Background Commonly used Asthma Predictive Index (API; Castro-Rodriguez 2000) has good negative but poor positive predictive value (PPV) which reduces its clinical usefulness. We aimed to develop, within a population-based cohort, an asthma risk prediction tool (ARPT) better suited for clinical use.
Method Children from a UK birth cohort attended follow-up at age 3, 8 and 11 years. Parents completed validated questionnaires. Children were skin prick tested (SPT) to common inhalant and food allergens. GP records were transcribed to collect information on physician-confirmed wheeze. An ARPT was developed using backwards logistic regression of data collected at age 3 years. We assessed its predictive performance using area under the receiver operating characteristic curve (AUROC). Repeated internal validation was performed.
Results Of 829 children were included in the analysis, 132(17.8%) had asthma at school-age (year 8 or 11). The ARPT was found to include 4 predictors at age 3 years yielding a total score of 5: wheeze ever (1), wheeze causing shortness of breath (1), wheeze after exercise (2) and eczema (1). In the whole population, children with a score of >4 had a significantly increased risk of having asthma at school-age odds ratio [OR] 25.3, 95% CI [11.8–54.1], p < 0.0001; PPV 80%; sensitivity 28%). API, when applied to our cohort, yielded an OR for asthma at school-age of 7.4 (95% CI [4.5–12.2], p < 0.0001; PPV 54%; sensitivity 31%). We then applied our ARPT amongst children who presented to their GP with wheeze within the first 3 years of life; those with ARTP score of ≥4 had a significantly increased risk of asthma at school-age (OR 15.6, 95% CI [6.1–39.9] p < 0.0001; PPV 82%; sensitivity 35%). API, when applied in this population, had OR of 4.7 (95% CI [2.53–8.60], p < 0.0001; PPV 57%, Sensitivity 44%). Internal validation showed good agreement (AUCpredicting 0.78 vs. AUCobserved 0.77).
Conclusion APRT is a simple tool based mainly on clinical history which could easily be applied in primary/secondary care to risk stratify children with wheeze symptoms in early childhood and predict asthma by school-age with a good positive predictive value.