The previous article dove deep into measuring pulmonary emboli as a potential business case and demonstrated an example system of how to track, treat and predict a condition. If you missed it, here is the link.

In this article, I will review some specific, non-clinical pain points and present thoughts on regulatory approval.

There is a chance you have never seen a fax machine and wonder if this is truly a problem in the health system. Sadly, they are still everywhere and it is a problem. 

There are benefits to the reliability and security of the decades old system and the onus is on industry to develop and implement a platform that is not marginally better, but 10 or 100 times better to get rid of faxes and their machines. 

In addition to the obvious waste of paper, every data point on that paper is not captured. Granted natural language processing (NLP) and optical character recognition (OCR) can gather much of the data when used properly, however the error rate is obviously higher than if the data was transmitted directly.

This is a segue to another legacy system of pens, paper and clipboards in the waiting room. Over the past few years kiosks and tablets have been making an entrance, but we still ask patients to remember old information and ask for it many times. With widespread smartphone use, there will be a better way to collect, review and transmit from the ‘patient side’ to the ‘health care side’.

As patients in the health care system are becoming much more skilled at smartphone use, digital collection of their data should follow. And it doesn’t need to be in a waiting room. If the patient was able to enter data the evening before, it could be processed prior to the virtual or in person clinic visit. Leveraging voice recognition to answer clinical questions is one more step in the right direction and more still is capturing public facing data to assist in creating a complete digital health predictive model. Specifically there are marketing companies that predict our purchasing behavior and clinically leveraging that data could be useful to predict outcomes.

To be clear, there are many companies working on almost all of the above, but it is not widespread and opportunities persist.

The final point for this series is positioning yourself and your team to simultaneously create a safe and useful platform as well as prepare for regulatory scrutiny.

It may seem premature to think about the regulatory approval of your new platform when you begin, but that would be wrong. There are safety and quality measures that will need to be documented eventually so creating those systems at the beginning is the best use of your time and resources.  

If you don’t know about FDA approval let’s begin now. Is your new platform going to need approval? This FDA site is a good place to start to find out. You can find a list of all of the 510k approved devices and platforms here. Also listed are the FDA approval letters and I recommend reading many of them to get an additional sense of what is required. 

Beginning with the end in mind is helpful on another front too. Your customers are experts and have high expectations. Dr. Harvey has been preparing the radiology community and has 5 questions you should be prepared to answer and the performance of your claim is critical. Measurements like sensitivity and specificity and the accompanying confusion matrix are expected to be current and defendable. His article is here to dig in more.

By considering these points and doing proper research, you will be prepared to present to investors an addressable market, as well as return on investment data. By including regulatory steps early, you can show competence during a rigorous due diligence process.

Together, entrepreneurs, medical providers and the investing community can collaborate to create useful and sustainable products to move closer to a more widespread use of AI to make a positive impact on the care of our health.