Vital signs should be maintained as continuous variables when predicting bacterial infections in febrile children

Area of Science:

• Pediatrics
• Infectious Diseases
• Biostatistics

Background:

• Febrile children require accurate prediction models for serious bacterial infections (SBI).
• Vital signs like heart rate and respiratory rate are crucial indicators.
• Optimal methods for incorporating vital signs into predictive models are debated.

Purpose of the Study:

• To evaluate the impact of different methods for including vital signs in prediction models for SBI in children.
• To compare the predictive performance of dichotomized versus continuous vital sign data.

Main Methods:

• Prospective observational study of 1,750 febrile children (<16 years).
• Compared seven strategies for incorporating heart and respiratory rates (dichotomized vs. continuous) into SBI prediction models.
• Utilized area under the receiver operating characteristic curve (AUC) to assess model performance.

Main Results:

• Dichotomized vital signs (tachycardia, tachypnea) showed limited predictive value (AUC 0.53-0.55).
• Models including age as a continuous predictor improved prediction (AUC 0.58).
• Models incorporating age and continuous vital signs yielded the highest predictive accuracy (AUC 0.60-0.63).

Conclusions:

• Heart and respiratory rates should be utilized as continuous variables in prediction models for SBI in febrile children.
• Dichotomization of vital signs leads to significant information loss and diminished predictive ability.
• Continuous vital sign data enhances the accuracy of SBI prediction models in pediatric populations.