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

Abstract

Purpose Data mining algorithms using electronic health records (EHRs) are useful in large-scale population-wide studies to classify etiology and comorbidities ( Casey et al., 2016 ). Here, we apply this approach to developmental language disorder (DLD), a prevalent communication disorder whose risk factors and epidemiology remain largely undiscovered. Method We first created a reliable system for manually identifying DLD in EHRs based on speech-language pathologist (SLP) diagnostic expertise. We then developed and validated an automated algorithmic procedure, called, Automated Phenotyping Tool for identifying DLD cases in health systems data (APT-DLD), that classifies a DLD status for patients within EHRs on the basis of ICD (International Statistical Classification of Diseases and Related Health Problems) codes. APT-DLD was validated in a discovery sample ( N = 973) using expert SLP manual phenotype coding as a gold-standard comparison and then applied and further validated in a replication sample of N = 13,652 EHRs. Results In the discovery sample, the APT-DLD algorithm correctly classified 98% (concordance) of DLD cases in concordance with manually coded records in the training set, indicating that APT-DLD successfully mimics a comprehensive chart review. The output of APT-DLD was also validated in relation to independently conducted SLP clinician coding in a subset of records, with a positive predictive value of 95% of cases correctly classified as DLD. We also applied APT-DLD to the replication sample, where it achieved a positive predictive value of 90% in relation to SLP clinician classification of DLD. Conclusions APT-DLD is a reliable, valid, and scalable tool for identifying DLD cohorts in EHRs. This new method has promising public health implications for future large-scale epidemiological investigations of DLD and may inform EHR data mining algorithms for other communication disorders. Supplemental Material https://doi.org/10.23641/asha.12753578

The Musical Abilities, Pleiotropy, Language, and Environment (MAPLE) Framework for Understanding Musicality-Language Links Across the Lifespan

“…Recent advances in data-driven approaches such as data mining algorithms and machine learning models have the potential to further accelerate progress in this research area, by automating the identification of cases with particular symptomatology in electronic heath records (e.g., automated phenotyping tool for DLD cases, APT-DLD: Walters et al, 2020 ; phenome risk classifier for stuttering: Pruett et al, 2021 ), the extraction of neural features from magnetic resonance imaging (MRI) data (e.g., toolbox for the automatic segmentation of Heschl’s gyrus, TASH toolbox: Dalboni da Rocha et al, 2020 ), the extraction of features in the genetic architecture of a trait (e.g., GWAS loci prioritization: Nicholls et al, 2020 ), and the integration of neuroimaging and genomic data to predict phenotypic outcomes (e.g., Shen & Thompson, 2020 ).…”

Association of Developmental Language Disorder With Comorbid Developmental Conditions Using Algorithmic Phenotyping

“…The DLD case set was previously identified by applying the APT-DLD algorithm to the 3.5 million clinical records in Vanderbilt University Medical Center's EHR system called Synthetic Derivative, as part of the APT-DLD development efforts; the details were reported in Walters et al 30 The APT-DLD algorithm identified 6013 pediatric records as DLD cases with available ICD codes, dates of the ICD codes, and relevant demographic information (Table 1). This cohort of DLD cases was used as the input for the large-scale comorbidity analysis performed in the present study.…”

“…[26][27][28][29] Developmental language disorder does not have a single unique diagnostic billing code; thus, to label DLD cases, we leveraged an existing electronic health record (EHR)-based tool with high positive predictive value (90%-95%) for DLD called the Automated Phenotyping Tool for Identifying Developmental Language Disorder (APT-DLD). 30 We then performed an enrichment analysis to identify comorbidities across the medical phenome. The aim of this study was to identify these clinically relevant comorbid conditions that co-occur with DLD using data-rich EHRs to better understand the epidemiological characteristics of DLD at a population level and potentially raise awareness of the DLD phenotype in the clinical community.…”

Test of Prosody via Syllable Emphasis (“TOPsy”): Psychometric Validation of a Brief Scalable Test of Lexical Stress Perception

“…While reliable phenotyping is currently available in biobanks and electronic health records for many health, cognitive, and psychiatric traits, speech-language and reading traits are generally underrepresented in these large-scale databases and are often superficially phenotyped ( Raghavan et al, 2018 ). Efforts are therefore underway to mine available speech-language phenotypes in existing electronic health records using sophisticated computational methods such as automated pipelines (e.g., for identifying cases with Developmental Language Disorder: Walters et al, 2020 ) and machine learning (e.g., for imputing stuttering cases: Pruett et al, 2021 ). Both approaches aim to decrease manual effort (e.g., clinical chart review) toward achieving the scales required for genetic investigations.…”