Using automated methods to detect safety problems with health information technology: a scoping review

Area of Science:

• Health Informatics
• Medical Device Safety
• Artificial Intelligence in Healthcare

Background:

• Health Information Technology (HIT) systems are critical for modern healthcare delivery.
• Ensuring the safety and reliability of HIT is paramount to prevent patient harm.
• Automated methods offer potential for proactive identification of HIT-related safety concerns.

Purpose of the Study:

• To systematically review and summarize existing research on automated methods for detecting safety problems in HIT.
• To categorize the types of automated methods employed and the HIT issues they address.
• To assess the performance and application of these methods in identifying safety concerns.

Main Methods:

• A comprehensive literature search was conducted across multiple bibliographic databases (MEDLINE, ACM Digital, Embase, CINAHL Complete, PsycINFO, Web of Science) from January 2010 to June 2021.
• Studies evaluating automated methods for HIT safety problem detection were included.
• Automated methods were classified as rule-based, statistical, or machine learning, and their performance was assessed against existing safety concern classifications.

Main Results:

• The review identified 45 studies focusing on automated detection of HIT safety problems.
• The majority of studies (60%) concentrated on use errors within electronic health records and order entry systems.
• Machine learning (22 studies) and statistical modeling (17 studies) were the predominant automated methods utilized.

Conclusions:

• A diverse range of automated methods, including rule-based, statistical, and machine learning approaches, are applied to detect HIT safety issues.
• Significant opportunities exist for further systematic investigation into the application and effectiveness of these automated methods in real-world healthcare settings.