When Climate, AI, and Equity Collide: Reflections from the IEEE SA Global Public Health Forum 2025

Last week, I was fortunate to have joined the IEEE Global Public Health Forum, which was an intimate and engaging gathering of health leaders, clinicians, researchers, and innovators. We all shared a curiosity for exploring and discussing how we can harness advances in artificial intelligence to address the threats of climate change, develop greater population-level intelligence and strengthen public health and frontline care. I came away with several insights that feel highly relevant to anyone shaping digital health transformation.

We heard that in addition to the social determinants that shape the health of a population such as housing, employment, education, are the environmental determinants, such as air quality, rising temperatures, water safety that are having a profound impact on health, driving respiratory disease, heat-related illness, and infectious disease patterns.

Artificial intelligence is beginning to redefine what's possible in public health. From early prevention to predictive modelling and population insight, data-driven innovation is turning complex information into actionable intelligence. The examples that follow show how this is already happening in real settings.

UK-Wide Vaccination Analysis

A national vaccination analysis programme led by Health Data Research UK linked 67 million electronic health records, from across primary care, hospital admissions, vaccinations, COVID testing, and prescribing data, accessed through trusted and secure research environments. The resulting analysis identified patterns in under-vaccination, with higher rates among younger males, ethnically minoritised groups, and people with fewer underlying conditions. Understanding patters like this, enables delivery of far more targeted public health campaigns. Public involvement was a critical part of this programme and helped to shaped how findings were communicated. The public members advocated for development of visual materials, and drove policy engagement priorities.

Public participation played a pivotal role in this programme, directly influencing how results were interpreted and shared and the direction of policy engagement. The public members pushed for clearer, more visual ways of sharing insights and helped set the agenda for where influence would matter most and the prioritisation of policy engagement activities.

The Use of Satellite Surveillance Data to Improve Public Health

We learned how satellite data can be used to measure, map, and action interventions at population scale. That's not just technically interesting, it will be strategically essential as climate impacts intensify.

Space East, part of the UK Government's Space Agency, used Copernicus satellite imagery to detect harmful algal blooms in Lough Neagh, which supplies 40% of Northern Ireland's drinking water. Early detection of contamination enables public health warnings and adjustments to water treatment processes.

Satellite-derived street-level urban heat imagery is being used to enable targeted risk mapping during heatwaves and drought conditions. This allows health services to identify and reach vulnerable populations proactively, particularly elderly residents and people with chronic conditions, before heat-related illness escalates, preventing hospital admissions.

NASA MODIS satellite data is combining data regarding rainfall patterns, temperature, and vegetation indices, which is being used to predict malaria risk hotspots. By identifying environmental conditions that support mosquito breeding and disease transmission, health services can target preventive interventions and vector control measures to highest-risk areas before outbreaks occur.

Norfolk and Norwich Virtual Frailty Care

Norfolk and Norwich University Hospital is also working with Space East, and seven GP practices in North Norfolk to deliver remote frailty monitoring and virtual ward care. The programme uses satellite connectivity to enable data movement and predictive analytics in rural areas with poor terrestrial broadband, addressing both clinical need and digital infrastructure gaps simultaneously.

Guy's and St Thomas' Ambient Documentation

Guy's and St Thomas' are using ambient clinical documentation which listens to the clinician and patient during a consultation, producing a structured notes and follow up communications in real time, which are then routed to relevant services and sent to the patient's GP before the patient leaves the building. The system has significantly reduced the administrative workload for clinicians, delivered faster communication across services, and improved the patient experience.

Ministry of Justice Longitudinal Data Linkage

The Ministry of Justice have carried out longitudinal analyses linking justice system data with health, housing, education, and social datasets for women in the justice system. This cross-sector data integration enables earlier identification of health and mental health risk factors which results in more targeted and whole-person approaches to preventative interventions, and supportive strategies.

Throughout the day, presenters also shared challenges they faced in implementing their AI driven projects ethically and responsibly. The barriers discussed across the various presenters were fairly consistent. Here are some of the themes that jumped out to me across the day which should be considered if you're commissioning or deploying AI in healthcare.

We need more integral public involvement

This means more than just running a consultation event and means having a standing and ongoing public involvement in shaping governance with remuneration, protected time, and decision-making authority. We need to fund community organisations to build data literacy so that the public understand AI implications and can engage usefully.

Data Governance and Trust

The experience of Space East, as with so many other organisations implementing innovation in the NHS, alignment with complex hospital systems was a major hurdle. Establishing safe, legal pathways for data sharing is essential but was often time-consuming and protracted.

Yet transparent data use is fundamental to building public trust. If we want to ensure we are creating trustworthy AI use in healthcare, a range of stakeholders including the public, need to collaborate in designing governance frameworks, consent models and communications strategies. This requires more than tokenistic 'inclusion', with resources, remuneration, and real shared decision-making.

System Fragmentation

Fragmentation across global systems is an endemic challenge and use of multiple electronic health record systems across care settings and organisations continue to pose barriers to interoperability, which is needed for effective deployment of AI. While national interoperability standards exist, enforcement remains inconsistent, with a lot of duplication and inefficiency. We need to solve the data interoperability challenge before we accelerate the scaling of AI.

Procurement and Adoption

Procurement emerged as a significant bottleneck. Historical tendering models are not suited to fast-evolving technology deployments. Current processes often delay adoption without improving safety proportionately. We need to demand transparency from vendors, such as training data provenance, safety evaluations, and human-in-the-loop controls should be standard procurement requirements. If a supplier can't or won't disclose these, that's a red flag. We also need shift to greater use of outcome-based contracting, where vendors are held to delivery milestones and outcomes, underpinned by contracts that can be easily terminated if safety or benefit thresholds are not met.

While there is broad consensus on the need for procurement reform, there are very few examples of trusts or integrated care systems deploying modernised approaches such as these.

Skills and Capacity Gaps

The gaps in skills and capacity for deploying AI safely and ethically are considerable across the NHS in both clinical and technical roles. Presenters highlighted that engineering and computer science curricula still do not cover ethics. We continue to train people to build intelligent systems without preparing them to understand the environments those systems will operate within and how to deploy them ethically.

From a clinician point of view, adding increasing data points such as environmental risk alerts, and remote monitoring data to clinical workflows is only beneficial if it reduces overall cognitive burden rather than increasing it. This requires configurable thresholds, summarisation, and decision support that provides rationale as well as recommendations and that integrates with real care pathways. Primary care, already under unsustainable demand, cannot absorb more data without fundamental redesign. Digital tools must lighten workload, not add to it.

We discussed whether AI technologies are advancing faster than our ability to redesign the systems that must use it? Without adaptive workflow design, the flood of new data risks overloading rather than reducing workload.

Cultural and organisational readiness are often underestimated. Anxiety about new technologies is natural, yet we don't often give our workforce the psychological safety to discuss and learn from failure. Staff need protected time to learn and adapt to new workflows and without this, resistance to technology will compound and create barriers to innovation.

IG Governance Maturity

Governance quality varies widely across organisations. We have national frameworks from the CQC, NICE, NHS England, and MHRA which provide strong direction, but local implementation is inconsistent. Some trusts have mature processes and protocols, multidisciplinary ethics boards with public representation for example. Other organisations lack such structured oversight entirely. Strengthening local capability as we move forward will be as important as refining national policy.

Redefine success metrics

We need to judge AI initiatives not just on throughput and productivity, but on quality measures such as disease prevention, equity, patient and staff experience, and workforce sustainability. Equity outcomes, in particular, are not being systematically measured. If AI tools improve outcomes for digitally confident, well-connected populations while excluding others, we risk widening inequity at scale. Evaluations must therefore include mandatory equity metrics and report disaggregated outcomes as standard practice.

Accountability and Evaluation

AI inevitably makes mistakes. Error rates need systematic monitoring, and accountability frameworks must be developed to be unambiguous about who is responsible when harm inevitably does occur. We need to build an AI safety and benefits registry which logs use cases, risk classification, evaluation plan, benefits achieved, harms identified, and mitigations applied. Not every tool will deliver on its promise. Knowing when to stop using an ineffective system is as important as knowing when to scale a successful one.

We need to tell the truth about what's working and what isn't. The NHS has a cultural tendency toward optimism when we talk about innovation. We celebrate launches and pilots. We're less good at publishing evaluations of stopped projects or where the technology failed to deliver. Creating psychological safety to learn from failure is how we get better.

Climate change is already reshaping disease patterns, service demand, and system resilience. AI offers genuine potential to improve how we deliver health services with greater prediction, prevention focus, and coordination.

The day for me demonstrated that we have the expertise and commitment across the system. The work ahead will be translating that into robust governance structures, investment decisions, and implementation practices that deliver accelerated deployment equitably, with transparency, and while bringing public trust along with us. As artificial intelligence moves from research to routine use, the challenge will be whether we can deploy these technologies at scale, responsibly, equitably, and at the speed that health systems and the climate crisis demand.