Artificial intelligence-based video monitoring of cytotoxic drug preparation in an ISO 9001-certified unit: a ten-year review of over 300,000 preparations

Background
Ensuring the safety of chemotherapy preparations is a major challenge for cancer treatment preparation units. Since 2014, our cytotoxic drug reconstitution unit has used a digital video monitoring system based on artificial intelligence and computer vision (Drugcam^®, Eurekam) to track and validate, in real time, the critical steps of the cytotoxic drug preparation process. This system replaced the double visual check previously performed by a pharmacy technician.

Objectives
To quantify and characterize the errors detected by the system, and to provide an overview of ten years of use.

Materiels and Methods
A prospective, single-centre descriptive study (2014-2024) was conducted, in which all detected errors were recorder in an Excel^® spreadsheet in accordance with the ISO 9001 standard. Each error was rated according to four severity levels (minor, moderate, major, and critical). Activity data were extracted from the preparation scenarios generated by the system’s control module.

Results
Of 321,093 preparations performed over ten years, 561 critical errors were detected. The main error categories were: active ingredient volume errors (n = 311; 55.4%), including 20.3% underdoses and 79.7% overdoses; solvent errors (n = 130; 23.2%); solvent volume errors (n = 68; 12.1%); vial errors (n = 34; 6.1%); labelling errors (n = 15; 2.7%); and administration device errors (n = 3; 0.5%). The overall rate of intercepted critical errors was 1.75 per 1,000 preparations.

Conclusion
This ten-year analysis represents the largest real-world study published to date on the use of a digital video monitoring system to ensure the safety of chemotherapy preparation. The Drugcam^® system enables the quantification, identification and prevention of potentially serious medication errors in cytotoxic drug preparation units. Beyond ensuring the safety of anticancer preparations at every stage of the process, this tool can also be used to assess operators’ practices and support the training of new pharmacy technicians.

See the poster