This post was originally published on

Medicine as Maintenance

Medicine, at its core, is human maintenance. It works to rectify deviations from health norms, restoring the body’s equilibrium, much like fixing malfunctions in machines to ensure optimal performance. From vaccinations that fortify the body against invasion, to medications that adjust biochemical imbalances, and surgeries that repair physical aberrations, the essence of medicine is to sustain the human body’s functionality and longevity, address wear and tear, prevent breakdowns, and correct anomalies, maintaining intricate human machinery in its optimal state. Or at least that is how it should be, currently the model is too reactive, predominantly dealing with illness and disease post-occurrence. Good diet and nutritional intake is deeply associated with health and prevention of disease yet doctors, to their admission, do not know enough about it. One study  has shown, pooling over 800 doctors and medical students, that most felt their nutritional training was inadequate with over 70% reporting less than two hours.

Digital Twins

As medicine is human maintenance, then it is natural to see digital twins, invented in the realms of manufacturing and maintenance, used in medicine as well. According to The Alan Turing Institute: “A digital twin is a high-fidelity, data-driven representation of a real-world system, which can be used to support decision-making” . In manufacturing, digital twins are pivotal in simulating, monitoring, and controlling the state and performance of machines. They provide a dynamic platform to foresee issues, analyse functionalities, and optimise performance without needing to interact with the physical entity, greatly reducing cost and risk.

Similarly, in medicine, digital twins can act as virtual counterparts of the human machine, enabling healthcare professionals to monitor, analyse, simulate, and optimise human health. They could allow for predictive analysis of medical conditions, personalised intervention strategies, and real-time monitoring of physiological states, mirroring the proactive and analytical maintenance strategies employed in manufacturing.

The Glorious Future of Medicine

One can imagine a future of unprecedented precision, personalisation, and proactive interventions where healthcare transcends the current (mostly) reactive model. Every individual will have a digital twin, constantly updated with real-time data from wearables, enabling instantaneous health monitoring and alerting at the sign of any potential health risk. The integration of artificial intelligence and machine learning with digital twins will predict and allow for prevention of diseases before they even think about manifesting and treatment will be hyper-personalised based on each patient’s individual metrics. For surgery, digital twins would allow surgeons (whether human or mechanical) to practise and refine procedures on the patient’s digital twin before real surgery, reducing the chance of complication and enhancing outcomes.

Current Barriers

So this future, deliberately optimistic, is not here just yet, and probably will not be for a while. Currently, we are witnessing the initial stages of digital twins in healthcare, primarily confined to specialised medical applications and research. What are the main issues – in both performance and adoption? Digital twins, by design, are incredibly data-hungry due to their need to accurately mirror real-world entities. In healthcare, vast amounts of data will need to be connected to create precise, individualised replicas of patients’ anatomies. And even with lots of data, that doesn’t tackle the inherent uncertainty involved in diagnosis and treatment. This will include, genetic information, medical history, lifestyle data, and real-time data monitored by wearables. Currently, however, this level of comprehensive data is not there, and where it is, data is often of poor quality or inaccessible in data silos.

There is also the problem of responsibility. In medicine, doctors are the lead decision-makers and can be held liable for making bad decisions. Therefore, it is highly important to have digital twins that can explain their decisions and give fidelity to their outputs, creating trust so that doctors are able to use them. The day may come when the capability of digital twins is so powerful that their explainability is not necessary or so complicated they are not understandable. But now with so much unknown, we need to build a framework that embraces uncertainty and helps make diagnoses and decisions understandable for clinicians and patients alike.

Healthcare as a System

Digital Twins in medicine have inherited from their manufacturing origins an overly physical perspective, concentrating predominantly on the tangible, biochemical aspects of human health. However, medicine is not merely the biological but a complex socio-technical system encompassing a multitude of intertwined dimensions, including psychological, social, and ethical considerations. As we advance, there is a growing need to transcend this overly physical view and approach medicine holistically, encapsulating the entire spectrum of socio-technical dynamics that influence human health. Which is a little ironic, because now Digital Twins for manufacturing are finding they need to corporate these dynamics.

A truly effective digital twin for medicine should integrate biological and physiological data alongside socio-environmental data, behavioural patterns, and psychological dynamics. This will add to the complexity and difficulty of making predictions. Addressing these challenges and progressing towards a more patient-centred, holistic view of medicine will enable digital twins to evolve as essential to healthcare, reflecting the multifaceted nature of humans and contributing to a more nuanced, individualised approach to health and wellbeing.

Our Approach

At Lone Star Analysis, we build digital twins that aim to capture many of the myriad socio-technical factors in healthcare to make our predictions. Our twins provide explainable outputs to help doctors make decisions, not tell them what to do. We address the problem of poor datasets by working both with the data and with experts to capture any uncertainty in the problem space. This allows us to make usable predictions in the here and now. Going forward, our models work like a torch, shining on the most effective areas for us to next collect data. This allows for a more organic growth of the twin not held back by a lack of data but growing with, guiding, and benefitting from its collection, helping to illuminate the journey to the glorious future of medicine.