The vast majority of patients who exhibit signs of tardive dyskinesia do not receive a formal diagnosis for the condition, underscoring a need for better recognition and diagnosis, according to findings presented by Holmusk recently at the Schizophrenia International Research Society annual meeting.

Tardive dyskinesia is a severe and persistent involuntary movement disorder which is often associated with long-term antipsychotic treatment. Although there are diagnosis codes assigned to tardive dyskinesia, these may not be formally recorded in electronic health records. Instead, clinicians may be more likely to document features of tardive dyskinesia within unstructured data fields, such as clinical notes and treatment plans.

The Holmusk study team, led by Kira Griffiths, PhD, applied natural language processing to extract information about patient psychomotor function that was recorded in unstructured clinical notes. The team then reviewed these notes for key words that could indicate a likely diagnosis of tardive dyskinesia.

The study included over 350,000 adults who had been treated with antipsychotics; the cohort was derived from Holmusk’s NeuroBlu Database, the world’s leading source for real-world data on behavioral health. From this cohort, over 4,000 patients had features of tardive dyskinesia recorded within unstructured clinical notes—however, only 54 patients were diagnosed with the condition.

“Our findings show that only 1 percent of patients who may have tardive dyskinesia receive an official ICD diagnosis,” Griffiths said. “These results highlight a pressing need for improved recognition and diagnosis of the disorder, in order to increase research efforts in this population and to ultimately improve access to treatment.”

“The NeuroBlu Database was the ideal real-world data source to conduct a study like this,” Griffiths continued. “The breadth and depth of the NeuroBlu Database allowed us to assemble a large cohort of patients who were treated with antipsychotics and identify a relatively uncommon and underreported side effect. This study also underscores Holmusk’s strength in the examination of unstructured EHR data in order to obtain a richer view of the clinical picture beyond structured diagnoses.”