Analysis of client data helps improve weight-loss program

Newswise — Department of Mathematics and Statistics researchers at South Dakota State are analyzing big data to help Profile by Sanford increase the effectiveness its individualized weight-loss program. What they learn will help more members achieve their goals.

“We are finding patterns that will help clients and the weight management team and coaches,” said Xijin Ge, an associate professor whose specialty is bioinformatics and data mining. He is working with statisticians Associate Professor Gemechis Djira and Assistant Professor Gary Hatfield, and Sanford Research senior scientist Paul Thompson, as well as Kurt Cogswell, a professor and head of the mathematics and statistics department head.

The researchers, who began the data analysis in 2014, received $155,014 in funding through a partnership between Sanford Health and South Dakota State University and with support from the South Dakota Board of Regents Research and Development Innovation Program.

“We know the program works,” said Stephen Herrmann, director of program development and training for Profile by Sanford. “Using big data to identify trends gives us another way to individualize the program, to personalize what we do to make sure that person is successful.” That will then allow coaches to test different approaches to help more members achieve their goals.

The research supports the nationwide effort to fight obesity and reduce health-care costs. Approximately 30 percent of South Dakota adults are considered obese, according to the Centers for Disease Control and Prevention.

Obese adults spend 42 percent more on health care than those who maintain a healthy weight, according to stateofobesity.org. Further, a 2008 study estimated that a $10 per person investment in community-based programs proven to improve diet, exercise and lifestyles could save the nation more than $16 billion annually within five years.

Using descriptive analytics

“This is an observational study. We take the data and look for associations,” Ge said.

Using data on more than 33,000 Profile clients collected from 2014 to 2016, doctoral student Valerie Bares, now a senior biostatistician at Sanford Research, identified characteristics and behaviors that may have influenced weight loss.

Members typically lose more than 10 percent of their original body weight and most reach their maximum percentage weight loss after six months on the program. However, by month six, more than half the members have stopped attending coaching appointments. From months nine through 12, members start to regain some of the weight.

“This is where they struggle,” Ge said. “We want to use math to help them stay healthy over the long term.”

Looking at associations, risk-to-event

Bares’ statistical modeling revealed a 140-percent increased risk of dropping out with each percentage point increase in monthly weight gain.

However, Ge pointed out, “those who meet with their weight-loss coaches more frequently tend to lose more weight.” Meeting with a coach one more time per month could result in an average 2.5 percentage points more weight loss for those who weigh themselves monthly.

Nevertheless, by month 12, only 16 percent of Profile members attend one or more coach meetings a month.

Doctoral student Runan Yao is now building on Bares’ work, using four years of Profile member data. “We will dig deeper into the analytics and implement what we find into something actionable,” Ge said.

As Profile expands, Herrmann hopes to use predictive algorithms to identify those at risk for dropping out and then test different approaches to retain those members.

 “The long-term goal is to move from a place where we have descriptive analytics to predictive analytics, which is what we are working on now,” Herrmann said. “Through this research, we can narrow the focus to the most meaningful predictive variables and use this information to help our members be more successful at losing weight and keeping it off.”