Access our support and expertise:
Mitigating the barriers to implementing AI in health and social care requires more than adopting new technologies; it demands careful consideration of regulation, workforce engagement, patient trust and clinical safety.
In this blog, our Digital Health Lead explores some of the most common challenges NHS organisations face when introducing AI solutions into healthcare settings, alongside practical approaches to help overcome them. Drawing on local examples and projects supported by Health Innovation East Midlands, Dawn highlights how thoughtful implementation can help unlock the potential of AI to improve patient care, support staff, and drive productivity across the NHS.
What is AI and how is it being used in health and care?
Artificial intelligence refers to technologies and systems that perform thinking and reasoning somewhat similar to a human. AI systems can be trained to conduct simple actions much quicker than a human or perform many specialised tasks to a high degree of accuracy.
Common uses of AI in the NHS include ambient voice technology to listen and summarise clinician-to-patient interactions, or clinical decision support tools to collate information on a patient’s health status and offer a clinician a range of diagnostics or suggested diagnoses to speed up care decisions.
The outputs of AI tools can result in faster decision-making or correlating otherwise undetected trends to solidify diagnoses and personalise a patient's treatment plan. These AI systems have huge potential to positively impact both the productivity challenge and staff burnout faced within the NHS today.
How are Health Innovation East Midlands involved in local AI projects?
Health Innovation East Midlands (HIEM) is currently supporting a number of AIprojects in our region, including the Heidi Health ambient voice technology (AVT) and Rapid Health triage tools at Nottinghamshire PICS, MyPreOp+ at University Hospitals of Derby and Burton and AI fracture detection tool RBFracture at Northampton and Kettering General Hospitals.
These projects will monitor a set of key success metrics to determine whether projects meet their expected outcomes and ultimately become business as usual in an organisation. HIEM will utilise the outcomes of these projects to influence the adoption and spread of successfully evaluated solutions across the wider health system.
We are also supporting some promising digital and AI technologies to become ready to tackle key challenges for our local NHS organisations, through the Grow Digital Health Midlands programme.
What are the barriers to AI implementation?
The success of an AI deployment is just as strongly impacted by how the AI is implemented within a health system as the accuracy and quality of the work that it performs. Even the best, most accurate solutions will suffer from an ineffective implementation plan, whether that is through a lack of socialisation with clinical teams, poor data collection to feed an evaluation or a lack of integration.
A paper published by the East and West Clinical Senates outlines the six barriers to AI implementation – ethical, technological, liability and regulatory, workforce, patient safety and social barriers. These barriers must be considered and mitigated to give any new technology deployment the maximum chance of success.
Here, I will outline some areas of consideration that can mitigate common barriers.
Deploying organisations are responsible for ensuring they perform comprehensive due diligence on AI solutions, how they work and how they are trained.
Healthcare organisations need to consider which regulatory requirements need to be in place for an AI solution. Most AI used in a healthcare environment will require both DTAC (Digital Technology Assessment Criteria) and MHRA software as a medical device (SAMD) classification at the correct level to cover the functionality and outcomes it claims to provide.
There is an assumption that clinicians using decision making technologies and tools are ultimately responsible for the decisions they take and the outcomes they record, even when this decision may have been supported, suggested or summarised by an AI technology. Early clinical engagement and clinical safety testing are key to considering the appropriateness of this assumption and to ensure that the correct safeguards are in place to protect and support NHS staff to use the AI tools available to them in the right way.
AI is often defined as a ‘black box’ technology, which suggests that it operates in a way that is often unknown to the user. Organisations deploying AI tools must be satisfied that the AI has enough research evidence for the use case it is intended to be used for, and that it has been trained on a rich and diverse dataset covering every eventuality that it may come across in the real world.
There is huge potential for bias in AI technologies and the ways they are trained. Datasets used to train or test AI technologies can often contain limited data from female patients or do not include them at all. AI trained on these datasets carry a risk that tools will either misdiagnose or undertreat women.
There is also a need to ensure that datasets are ethnically and culturally diverse. For example, we would not want to see an AI technology deployed to detect skin conditions that has only been trained to recognise these on light skin tones. It is therefore important to understand how AI technologies have been trained and tested to ensure that they are equally as effective across our populations.
How and when you involve clinical teams impacts on overall success
Where AI deployments will affect their work, clinical teams must be involved in the project as early as possible to mitigate misunderstanding around clinical risks and job displacement. Clinical champions who are experts in the pathways where the solutions will be used can provide valuable insight into how best to deploy the solution and the issues to avoid, ultimately optimising training and roll out. Clinical champions are also the best cheerleaders for a solution, if they believe in the benefits of a particular product they will be best placed to convince their peers to think similarly.
The onus to check the output of AI systems often falls to clinicians, particularly those which aid with clinical decision making or summarising clinician to patient interactions, it is important that clinicians are satisfied with the accuracy of these solutions and agree with the proposed benefits of implementing them. Clinicians must be satisfied that AI solutions will not negatively impact patient safety and need to feel confident and informed in order to consent their patients regarding the use of AI in their care.
Engage patients and the public to test acceptability and gain informed consent
We have considered how best to engage clinical teams, but it is just as important to consider the acceptability of an AI solution to our patients. We must consider how patients will be informed about the use of AI in their healthcare and give them the right information to provide their consent, using the correct language and level of detail.
For patient-facing AI such as ambient voice technology, it is important that the patient is consented before their appointment is recorded and that they have a choice to opt-out if they are uncomfortable with its use. Patients can be reassured by being provided information on how the AI records and stores their data, for how long and how (and who) can access this. Often, AI notetakers only hold the recording data for the time it takes the clinician to check and approve the notes and/or templated letters to be stored in the patient's record. Knowing this may make patients feel comfortable in the use of AVTs.
For AI solutions that provide a second opinion on a patient scans or symptoms, information could be provided on what the AI is replacing, for example a second human, and how accurate the AI is. Patients may feel reassured to be informed about the accuracy of the platform against a human clinician as determined by research and evidence generation already conducted.
How Health Innovation East Midlands can help
Health Innovation East Midlands have a range of support on offer to local organisations to help identify the right AI based solutions to meet their needs, whilst offering advice and guidance to tackle of a number of common challenges. Our support offers include:
Contact This email address is being protected from spambots. You need JavaScript enabled to view it. for more information on how to access our support.
How to Implement Ambient Voice Technology in Health and Care
Transforming healthcare through AI
Making AI in healthcare work for women
AVID AVT community
The AI readiness checklist
Digital Health Innovation Lead and Deputy Head of Innovation Pipeline Development