Many individuals, particularly in elderly populations, take a variety of prescription and over-the-counter medicines and dietary supplements each day. When taken correctly, many of these medications help patients live healthier, fuller lives. However, when used in combination, they can potentially cause more harm than good.

PRA Insights Team
PRA Insights Team

Polypharmacy is the simultaneous use of multiple drugs by a single patient to treat one or more conditions. Not only can polypharmacy be an additional burden for patients, it can also be dangerous. This is because the more drugs one takes, the higher the risks of a drug interaction occurring.

According to the FDA, “drug-drug interactions occur when two or more drugs react with each other.” Drug interactions can vary from inconvenient to incredibly harmful to a patient, including making a drug less effective, causing unexpected side effects, or increasing the action of a particular drug. For example, taking a sedative with an antihistamine can slow your reactions, making it dangerous to drive a car or operate machinery.

The use of multiple products increases the risk of adverse reactions and even morbidity in older adults. With that in mind, PHUSE and the FDA challenged the healthcare industry to develop innovative approaches for predicting drug interactions.

Our multidisciplinary team of data scientists, machine learning/artificial intelligence experts, epidemiologists, and clinical specialists understood we needed to keep patients at the center of this challenge. We also knew that we could leverage our expertise, data, and technologies to gather evidence that would help fill in the gaps in understanding the true safety profile of treatments and give us a deeper, more holistic view of what happens when patients are taking certain treatments together.

Focusing primarily on geriatric populations, patients undergoing immunosuppressive therapy, and individuals receiving prescriptions from a high number of prescribers, we leveraged machine learning (ML), pattern recognition models built on publicly available data sets, Symphony Health claims, and EHR data to predict and/or identify side effects or adverse events related to drug-drug interactions. Our team applied multiple ML and pattern recognition models to develop our recommendation system, including disproportionality analysis, frequent pattern analysis, neural network, collaborative filtering, and tree models. This system generates possible drug interactions that could produce observable adverse events or side effects as a subsequent diagnosis in the patient’s health record.

This approach allows us to see healthcare transactions happening with patients and evaluate these trends. When combined with our social listening capabilities, we can not only better understand these risks, we can design comprehensive studies that address them head on.

PRA Health Sciences was one of just five teams invited to share their work in this category at the 2020 PHUSE/FDA Data Science Innovation Challenge.

You may also be interested in: