Flagler’s AI Software: Less A more I

Can’t teach an old dog new tricks, but an AI can just teach itself

After a recent data analysis with one of our partner pain clinics, we discovered something super cool – our AI diagnosed a rare disease with incomplete information! Beyond pattern recognition, our AI has developed the critical thinking skills necessary to make sophisticated predictions.

.

How did it do that?

We recently signed a large interventional pain clinic and ran their data through our algorithm. When our CTO and I were preparing a report to present to our client, we noticed that the AI was not just making treatment recommendations for MSK diseases that physicians were reporting in the notes but also discovering medical conditions not recorded in the notes. It was diagnosing patients based on other symptoms, physical exam findings, and imaging that were present in the patient’s health records. It was successfully diagnosing patients with a rare pain condition called Complex Regional Pain Syndrome (CRPS). This condition is very difficult to diagnose because there is no lab test for it, and the diagnostic criteria is very subjective and varies patient-to-patient.

At first, we thought it was a mistake. But after reviewing the results with the clinic, they agreed with our findings. Our AI had successfully diagnosed and recommended treatments for a difficult to diagnose condition. To give some context, less than 1% of the US population suffer from CRPS, and even at a pain clinic that specializes in treating patients with pain conditions, CRPS patients only comprise 1-3% of the patient population.

What does this mean?

This finding reaffirmed the importance of using high-quality data as opposed to clearinghouse data for healthcare AI. Ground level clinical data is much harder to get since it requires gaining provider trust, setting up data funnels, and building large language models, which is why many companies resort to clearinghouse data that is less accurate and not comprehensive enough to help the AI gain deep insights. 

Luckily, Flagler has access to this comprehensive and accurate data because it is input by the provider at the point of care. This is essential for the Flagler AI to get smarter and more precise with its insights. The data foundation that we have created and continue to create is paying significant dividends in ways we could never have expected.

Future Directions:

The implications of this finding are significant. It is proof that our AI, once trained on real-time, quality data, formulates insights about a patient’s condition to a degree that we never thought possible. Beyond MSK, what if our AI can start identifying other neurologic issues among patients being treated in pain clinics? This can disrupt how clinical research is conducted and, ultimately, how patients engage with the health system. In the future, patients will be viewed not just as pain or cardiac or neurology patients that need to be evaluated by different specialties; they will be viewed by AI as a unique entity without any restrictions as to what can be diagnosed or treated. Our team is galvanized by the infinite prospects Flagler holds and is excited to bring you along on our journey. 

Fun Flag Facts

There are only 3 national flags that differ on their front and reverse sides

They are:

• Moldova – Only the front displays the Coat of Arms

• Paraguay – The reverse side contains the treasury seal

• Saudi Arabia – has a sword and the Arabic inscription for “There is no God but Allah, and Muhammad is the messenger of Allah.” Since the flag contains sacred writing, the reverse side of the flag is a duplicate of the front, and two flags are usually sewn together

Best of Health,