Predictive Analytics Case Study: Lillian L. Dittrick, FSA, MAAA
Since this interview, Lillian Dittrick has taken on a new role as director of strategic analytics at Henry Ford Health System’s Health Alliance Plan.
Lillian Dittrick's strategic analytics team at UnityPoint Health® has developed an application to automatically cull unstructured doctors' notes in electronic medical records.
They use a natural language processer to scan for valuable information, such as past and current ailments and family history and are also working to include social determinants. They then analyze that data to predict patients' medical needs and match them with appropriate, personalized care.
One of the initial use cases is identifying chronic conditions that haven't been recorded in structured fields in the electronic medical records. Their approach will have the added benefit of simultaneously improving coding and billing, which helps keep healthcare costs at an appropriate level given each patients' illness burden.
Dittrick's team is part of a strategic analytics department that did not exist three years ago.
UnityPoint Health®, one of the nation's largest nonprofit health systems, has assembled a team of people who use advanced analytics to build more complete patient risk stratification models by mining claims, clinical and social determinant data in new ways.
Dittrick's team includes an engineer, actuaries, former health insurance underwriters, certified public accountants (CPAs), data scientists and a statistician. The actuaries bring expansive know-how with a combination of modeling, claims and risk management training.
An actuary's skill set is naturally aligned with current and emerging trends in healthcare, Dittrick said, especially as providers move away from the fee-for-service model and take on more risk.
"You have to find someone who understands the data and how to analyze it, recognizes its shortcomings, and is able to communicate the findings," Dittrick said. "Actuaries get that kind of specialized, broad training."
Dittrick's actuarial training drilled into her the importance of accurate and valid data.
She appreciates precision. But she desired a greater connection to results, which is why she began incorporating predictive modeling in her work.
"I wanted to have a more direct connection to bringing value," she said. "Healthcare was appealing to me because you can see the direct impact you have on the patient population. You're driving positive patient results."
While Dittrick does little traditional actuarial work today, she is still involved with the development of regression algorithms and uses statistical and financial modeling techniques. There is, however, a greater concentration on wrapping the appropriate care around patients and identifying variation in practice patterns.
"It comes down to the way you look at things and the way you attack a problem or identify an opportunity," Dittrick said. "It's a whole new field in a way."
Predictive modeling has also given her a more complete picture of patients. Dittrick's team plans to use the natural language processor to identify diabetics who do not have a corresponding diagnosis code recorded.
The team is also working on a length of stay model to help predict how long someone will be in the hospital. Clinicians can then use this information to decrease the amount of variation.
The approach has been so successful that the department has received awards for its blood utilization model, which predicts how many units of blood are needed during transfusions.
"Existing rules and standards for blood transfusions have been in place for years. There are so many things that people are used to doing the same way for so long, they don't realize that there may be better way," she said.
"I feel very strongly that actuaries belong in this field. There is a real and almost urgent need for them to help manage all kinds of risk in the provider arena."