Statistical Analysis
Statistical analysis was performed with Matlab® R2007b (MathWorks Inc.,
Natick, USA). Qualitative variables were reported with number,
percentage and quantitative variables with mean and standard deviation.
We calculated sensitivity, specificity, positive predictive value (PPV),
negative predictive value (NPV) and their respective 95 % confidence
interval for the diagnosis of a COVID-19 infection with chest CT versus
RT-PCR (reference).
The associations between the mortality or morbidity (invasive
endotracheal ventilation or ICUH) and chest CT signs of COVID-19
infection (ground glass opacities, crazy paving, consolidation area,
subpleural curvilinear bands, bronchial distortion, vascular dilatation,
air bubble sign, interlobular thickening, total volume [50-100 %
versus < 50%], compatible aspect of organized pneumonia,
compatible CT aspect of ARDS, distribution [subpleural, central, and
dual i.e. central and subpleural], inferior localization, number of
involved lobes (0 to 3 VS 4 or 5 involved lobes) were evaluated using
Chi2 tests when applicable (all theorical numbers superior to 5), and
Fisher’s exact test otherwise.
All test results with a p-value < 0.05 were considered
statistically significant.