Using analytics to manage diabetes – Allison O’Hair

MIT Sloan Lecturer Allison O'Hair

MIT Sloan Lecturer Allison O’Hair

What do data, analytics, and optimization have to do with diabetes? I believe they have a great deal to do with each other. These operations research tools may be the key to solving this major health crisis.

First, let’s examine the problem:

- About 25.8 million people in the U.S. – or 8.3% of the population – have diabetes today.

- By 2050, it’s projected that one in three U.S. adults could have it.

- In 2010, 285 million people worldwide had diabetes, with that number projected to increase to 438 million by 2030.

- If not controlled, diabetes can cause serious complications, such as heart disease, kidney disease, blindness and amputation.

- Diabetes cost the U.S. $245 billion in 2012 alone.

While there are several factors causing this epidemic, obesity is certainly one of them. Obesity is linked to Type II diabetes, which is triggering the majority of new cases. Back in 1990, no state in the U.S. had an obesity rate greater than 15%. Today, no state has an obesity rate less than 20%, with many states reporting rates greater than 30%. It’s become a national priority to reduce the number of new cases of diabetes and obesity.

Given the severity of this issue, MIT Sloan Prof. Dimitris Bertsimas and I set out to see how we could use techniques in our field of operations research to improve the situation. More specifically, we wanted to determine if tools like data analytics and optimization could be used to create a new and more approachable diabetes management system.

Most patients with diabetes are given a one-size-fits-all approach to meals and exercise. Not surprisingly, patients are often prone to regimen adherence problems. After all, if you don’t like what you’re told to do, you’re less likely to follow the plan. Our goal was to create a personalized meal and exercise plan that is more appealing and, consequently, easier to follow.

Our online system, which we call “LiA” for Lifestyle Analytics, follows a multi-step approach. We try to make it fun and fast for participants to utilize. Here’s how it works:

  1. The system begins by asking patients for basic information like age, gender and height to generate nutritional requirements.
  2. Then it analyzes personal preferences based on responses to carefully selected comparison questions. The preference algorithm combines ideas from conjoint analysis, integer optimization, and robust optimization to learn preferences as quickly as possible.
  3. The system also asks patients questions about things like the time they want to spend preparing meals, how much money they want to spend, and the types of food they prefer.
  4. It models how a person’s blood glucose level will change after eating and exercising.
  5. Taking all of this information together, the system uses analytics and optimization to create a daily meal and exercise plan customized to preferences, attributes and blood glucose measurements.

Currently, there is no other system available that is this comprehensive, combining preferences, modeling of glucose dynamics, and suggestions for food and exercise plans.

While LiA is still undergoing testing, we are planning to make it accessible to the public online as well as through smart phones and tablets. LiA could be used by patients independently, or it could be utilized in conjunction with healthcare providers, nutritionists, and even at senior care facilities.

Operations research tools have an enormous number of applications in the healthcare arena. So far, we’ve only scratched the surface.

Allison O’Hair completed this work as part of her doctoral dissertation at the MIT Sloan School of Management.  A lecturer of operations research and statistics, she is the coauthor of “LiA: An Online System for Diabetes and Weight Management” with Prof. Dimitris Bertsimas.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>