Background
Artificial intelligence (AI) is becoming more common in healthcare, especially in decision support systems. Some AI systems provide predictions along with confidence levels (uncertainty-aware AI), while others only give predictions (black-box AI). It’s unclear how these different types of AI affect healthcare providers’ performance and response times.
Study Objective
This study explored how black-box AI and uncertainty-aware AI affect pharmacists’ decision-making and response times.
Methods
Pharmacists were invited to participate in a web-based study where they completed tasks with and without AI assistance. They were divided into two groups, using either black-box or uncertainty-aware AI. Each pharmacist worked on 100 tasks, which involved identifying whether prescriptions were correct or incorrect.
Results
The study included 30 pharmacists. The findings showed:
- With accurate AI advice, pharmacists rejected incorrect medications 96.1% of the time using uncertainty-aware AI compared to 91.8% with black-box AI, versus 81.2% without any AI.
- Correct medications were accepted 99.2% of the time with black-box AI help, 94.1% with uncertainty-aware AI, and 94.6% without AI.
- Uncertainty-aware AI helped prevent poor decisions when the AI incorrectly suggested accepting a wrong medication (83.3% vs 76.7% for black-box AI).
- When the AI advised rejecting a correct medication, pharmacists without AI correctly accepted it 94.6% of the time, compared to 86.2% with uncertainty-aware AI and 81.2% with black-box AI.
- Uncertainty-aware AI also led to faster decisions than black-box AI and no AI, except when the AI rejected the correct medication.
Conclusions
The type of AI used influenced pharmacists’ decisions and response times. Uncertainty-aware AI improved speed and offered a safeguard against faulty advice, whereas black-box AI resulted in slower reactions and poorer decision outcomes when it provided bad guidance. Choosing the right AI tool is crucial for enhancing performance while minimizing overreliance on technology.
Practical Solutions
Define Measurable Outcomes
Set clear goals for how AI can improve pharmacists’ response times and decision-making in medication verification tasks.
Select AI Tools That Fit Clinical Needs
Choose AI solutions that are specifically designed for the tasks at hand.
Implement Step by Step and Expand
Start with a pilot project using AI solutions, monitor the results, and assess the real-world impact based on the trial data.
Contact Us
For AI solutions in medical management, reach out to us:
- Telegram: https://t.me/itinai
- X: https://x.com/vlruso
- LinkedIn: https://www.linkedin.com/company/itinai/
