Data plus AI = ?

Before we begin today, I’d like to take a moment to mourn the tragic passing of Kobe Bryant, his daughter and the other people on the helicopter. Kobe was more than one of the greatest basketball players of all time and his presence will be sorely missed. Our prayers go out to his family and sports fans across the globe. We have lost an all-time great.

Over the weekend besides being stunned by the Kobe news we read several articles on what else, but data combined with artificial intelligence. Digital health and more specifically digital diabetes continues to suck the air out of the diabetes management room. Everyone and yes, we mean everyone seems to believe that data combined with AI is the answer. That at long last after decades of poor outcomes this trend will be reversed.

Not that we don’t see the POTENTIAL of digital diabetes we are however extraordinarily skeptical that digital diabetes is the answer. As we have noted many times the reality is digital diabetes is NOT a new idea, it’s an old idea updated with new technology. An old idea that didn’t work when patients were called on their phones and given coaching.

Yes, we know smartphones have become an essential part of our lives but whether the coaching comes via a call or text message one thing hasn’t changed. There is a real live human receiving that text message. A real live human who must then act upon that message. A real live human who has NOT made their diabetes management the centerpiece of their very busy life.

To help everyone fully appreciate where data plus AI can work versus where it runs into problems let’s look at hybrid closed insulin delivery systems and Tyler. We shall define a hybrid closed loop insulin delivery system as a system which contains a CGM and insulin pump. Tyler is a CGM, connected insulin pen and app. At the moment we have two hybrid systems the 670G from Medtronic and the Control IQ from Tandem. We do not YET have a Tyler, but everyone is working on one.

What these systems have in common is that data collected by the CGM is transformed into actionable information by an insulin dosing algorithm. The main difference comes with who performs the action. With a hybrid system the pump executes the directions given by the algorithm. Patients using a Tyler must still inject, they must do what the system is telling them to do for the system to deliver better results. Patients using a hybrid system can make adjustments to what the system is recommending but they cannot skip doses.

If there is a flaw with Tyler it’s the patient has to participate, they have to do something. One thing we have learned from talking with patients using either the 670G or Control IQ is these systems when they work as designed make their diabetes management much easier. While their diabetes management never leaves their mind, they do think less about their diabetes management and feel less worry as the system is protecting them from severe hypoglycemic events.

Since there are no Tyler’s on the market yet we don’t know whether patients will follow the dosing recommendations given by the app. There is plenty of empirical evidence they will but until we have an FDA approved Tyler there is no way of knowing for sure.

But let’s step away from insulin using patients and look at patients on a GLP-1. How does data combined with AI help them? Patients using a GLP-1 inject either once a day or once a week. There is no need to monitor glucose levels as GLP-1 are not dose dependent they are fixed doses. There is little worry of hypoglycemia as a GLP-1 only works when it needs to work it is not like insulin which is active all the time once in the body. There is the additional benefit of weight loss.

GLP-1 therapy is as close to stupid as it gets and it’s about to get even dumber. Once the FDA approves the Intracia exenatide micropump patients won’t even inject. Once inserted in the patient, an easy and relatively painless process, the patient does NOTHING. No glucose monitoring, no injections no nothing. About the only worry the patient has is system malfunction, a common worry with any medical device. However baring a malfunction this system is about as stupid as it gets.

Yes, there is now an oral GLP-1 but until it’s been used by patients in substantial numbers, we will not know whether its rather complex dosing regimen interferes with patient adherence. Frankly we think it’s easier to inject once weekly than to make sure these dosing instructions are followed each and every day but as we have said time will tell.

Still the fact is, yes, those pesky facts again, a patient using a GLP-1 has little need for coaching.

So here we are once again looking at the biggest obstacle of every patient coaching tool invented. It does not matter if this tool used a phone call or a way cool whiz bang app. It does not matter whether these patients used an old-fashioned finger stick monitor or a CGM. These patients by the nature of their therapy regimen – pills alone or pills plus insulin – are not highly engaged with their diabetes management. Never have been and likely never will be unless some major changes are made.

All the data in the world, whether its combined with AI or not will change one indisputable fact. When it comes to this group the difference between good or not so good outcomes comes down to a simple fact, are they adherent with their therapy regimen. All the coaching in the world cannot make these patients take their pills. The fact is, yes, those pesky facts again is that everyone and we do mean everyone has attempted to climb this mountain and so far, no one has reached the peak.

Sure we’ve seen some small improvements in a small number of patients over a short period of time. But we have yet to see any real sustainable improvements in a large number of patients. The digital diabetes companies use this fact to say there is a large untapped market for their services. This is what investors are hanging their hats on, that these way cool whiz bang systems will do what has never been done before namely to get these patients adherent with their therapy regimen.

Let’s be crystal clear here we are not saying that these way cool whiz bang systems cannot help. What we are saying, what history has shown is that all data in the world, all the AI on the planet cannot change one huge indisputable fact. That when it comes to achieving real long-term sustainable improvements in outcomes the HUMAN involved must do their part. That without their participation all this data, all the AI doesn’t mean a damn thing.