Clinical Research Coordinator

University Health nurse highlights Overdose Awareness Day. Posted: Aug 31, 2026 / 10:14 PM CDT Updated: Aug 31, 2026 / 10:29 PM CDT KANSAS CITY, Mo. - Monday marks the 25th anniversary of Overdose Awareness Day. The latest data suggests that prevention and awareness are working. Deaths from overdoses have been going down nationwide. This is also the case in Kansas and Missouri. "We've had a reduction of lives lost to overdose last year and this year, which is excellent; it's showing we're making some progress," Kris Trigoso, a mental health nurse practitioner with University Health, said. There has been a steady reduction in overdose deaths over the last three years in the United States. In 2024, 79,000 people died of overdoses, with 54,000 of them linked to opioids. "We attribute it to an increase in education in the community and increased awareness of the problem," Trigoso explained. "And an increase in access to naloxone, which is the opioid reversal drug that's saving a lot of lives." University Health has added an entire building dedicated to drug use treatment. This exemplifies the increased awareness of the problem in Kansas City and ways to treat it. The leading cause of opioid deaths is fentanyl, which frequently happens without a user's knowledge. Users are often exposed through their drug supply. "Those people don't have a tolerance to opioids compared to someone who has a use disorder," Trigoso said. "When they're accidentally exposed, they're at very high risk of overdose and death, which is scary." The best defense for an overdose is naloxone, also known under its brand name, Narcan. It is available over the counter, and even free in vending machines in Jackson County libraries.

Pharmacy technology report. AUGUST 2, 2026 AI emerges as powerful PA and 340B compliance tool. University Health is using artificial intelligence to transform two of pharmacy's most labor-intensive functions: navigating prior authorizations (PAs) and ensuring 340B compliance, one of the program's chief architects noted at the ASHP Pharmacy Futures 2026 meeting, in St. Louis. The PA initiative, developed at University Health, in Kansas City, Missouri, has cut PA review times by more than half, reduced payor decisions from days to hours, and will soon enable its 340B team to continuously audit 100% of claims instead of a 10% sample, creating what leaders described as a more efficient, compliant, and audit-ready operation. For the PA effort, "we wanted to focus on denial prevention, not denial management," said Kristi Stice, PharmD, MBA, BCPS, DPLA, the director of outpatient pharmacy services and 340B compliance at University Health. "I want the denials team to go away because we've done it right the first time." A frequent driver of denials is the payor's "unasked" questions that are not listed on any form, Dr. Stice noted. "You don't learn what these questions are until you receive all the denials," she said. The tool now banks those questions, so technicians are prepared to answer them up front, she explained. Before centralizing the work, PA submission consumed about 33% of specialty technician time and was split across nursing, revenue cycle, and pharmacy. In one endocrinology clinic, nurses were handling PAs one half day per week. "That was so scary to us," Dr. Stice said, with patients left waiting on claims status. About 65% of the specialty team's PAs were approved at baseline, and 27% went to pharmacist-written appeals. After the AI tool went live in late October 2025, specialty PA review fell from 10 to 15 minutes to about 3.5 minutes, community retail to 5.8 minutes, and buy-and-bill to nine to 12 minutes. The software reads the chart and proposes answers to each form's questions for technicians to confirm. "Our team members are now data validators and not data hunters," said Christa Tetuan, PharmD, BCACP, University Health's ambulatory specialty operations manager. Turnaround, measured from submission to payor determination, dropped as well, falling for specialty PAs from two to three days to about a half day, she noted. When deploying AI, start where the payoff is clearest, Dr. Tetuan suggested. "Our highest return on investment in our specialty pharmacy PAs. So that's where we began." Community-administered (retail) and clinic-administered (buy-and-bill) PAs were phased in later, Dr. Tetuan noted. Consolidating submission onto a single platform and replacing the many insurer portals and passwords technicians had juggled also lifted capacity. PAs submitted per team member rose after the go-live date. "We are able to do more work with the same amount of people," Dr. Tetuan said. The tool also made cross-training easier, she said, since technicians no longer need to memorize each specialty's PA questions. Dr. Stice said the tool paid for itself with more approvals of specialty medications. "That is revenue for the organization." Always 340B Audit-Ready University Health has also been applying an AI tool to help its small 340B Drug Pricing Program compliance team audit every claim instead of a 10% sample without adding staff, Dr. Stice noted. A compliance coordinator and three analysts currently audit about 10% of the system's roughly 325,000 claims per year. "We set a very lofty goal with our 340B team. We wanted to go from that 10% to 100%," Dr. Stice said. "So, we should have to add 10 new team members, right? We can't do that in our safety-net organization." 5 best practices for applying artificial intelligence. * Categorize Before You Implement. Determine whether the tool is predictive or generative AI and assess its distance from patient care. * Evaluate the Right Outcomes. Predictive AI requires performance metrics; generative AI should be judged by workflow, quality, and operational outcomes. * Scale Safety to Risk. Clinical decision-support tools require far more rigorous oversight than administrative applications such as email drafting. * Train Users, Not Developers. Most pharmacists need AI literacy - not coding skills. Focus on appropriate use, limitations, and risk awareness. * Keep Humans in - and on - the Loop. Humans should review recommendations, make final decisions, monitor performance over time, and participate ingovernance. The 340B build came last and is not yet live. Staff have been writing customized rules and feeding the tool the system's patient definitions, and it is now training on University Health's data. "We are in the very final stages," Dr. Stice said. Analysts will validate its output at first and lean on it over time to flag only the claims that need a closer look. "Everything that passes our logic goes into one bucket, and then whatever does not pass our logic, that's what we will dig deeper into," Dr. Stice said. She expects analysts to manually examine about 5% of claims rather than 10%, on top of the review the system's third-party administrator already performs. The shift may not immediately free up staff, she acknowledged. Sorting through flagged claims could take more team time at first, with the payoff coming as the group fixes the root causes behind them. "Over time we can address those root causes and then ultimately have a more compliant program." Rather than pick a specialist for each AI domain, University Health wanted one vendor across them all, from prior authorization to 340B. "We wanted one partner with many abilities to help solve our problems," Dr. Stice said. The team vetted the tool against its own specific population, asking how the vendor would handle Missouri Medicaid, which covers about half of University Health's patients, and made reference calls to other organizations already using the tool. Dr. Stice ensured that a clinician had a seat on the AI committee, along with the chief medical information officer, compliance, and internal and external legal counsel. "It's just as it is for clinicians: No medication comes without risk, so how do we balance that for this patient? It's the same thing in this context. No AI tool comes without risk, so how do we balance the risks and benefits?" To contain the attendant risks, the committee chose HIPAA-compliant de-identification, data ownership, and indemnification. The tool can access only the records of patients who need a claim audited or a PA submitted. Any data sent for model training is de-identified, and retained data must be destroyed after six months to prevent reidentification. Human in the Loop Steven Smoke, PharmD, an informatics pharmacist at RWJBarnabas Health, in New Jersey, echoed the importance of having clinicians and other end users serve as the "human in the loop," interpreting AI-generated recommendations and making final decisions. The goal, he noted in an exclusive video interview with Pharmacy Practice News, is to continuously monitor AI performance to identify any decline in accuracy or effectiveness over time. Dr. Smoke also stressed the need for interdisciplinary collaboration among governance, clinical, and operational teams to ensure AI tools are integrated into workflows appropriately, with recommendations delivered to the right individuals at the right time. The sources reported no relevant financial disclosures.