Case Study: CSL Behring doubles MedDRA autocoding and improves coding quality with IQVIA NLP

How CSL Behring Uses Natural Language Processing To Improve MedDRA Coding

CSL Behring, the fifth largest biotech and a leader in protein therapies, faced a time-consuming pharmacovigilance challenge: coding adverse event verbatim reports to MedDRA. Autocoding covered only about 30% of reports (with 70% manually coded), and because CSL Behring works in rare diseases more than 90% of verbatims were unique, making manual coding slow, inconsistent, and resource intensive. CSL Behring engaged IQVIA (Linguamatics) to explore an NLP-based solution using the IQVIA NLP platform and i2e extraction tools.

IQVIA developed NLP queries on randomized training and test sets and ran a blind proof-of-concept, which doubled autocoding from 30% to 60% while keeping mismatches very low; 62% of automatic codes matched prior manual codes, 34% were left unencoded, and only 4% were mismatches—many of which showed NLP provided better, more up-to-date MedDRA codes. The IQVIA solution improved coding quality and consistency, reduced manual effort, and set CSL Behring on a path to integrate a beta tool with ongoing machine‑learning enhancements for continual improvement.

CSL Behring

Martin Menke

Global Medical Coding Lead