Fighting Diabetes with Data

Over 25 million children and adults in the United States suffer from diabetes. And nearly three times that number, 79 million people, are at risk of becoming diabetic, a condition known as prediabetes.1 In other words, nearly one in three Americans has or is at risk of developing diabetes.

It’s a staggering number with an equally staggering price tag. In 2012, diabetes cost the United States a combined $245 billion in medical costs ($176 billion) and lost worker productivity ($69 billion).2 The presence of diabetes more than doubles an individual’s medical costs; on average, those individuals diagnosed with diabetes spent 2.3 times as much on medical costs as the rest of the general population.

At one health benefits company, doctors are fighting Type 2 diabetes with an unusual weapon: data. The company is using advanced data analytics to pore over billions of medical claims in search of ways to better identify and treat diabetes in their patient population. Today, the company protects nearly 71 million people through its national health network, including more than a dozen health insurance affiliates. The size of the data collected by the company is vast, covering more than 10 million unique healthcare providers and reaching back as far as three years. Think billions and trillions of bytes of data.

Healthcare data growing by a quarter year on year

The health benefits company, like many in the healthcare and pharma industry, understands the value of data analytics. The industry as a whole spent $5.8 billion on healthcare and medical analytics last year, and that spending is expected to grow over 25% year over year to ultimately reach $18.7 billion by 2020.3 The growth of healthcare analytics is being driven in large part by the growth in healthcare data, which is also estimated to grow by more than 25% each year.

Not surprisingly, data analytics is not a new discovery for the company. They have used analytics for years to help improve care and reduce costs, including the treatment of Type 2 diabetes. What the company lacked, however, was truly actionable data. The scale of the data involved in diabetes analysis, which included billions of rows of data, required their data scientists to split the problem across sixteen distinct data analytics servers, and subsequently stitch the results together—a process that took them 22 days from start to finish.

A dynamic patient population and the need to monitor diabetic conditions at frequent intervals meant that the company had to re-analyze its data every month. A delay of 22 days for results left the company little time to act upon data before it became outdated, marginalizing the value of their analytics. Clearly, a better solution was needed to effectively manage diabetes in their patient population.

Ten million people, ten thousand rules, twenty-two hours

Diabetes can lead to a host of complications including high blood pressure, heart disease, kidney disease and blindness. The earlier that physicians can identify patients in a prediabetic state, the longer they can delay the onset of diabetes—by as much as ten years. That results in a better quality of life for patients and lower medical costs for insurers. In addition, health insurers need to verify that patients are actually taking their prescribed medications. According to the National Institute of Health, nearly half of all Americans do not take their medications as prescribed.

The challenge of finding these answers could be summed in two numbers: 10,000 and 10 million. In tandem with consultants, the health benefits company and its partner physicians developed 10,000 unique rules to screen patients for diabetes and prediabetes. These rules identified patients who were either not at risk of diabetes or whose A1C blood tests might show a false positive such as pregnant women, patients who had undergone major surgery in the last two to three months, patients on medications that could elevate glucose levels, and so on. After screening, the company was still left with ten million patients and billions of medical claims to analyze.

In order to perform this analysis in a more timely manner, the company turned to Fuzzy Logix and its DB Lytix solution. DB Lytix offered a much different analytic approach than their legacy analytics system. Instead of moving terabytes of data out of their database and dividing the problem across more than a dozen analytic servers, DB Lytix allowed the company to analyze the data directly in the database without moving anything. Their data scientists didn’t need to change the way they queried their data—the same Structured Query Language (SQL) commands could be employed as before—but DB Lytix changed the way those queries were processed by using massively parallel processing algorithms developed by Fuzzy Logix to divide the problem logically rather than physically.

The results were impressive. The same predictive and care gap analytics that consumed 22 days under their old analytics platform were now performed by DB Lytix in 27 hours. And that resulted in much more actionable information that the health benefits company could use to diagnose Type 2 diabetes and prediabetes sooner as well as identify patients who were not following their treatments and proactively address that behavior earlier.

The Company
  • One of the country’s largest healthcare benefits companies, with more than 70 million subscribers.
The Challenge
  • Accelerate predictive and care gap analytics in the treatment of Type 2 diabetes in order to provide more actionable information
The Solution
  • DB Lytix from Fuzzy Logix, an in-database analytics solution with over 700 unique algorithms
The Results
  • Reduced analytics time from 22 days to 27 hours
  • Identified patients who were not taking their medication and created a medical adherence program to ensure patients were following their prescribed regimens
Sources
  1. American Diabetes Association.
  2. American Diabetes Association.
  3. “Healthcare Analytics/Medical Analytics Market – Global Forecast to 2020,” MarketsandMarkets.

Diabetes Prevention