Optimizing workload by balancing antagonist protocol retrievals using artificial intelligence algorithm for antagonist trigger day decision

This study aims to develop a real-time balancing algorithm using a trigger day suggestion algorithm, which can evenly distribute retrievals throughout the month, thereby reducing the variations of retrievals per day, without compromising clinical outcome. A retrospective cohort study including data of 5,317 antagonist protocol cycles performed in a large center serving over 50 physicians, from November 2021 to October 2022. A trigger suggestion algorithm was utilized to provide optional trigger dates, out of three consecutive days, and by that provide optional retrieval dates. This algorithm was designed specifically for antagonist protocol cycles to suggest one, two or three optional trigger dates, with similar total and mature oocytes retrieved. Analysis of the suggested trigger dates for the antagonist cycles showed that 65.62%, 25.74%, and 8.64% of the cycles had one, two, and three trigger options, respectively. To stimulate real-life conditions, the real time balancing algorithm selects the optimal trigger days using only cycle information available up to the decision day. The algorithm was evaluated by the standard deviation (STD) of the number of retrievals per day, and was compared to the actual STD of retrievals performed. In addition, it was compared to a retrospective balancing algorithm (in which the real number of future retrievals performed in each day is known on decision day) to assess how well the algorithm performed compared to an optimal retrieval balancing model. The implementation of the real time balancing algorithm in a clinic with an average of 14.7 antagonist protocol retrievals per day, reduced the STD of daily retrievals from an average of 6.45 to 3.44, narrowing the range of daily retrievals (average ± 2 STD) from 2-28 to 9-22 while keeping the same average number of daily retrievals. Assuming the clinic has the capacity to handle the pre-balancing upper range of daily retrievals, the algorithm allowed for an approximate 27% increase in the average number of daily retrievals to 18.68, while clinic's capacity remained unchanged. The retrospective balancing algorithm achieved an increase of about 33% to an average of 19.55 retrievals per day with a 2.63 STD and a range of 9-20 retrievals per day. The use of a real time balancing algorithm, utilizing a trigger suggestion algorithm, can reduce the STD of daily retrievals, resulting in a narrower range of daily retrievals and enabling an increase in the average number of daily retrievals without compromising clinical outcome.