Premium
Validity of laryngomalacia classification systems: A multi‐institutional agreement study
Author(s) -
Sivarajah Shanmugappiriya,
Isaac Andre,
Anderson Shelaina,
Mehta Deepak,
Hong Paul,
Alrajhi Yaser,
ElHakim Hamdy
Publication year - 2020
Publication title -
clinical otolaryngology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.914
H-Index - 68
eISSN - 1749-4486
pISSN - 1749-4478
DOI - 10.1111/coa.13530
Subject(s) - medicine , laryngomalacia , kappa , cohen's kappa , pediatrics , referral , inter rater reliability , surgery , airway , family medicine , psychology , linguistics , philosophy , rating scale , developmental psychology , machine learning , computer science , stridor
Objectives To determine the agreement of paediatric otolaryngologists on classifying laryngomalacia (LM). Design Intra‐ and interobserver agreement study of two classification systems. Setting Three tertiary referral paediatric centres. Participants Three paediatric otolaryngologists, who were blinded to any clinical details, interpreted the videos of children diagnosed with LM using the Holinger and Olney classifications independently. They rated the videos twice with a washout period of at least 2 weeks. The main outcome measures Inter‐ and intra‐observer agreement measured by overall Fleiss kappa and unweighted Cohen's kappa coefficients. The secondary outcome measures were inter‐ and intra‐observer agreement on the individual anatomical subunits of the supraglottis affected by LM, characterised by the subcategories of the classifications. Results Video records of infants and children <18 years who had an endoscopic diagnosis of LM from 2012 to 2017 were retrospectively chosen for inclusion (n = 73). The overall Fleiss kappa coefficient was 0.25 (95% CI 0.18‐0.32) amongst the raters using the Holinger classification and 0.31 (95% CI 0.21‐0.42) for the Olney classification. Intra‐observer agreement using the Holinger classification was 0.30 (95% CI 0.18‐0.42), 0.62 (95% CI 0.23‐0.85) and 0.84 (95% CI 0.75‐0.94], whilst the Olney classification yielded values of 0.41 (95% CI 0.26‐0.56), 0.51 (95% CI 0.29‐0.63) and 0.63 (95% CI 0.48‐0.78). Conclusions The agreement on types of LM between expert observers is modest using the Holinger and Olney classifications. This has significant implications for accurately diagnosing LM and exposes potential obstacles against credible pooling of data and extrapolation of information.