A novel method to retrieve alerts from a homegrown Computerized Physician Order Entry (CPOE) system of an academic medical center: Comprehensive alert characteristic analysis

Background The collection and analysis of alert logs are necessary for hospital administrators to understand the types and distribution of alert categories within the organization and reduce alert fatigue. However, this is not readily available in most homegrown Computerized Physician Order Entry (CPOE) systems. Objective To present a novel method that can collect alert information from a homegrown CPOE system (at an academic medical center in Taiwan) and conduct a comprehensive analysis of the number of alerts triggered and alert characteristics. Methods An alert log collector was developed using the Golang programming language and was implemented to collect all triggered interruptive alerts from a homegrown CPOE system of a 726-bed academic medical center from November 2017 to June 2018. Two physicians categorized the alerts from the log collector as either clinical or non-clinical (administrative). Results Overall, 1,625,341 interruptive alerts were collected and classified into 1,474 different categories based on message content. The sum of the top 20, 50, and 100 categories of most frequently triggered alerts accounted for approximately 80, 90 and 97 percent of the total triggered alerts, respectively. Among alerts from the 100 most frequently triggered categories, 1,266,818 (80.2%) were administrative and 312,593 (19.8%) were clinical alerts. Conclusion We have successfully developed an alert log collector that can serve as an extended function to retrieve alerts from a homegrown CPOE system. The insight generated from the present study could also potentially bring value to hospital system designers and hospital administrators when redesigning their CPOE system.

Automated summary

A novel method was developed to collect alert information from a homegrown CPOE system in Taiwan, providing insights into the number and characteristics of triggered alerts in an academic medical center. The study found that 80.2% of the collected alerts were administrative, while 19.8% were clinical alerts, highlighting the importance of understanding alert categories to reduce alert fatigue.