The simple 10-item PARC tool to predict childhood asthma – an external validation

Publication date: Available online 9 October 2018

Source: The Journal of Allergy and Clinical Immunology: In Practice

Author(s): Eva S.L. Pedersen, Ben D. Spycher, Carmen de Jong, Florian Halbeisen, Alban Ramette, Erol A. Gaillard, Raquel Granell, A John Henderson, Claudia E. Kuehni

Abstract

Background

External validation of prediction models is important to assess generalisability to other populations than the one used for model development. The Predicting Asthma Risk in Children (PARC) tool, developed in the Leicestershire Respiratory Cohort (LRC), uses information on preschool respiratory symptoms to predict asthma at school age.

Objective

We performed an external validation of PARC using the Avon Longitudinal Study of Parents and Children (ALSPAC).

Methods

We defined inclusion criteria, prediction score items at baseline and asthma at follow-up in ALSPAC to match those used in LRC using information from parent-reported questionnaires. We assessed performance of PARC by calculating sensitivity, specificity, predictive values, likelihood ratios, area under the curve (AUC), Brier score and Nagelkerke's R-squared. Sensitivity analyses varied inclusion criteria, scoring items and outcomes.

Results

The validation population included 2690 children with preschool respiratory symptoms of which 373 (14%) had asthma at school age. Discriminative performance of PARC was similar in ALSPAC (AUC=0.77, Brier score 0.13) as in LRC (0.78, 0.22). The score cut-off of 4 showed the highest sum of sensitivity (69%) and specificity (76%) and positive and negative likelihood ratios of 2.87 and 0.41, respectively. Changes to inclusion criteria, scoring items or outcome definitions barely altered the prediction performance.

Conclusion

Performing equally well in the validation cohort as in the development cohort, PARC is a valid tool for predicting asthma in population based cohorts. Its use in clinical practice is ready to be tested.