Gloucestershire Hospitals NHS Trust detects 66% of long-stay patients with AI risk stratification tool

Gloucestershire Hospitals NHS Foundation Trust serves a population of 660,000, and 4% of all admissions become 'long stayers' (21+ days), consuming 34% of all bed days. Long stayers face an 11% mortality rate (vs. 5% average), a 23% risk of readmission, and severe deconditioning — yet many extended stays have no medical necessity and could be prevented with early intervention.

ACE partnered with Polygeist to develop an AI-powered long-stay stratification tool trained on 460,000 anonymised patient records. The model generates an immediate risk score at initial data collection, visible to reception and clinical staff. Clinicians can then act early — avoiding catheterisation, selecting a different ward, or immediately referring high-risk patients to geriatricians or physiotherapists. The PoC was delivered in 12 weeks and subsequently integrated with the trust's electronic health record system via APIs.

The tool detected 66% of long stayers within the highest-risk categories. A single-day reduction in average length of stay yields £1.7 million in savings for Gloucestershire Hospitals alone, giving the tool significant economic as well as clinical value. Following the PoC, the tool moved into a closed Alpha phase and was tested against Covid-era datasets with continued high accuracy.

• AI trained on routinely collected admission data can reliably stratify long-stay risk at the point of first contact, enabling truly preventive clinical decisions.
• Even a proof-of-concept with a 66% detection rate in the highest-risk tier creates material financial upside: £1.7M per day-of-stay reduction for a single trust.
• EHR integration via APIs is essential to operationalise the risk score — without it, the model insight never reaches frontline staff.