KLiP: A neural network-based surveillance tool for early identification of listening difficulties in children aged...

To develop and validate a neural network-based Kid's Listening Performance Checklist (KLiP) for early identification of listening difficulties in children aged 3-6 years, and to evaluate its effectiveness as a surveillance tool in real-world settings.

Development followed a multi-phase process: (1) systematic resource review to create preliminary items, (2) expert evaluation for content validity (7 experts), (3) assessment of inter-rater reliability between parents and teachers (182 parent-teacher pairs), (4) discriminant validity testing, (5) neural network model development trained on data from 341 children (202 with typical hearing, 139 with hearing loss), and (6) real-world validation testing with 71 parent-child pairs who completed both KLiP and hearing screening.

The KLiP checklist comprises eight discriminating behavioral items and six risk factors. The neural network model achieved 90% sensitivity and 97% specificity in distinguishing between children with and without hearing loss in the development dataset. In preliminary real-world testing, high KLiP scores (≥0.5) appeared to identify children requiring further audiological attention, while low scores (<0.5) strongly indicated typical hearing abilities in children without developmental concerns.

This preliminary study suggests that KLiP may demonstrate potential as a first-line surveillance tool for identifying listening difficulties in preschool children. Beyond surveillance, its implementation as an online platform with automated risk assessment and educational resources promotes awareness of listening difficulties among parents and teachers, supporting early identification within the pediatric hearing healthcare framework.