The Power of Predictive Analytics

In the last article, we explored big data and decision support as a way to help patients ease the cognitive burden associated with chronic disease and do so in a way that leads to an actionable insight.  One approach is to use the data to predict events and present them to a patient in a way that can help them proactively manage their disease.  A proactive actionable insight, if you will.

If you think about it, humanity has been in the business of prediction for a long time.  I will stick to a few relatively recent examples.  Meteorologists have in fact been trying to predict the weather since 1861.  The Old Farmer’s Almanac used to try to do the same and its weather forecasts and planting charts actually date back to 1792!  The sports books in Vegas are powered by some of the most sophisticated statistical algorithms in an attempt to predict the winner, the score and even how specific players will do in sporting events.  Wall Street analysts and traders attempt to predict the performance of stocks, bonds and commodities.  Companies spend millions trying to predict the buying behaviors of their customers.  Today, you can turn on your car and your phone can predict that you are driving to work or about to return home.  Over the ages, just about everywhere you turn there has been someone trying to predict something.

In a world where the art and science of prediction has been around for ages, how come it is only just starting to emerge in healthcare?  Healthcare is a field where the discipline of prediction should be the most mature, but it’s unfortunately quite nascent.

Here is the typical scenario of how things play out in healthcare:  A patient visits their doctor.  The patient highlights something that’s not quite right, or a test signals an abnormality and the process of understanding and treating starts.  This isn’t predictive medicine, it’s 100% reactive medicine.

However, things are starting to change.  Gene sequencing, for example, is starting to highlight our predisposition to certain diseases.  In my recent 23 and Me report, for instance, I learned that I have a high likelihood of developing macular degeneration when I get older.  For me it wasn’t an entirely shocking revelation since both my parents have the disease but, nevertheless, it’s a good start and allows me to think about what I can do now to either prevent or delay the onset of the disease.

But what about chronic disease management?  More specifically, can we use the power of prediction to provide proactive actionable insights so that a physician can better understand a patient, and patient can better manage their disease?  Let’s unpack this a bit.

In the management of many types of chronic diseases, it’s often important to know when and what someone ate; if and when they exercised; when and how much they slept; and if and when they took their prescribed medications.  For physicians, these behaviors, while certainly not all inclusive, inform a better understanding the risks of disease progression, while compliance with proper sleep, nutrition, exercise and medication management are vital for the health of the patient.  

Miniaturization, coupled with an explosion in sensor technology, could help play a role.  If you are wearing a smart watch, it is likely that right now on, right there on your wrist, you have an accelerometer, thermometer, barometer, microphone, GPS sensor, and gyroscope – perhaps even more.  With the advent of these sensors, we have entered the world of real-time measurement.  Activity trackers can tell us if someone exercised, when they exercised and, increasingly, their level of exertion during exercise.  All cycles of our sleep can now be tracked.  Understanding sleep and activity are clearly important pieces in the management of chronic disease.  

These sensors, however, have some limitations.  Other elements associated with chronic disease management are much harder to measure and track, like food or medication consumption.  Today, patients are asked to manually track this information.  This is sub-standard, at best, because it is well known that patients are notoriously bad at entering this information on their own – even when there are apps and other technologies out there that make it pretty simple.  Today, most patients use post-it notes and plastic pill organizers sorted by the day of the week to help remind them to take their medicine.  Yet we expect them to consistently and accurately log these things?  On top of that, we also want them to log that they’ve been sneaking chocolate chip cookies at night?  Unlikely.  Food and medication management are areas where a different approach is needed.  These are also areas where the power of prediction could play an important role.

So how could this work?  Let me give you an example, starting with a glimpse into my daily routine.  Each morning when I wake up, I follow a similar pattern.  After my alarm goes off, I brush my teeth and then usually go to the kitchen and start making coffee.  Once the coffee is done, I sit down, have a little breakfast, take my multi-vitamins and read the paper.  After about 30-60 minutes, I go to the gym.  Once I return from the gym, I drink a protein shake.  I shower, go to work and about 4 hours later, I have lunch.  Let’s stop there.  

With just the data from my smartphone, smart watch or my activity tracker, someone could very easily begin to use predictive analytics to determine, with fairly high probability, when, during the course of the day, I would be eating.  If you combine this with technologies like body temperature sensing, heart rate monitoring, geo-positioning, etc. you can increase the probability further.  Finally, if you combined all this data with a glucose sensor, you could even begin to make some predictions about the nutritional content of the food I was eating and even predict whether what I just ate was about to cause my blood sugar to spike.  

Now, let’s also say that one day the watch also shows that I didn’t exercise – in other words, a deviation from my regular routine.  By using all of the data I mentioned before and some math, the probability of me having my post work-out protein shake on a day when I didn’t exercise could also be assessed.  This could prompt a haptic response on my watch which asks me to confirm whether I drank a shake or not.

Predictive analytics centered around food will prove to be incredibly powerful in the management of chronic disease.  But it doesn’t need to stop with meals.  The same is true with medication management.  Between the watch on your wrist and perhaps your activity tracker, we may be able to predict how compliant you have been taking your prescribed medication each day.  There are technologies today that use the sensors in a watch or an activity tracker to predict if a person has been eating, drinking, or even taking medication.  With this information, again, the watch or phone may be able to prompt the patient to confirm that they have, in fact, taken their medicine – in near-real-time!  Over time, and with more data, the step of prompting the patient to confirm an action may also become unnecessary.  Using exogenous data to determine medication management – without having to rely on the patient for this information – could be a game-changer to help doctors better care for their patients.  Imagine how liberating taking down the post-it notes and throwing away the plastic pill containers could be!   

Today, we are just scratching the surface on predictive analytics and there are many different ways we can take the power of prediction.  The sad reality though is too few are taking advantage of this.  This is a real missed opportunity because many firms have devices and digital content that are rich with data that can make predictive medicine not only possible, but also integral to how we think about healthcare technology.  Companies developing technologies to help their customers with disease management would be very well served to ensure that they are paying enough attention to how they may be able to use data to not only inform, but also predict.  This will certainly help drive a greater degree of value for their technologies and solutions and, ultimately, help physicians better manage their patients, and help patients better manage their disease.