How digital twins can drive real change for the healthcare sector
Modern healthcare has myriad opportunities from digital technologies, including command centres to optimise flow, AI and machine-learning to support human insight and decisions, IoT and connected devices to provide real-time data, and robust digital infrastructure to manage and secure large dataflows. Bringing much of this together is the concept of creating digital twins and using virtual technologies in healthcare to drive real world value.
Real-world Value of Digital Twins in Healthcare
A digital twin is a virtual replica of a physical object or process that learns and evolves through simulation and feedback from its physical counterpart. It deploys AI and machine learning alongside dynamic system modelling, and is applicable in both healthcare and life sciences environments. A digital twin creates the opportunity to model and predict the impact of interventions, pathway changes and operational improvements on systems before implementing them, to maximise benefits and minimise risks.
Such a simulacrum creates the opportunity to: test scenarios to predict impact and aid decision making (e.g., in system design and patient treatment); identify inefficiencies, bottlenecks and opportunities and model benefits/disbenefits (e.g., in process improvements); automate responses and decision making (e.g., in environmental control); and increasingly enable testing in a virtual environment (e.g., in silico research – both US and European regulators are exploring use of such “digital evidence” in approvals for new medical drug and technologies).
This digital twin can operate at different levels: as a replica of a whole system or organisation, an organism, or of a building.
Practical Applications of Digital Twins
For example, at a system level, Mater Hospital, Ireland, created a digital twin of its radiology department to test different scenarios and effectively allocate capacity to meet demand. This reduced waiting and turnaround times by up to 24 minutes, improved machine utilisation by up to 32%, and reduced the time needed for scans by up to 50 minutes per day.
Hamilton Health Sciences, Canada, developed a digital twin for their patient flows through hospitals, resulting in a 9X improvement in patient flow. And Leeds Teaching Hospital created capacity for hundreds of additional operations with its existing theatre capacity by dynamically linking theatre scheduling to waiting lists.
For an organism, Heidelberg University Hospital, Germany, is developing a digital twin of the human heart, using AI to model the electrical and physical properties, as well as the structure of the heart. The model is now being tested for its predictive power in treatment of heart failure. The Living Heart project is developing a similar twin, a 3D model of the human heart intended to support in silico trials.
And in the built environment, Frasers Tower, Singapore, utilises a digital twin and real-time sensors to optimise space utilisation, air conditioning and lighting. This has improved energy efficiency while ensuring a comfortable place to work.
These examples all result in significant benefits and reduce the risk involved in intervention, treatment or change by enabling virtual testing before real-world decisions.
However, such digital twins in healthcare do have significant requirements. They need a large dataset of the subject (e.g., the virtual human heart required 250m images, reports and datapoints to generate); in an open system this needs to be supplemented by real-time, real world data, supported by physical sensor technology (e.g., Frasers Tower utilises >1000 sensors to provide real-time data in a relatively closed system). From this, sophisticated AI and statistical and mechanistic models are needed to both learn the system and offer insight and outputs, supported by the digital infrastructure and processing power that fuels this. Digital twins then need to be asked the right questions and test the right hypotheses to generate useful, valuable answers. Finally, and critically, the owners of the twin need to be able to implement and realise the changes hypothesised, as the examples above were able to do.
All this suggests digital twins offer a significant opportunity to improve healthcare, from patient treatments, operational and system improvements, and the management of the built environment. Where organisations are already making large infrastructure investments, such as the UK’s new hospital programme, there is an even bigger opportunity to embed the infrastructure needed and develop the digital capabilities to maximise the benefits of digital twins.
Given the expected benefits, there would be a case for every hospital to have three suites of digital twins: operational twins for process improvement, buildings twins for managing the environment, and human twins for research and treatment support.
To achieve this, healthcare organisations should invest in their enabling infrastructure (data supply), build their organisational digital capabilities (data demand), and ensure that any digital twins are clinically-led to ensure maximum benefits to patients. These are critical; a digital twin in isolation will not deliver benefits without the supporting environment and infrastructure it needs.
Organisations need to invest in enabling digital infrastructure that underpins data-rich, insight-driven organisations, and in particular, ensuring a robust data supply. Many hospitals still rely on semi-manual data collection and analysis, which can be automated to offer real-time data and both retrospective and prospective analysis. There is a spectrum of solutions available, but this can include remote monitoring and sensor technology, cloud storage, holistic data collection, and robust connectivity and security standards. This is aligned to the infrastructure that supports most modern healthcare technologies; digital twins (and associated machine learning, artificial intelligence and simulation algorithms) then need to work effectively with this infrastructure to provide value.
In parallel, they should build the organisational capabilities to run digital twin models and implement the insights, including developing high levels of organisational digital maturity, executive involvement and ownership, and effectively delegated decision-making, which may be automated. This generates the data demand to ensure effective use of a digital twin. Similarly, this can be developed in parallel with the enabling infrastructure and will bring benefits from other digital investments (e.g., electronic patient records, virtual care and digital insights) as well as enabling digital twins.
Any digital twin must be clinically-led. This includes designing twins around the needs of clinicians, empowering clinical decision-making and research, and then supporting clinicians to embed digital twins into their practice. In this way, clinicians can be empowered to utilise digital twins to the benefit of their patients.
These measures will bring benefits before implementing digital twins, as they are critical to being digitally mature and data-driven organisations. They can then build the platform that digital twins can be placed upon; digital twins offer greatest value in organisations that are already digitally mature and have both the digital infrastructure and the change capability needed to realise all the advantages from such a tool.
All organisations can make steps along this road by incorporating digital twins into their digital strategy and investments, realising benefits along the way.