Natural Language Processing models to determine the context of references in free text data and assign clinical codes (SNOMED, ICD-10 and others).
Not intended for use as a medical device.
CogStack works with all text data in any format, including electronic records, Word documents, plain text and PDFs
CogStack's NLP models output standardised clinical concepts (SNOMED CT, ICD-10, etc.). Further CogStack tools can update to a common database and can write-back to the EHR.
Models are cotinually validaed and fine-tuned at across a range of clinical concepts. Common commorbidities, symptoms, findings, procedures and medications are validated to over >.95 F1 score.
CogStack NLP models have run over millions of records and extracted / contexualised billions of distinct concepts. Outputs have improved in-patient coding accuracy by 25%; reduced time taken for a patient medication review from 2 hours to 5 minutes.
CogStack increases clinical coding productivity by 25% and can detect $100,000s in unbilled activity.
CogStack has been deployed in healthcare settings in the UK and around the world. Models are locally validated using our human-in-the-loop interface, attesting to the flexibility of the solution.
CogStack has published more than 100 papers. See our research archive for links: https://cogstack.org/publications/
