Researchers from Holmusk recently presented two studies at the American Psychiatric Association’s annual conference.
Both of the studies were conducted using NeuroBlu, Holmusk’s data and analytics platform that generates new insights on a range of mental and behavioral disorders by analyzing real-world data derived from the NeuroBlu database, an industry-leading behavioral health database that contains thousands of deidentified electronic health records (EHRs).
Real-World data reveals differences between diagnoses and phenotypes
Traditionally, providers use the ICD/DSM classification systems to diagnose mental health disorders. However, because these diagnoses are based on a specific set of criteria, they may not be representative of the full range of potential presentations of the disorder.
Currently, providers include clinical information via notes in patients’ electronic health records (EHRs). A Holmusk study investigated whether this unstructured free text might be useful to better characterize clinical presentation beyond traditional diagnostic classification.
The study, which included data from 543,849 patients, used Holmusk’s natural language processing models to extract discrete mental state examination (MSE) data from the EHRs. The extracted data included variables corresponding to mood, cognition, and presence of delusions or hallucinations. Researchers then mapped and examined the distribution of these MSE data by original diagnosis.
Some of the findings were consistent with the original diagnosis; for example, delusions and hallucinations were most common inpatients with schizophrenia or schizoaffective disorder, while problems with cognition were seen most often in patients with dementia or Alzheimer’s disorder. However, some of the findings shed light on potential disease presentations that do not correspond with the original diagnosis. For example, impacts on mood were seen across non-mood related disorders (e.g.schizophrenia) in addition to mood-related disorders (e.g. major depressive disorder (MDD) and bipolar disorder).
“These findings highlight the potential for rich clinical information drawn from real-world data to better understand the clinical presentation in individual patients,” said Rashmi Patel, MD, PhD, Holmusk’s Vice President of Scientific and Medical Affairs and lead author of this study. “In this case, real-world data have suggested that a transdiagnostic approach to care could help to better characterize individual clinical presentation beyond traditional diagnostic classification to enable more effective, personalized care. Moving forward, NLP-derived measures using EHR data present a key opportunity to support real-time clinical decision making.”
Additional Holmusk contributors to this study included Soon Nan Wee, Jesisca Tandi, MS; Joydeep Sarkar, PhD; and Scott Kollins, PhD.
How polypharmacy affects functioning for children with ADHD
Kollins, Holmusk’s Chief Medical Officer, also presented a poster as an encore presentation to the American Academy of Child and Adolescent Psychiatry meeting in fall 2021, where the study was accepted as an abstract and published in Journal of the American Academy of Child & Adolescent Psychiatry.
The study used the NeuroBlu analytics platform to examine real-world data and study the relationship between polypharmacy and clinical functioning in patients.
“We were pleased to be able to present this work at an important conference like the American Psychiatric Association,” Kollins said. “We were glad to have the opportunity to highlight the importance of accounting for and leveraging real-world data when studying a range of behavioral and mental health conditions, from ADHD to schizophrenia to major depressive disorder.”
Additional Holmusk authors who contributed to this study are Matthew Valko, MS, and Miguel Rentería, PhD.
