) and SARS-CoV-2 negative patients (n=53). The patients in two cohorts all had similar symptoms, mainly fever and respiratory symptoms. The rates of patients with leukocyte counts (normal or decreased number) or lymphopenia (two parameters suggested by current National and WHO COVID-19 guidelines) had no differences between these two cohorts, while the rate of eosinopenia (decreased number of eosinophils) in SARS-CoV-2 positive patients (79%) was much higher than that in SARS-CoV-2 negative patients (36%). When the symptoms were combined with eosinopenia, this combination led to a diagnosis sensitivity and specificity of 79% and 64%, respectively, much higher than 48% and 53% when symptoms were combined with leukocyte counts (normal or decreased number) and/ or lymphopenia. Thus, our analysis reveals that eosinopenia may be a potentially more reliable laboratory predictor for SARS-CoV-2 infection than leukocyte counts and lymphopenia recommended by the current guidelines.

Matching journals

The top 7 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.