CogStack
Talk to Our Team

Resources

Case Studies

How Healthcare providers and research teams put CogStack to work.

Critical care planning

Flagging 9 in 10 unplanned intensive care admissions before elective neurosurgery at UCLH

How University College London Hospitals used CogStack AI to read pre-operative notes and spot the patients most likely to need intensive care after surgery.

89%Unplanned ITU admissions flagged from pre-operative notes
Mental health

Mental Health Records Made Usable: Medication Reviews Cut from Hours to Minutes

Mental health clinicians spend hours piecing patients’ stories together from notes. What if the whole picture was one click away?

10–20 minper medication review, down from 1–2 hours
Trial recruitment

Finding Septic Shock Trial Patients Sooner with Clinical NLP with 90% sensitivity

In critical care trials, enrolment windows can be just hours long. Can NLP on ICU records flag eligible patients in time or earlier than manual screening?

90%sensitivity against the trial’s manual screening
Clinical coding

AI-Assisted Clinical Coding Recovered up to £5.5m in Missed Outpatient Activity

The procedures were done and the notes were written, but the codes never arrived. Here’s how one London Trust closed the gap.

Up to £5.5madditional outpatient income identified, FY 2023/24
Health equity

Health equity in clinical trials: revealing who heart disease trials leave out

How King’s used CogStack AI to measure health equity in its clinical trials, and found women, minority and deprived patients underrepresented.

36% vs 21%Women among heart disease patients vs trial participants
Real-world data

From clinic letters to a real-world cancer dataset: AI extracts head and neck cancer data

How Guy’s and St Thomas’ NHS Foundation Trust taught CogStack AI to read oncology notes, for real-world cancer research.

2,895Head and neck cancer patients in a new real-world dataset
Deep phenotypingReal-world evidence

Uncovering hidden heart failure: AI phenogrouping reveals early and late-stage HFpEF

Almost 9 in 10 patients meeting HFpEF criteria at King’s had no recorded diagnosis. Here’s what CogStack AI found when it looked.

10.4%HFpEF patients with a clinician-recorded diagnosis
Clinical data platform

Turning 44 million+ clinical records from 3.4 million patients into a hospital-wide clinical intelligence at UCLH

Most of a patient’s story is written, not coded. See how University College London Hospitals made all of it usable, from Epic notes to scanned archives.

44M+Clinical notes, letters, imaging reports and archived records processed
Real-world data

Epilepsy-related real world data from neurologist letters using large language models

How King's College Hospital used local AI to read patient letters for disease registries

0.92F1 for current medications (experts: 0.95)