COVID-19 in Wuhan: Immediate Psychological Impact on 5062 Health Workers

Zhou Zhu; Shabei Xu; Hui Wang; Zheng Liu; Jianhong Wu; Guo Li; Jinfeng Miao; Chenyan Zhang; Yuan Yang; Wenzhe Sun; Suiqiang Zhu; Yebin Fan; Junbo Hu; Jihong Liu; Wei Wang

Continuous variables were divided as categorical variables firstly and all variables were shown as the counts and percentages. Variables with P values＜0.05 in univariate analyses were subjected to multivariate logistic regression analysis with a stepwise backwards elimination procedure. Statistical analyses to identify influencing factors were performed using SPSS 22.0 (Statistical Package for the Social Sciences) for Windows (SPSS, Chicago, IL) . The Depression, Anxiety and Psychological Stress nomograms were formulated based on the results of multivariate logistic regression analysis using the R packages "rms", "Hmisc" and "ggplot2" in R 3.5.2 (http://www.r-project.org/). The performances of the nomograms were measured by Concordance statistics (C-statistics) and assessed by calibration curves. Bootstraps with 1000 resamples were applied to these activities. The higher C-statistics indicate better ability to distinguish HWs with different risks of three outcomes. We considered the risk screening models as a useful clinical tool particularly when the C-statistic is higher than 0.70 14 . The calibration curves were used to compare the observed probability with the predicted probability. Dots on the calibration plot would be close to a 45° diagonal line if the model calibration is correct.