Given the clinical processes I described in Article 2, from patient/doctor interaction to tests, treatment and wellness, consider looking very big or very small. What I mean by that is look at the system to either solve one clinical problem from beginning to end or very a specific process that happens frequently. I will give examples of small specific processes in the final article next week and in this article I will do a very deep dive on one specific clinical example. 

If you need brainstorming ideas on 10k + problems to solve, you can review Article 1.

At the end of this clinical example, I will also review three of the important questions to answer from Article 3: Is the problem solvable at this time? Is this solution helpful and is it sellable?

A pulmonary embolism is traditionally thought of as a clot in the lungs. More specifically, it is usually a clot from the deep veins of the legs that travels to the pulmonary arteries. These likely occur much more frequently than we are aware. The small clots go to the pulmonary arteries and the smaller vessels act as a filter to stop the clot from going any further. In one sense, this system prevents clots from going to the brain where even very small clots can cause harm.

So back to the lung. What sized clot is bad? How do you measure the size of a clot? Since it usually starts in a leg vein, it is reasonably tubular in shape but by the time it gets to the pulmonary arteries it may have coiled up, broken into pieces or sometimes even stays looking like a tube. It is possible to see clots in the pulmonary arteries by injecting a special dye in the blood and timing it perfectly so this region of dye is concentrated in the arteries we are trying to see. So if the dye fills the arteries entirely we can presume there is nothing abnormal in there. However if there is a defect, especially if it is tubular in shape, we can be fairly certain that it is a clot.  

We see these clots on a CT scan in slices. As we scroll up and down and side to side, we can see these clots in the pulmonary arteries. Sometimes the clot is so big that the right ventricle is having trouble pushing blood around the clot. 

So there is a spectrum of clot size from very big to very small. If the clot is very big, the body cannot break the clot down before the backup is so severe, the patient dies. Very small and the patient may only be uncomfortable for a short time, if at all, and recover fine while the patient’s body clears the clot. 

The really big and worrisome clots cause the patient to get admitted to the hospital, sometimes specialists try to get the clot out with a procedure. But most are treated with a medicine to prevent future clots and allow the body to break down the clot. Sometimes they are in the ICU. But there is no great way to look at this clinical scenario to assess the size of clot and expected outcome. 

However, this scenario allows us to set up some hypotheses to test. Something like, a single clot, less than 10 ml in volume, in the right or left main pulmonary artery, in an otherwise healthy patient will be hospitalized for 2 days. Or another scenario, a clot that straddles the bifurcation of the right and left pulmonary arteries (saddle embolus) that measures between 12 and 15 ml will result in 4 days of ICU care.

The purpose of this exercise to show the CT scan findings are very important, but using the collected data to make clinical predictions can be even more useful.

So is this solvable? Do the pieces exist to put this platform together?

They do. It is very reasonable to train an algorithm to identify and determine the volume of pulmonary emboli. It is also reasonable to apply the algorithm, retrospectively, to scans that have already happened, and find a few thousand cases. Here is an example of a study that shows an AI algorithm that can find clots. 

It is then perfectly reasonable to use the medical record number from the scans to reference the electronic health record to find the patient’s outcome data. Unfortunately for us, ‘days admitted to the ward or ICU’ may or may not be specific mineable data so there may be some manual data cleaning, but this process is definitely doable.

Is it helpful? Once this loop is created, we can potentially do real time problem solving. Does being more aggressive help? Are there patient populations that respond to one treatment or another? This is the heart of precision medicine determining which is the best treatment for the right patient. Additional resource management, while not front of mind in patient care, is certainly important during times of scarcity as well as addressing the bottom line. If there is waste, it may as well be trimmed. Finally, it can also help providers when communicating with patients’ families and possibly give a more guidance on prognosis.

Is it sellable? This is where the business side of your operation will dig in. What is the addressable market? How many of these emboli occur per month? Per week? What is the current treatment cost? Is the current market serviceable or do you need to be creative? Would self insured systems be interested or large payers? I believe this is sellable, but your due diligence will be much more important in building your pitch deck and presenting to investors.

This is just one clinical entity, pulmonary embolism, and just one specific scan. An entire company could be built around this concept. And there are 10,000 more. 

The next article will dive into smaller pain points in the system and opportunities for improvement as well as benefits to understanding regulatory processes early.