SciMl/Differential equations for modeling clinic medicine

After watching numerous lectures by Dr. Christopher Rackauckas, i am trying to develop the theory/understanding of using SciMl/differential equations vs biostats in modeling clinical medicinal problems? Biostats/evidence based medicine seems out dated in todays healthcare environment. We should be able to predict a patients likely clinic course and what resources are most likely needed at the time of initial clinic encounter with rare/complex event monitoring,
Being 30 years removed from differential equations is was hoping someone has a bibliography Of articles on modeling biological/medical modeling with Julia/Sciml.
To the math purist is there an article/blog post on the differences between biostats modeling vs differential equation modeling.


1 Like

For modeling clinical medicine we created and its Pumas tool for nonlinear mixed effects modeling.

I understand that you are probably just playing for now. But I suggest these two articles to see how/why you we should be able to translates into the clinic.


Nice articles, but i am more interested in differential equation based modeling as a better modeling that standard evidence based medicine guidelines. Many of the clinical problems facing pediatrics have rates of less than 1 in a 1000 so i think the guidelines/modeling need to have better resolution.
As an example,
What is the number needed to screen and treat to prevent one case of kernicterus.


Is there a replacement for

We at Julia Computing are working on adding compute capabilities to, a bit like, but with a more seamless transition from desktop to the cloud. Happy to have the team involved to discuss these things further.


1 Like

My fundamental idea is to infuse the front end/practitioner EHR with computational power that Julia provides. The last EHR i worked with worked off of MUMPS which the best i could tell does not have numerics baked into the core of it language although the proprietary code base is hidden.
Examples: it has been known for 20 years that loss of heart rate to heart rate variability is a sign of impending death i have never seen that calculation built into an EHR.
Predicting neonatal sepsis.

Julia seems to be a language that could these calculation with ease and tremendous speed. So building a healthcare calculation engine in Julia seems to be a great idea for a proof of concept.
But being a physician not a great Mathematician or computer scientist finding the right mechanism/ system for development and testing/usage across/regardless of institutional barriers.
Juliabox allowed web based development and easy sharing. Almost all physicians can not install software on our work computers but can used web based tools.

Healthcare knowledge needs to be more open source than the current systems that have monopolize the healthcare industry. My vision is that the same equation that can predict/model global climate phenomenon can predict/model a patient disease course and cost effective treatments and interventions. I am sure a good Mathematician would use the same differential equation infrastructure to build the model and infuse it into the UI of a clinician. A patient regardless of being at a top children’s hospital or in a critical access clinic in rural america, or rural Africa for that matter can have access to AI in their healthcare.

You’ve come to the right place! There are literally more papers than I can keep track of (these 8 are the ones I could remember/find in 10 minutes!) trying to apply some form of neural differential equation to clinical (usually EHR) data.

1 Like

And we’re automated the use with EHR in Pumas, with more information on that coming out soon.