Can a machine identify DLD?

Walters Jr, C. E., Nitin, R., Margulis, K., Boorom, O., Gustavson, D. E., Bush, C. T., ... & Gordon, R. L. (2020). Automated Phenotyping Tool for Identifying Developmental Language Disorder Cases in Health Systems Data (APT-DLD): A New Research Algorithm for Deployment in Large-Scale Electronic Health Record Systems. Journal of Speech, Language, and Hearing Research, 63(9), 3019-3035. Aim of the paper: Large-scale health record databases can provide rich data for improving our understanding of disorders. This paper aims to create an automatic programme that can identify individuals with DLD from large electronic health record (EHR) databases. The automatic programme, called Automated Phenotyping Tool for identifying DLD (APT-DLD), was based on speech-language pathologist diagnostic principles. The APT-DLD was then tested to see how well it matches the results of manual identification by trained research assistants and speech-language pathologists. What they found:

  • In a sample of 973 children’s records, classification of DLD cases using APT-DLD shares a 98% similarity with manual identification by trained research assistants, and a 95% similarity with manual identification by speech-language pathologists.

  • A replication study was conducted with a larger sample of 13,652 children’s records. Here, classification of DLD cases using APT-DLD shares a 90% similarity with manual identification by speech-language pathologists.

What does this mean? The APT-DLD is a reliable tool for identifying individuals with DLD from large health record databases. Due to its automatic feature, such a programme can be applied to multiple databases at once, allowing future large-scale investigations of DLD. The authors of this paper suggest future research to apply APT-DLD to other EHR databases to: -> further refine the programme -> widen the demographic base (the characteristics of the people included in the health records) -> improve our understanding of the risk factors and underlying cause of DLD -> generate large samples for potential genetic investigations Where can I read this paper? This paper is open access, which means everyone can read it. Please click here to find the full paper.