Analysis of electronic medication orders with large overdoses: opportunities for mitigating dosing errors.

Area of Science:

• Health Informatics
• Clinical Decision Support
• Medication Safety

Background:

• Electronic health record (EHR) systems often present dosing alerts that users ignore.
• System design flaws contribute to medication dosing errors.
• Analyzing large overdoses can identify necessary EHR system improvements.

Purpose of the Study:

• Develop database techniques to detect and extract large overdose orders from EHRs.
• Characterize user responses to large overdose alerts.
• Identify and mitigate causes of large-overdose errors.

Main Methods:

• Constructed a data mart of medication orders and dosing alerts from a pediatric hospital (2011-2013).
• Analyzed order processing and alert generation for large (>500%) and extreme (>10,000%) overdoses.
• Characterized user response using alert salience rate (corrective action proportion).

Main Results:

• Evaluated over 5.4 million medication orders and 2.2 million dose alerts.
• 8% of orders generated alerts, with 25% related to overdosing.
• Alert salience rates were low (4-10%), decreasing with larger overdoses; trainees had higher salience than senior staff.

Conclusions:

• Novel analytic systems are needed to understand prescriber interaction with clinical decision support (CDS).
• A new system detected large overdoses (≥500%) and identified sociotechnical error factors.
• System design changes can mitigate many large overdose errors, and EHRs should be designed to prevent them proactively.