Certification in all its forms – from gaining regulatory approval to receiving medical device status – is the authoritative stamp of approval when it comes to whether a digital health solution meets patient safety requirements. It’s often the gate-keeper for new solutions to enter the market – and rightly so. By providing a strong indication of a product’s safety, it can subsequently help build trust among users in adopting new technologies. But sadly, it’s not a catch-all. Too often, despite rigorous testing and regulatory procedure, unsafe solutions can still slip through the cracks.
Large-scale med tech recalls – such as the recent reclaiming of 5.5 million Philips respiratory devices – prove that certification, although a crucial element of patient safeguarding, can still be prone to pitfalls. In a world where innovation is happening fast, regulation can’t always keep pace with progress. Take AI as an example: it’s a fast-growing technology already being used within healthcare, but it’s largely unregulated. As a result, it’s often down to innovators themselves to guarantee patient safety and lead the charge on shaping what best and safe practice should look like.
As a machine learning practitioner, I know all too well the dangers of new and emerging technologies if developed or used in the wrong ways. But implemented safely, they have an inimitable power to unlock revolutionary improvements to healthcare delivery right across the globe. To ensure we’re innovating responsibly, and protecting against products doing harm, we must consciously develop solutions with patient safety at the core of their design. This means identifying from day one where the risk factors lie, and ensuring every member of your team is actively working to mitigate these and build in a safe, and effective, way.
Prioritising patient perspective
First and foremost, we must prioritise the patient perspective and work backwards from there. Before starting to design or develop a new solution, it’s crucial to stop and build a thorough understanding of exactly how and where your technology will have an impact on patient experience and outcomes. At every subsequent stage of design this consideration should remain front of mind.
It’s as simple as asking the right questions: How will the patient be interacting with your product? What potential impact could each design feature have on care quality or outcome? How might the product affect different groups of patients in different ways? Could any of the content or insights you’re presenting be misinterpreted?
Asking these questions throughout the development of your solution will enable you to flag any safety concerns early and adapt the product to overcome them. It is essential that this is a continuous process that begins from day one of design and accompanies each stage of development as your product grows. By taking this preventative approach to patient safety, rather than trying to retrofit your design, you will ensure the finished product is safe and ready to enter the market sooner.
Presenting peer-reviewed evidence
Another way of testing the safety of your product, and building external trust with end-users, is through publishing your own research. By presenting the ways in which you have consciously developed your solution to account for patient safety, and having these peer-reviewed, you can hold yourself and your team to account and gain external assurance that you are innovating responsibly. This is something that we have found extremely useful at thymia – having recently published research on the use of multimodal machine learning to detect depression symptoms, we’ve been able to scientifically validate our approach and build trust in the tools that we are creating.
Being transparent
While sharing best practice on patient safety is vital, it is equally important to acknowledge your product’s limitations. There are countless measures we can, and must, use to build safety into digital healthcare design, but there will always be areas which pose a certain level of risk and cannot be as easily overcome. Within AI, one of the biggest of these limitations is the margin of subjectivity that results from the discrete number of data sets on which it has been trained. Bias should always be actively avoided and a diverse range of data used at every stage – but we must acknowledge that it is not a watertight solution.
To guarantee patient safety, it is therefore essential to identify these risk factors and communicate the parameters of new technologies with the clinicians and patients who will be using them. This will ensure that they are armed with the knowledge to use the product safely and effectively, and are able to identify any potential areas for concern before they can have an impact on the patient’s care.
There is no doubt that certification is, and will continue to be, an integral part of patient safeguarding in digital health care. But a conscious effort is needed on an individual level from innovators to self-regulate and build responsible innovation into every stage of their product design. Trust is more than a tick-box exercise, and certification alone for digital health cannot guarantee patient safety. By working together to share best practice, and being transparent in our efforts to build responsibly, we can ensure a safe, innovative and transformative future for digital health.
About the author
Dr Stefano Goria is a machine learning expert and CTO and Co-Founder of thymia, a health tech platform using AI to help objectively diagnose and monitor depression.