Version: 1.0.0 | Published: 2 Nov 2026 | Updated: 0 days ago
Documentation
Description:
Antimicrobial resistance, or AMR, can stop antibiotics from working properly. This means patients can take longer to recover, and in serious infections patients may have a higher chance of dying. This study uses AI to predict which patients with infections are at risk of AMR. This approach will rapidly scan and check detailed medical records from hospitals and GPs to help select the best antibiotics for each individual patient.
The focus is on two common infections:
1. Bloodstream infections, which can be life-threatening. Sepsis, which is closely linked to bloodstream infections, kills 48,000 people in the UK each year.
2. Urinary tract infections (UTIs), which are very common and lead to millions of GP and pharmacy visits each year.
Firstly, to develop models for the most serious group of infections, bloodstream infection. There are 48,000 deaths from sepsis annually in the UK. Currently AMR means as many as 30% of patients with one of the most common and life-threatening types of sepsis (Gram-negative bloodstream infection) receive antibiotics that may not work when they are initially admitted to hospital. In contrast, the proposed approach will ensure as many people as possible receive treatment that works first time. Patients with cancer are at increased risk of harm from bloodstream infection and AMR, as are those with immunosuppression and diabetes; these patients could benefit most from these tools.
The second focus will be urinary tract infection (UTI), addressing a condition that results in approx. 3 million primary care consultations per year.
By identifying patients at low risk of AMR who can be safely given narrow spectrum treatment across both applications, these models could also substantially contribute to reducing the spread of AMR and addressing National AMR Action Plan targets.
A major barrier to the implementation of healthcare AI tools will be addressed by demonstrating the consistent performance of our models over time and across different settings through validation of the bloodstream infection models.
Coverage
Spatial:
West Midlands
Typical Age Range:
4-88
Follow Up:
Other
Provenance
Origin
Purposes:
Care
Sources:
- EPR
- LIMS
Collection Situations:
- Clinic
- Prescribing - Hospital
- Secondary care - Accident and Emergency
- Secondary care - ICU
- Secondary care - In-patients
- Secondary care - Outpatients
Temporal
Accrual Periodicity:
Static
Distribution Release Date:
10 September 2026
Start Date:
01 January 2015
End Date:
31 December 2025
Time Lag:
Not applicable
Accessibility
Access
Access Rights:
Information Governance and Ethics - West Midlands Secure Data Environment (https://westmidlandssde.nhs.uk/information-governance-and-ethics)
Access Service:
Data Request Process - West Midlands Secure Data Environment
(https://westmidlandssde.nhs.uk/research/data-request)
Access Request Cost:
Please email wmsde@uhb.nhs.uk
Delivery Lead Time:
Variable
Jurisdictions:
GB
Data Controller:
University Hospitals Birmingham NHS Foundation Trust
Data Processor:
University Hospitals Birmingham NHS Foundation Trust
Usage
Data Use Limitations:
Research-specific restrictions
Data Use Requirements:
Project-specific restrictions
Resource Creators:
This publication uses data from WMSDE, an ethically approved NHS Secure Data Environment (NRES Reference 24/YH/0022)
Format and Standards
Vocabulary Encoding Schemes:
- ICD10
- OPCS4
- SNOMED CT
Conforms To:
LOCAL
Languages:
en
Formats:
CSV