Artificial intelligence management system : what impact on the availability time of chemotherapy ?

Objective
Artificial Intelligence management system (AIMS) has been installed in our centralized cytotoxic preparation unit to secure the chemotherapy circuit. Security tools are reported as impacting the preparation availability timelines ^(1),(2). The goal is to ensure that this security does not come at the expense of preparation availability times.

Materials and methods
The study weeks (respectively W1 and W2) were selected in order to be representative of our production in terms of volume and daily average.
Production data are exported from the prescription assistance software toward a spreadsheet. Only patients coming to the Day Hospital, excluding clinical trials, are included in this study. For each patient, only the parameters related to the first chemotherapy of the protocol is studied.
The data contained in the spreadsheet include : the day of the chemotherapy, the patient’s identification number, the time of prescription, the patient’s arrival time, the time of chemotherapy release, and the delay between the time of the prescription and the time of release. In case of an anticipated prescription, the time considered will be the theoretical patient’s arrival time.
Several parameters are calculated : the mean, the standard deviation, the variance, and the quartiles. To compare the obtained results, a Z-test for comparing means with independent samples is performed.

Results
The annual average is 62.3 chemotherapies per day. For the selected weeks, the daily average are 62.2 and 63.8 chemotherapies respectively.
For W1, the results are : number of patients 147, mean of 17.4 minutes (min), standard deviation of 23.4 min, variance of 550.4 min². The quartiles are determined : Q1 is 0 min, Q2 is 7.5 min, and Q3 is 27 min.
For W2, the results are : number of patients 151, mean of 17.0 min, standard deviation of 19.1 min, variance of 363.6 min². The quartiles are determined: Q1 is 0 min, Q2 is 14 min, and Q3 is 29.2 min. Since both samples are greater than 30, the Z-test follows a standard normal distribution under H0. The result of the Z-test, with an alpha risk of 5%, is 1.58 (H0 being that the means of the two groups are not significantly different). Since the threshold value is 1.96, H0 is not rejected: thus, the means are not significantly different.

Discussion - Conclusion
Thanks to the addition of AIMS as a control method, 99.3% of the production is secured without a negative impact on the availability time. Some preparations are not yet feasible through AIMS due to their complexity. The results for W2 were obtained 2 months after the installation of AIMS. It will be interesting to conduct this study again in 6 months to measure the impact of the learning curve on performance.

1. Carrez L, Falashi L, Cingria L, Sadeghipour F, Bouchoud L, Bonnabry P. Organisation et sécurisation du circuit des chimiothérapies : Exemple de la pharmacie des Hôpitaux Universitaires de Genève. 13 avr 2014 [cité 11 avr 2024];47(2). Disponible sur: https://pharmacie.hug.ch/sites/pharmacie/files/rd/publications/2014_02rqs.pdf
2. Descout J, Grignon C, Coulon S, Dauphin A. Gerpac.eu. 2011 [cité 11 avr 2024]. Analyse des différentes méthodes de contrôle des chimiothérapies anticancéreuses de 30 hôpitaux français. Disponible sur: https://www.gerpac.eu/analyse-des-differentes-methodes-de-controle-des-chimiotherapies-anticancereuses-de-30-hopitaux-francais

See the poster