Agreement between a health claims algorithm and parent‐reported asthma in young children
Omand JA, Maguire JL, O'Connor DL, Parkin PC, Birken CS, Thorpe KE, Zhu J, To T. Pediatr. Pulmonol. 2019; Jul 22 [Epub ahead of print]. DOI: 10.1002/ppul.24432.
Introduction — Asthma prevalence is commonly measured in national surveys by questionnaire. The Ontario Asthma Surveillance Information System (OASIS) developed a validated health claims diagnosis algorithm to estimate asthma prevalence. The primary objective was to assess the agreement between two approaches of measuring asthma in young children. Secondary objectives were to identify concordant and discordant pairs, and to identify factors associated with disagreement.
Study Design and Setting — A measurement study to evaluate the agreement between the OASIS algorithm and parent‐reported asthma (criterion standard). Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated. Multivariable logistic regression was used to determine factors associated with disagreement.
Results — Healthy children aged 1 to 5 years (n =3642) participating in the TARGet Kids! practice based research network 2008‐2013 in Toronto, Canada were included. Prevalence of asthma was 14% and 6% by the OASIS algorithm and parent‐reported asthma, respectively. The Kappa statistic was 0.43, sensitivity 81%, specificity 90%, PPV 34%, and NPV 99%. There were 3249 concordant and 393 discordant pairs. Statistically significant factors associated with asthma identified by OASIS but not parent report included: male sex, higher zBMI, and parent history of asthma. Males were less likely to have asthma identified by parent report but not OASIS.
Conclusion — The OASIS algorithm identified more asthma cases in young children than parent‐reported asthma. The OASIS algorithm had high sensitivity, specificity, and NPV but low PPV relative to parent‐reported asthma. These findings need replication in other populations.