How a father's death inspired a new venture

Using predictive analytics to treat patients
By Dean Sawyer
12:50 PM
Share

Of all the things my Dad taught me during his lifetime, none was more influential to my career than the lesson of his death. It’s what led me to launch a new business with the goal of ending all preventable hospitalizations. And if we succeed, it will be because my Dad’s needlessly early and completely preventable death led me to see the power of Big Data and predictive analytics.

The story of his death and how it inspired the launch of a business begins in 2011, when I attended Singularity University’s inaugural FutureMed executive program (now called Exponential Medicine), an intensive program for physicians, healthcare executives, innovators and investors focused on exploring the impact of rapidly developing technologies on the future of health and biomedicine.  One of the many amazing trends I saw was that advances in consumer and medical grade biosensors were going to make it possible to collect health data outside the four walls of a hospital or clinic and that we would soon be able to collect this data continuously and passively.

A year later I met Dr. Jack Kreindler, a physician and innovator in healthcare and health tech, and alumnus of the second Exponential Medicine program had an idea to create a company to make sense of sensors and other types of remote health data.  Jack had some seed funding from Frost Data Capital, a new type of investment vehicle and incubator of big data companies and asked me to join him as CEO and co-founder. Jack pointed out that as medical sensors become more generally available, the stream of information available to clinicians would completely overwhelm their ability to understand and react and that there was an enormous opportunity to solve this problem.

So using the Lean Startup methodology, we set out to develop a business model around this idea and eventually postulated that ubiquitous smartphones and connected devices, and breakthroughs in machine learning and big data analytics could be used to transform remote health data into actionable information to solve real problems.  But, what problems?  What would be the highest and best use of this technology  -- assuming we could build it?

As we were working through these issues, I often received calls from my sister Kimberly about my Dad, Terry Sawyer, 69, who had congestive heart failure.  He lived in an assisted living facility near my sister, making it easy for her to drop in on him every couple days to see how he was doing.  She also picked him up and brought him to her house most weekends so he could visit his grandkids, eat a home cooked meal and snack on his favorite desert, vanilla ice cream. Because my sister was paying attention, she often detected subtle changes in his health and would call me to collaborate. On one occasion, she noticed that despite being at her house all weekend, dad had not eaten any ice cream. This prompted her to inquire about how he was feeling and he begrudgingly told her he was having some shortness of breath and was feeling fatigued.  I advised my sister to make an appointment with my dad’s cardiologist. At the appointment the next day, the cardiologist explained that my dad was in the early stage of heart failure decompensation but because she caught it early it could be treated with a simple change in medication.

I would never have thought that a change in the consumption of ice cream could predict heart failure decompensation but in this case, it did.  I have since learned that the power of big data is it can often detect patterns that are not obvious, even to trained professionals. 

A couple of months later my dad was not as fortunate.  Kimberley had to go on a 10-day business trip and obviously could not check in on Dad during that time.  When she arrived back in town, she visited Dad at his assisted living facility, took one look at him and called me. She said he looked terrible. His face and ankles were swollen; he complained of shortness of breath, he was extremely fatigued and had not left his easy chair in three days. I immediately told her to call 911. Dad was admitted to the hospital with acute decompensated heart failure resulting in severe congestion of multiple organs by fluid that is inadequately circulated by the failing heart. In other words, because the decompensation was caught so late this time, his organs were failing. Despite very aggressive treatment over the next two months including dialysis and a heroic fight by Dad, he died on July 26, 2012. 

I was angry because I knew his death could have been prevented.  How could the staff at his assisted living facility come into his room twice a day to give him his medicine and not notice that something was wrong?  I learned in the following days that the pathophysiology of heart failure decompensation starts up to 14 days before the patient becomes symptomatic. In discussing this with Jack, he postulated that advances in machine learning and big data analytics could detect these subtle changes in the way my sister did when she was really paying attention, or the way a team of clinicians does as they manage patients in the ICU. Upon further research, I found out my dad was not alone. Chronic disease is responsible for 75 percent of all healthcare costs and one-third of these costs are for hospitalizations and according to the Agency for Healthcare Research and Quality, millions of these hospitalizations are preventable. 

We now knew the highest and best use of the company we were building, which is now called Sentrian. Our aspiration or “massive transformative purpose” – a phrase coined by Salim Ismael, author of “Exponential Organizations” – would be to eliminate all preventable hospitalizations.

It was too late for my dad, but our team was energized by the idea that millions of hospitalizations and an unquantifiable amount of human suffering, not to mention billions of dollars in annual cost, could be avoided if we could leverage the revolution in biometric devices and machine learning to detect health deterioration in patients earlier and with higher accuracy.

One of the newest team members Jeff Haggard, our Vice President of Customer Success told me that his 80-year-old father lived in an assisted living facility and was hospitalized nine times in 2014 for CHF-related acute events and asked if we could use Sentrian to try to prevent additional hospitalizations. 

We enrolled Jeff’s dad in January of 2015, Sentrian has detected early CHF decompensation twice so far.

In both cases, since we could clearly see the change in patterns in his weight, Peak Expiratory Flow, SpO2, and blood pressure, relative to his established baselines, his care team was able to intervene and prevent hospital admissions.  

In the first case, culminating on March 26, 2015, the intervention involved a three-hour trip to the ED for IV Diuretics, but it was clear to the care team that the early detection prevented what would surely have been an admission to the hospital. The second case occurred April 23, 2015 and in this case, his doctor was able to adjust Jeff’s Dad’s oral diuretics and prevent a trip to both the ED and the hospital.

As I remember back to the last moments with Dad as I held his hand and he slipped away, I’m encouraged to know that at this very moment, Jeff is spending time with his father and imagine Jeff holding his dad’s hand while they share a story or a laugh knowing that in some small way, my Dad helped make that moment possible.