Development and Validation of a Predictive Model for Wheezing Illness Following Human Bocavirus 1 Infection in Children

Figures

• 

  Figure 1. LASSO coefficient paths for predictor selection. This plot visualizes the shrinkage of regression coefficients as the penalty parameter (Log(λ)) increases. Each colored line represents the trajectory of a specific clinical or laboratory variable. The numbers across the top axis indicate the number of non-zero variables remaining in the model at that specific λ.

• 

  Figure 2. ROC curves of the logistic regression model for training and test sets. Receiver Operating Characteristic (ROC) curves illustrate the diagnostic performance of the multivariate logistic regression model. The blue line represents the training set, with an AUC of 0.904. The red line represents the independent test set, showing an AUC of 0.876. The x-axis represents the false positive rate (1–specificity) and the y-axis represents the true positive rate (sensitivity). The diagonal gray line represents the performance of a random classifier (AUC = 0.5).

• 

  Figure 3. Simplified clinical scoring tool for early identification of children at risk of HBoV1-induced wheezing. This nomogram-style scoring chart translates the three independent predictors (abnormal NK cell percentage, personal history of allergy, and preterm birth) into an integer point system. Points are assigned as follows: abnormal NK cell percentage (present = 1 point), personal history of allergy (present = 2 points), and preterm birth (present = 4 points). The total score ranges from 0 to 7. Based on the Youden index-derived cutoff (dash line) of ≥3 points from the training set, patients are stratified into low-risk (total score 0–2) and high-risk (total score 3–7) groups. In the independent test set, the high-risk group had a wheezing rate of 84.2%, compared with 19.5% in the low-risk group (p < 0.001). The simplified scoring system achieved a sensitivity of 0.81 and a specificity of 0.79 for predicting post-HBoV1 wheezing illness.