Analysis of non-conformities after 7 months of video control use by DrugCam®
DrugCam® was deployed in the Oncology Pharmaceutical Unit (UPO) in September 2017. After a progressive introduction, 7 molecules are now prepared with Drugcam®, representing 13% of the activity, that is 2226 preparations on 4 equipped stations. An inventory was carried out with the aim of analysing and evaluating the data and the number of non-compliance enabling an intermediate assessment to be made of the implementation of this new control method.
Data extraction was performed by EureKam® between September 2017 and April 2018. From these data, the detection errors and manual interventions were analyzed according to the different workstations.
The distribution of activity between the 4 workstations is not equivalent: 2 workstations represent 81% of the activity. 22% of preparations have at least one step validated manually during production: label presentation (6%), bottle presentation (6%), syringe presentation (13%). Manual reading varies according to the station used, from 0 to 9% (p<0.05) for vials and from 3 to 15% (p<0.05) for syringes. In addition, reading errors are detected for 11% of preparations. These errors concern vials in 3% of cases with 0.09% of true positives and apply only to vials without datamatrix. For 9% of preparations, at least one error is detected when reading samples from syringes with 1% of true positives. Depending on the station, the detection of bottle errors varies from 0 to 6% (p<0.05) and from 11 to 20% (p<0.05) for sampled volumes. We also have 0.27% of the preparations for which a production step was cancelled and redone and 0.13% having had a manual indexing step. Finally, 30% of the preparations required a pharmaceutical validation step after the production (manual step or detection error).
This analysis shows that the number of detection errors is low and could be improved by various actions such as the use of test modules proposed by the company (analysis of detections by stations). The variation in brightness between workstations could be an explanation for the differences observed. The presence of datamatrix on the bottles also seems important to promote a good reading. Their recent addition to the solvent bags now automates and secures the entire preparation. This data is also interesting for evaluating and encouraging the training of preparers as well as increasing their trust in this video control compared to double human control. These results are promising but the extraction will have to be renewed to confirm them with a more important production.