The digital eye: an original pharmaceutical technology supporting cytotoxic drug production. Focus on one year of artificial intelligence.
7 October 2016E. Girault, M. Laplace, B. Dalifard, B. Le Franc, M-A. Cadeac Service pharmacie, Groupe Hospitalier de la Rochelle-Ré-Aunis,
Rue du Docteur Schweitzer, 17000 La Rochelle, Paris
Since June 2015, all five manufacturing stations (isolators and filtering hoods) belonging to our centralized cytotoxic drug reconstitution unit are equipped with DrugCam®, the first digital eye system, an intelligent video system. By providing a rapid and reliable detection of anticancer drug vials and syringes volumes, this real-time approach for controlling chemotherapy preparations was designed to assist pharmacy technicians or pharmacists throughout the entire drug preparation process.
We first aimed to assess the impact of DrugCam® on our production of cytotoxic drugs. Our second objective was to review all manufacturing mistakes that required a discontinuation of the preparation process by the AI system.
One-year data were retrospectively extracted from the DrugCam® server and subsequently analysed. Additionally, an error analysis was conducted reviewing all DrugCam® video records of the preparations.
In one year, 23 118 drug preparations (100% of our drug production to date) were produced using Drugcam®, including all medicines, solvents and syringes (1 to 50 millilitres). The median manufacturing time was 270 seconds, peel and bleed time included. On average, the 5-Fluororuracile had the fastest manufacturing time 90 seconds and the Bendamustine the largest (668 +/- SD seconds). The highest productivity period occured between 11:30am and 12:30pm, reaching a peak of forty preparations per hour and requiring four technicians.
35 mistakes were detected and corrected immediately including: 31 volume errors [-91%-+500%], one labelling problem and two errors related to solvent volumes. All technicians made mistakes, particularly when they were under heavy workloads. These mistakes usually happened during the first fifteen drug preparations and dosing errors were linked to the residue in more than 30% of cases.
All drug preparations were made under the guidance of the DrugCam® system over a year. This AI proved to be useful for improving the department productivity and helped to better control and secure our drug production process. It also provided valuable insights on manufacturing mistakes which makes it an effective tool for further risk assessment and management initiatives. DrugCam® will be soon available for nurses within wards, to ensure the safety of at risk cytotoxic drug preparations.