UB pharmacy professor develops AI model to predict hospitalization of at-risk cardiac patients

BUFFALO, N.Y. — When Arinze Nkemdirim Okere, PharmD, MBA, worked as the pharmacist for a hospital in Tallahassee, Fla., he noticed that discharged patients would regularly return, often for issues that could have been easily treated.

“The reasons behind hospital readmissions are multifactorial,” says Okere, who joined the University at Buffalo School of Pharmacy and Pharmaceutical Sciences as clinical professor and head of the Divisions of Outcomes and Practice Advancement in September 2025.

“One big factor is medication adherence issues,” he says. “This is especially true among patients with cardiovascular disease.”

Readmission can be costly to both the patient and the hospital and can put vulnerable patients at risk for infections and even death, Okere adds.

Okere, who specializes in treatments for cardiovascular disease, wanted to find ways to prevent unnecessary readmissions from occurring. From July 2021 to December 2022, he worked with a small team of researchers to develop a machine-learning model (artificial intelligence) that predicted patient readmissions to the hospitals within 90 days with an accuracy rate of 95%.

Their findings were published in the December 2025 issue of British Medical Journal Health and Care Informatics.

Co-authors are Md. Mohaimenul Islam, research assistant professor in UB’s Department of Pharmacy Practice, and researchers at the University of Florida and Florida A&M University, where Okere previously taught.

AI identifies patterns of risk in surveys

They knew that existing tools for hospitalization predominately relied on retrospective clinical data that identified risk only after adverse risks occurred.

Integrating patient-reported behavioral data into electronic health records allows the AI model to accurately identify people at higher risk of re-hospitalization early on.

By recruiting participants from community pharmacies, outpatient clinics and social media platforms, the team ended up with a wide sampling of more than 1,300 adults from across the country with at least one cardiovascular heart factor, such as high blood pressure, high cholesterol or Type 2 diabetes.

From the surveys, AI was able to find patterns in the answers tied to the participants’ risk of returning to the hospital both in a linear and nonlinear manner, Okere explains.

Of the participants, 35% reported at least one hospitalization and 10.4% reported a 90-day readmission. Heart disease, multiple medications, race/ethnicity, employment and insurance status were the most influential predictors of who would return within that three-month timeframe.

Many missteps can land patient back in the hospital

Factors that lead to a patient with heart risk returning to the hospital include medication allergies, missing medications, adverse interactions among multiple drugs and misunderstanding of how to properly take medications.

For instance, some patients who should have been taking medications shown to reduce the risk of rehospitalization associated with heart failure were not prescribed those medications, Okere points out.

“I have seen this mostly among patients living in underserved communities and ones who do not have a primary care physician,” he says.

Adverse drug interactions, particularly in older adults with multimorbidity, can precipitate hypertension, renal dysfunction, arrhythmias or bleeding—common causes of readmission. Related to that is poor medication reconciliation and transitions of care, where discrepancies between inpatient and outpatient regimens lead to omissions, duplications or inappropriate continuation of high-risk medications.

“I recall a patient discharged on an appropriate antibiotic therapy for pneumonia who continued to experience shortness of breath,” he says. “During a follow-up transition-of-care call, I identified a history of mild COPD and promptly collaborated with the physician to initiate oral prednisone. This intervention prevented an unnecessary clinic visit or hospital admission.”

Finally, misunderstanding how to properly take medication is a problem that can land a patient back in the hospital.

“In an earlier study, we realized that some patients didn’t understand that you can’t just take a medication until you feel better and then stop,” he says. “They are more likely to be adherent if they understand that not taking the medication can lead to problems. And once they understood how to properly take their medications, they were less likely to go back to the hospital.”

Future collaborations with area health care providers

“AI has the potential to help identify high-risk patients and prompt timely, targeted interventions to prevent adverse drug events and avoidable readmissions,” he says. “My team and I are making progress toward developing an AI-enabled approach to support this goal, and we plan to establish broader collaborations to bring this work to fruition.”

Okere, who also serves on the adjunct faculty at Roswell Park Comprehensive Cancer Center, hopes to collaborate with Buffalo hospitals and identify patients who have the highest chance of being readmitted.

“Being new to Buffalo, I’m trying to build relationships where we can try to implement what we’ve done and test it using their own electronic system,” he said. “We are hoping that physicians, and even the nursing assistants and triage nurses, can use this AI system to quickly flag a patient who might be at risk of being readmitted. We want to actively be involved in those patients’ care for better outcomes in the short and long term.”