ABSTRACT
BackgroundDiagnostic tests for SARS-CoV-2 infection (mostly RT-PCR and Computed Tomography) are not widely available in numerous countries, expensive and with imperfect performance MethodsThis multicenter retrospective study aimed to determine a pre-test probability score for SARS-CoV-2 infection based on clinical and biological variables. Patients were recruited from emergency and infectious disease departments and were divided into a training and a validation cohort. Demographic characteristics, clinical symptoms, and results of blood tests (complete white blood cell count, serum electrolytes and CRP) were collected. The pre-test probability score was derived from univariate analyses between patients and controls, followed by multivariate binary logistic analysis to determine the independent variables associated with SARS-CoV-2 infection. Points were assigned to each variable to create the PARIS score. ROC curve analysis determined the area under the curve (AUC). FindingsOne hundred subjects with clinical suspicion of SARS-CoV-2 infection were included in the training cohort, and 300 other consecutive individuals were included in the validation cohort. Low lymphocyte (<1{middle dot}3 G/L), eosinophil (<0{middle dot}06G/L), basophil (<0{middle dot}04G/L) and neutrophil counts (<5G/L) were associated with a high probability of SARS-CoV-2 infection. No clinical variable was statistically significant. The score had a good performance in the validation cohort (AUC=0.889 (CI: [0.846-0.932]; STD=0.022) with a sensitivity and Positive Predictive Value of high-probability score of 80{middle dot}3% and 92{middle dot}3% respectively. Furthermore, a low-probability score excluded SARS-CoV-2 infection with a Negative Predictive Value of 99.5%. InterpretationThe PARIS score based on complete white blood cell count has a good performance to categorize the pre-test probability of SARS-CoV-2 infection. It could help clinicians avoid diagnostic tests in patients with a low-probability score and conversely keep on testing individuals with high-probability score but negative RT-PCR or CT. It could prove helpful in countries with a low-availability of PCR and/or CT during the current period of pandemic. FundingNone Putting research into contextO_ST_ABSEvidence before this studyC_ST_ABSIn numerous countries, large population testing is impossible due to the limited availability and costs of RT-PCR kits and CT-scan. Furthermore, false-negativity of PCR or CT as well as COVID-19 pneumonia mimickers on CT may lead to inaccurate diagnoses. Pre-test probability combining clinical and biological features has proven to be a particularly useful tool, already used in clinical practice for management of patients with a suspicion of pulmonary embolism. Added value of this studyThis retrospective study including 400 patients with clinical suspicion of SARS-CoV-2 infection was composed of a training and a validation cohort. The pre-test probability score (PARIS score) determines 3 levels of probability of SARS-CoV2 infection based on white blood cell count (lymphocyte, eosinophil, basophil and neutrophil cell count). Implications of the available evidenceThis pre-test probability may help to adapt SARS-CoV-2 infection diagnostic tests. The high negative predictive value (99{middle dot}5%) of the low probability category may help avoid further tests, especially during a pandemic with overwhelmed resources. A high probability score combined with typical CT features can be considered sufficient for diagnosis confirmation.
ABSTRACT
Improving screening, discovering therapies, developing a vaccine and performing staging and prognosis are decisive steps in addressing the COVID-19 pandemic. Staging and prognosis are especially crucial for organizational anticipation (intensive-care bed availability, patient management planning) and accelerating drug development; through rapid, reproducible and quantified response-to-treatment assessment. In this letter, we report on an artificial intelligence solution for performing automatic staging and prognosis based on imaging, clinical, comorbidities and biological data. This approach relies on automatic computed tomography (CT)-based disease quantification using deep learning, robust data-driven identification of physiologically-inspired COVID-19 holistic patient profiling, and strong, reproducible staging/outcome prediction with good generalization properties using an ensemble of consensus methods. Highly promising results on multiple independent external evaluation cohorts along with comparisons with expert human readers demonstrate the potentials of our approach. The developed solution offers perspectives for optimal patient management, given the shortage of intensive care beds and ventilators1, 2, along with means to assess patient response to treatment.
