It’s been an intense week. A week straight of 8am to 8pm every day. We are working on the medical simulation software for London hospitals now.
So why are we really pushing this? To understand the motivation, we need to understand how to revolutionise a field of science. One of the clearest examples is Haley’s comment. Before Haley, people just thought that stars and comets were messages from the gods. The word disaster means bad star. Mystics gripped society with wild guesses on what they meant and even collected botched statistics. For instance, the Chinese tried to predict the type of disaster based on the number of tales a comet had as they were logging every appearance. When Haley predicted the path of a comet with Newton’s laws of mechanics (at cutting edge theory at the time), mysticism and theories pulled from shoddy statistics were defeated and society now doesn’t waste it’s time looking for correlations of stars and world events. Astrology became astronomy, and using the math to describe planets helped us understand tides with the moon, weather patterns, and have even achieved space travel.
When a field achieves real mathematical laws, the field is propelled into a new pace of advancement and completely revolutionised. This leads to a concept called “physics envy”. Not all concepts are as easy to boil down into mathematical laws as a path of a comet. As science is measurement of outcome against a theory, properly educated scientists know that achieving a mathematical theory of a concept is the gold standard for not just understanding, but measuring. However, you need good quality data to achieve these mathematical models. Physics envy is when a person or group merely makes up a load of coefficients and assumptions to come up with a theory. They are trying to make a shortcut. Economics famously went through a phase of this.
So, what are we doing? The data around clinical decision making is terrible. A clinician might not write notes on a patient until hours later. Different systems like the imaging system and patient notes don’t talk to each other. Hospital policies and layout also alter clinical decisions such as referral policies, patient load, politics etc affects the data being stored. Frankly, the quality of data you can collect from a hospital is too noisy to see the underlying mechanics. Our system is giving users multiple unique, realistic patients where they can order any test, write notes, prescribe medication/treatments, diagnose, and discharge patients to different destinations. Each of these things takes time for results or effects to happen and the user must handle multiple patients. The moment the user orders a test, we log it. The moment the user sees the obs, we log it. There is a unique chatbot trained for each patient with their past med history etc, and every conversation a user has with the patient…. We log it. Every user has the same hospital layout.
Because of this, we have clean data that is all timestamped. We’ve been trialling this in the German medical schools so a standard session of 70 students gives us data of 345 patients. We can apply game theory mathematics to accuracy of diagnosis. We can see how tests affect the decision making, and what happens over time in a process. Are there certain questions that prevent the need for tests? We can start to mine the data to find out. With this pure math theory, we can then move more into the engineering approach. This is where you get math theory and apply it to real world situations. Of course, the outcome of the theory will not be the same as outcomes in hospitals. This is where you get your coefficients. For instance, pure physics doesn’t consider friction. Engineers measure outcomes, and then apply pure physics. The difference in the real-world outcome compared to the pure physics would be the friction. You then have a friction coefficient tacked on. You can then test different materials to get different friction coefficients, and thus know which materials are the best. With pure mathematical models for clinical decision making, we can then apply them to real world data of hospitals to see how much effect particular environmental factors have on the outcome.
This may all fall flat on its face. Nobody can guarantee a scientific revolution. However, with the software we have built I don’t think any other group is as close as we are in dragging the science of clinical decision making from astrology to astronomy.