Digital Health Coffee Time Briefing ☕

Original article here

North Cheshire and Mersey NHS Foundation Trust is inviting technology providers to an Elevate showcase at Warrington Hospital on 27 November. The newly-formed trust wants staff to explore AI, automation, remote monitoring, connected devices, virtual care, patient access and workforce-productivity tools, with feedback feeding into its digital and clinical strategies.

☣ NHS England’s cyber team has issued a high-severity alert after Citrix reported two actively exploited zero-day vulnerabilities in NetScaler ADC and Gateway. Health organisations are being urged to perform a compromise assessment before patching, upgrade affected systems to fixed releases, and report evidence of compromise to the NHS England National CSOC.

The Allergy Centre of Excellence (ACE) has launched its new Allergy Assist service. Allergy Assist is an app-based support service, led by specialist nurses at ACE, which provides families with direct access to expert guidance, supporting the management of food allergies, eczema, asthma and allergic rhinitis.

Medway NHS Foundation Trust has begun using a digital system to manage fire safety information across Medway Maritime Hospital, with the aim of making it easier to identify risks and prioritise investment. The system, BIS Lens, developed by consultancy BIS Consult, brings together fire risk assessment data from multiple sources into a central digital format for the trust’s Fire Safety Group.

NHS England has outlined plans for an £838.7m Digital Primary Care Open Framework covering foundation GP electronic patient records (EPRs), clinical and operational software, and research and development. The procurement is expected to begin in November 2026, with the framework going live in 2027.

? Did you know that?

Leakage and Interpretability in Concept-Based Models proposes an information-theoretic framework for testing whether supposedly explainable AI is genuinely transparent.

The document, published on 9 September 2026 by Enrico Parisini, Tapabrata Chakraborti, Chris Harbron, Ben D. MacArthur and Christopher R.S. Banerji, outlines how concept-based systems can still hide extra information inside clinically meaningful features. As the paper puts it: “Models exploit unintended information encoded within the learned concepts.”

The authors introduce concepts-task leakage and interconcept leakage scores, and show that these measures predict model behaviour under intervention better than existing alternatives. This matters for the NHS because clinicians must be able to understand, challenge and safely oversee AI recommendations, they state.

A model that looks interpretable but relies on hidden signals can create misplaced confidence, particularly in diagnosis or risk prediction, so leakage testing should form part of technical assurance and post-deployment monitoring, the document adds.

What we’re reading

New research has evaluated detection of apoptosing retinal cells (DARC) as a predictor of directional macular atrophy (MA) growth in age-related macular degeneration (AMD).

Published on 23 September 2026 by Wei Wei, Panos Gartaganis, Jonathan Young and colleagues, Predicting directional macular atrophy progression using angular plot analysis of DARC reports an exploratory study of seven eyes followed for an average of 48.4 months

The authors wrote: “In this small preliminary study, DARC appears to provide the strongest, most consistent predictor of directional MA growth.

“Angular plot analysis provides a robust approach for evaluating progression and supports DARC as a promising biomarker for identifying patients at high risk of MA expansion.”