Bart's Health NHS Trust (3 of 3)

A SNOMED driven clinical dashboard system to support the identification and management of diabetic patients with possible foot disease in a large academic health system in east London

EMEA

Artificial intelligence, Clinical Practice, Data analytics, Implementation, Research

Diabetic patients require constant monitoring and foot screening on hospital admission by clinicians to identify injuries that might lead to infection and ulceration and to prevent progression of foot disease. Studies show that the correct management of diabetic foot patients (DFPs) in the UK could lead to a decrease in amputation rates, hospital admission avoidance, and a drastic reduction in the NHS cost of managing patients with diabetic foot problems1. The aim of this work is to develop and deploy a fully automated, API-based dashboard system that supports management of DFPs at Barts Health NHS Trust.

Our algorithm extracts a list of patients from the Trust Data Warehouse who have been admitted to the hospital. For each patient, the free-text clinical notes created in the past 6 months are identified and processed with CLiX, a commercial natural language processing tool developed by Clinithink which uses post coordinated SNOMED-CT expressions to extract clinical terms from unstructured clinical notes. Demographic, clinical, and automated risk assessment data are presented in a secure and customizable dashboard. Preliminary results at the Royal London Hospital in east London showed that, in one month, the dashboard correctly identified 42 DFPs. Of these, 33 received an immediate clinical intervention and 9 had no recorded intervention. There were 7 incorrect identifications, mostly involving gestational diabetes, congenital conditions or non-diabetic patients. Our data-driven clinical tool continues to identify inpatients with diabetes and shows strong potential for improving early clinical intervention for DFPs.

1 - Guest JF et al., 2018.

Description

This project aims to develop a simple, user-friendly dashboard system to support clinicians at Barts Health NHS Trust in managing diabetic foot patients. The system helps identify patients who are currently in the Trust and highlights diabetic patients who may be at risk of developing foot infections or ulcers. The system extracts structured and free text information from patient EHRs, which are routinely exported from Oracle Millennium and stored in the Trust's Data Warehouse. The extracted notes are then sent to CLiX, a commercial natural language processing tool developed by Clinithink which uses SNOMED-CT to identify clinically relevant text and match to cohorts of interest. Patient clinical summaries with previously defined risk stratification obtained from CLIX are then displayed in easy-to-understand charts and tables that staff can personalise. By making it easier to spot issues early, the tool helps healthcare providers improve care, reduce the number of amputations, and save NHS resources.

Scope

SNOMED CT is a standardized terminology system able to identify and categorize clinical concepts and key terms related to diabetic foot disease. This not only improves the quality of data extracted from unstructured medical notes but also facilitates interoperability across different systems in healthcare. We used SNOMED CT because we wanted to extract granular, relevant clinical information consistently and comprehensively from clinical reports. By using SNOMED CT, we are able to align with national and international standards, making our solution scalable and adaptable to various clinical environments and a range of clinical conditions.

How SNOMED CT will be used

We used SNOMED CT Expression Constraint Language (ECL) to define cohorts of patients with diabetes and diabetic foot disease. These expressions are used by the CLiX NLP tool to identify and categorise relevant patients based on post coordinated SNOMED expressions extracted from unstructured clinical notes.

Granular symptom information is automatically extracted by the NLP tool in SNOMED CT format. This level of detail allows us to

* stratify patients based on their risk of developing diabetic foot disease, supporting early intervention and more personalised care pathways
* generate a large SNOMED CT coded dataset used for analytics projects to better understand patients with diabetic foot disease in east London

Finally, confirmed cases identified using the dashboard have the appropriate SNOMED CT code added to the patient's electronic care record by the clinical team. This ensures the information is available for future care and planning decisions, embedding the tool's outputs into routine clinical workflows.