Fig. 1

Filtering variables for logistic regression by using LASSO regression. (A) Model coefficient trendlines of the total 47 variables for clinical deterioration. The y-axis represents the coefficient of the variable, the lower x-axis indicates the log (λ), the upper x-axis indicates the number of variables. As log (λ) increases, the coefficient of variables tends to zero, and the number of corresponding variables gradually decreases. (B) Cross-validation curves for LASSO regression. The y-axis represents binomial deviation. The dashed line on the left shows the log (λ) when the model deviation is minimal, and the dashed line on the right shows the log (λ) when the model deviation is minimal plus one standard error. We selected the log (λ) and the variables corresponding to the dashed line on the right