Cutting Through the Clutter: Helping Doctors Choose the Right Medications
For doctors who treat patients with epilepsy, deciding on the best course of treatment is a complex task. There are hundreds of variables at play, all of which could affect their treatment decision, and the decisions can be particularly daunting.
Doctors at Cincinnati Children’s wondered whether the experience they had gathered could be extended to help other physicians.
“I thought, how do people make decisions? What variables do they use?” says Tracy Glauser, MD. “How do you weight those variables and use those data to support a decision? How do you replicate the decisionmaking process that an expert goes through?”
Glauser joined forces with John Pestian, MBA, PhD, and Alexander Vinks, PharmD, PhD, to understand how drugs and dosages are selected. These questions ultimately led the three to create a decision support system to help clinicians make decisions more effectively. They named the system CHRISTINE, for Children’s Hospital Resource in Selecting Therapy Individualized Expert.
Simplifying a Complex Process
The idea behind the system builds on the pioneering work in advanced clinical decision support that is being done in Pestian’s lab. For CHRISTINE, they used a tool called neuro-cognitive computing (patent pending) developed in the lab. Neuro-cognitive computing replicates the human decision-making process by determining which data exist, which are important, and then identifying the most important choices.
In this case, information is gathered from the clinical, environmental and genetic areas. Doctors take data gathered from examinations and patient interviews and from previous laboratory, genetic, radiology, and other types of visits. They then answer a series of questions about the patient.
While these clinical and environmental factors are essential, CHRISTINE is novel because of the way it incorporates pharmacogenetic factors; it includes knowledge about how a patient will respond to a drug based on the patient’s genetic structure.
The system then considers all the factors – clinical, environmental, genetic, practical — and generates a report that lists the top five drug recommendations. The recommendations are ordered based on the relative weight of factors such as how the drug will be metabolized based on the patient’s genetic profile, drug-drug interaction and the drug delivery method. “This process,” Pestian explains,“ relies on an evidence-based approach by including clinical and environmental factors, and we introduce genetics for personalization.”
CHRISTINE provides a framework for the theory of neuro-cognitive computing developed by Pestian and his colleagues for advanced clinical decision support research. It has been designed with significant attention by Malik Spencer and Pawel Matykiewicz, members of Pestian’s lab, to allow for processing of natural language. For example, the lab is developing an artificial expert that can read the PubMed literature and make suggestions about what variables should be considered in the epilepsy database. In another example, Pestian’s lab is collaborating with Emergency Medicine and Psychiatry to develop a measure for assessing the likelihood of a repeated suicide attempt.
Decision Support, Not Decision Making
Pestian points out that this is not a decision-making tool, but rather a decision support tool. “It’s not, nor will it ever be, our intent to make decisions for physicians,” says Pestian. “That is the art of medicine. Computers cannot replace that art. Rather, in a morass of data, our intent is to provide the right information at the right time.”
CHRISTINE is currently in the pilot stage with two groups of physicians in the Cincinnati area who care for patients with epilepsy and attention deficit disorder. The system will be made available to general physicians following the pilot. Plans are underway to adapt it for a number of other illnesses in the future.
