Background: Spirometry is the most commonly used lung function test in pediatric pulmonology. Valid outcome values require correct execution of breathing maneuvers, which can be challenging in young children. To ensure valid outcomes, quality control (QC), a time-consuming, subjective process based on numerical and graphical criteria from ERS/ATS guideline*, is needed.

Aims: To develop an automated QC algorithm for spirometry and test its application in the outpatient setting of a tertiary children`s hospital.

Methods: In this pilot study we developed an automated QC algorithm based on numerical and graphical criteria* that classifies spirometry curves into ?acceptable?, ?unacceptable? and ?questionable?. Graphical criteria are used to describe common errors (like coughing, sighing). Pediatric spirometry QC results of two independent reviewers were compared to the automated QC algorithm.

Results: It was possible to develop an automated QC using a training set. A total of n=137 spirometry curves from n=28 children were available for algorithm testing. Accuracy between the reviewers was 0.69 (?=0.53), between reviewer 1 and automated QC 0.71 (?=0.56), and between reviewer 2 and automated QC 0.70 (?=0.50).

Conclusions: Automated QC of spirometry using an algorithm was feasible with comparable accuracy to QC done by experienced reviewers. Discordant classification will be used to further improve the algorithm for future application in larger test sets.

*Graham, B.L. et al. AJRCCM 2019;200(8):e70-e88.