New Model to Understand Phenotypic Presentation of Mendelian Disease in Electronic Health Records

A team of researchers from Vanderbilt University Medical Center develops a computational model to study phenotypic manifestations of genetic disorders using data from electronic health records (EHRs). Understanding the phenotypic expression in the EHR is key to developing algorithms to detect undiagnosed patients, the researchers say. For this study, published in Genetics in Medicine, the authors focused on nine Mendelian diseases because of their multisystem phenotypes, varied reported age of onset, and evidence in the literature of significant diagnostic delay. They identified 896 individuals with genetically confirmed diagnoses, of which 24 percent had fully ascertained diagnostic trajectories. The findings show that phenotype ascertainment is, in large part, driven by clinical exams and studies prompted by clinical suspicion of a genetic disease, which the authors termed as diagnostic convergence. "While diagnostic convergence is consistent with good clinical care, it poses a challenge for designing EHR-based algorithms to detect undiagnosed disease," the authors write. They noted that phenotypic attributes that were likely to have been present since birth such as congenital heart defects were often mentioned only after clinical suspicion. Meanwhile, this study also highlights the importance of censoring data prior to suspicion to avoid data leakage.