Automation bias in electronic prescribing.

Area of Science:

• Medical Informatics
• Human-Computer Interaction
• Patient Safety

Background:

• Clinical decision support (CDS) in e-prescribing enhances safety by flagging potential errors but can introduce new risks.
• Automation bias (AB) is the tendency for users to over-rely on automated systems, potentially reducing vigilance.
• While AB is documented in other clinical tasks, its presence in e-prescribing has not been extensively studied.

Purpose of the Study:

• To investigate the presence of automation bias (AB) in the context of electronic prescribing (e-prescribing).
• To assess the impact of task complexity and interruptions on the manifestation of AB during e-prescribing.

Main Methods:

• 120 medical students used a simulated e-prescribing system across nine clinical scenarios.
• Conditions varied in CDS quality (correct, incorrect, none) and task complexity (low, low with interruption, high).
• Omission errors (missed prescribing errors) and commission errors (accepted false alerts) were measured to quantify AB.

Main Results:

• Correct CDS significantly reduced omission errors across all complexity levels.
• Incorrect CDS significantly increased omission errors compared to no CDS.
• A substantial proportion of participants made commission errors, indicating over-reliance on CDS.
• Task complexity and interruptions did not significantly influence the degree of automation bias observed.

Conclusions:

• The study provides evidence of automation bias (both omission and commission errors) in e-prescribing.
• Verifying CDS alerts is critical to prevent AB-related errors, though interventions have shown limited success.
• Clinicians must maintain vigilance regarding potential CDS failures and actively verify system recommendations.