Google Fitbits May Be Able to Predict Complications in Pancreatic Cancer Surgeries

Fitbits aren't just for tracking steps: A new study shows that they can be used to help predict the outcomes of pancreatic surgery.

Dr. Chenyang Lu, the Washington University in St. Louis professor behind an acclaimed study on detecting depression and anxiety with Fitbits, has won a distinguished paper award after exploring what Fitbits can do for pancreatic cancer patients. He and his team, including first author doctoral student Jingwen Zhang, used data from Google’s Fitbit trackers to predict how pancreatic cancer patients would fare after a complex surgery–if they would have a successful outcome or experience complications serious enough to be readmitted to the hospital.

The team's machine learning model ended up foreshadowing complications far better than industry standard calculators.

The problem

Lu began putting the study together after he went to the WashU School of Medicine to give a talk, he told The Messenger. Dr. Chet Hammill, a surgeon at Barnes-Jewish Hospital who would end up collaborating with Lu on this project, came up to Lu and the two spoke at length about the tough decision pancreatic cancer patients have to make: Pancreatic surgery is extremely high risk, but it’s the only cure for the illness. 

If a patient undergoes the surgery but has complications afterward, they have a higher mortality risk within 90 days after the procedure than they had before it. Patients with successful surgeries, on the other hand, have a longer life expectancy than those who don’t get it at all. 

“We always have this dilemma, this difficult discussion, between the surgeon, the patient, and the family,” Lu said. “It would be helpful if we could have some level of prediction for the risk of this surgery.”

To that end, Lu and his team launched a case study with 61 pancreatic cancer patients. They gave each patient a Fitbit that would track their steps, sleep patterns, and heart rate for a month before the procedure. For reference, The Messenger tested the sleep tracking on the latest Fitbit Charge 6 and found even the free version to be on par with paid tracking reports from other wearable devices that are higher in price. Lu confirmed that the team went with a Fitbit Inspire because it was the most affordable wearable that could still accurately track steps, sleep and heart rate. 

Missing data 

The stakes were high, the case study pool was small, and the researchers admitted that gaps in data were prevalent, writing that 23.95% of all heart rate data in the study was missing. Out of the 1,830 days of data from all patients in the study, 94 days were completely missing and 1,099 days were partially missing.

Lu stated that gaps in data were a “fact of life” with wearables, especially because they relied on users charging them and wearing them constantly.

“We tackled this problem head-on,” he said. “Even if you lose a significant amount of data, you still have a lot of it.”

The Fitbit supplied the researchers with a wealth of long-term, longitudinal data, recording heart rate minute-by-minute over one month. To fill in the gaps, Lu and his team created machine learning algorithms that used steps walked before, after and during the gap to infer heart rate. The algorithms also looked at what the user's heart rate was before and after the gap to arrive at an educated guess.

On a higher level, the team had tools and dashboards to keep track of the data yield from each patient. If a patient wasn’t using their Fitbit, the study coordinator would contact them. That level of monitoring and intervention also helped keep the data consistent, according to Lu.

Step count

Of the 61 patients who participated in the study and underwent a complex surgery to remove tumors, 36 had a normal outcome, and the remaining 25 had complications or had to be readmitted. The complication rate was comparable to the national average. 

Patients who ultimately experienced complications after the surgery had less stable sleep, lower daily step averages and more significant fluctuations in heart rate in the month before they went into the operating room.

With detailed data, the researchers were able to draw correlations; a related paper from Lu's lab isolated daily step counts and found that patients who walked less than 4,300 steps on average were more likely to have severe complications after pancreatic cancer surgery.

Health anxiety?

Could stress from sleep and activity tracking, which must have been new to some study participants, have affected patients? Some studies show that wearables could lead to health anxiety. 

Lu and his team created accounts for patients, so they were, in principle, able to access their health-tracking insights. The researchers also collected data from their end for the study, with consent. 

“I’m not aware of reported cases of stress. In general, patients are willing to help surgical care for pancreatic cancer.” Lu said, pointing to the high rate of patient compliance with the study.

Fitbit as a standard of care

Under the Surgical Prehabilitation and Readiness (SPAR) Program, when WashU surgeons have a potentially high-risk patient, they now give them Fitbit devices and encourage the patient to be more active. 

“We’re really excited by the fact that our department of surgery is using Fitbit now as a standard of care,” Lu announced. “This is not even clinical trial research anymore.”

Other wearable applications

Looking forward, Lu and his team want to use other wearables in research, including the electrocardiogram (ECG) data found in Apple watches.

They’re also working on using Fitbit devices for everything from forecasting recovery for spine surgeries to furthering mental health studies. 

“There is indeed an amazing amount of clinical information that we can extract from the imperfect but longitudinal data that we can collect from people’s daily lives,” Lu said, referring to Fitbit data. “Of course, we need in-person clinical studies to leverage it.”