Looking into the future: a machine learning powered prediction model for oocyte return rates after cryopreservation.

Area of Science:

• Reproductive Medicine
• Biostatistics
• Health Informatics

Background:

• Oocyte cryopreservation is increasingly utilized for fertility preservation.
• Predicting oocyte return rates is crucial for managing storage programs and patient counseling.
• Existing models may not fully capture the complex factors influencing oocyte utilization.

Purpose of the Study:

• To develop and validate a predictive model for estimating the likelihood of patients using their stored oocytes.
• To identify key demographic, medical, and social factors associated with oocyte utilization.
• To enhance decision-making in fertility preservation and gamete storage programs.

Main Methods:

• Utilized a large dataset of 77,631 oocyte-cryopreservation cycles from US fertility clinics (2014-2020).
• Employed multiple machine learning algorithms, including ensemble methods combining bootstrap aggregation, gradient boosting, and linear discriminant analysis.
• Analyzed patient demographics, medical and fertility diagnoses, partner status, and geographic location.

Main Results:

• An ensemble model achieved high predictive accuracy (balanced accuracy: 0.83, ROC AUC: 0.90).
• Key predictors for oocyte use included patient age, partner status, race/ethnicity, clinic region, and cryopreservation indication.
• Treatment indications varied, with planned cryopreservation being the most common (35.6%).

Conclusions:

• The developed predictive model demonstrates significant accuracy in estimating oocyte return likelihood.
• This tool is valuable for patient counseling regarding oocyte cryopreservation outcomes.
• The model can optimize healthcare decisions and improve the efficiency of fertility storage programs.