Pediatric asthma risk score: A new robust metric for predicting asthma in children

The global prevalence of asthma was as high as 262 million cases, causing 461,000 deaths in 2019.1 Early identification and primary prevention of asthma in young children are crucial to lowering mortality and morbidity caused by this condition.2 The currently used predictive tools for asthma, such as the Asthma Predictive Index (API), do not reliably predict the development of asthma in particular children.2 Hence, a new tool named Pediatric Asthma Risk Score (PARS) was developed to address the shortcomings.2 However, this tool was tested in a relatively homogeneous population, and therefore its generalizability was limited.3 Results of a recent study that evaluated PARS in a large heterogeneous population consisting of 10 different cohorts were presented at the recently held American Academy of Allergy, Asthma, and Immunology (AAAAI) Annual Meeting in February 2023.3

This study evaluated 5,674 children from 10 cohorts from the Children’s Respiratory and Environmental Workgroup (CREW) and compared the results with the currently used predictive tool, API.3 The purpose of this study was to assess PARS performance in predicting asthma in each cohort by calculating areas under the curve (AUCs) stratified by race, ethnicity, gender, cohort type, and birth decade.3 The AUCs were also determined in harmonized data across all the cohorts.3 The percentage of African-American ranged from 1.6% to 72.2% across the cohorts.3 The mean prevalence of asthma throughout the 10 cohorts was 18.6%, ranging from 9.9% to 37.3%.3 The age at the time of asthma diagnosis ranged from 5.1 to 10.2 years.3

The PARS AUC was found to be significantly higher than the API AUC (p-value ranges from 0.01 to <0.001) in 9 cohorts.3 The AUC for harmonized PARS was likewise significantly higher than the AUC for harmonized API (0.76 vs. 0.70, p<0.001).3 Though PARS and API were both capable of detecting >99% of the reported asthma in patients at high risk, API failed to predict 45.9% of the reported asthmatics with a low-moderate risk.3 The median AUC did not change by cohort type (p=1.0), decade of enrollment (p=0.16), parent-reported race (p=0.31), ethnicity (p=0.58), sex (p=0.96), or missing PARS components (p=0.83), demonstrating that the PARS model was resistant to these strata.3 The odds ratios (ORs) revealed high heterogeneity, and not all PARS variables were significantly associated with asthma in all the cohorts.3 The meta-analysis showed that all 6 PARS factors were independently and significantly associated with asthma, and the weights for 4 of these factors were similar to the original weights.3

In conclusion, PARS outperformed the API in 9 out of 10 CREW cohorts.3 With excellent predictive capability and robust performance in populations with significant variations, PARS offers a quantitative risk assessment that is absent from other predictive techniques.3 These findings also confirmed that the 6 original PARS factors accurately predict the asthma risk over a wide range of populations, timelines, and datasets.3 The predictive capability of PARS is unaffected by the missing data.3 All 6 PARS factors are supported by the meta-analysis in their ability to predict the asthma risk, making these the minimum set of necessary factors.3 The clinical implication of these findings is that PARS can be used in research to risk-stratify people for preventative intervention studies.3

Get access to our exclusive articles.
Related Articles