ASHP Midyear: Pharmacists Must Proactively Consider and Implement AI Tools

Although there is still a lack of strong clinical research on applications of artificial intelligence (AI) in pharmacy, pharmacists must be proactive when researching and considering the use of these tools, particularly in small and rural health system pharmacies.¹

There are many potential uses for AI in hospital pharmacies, including medication management and dispensing; drug interaction and adverse event (AE) prediction, inventory and supply chain optimization; personalized medicine; clinical decision support; automated data entry and administrative tasks; and training and simulation. However, presenters in the session at the American Society of Health-System Pharmacists 2024 Midyear meeting also acknowledged that there are many concerns.¹

“While the integration of AI offers numerous benefits, some concerns include data privacy, ethical considerations, and ensuring that the technology complements, rather than replaces, the clinical expertise of pharmacists,” said presenter Kyle Johnicker, PharmD.¹

Illustration of AI tools in health care | Image credit: LALAKA | stock.adobe.com

Rural Health-System Pharmacy

One in 5 individuals in the United States live in a rural area, Johnicker said. These individuals are physically further from health care services, often older, lack insurance, and have a higher risk of death due to heart disease, cancer, stroke, unintentional injury, and chronic lower respiratory disease.¹

Furthermore, over 20% of counties in the US are hospital deserts, in which residents have to drive more than 30 minutes to reach the closest hospital. More than 40% of US counties do not have adequate access to a pharmacy, defined as needing to drive more than 15 minutes.¹

Johnicker and Gretchen Brummel, PharmD, BCPS, tried to find literature on AI with a high level of evidence, particularly randomized controlled trials. However, they said there were very few studies meeting these criteria. Notably, an article published in Nature Medicine found that just 4% of AI health tools with regulatory authorization had undergone a randomized controlled trial.² Therefore, Brummel said it is incumbent on clinicians to careful evaluate these tools and question vendors about how their tools are validated.

AI Use Case: Type 1 Diabetes

In one randomized controlled trial, Brummel said investigators examined the use of AI management versus clinician management for patients with type 1 diabetes who use an insulin pump. In the trial, 108 patients between the ages of 10 and 21 were randomized 1:1 to AI management or clinician management, with the primary outcome of time in target serum glucose range.AI management involved remote insulin dose adjustment every 3 weeks guided by either an automated AI-based decision support system (AI-DSS) or by physicians.³

The results showed that the AI-DSS was non-inferior to clinician management. The percentage of readings below 54 mg/dL within the AI-DSS arm was statistically non-inferior to that in the physician arm. Furthermore, 3 severe AEs related to diabetes were reported in the physician arm, compared with none in the AI-DSS arm.³

“We know that specialists are in shortage, primary care is in shortage, so if we’re able to leverage a tool like this to help care for this population, that could definitely be helpful,” Brummel said.¹

AI Use Case: Clinical Decision Support

In another trial, investigators aimed to assess the effectiveness and efficiency of an Asthma-Guidance and Prediction System (A-GPS) utilizing AI in optimizing asthma management. Children diagnosed with asthma were enrolled along with their 42 primary care providers before being stratified into 3 strata based on asthma severity, asthma care status, and asthma diagnosis.⁴

The intervention involved a quarterly A-GPS report sent to clinicians, including relevant clinical information for asthma management using the electronic health records (EHRs) and machine learning-based predictions for risk of exacerbation. The primary end point was the occurrence of an AE within 1 year and secondary outcomes included time required for clinicians to review EHRs for asthma management.⁴

According to the study results, although the proportion of children with AE in both groups decreased from the baseline (P=0.042), there was no difference in AE frequency between the 2 groups (12% for the intervention group vs 15% for the control group) during the study period. However, the A-GPS intervention significantly reduced time for reviewing EHRs for asthma management of each participant (median: 3.5 min, IQR: 2-5), compared to usual care without A-GPS (median: 11.3 min, IQR: 6.3-15).⁴

“There are a couple implications here for rural pharmacy,” Brummel said. “[This could] decrease charting time…and increase disease state awareness.”¹

AI Use Case: Telehealth

Finally, Brummel presented a case study of a virtual care platform leveraging AI to manage treatment-resistant hypertensive adults via telehealth. The study involved pharmacist and physician co-management during the first part, followed by pharmacist management under a collaborative practice agreement. Centralized monitoring used data from patients’ phones.⁵

According to the study results, 67% of patients achieved blood pressure control of <140/90 mm Hg at 6 months, and 74% achieved control by 12 months. Furthermore, systolic blood pressure was lowered by an average of 3.3 mm Hg/month for those with initial blood pressure readings >150/90 mm Hg; lowered by 2.4 mm Hg/month for those with initial readings between 140-149 mm Hg and 90-99 mm Hg; and lowered by 0.6 mm Hg/month for those with initial readings below 140/90 mm Hg.⁵

Pharmacist telehealth encounters were documented in 65% of patients and pharmacist interactions were associated with a 1.3 mm Hg/month decline in systolic blood pressure over time. Additionally, during the 12-month study period, 46% of patients had a blood pressure medication adjustment and 37% were prescribed new blood pressure medication.⁵

Implementation

Despite widespread excitement about the potential for AI-based tools, implementation can be a huge hurdle, particularly in small and rural pharmacies with very limited budgets and resources. Brummel encouraged attendees to think long-term, planning ahead and preparing themselves and their teams for these tools.¹

“We really need to be thinking about this in terms of the long range, and I think it’s everyone—not just the specialists—who are going to need to be in this space,” Brummel said.

Johnicker said there are several key things pharmacists can do now in order to prepare for AI tools and successfully argue for their implementation. First, pharmacy teams should evaluate what tools are already built into their EHRs. Johnicker noted that the top EHRs (Epic, Oracle Health, Meditech, and TruBridge) already host AI pages on their websites.¹

Pharmacy teams can also proactively identify gaps in care and challenges in their workflows, helping to establish which tools might be best. Johnicker also encouraged all pharmacy teams to recognize opportunities, educate themselves, and prepare to learn new things.

“If you aren’t prepared, it’s going to be like drinking from a fire hose,” he said. “The information is coming whether you like it or not.”¹

REFERENCES

1. Brummel G, Johnicker K. Do Whatcha Wanna With Artificial Intelligence in Small & Rural Pharmacy. Presented at: American Society of Health-System Pharmacists 2024 Midyear Clinical Meeting. New Orleans, LA. December 8, 2024.

2. Lenharo M. The testing of AI in medicine is a mess. Here’s how it should be done. Nature. August 21, 2024. Accessed December 8, 2024. https://www.nature.com/articles/d41586-024-02675-0

3. Nimri R, Battelino T, Laffel LM, et al. Insulin dose optimization using an automated artificial intelligence-based decision support system in youths with type 1 diabetes. Nature Med. 2020;26:1380-1384. doi:10.1038/s41591-020-1045-7

4. Seol HY, Shrestha P, Muth JF, et al. Artificial intelligence-assisted clinical decision support for childhood asthma management: a randomized clinical trial. PLoS One. 2021;16(8):e0255261. doi:10.1371/journal.pone.0255261

5. Remote monitoring and pharmacist helped improve hard-to-control blood pressure. News release. American Heart Association. September 5, 2024. Accessed December 8, 2024. https://newsroom.heart.org/news/remote-monitoring-and-pharmacist-helped-improve-hard-to-control-blood-pressure