The NHS operates in one of the most challenging environments since its inception.
The complexity of diseases is increasing and demand continues to rise sharply after the pandemic.
In addition, the population is aging. Hospitals must prepare for a 40% increase in demand by 2035. To keep pace with this development, healthcare spending must increase by 3.3% annually over the next 15 years, according to the IFS.
But it’s not all doom and gloom; the healthcare system is still made up of the best and brightest doctors and administrators. But for these professionals to reach their full potential, they need to be supported by the most innovative and practical technologies.
Companies across all industries are automating their processes with AI. And in healthcare, it is already making waves in diagnostics and contributing to life-saving medical advances. But it is also being used in other ways across the industry.
The Transformer architecture, which powers widely used Large Language Models (LLMs). The Transformer architecture is unique in that it can be used to understand and create text, image, and audio content.
However, this type of generative AI model could be used for much more. As a copilot, AI agents optimized for each problem area can help doctors and nurses manage multiple patients simultaneously, requiring fewer resources and contributing to better patient outcomes.
Solution to the communication issue.
According to a report by the Patients’ Association, more than half (55%) of people have experienced poor communication from the NHS in the past five years, with one in ten saying it has affected their care.
These communication problems have a number of causes – from letters being sent to the wrong address or lost in the post, to patients having difficulty reaching their GP or doctor by phone. It’s no wonder patients feel like their care is behind a barrier.
LLMs used in the NHS can summarise, categorise, transcribe, translate and interpret voice notes. When a patient sends a message, these tools help alert staff to relevant information, so they don’t have to switch between systems and sift through screen after screen to find the right appointment. Simple queries (the most common for booking teams is ‘Have you got my referral?’) can be answered automatically – both by text and voice. This reduces pressure on overstretched booking teams and improves the patient experience by leaving more time for complex calls with those who need them. When the healthcare system embraces digital communication, productivity will increase and patients will feel more supported.
Analysis of electronic patient records.
Transformer-based LLMs quickly adapt to the amount of medical information the NHS processes per patient, on a daily basis. The size of the ‘context windows’ or input is expanded to accommodate larger patient records, which is critical for rapid analysis of medical records and more efficient decision-making by clinical teams.
In addition to speed, these models also have advantages in terms of the quality of the outputs, which can lead to more optimal patient care. An “attention mechanism” learns how different inputs relate to each other. In a medical context, this might include the interactions of different medications in a patient’s record. It can find relationships between medications and certain allergies, and predict the impact of that interaction on the patient’s health. As more patient records become electronic, the larger training sets can make LLMs more accurate. These AI models can do what takes humans hours of manual work – review patient records, interpret medical records and family history, and understand relationships between past conditions and treatments.
The benefit of this system is that it creates a complete, contextual picture of a patient, helping clinical teams make quick decisions about treatment and advice, but it can also go so far as to suggest responses and next steps.
Putting it all together
There are two ways LLMs could evolve. The generalist approach is to create a high-performance model that can understand text, images, and speech and is capable of performing a wide range of tasks (think of the large, consumer-facing LLMs). These may consist of multiple models, often for cost and efficiency reasons, but their goal is to be used for as many tasks as possible. The other approach is to have highly specialized models that excel at a specific function and are connected together by another LLM. In healthcare, the specialized models will prove valuable because they can be fine-tuned using medical data sets and are easier to regulate and approve as medical devices.
LLMs can now understand our intent. This makes them more human assistants and gives them the ability to reason and “chain thought.” This method of breaking a problem down into subtasks works well for healthcare problems when equipped with specialized models that can understand health records, APIs for booking appointments or patient messages, and integration with existing medical software to interact with doctors.
This will lead in the near future to LLMs managing referrals, requesting diagnoses, waiting for results and gathering the available information to present to a physician with a set of next logical steps. Once approved, they will coordinate and arrange further appointments while keeping the patient informed, making them a comprehensive clinical assistant.
Maintain trust and transparency.
Traditional AI models can predict sequences and categorize data, but their success is measured by how often they get it right. Transformer-based LLMs work by predicting an outcome, which of course can be wrong. For the technology to be economically and legally scalable, it needs to have a much lower error rate than humans. This is especially true in healthcare—and it’s important to remember that not all decisions carry the same risk.
AI models are evolving at a rapid pace, and many industries are new to the widespread use of AI. Registered nurses and physicians are aware of the need to challenge AI analysis, both for existing diagnostic models and the new generation of LLMs. With more time and data-backed confidence that an AI copilot can provide, they find it much easier to assess the AI model’s interpretation of the problem.
The priority across the NHS should be to deploy transparent AI models armed with huge amounts of patient data. It’s true that LLMs are only as good as the data they are powered by. But once trained, they have the power to transform healthcare as we know it. The more healthcare systems embrace this technology, the steeper the benefits will be.
About the author
Perran Pengelly is co-founder and CTO at DrDoctor. At DrDoctor, we are shaping the future of outpatient care. DrDoctor is a patient engagement platform that helps the NHS engage patients in their care and increase capacity through waiting list validation and remote management. Our mission is to improve efficiency in healthcare and ultimately enable the NHS to catch up with the huge backlog of patients needing care. How do we do this? We are innovators. Our cloud-based patient management tools automate and virtualise processes and care, allowing doctors to focus on their patients and patients to be engaged as true partners in their care.
