Using predictive analytics to help patients take and stick with medication | HIMSS 2025

Las Vegas – Even as hospitals and health systems devise new ways to improve patient care, they are often stymied by a persistent problem.

As many as half of all patients don’t take their medications correctly, whether it’s the right dosage or not taking their prescriptions as scheduled, federal officials say. As many as one in five patients don’t even pick up their prescriptions.

Melissa Robinson, population health pharmacy program manager with the Piedmont health system in Georgia, says medication adherence can set back recovery and lead to unwanted complications. It’s especially problematic if patients don’t even start their prescribed medications.

“One of the problems we were trying to solve was to identify patients who maybe are not picking up their medications,” Robinson says. “Because once they miss doses, you can't make that up. You can't reverse time.”

To try and address that problem, Piedmont developed a predictive analytics model for patients with hypertension, aimed at identifying patients who may be more likely to struggle with adhering to their medication plans or simply may not pick up their prescription.

Robinson talked about the program at the HIMSS Global Health Conference & Exhibition and also shared some insights in an interview with Chief Healthcare Executive®. (See part of our conversation in this video. The story continues below.)

• Read more: Thinking about mental health differently, with technology | HIMSS 2025

Piedmont developed its pilot program more than a year ago. So far, the model has demonstrated good results, Robinson says. The program has identified patients who may need additional interventions, and as a result, Piedmont has seen improved adherence to medication programs, including patients on commercial insurance and those on Medicare.

The model analyzes patients holistically from Piedmont’s electronic medical record to predict if the patient will pick up or won’t pick up their medication “so that we can have a proactive strategy to intervene,” Robinson says.

If patients aren’t near a pharmacy or don’t have reliable transportation, Robinson says the system can plan to have medications delivered. The health system can also help patients if they need transportation for follow-up visits with their doctor.

The model examining patients with hypertension found a common denominator for patients who may not get their medication.

“For the hypertension model, specifically, zip code was the number one data feature that helped us to predict if they would pick it up or not, which sort of points to social determinants of health and health equity,” she says.

Typically, patients who are above the age of 65 are more likely to pick up their medications and take them as prescribed, Robinson says.

Younger patients may be more likely to skip picking up their medication or following up as prescribed. Robinson cautions against making assumptions that younger adults aren’t as interested in taking care of themselves.

“What we like to say is that medication non-adherence isn't something that patients choose, it's something that happens to patients,” Robinson says. “So, I think the younger populations are very busy. They are working families, they have children, they have a lot going on. And I think there may be other factors at play than just, you know, the ‘I'm invincible’ kind of mindset that younger people may or may not have.”

There have been some differences with women struggling a bit more with adherence than men, but Robinson says the data isn’t conclusive yet.

Beyond the development of the predictive model, Robinson praises the work of clinicians in stepping up outreach efforts to get patients to take their prescriptions and see what they need. If the outreach isn’t effective, the performance of the predictive model is less important.

“We already have hundreds of success stories. My staff are fantastic about writing up their success stories,” she says.

Even with the encouraging results, it’s impossible to discuss medication adherence without referencing the high price of many prescription drugs. Robinson notes that some patients absolutely want to get their medications, but they simply can’t afford them.

“If they're expensive, then patients are less likely to take them, especially if they have other social determinants,” she says.

Robinson cites collaboration across the organization as being key to the success of the project, including data scientists, clinicians, and leaders ensuring that it was a priority.

“We had a lot of buy-in from the organization that helped us be successful,” Robinson says.