Univfy, a fertility and health AI company, has published new research in Nature Communications validating the effectiveness of its artificial intelligence and machine learning platform in predicting IVF success rates. The peer-reviewed study demonstrates that Univfy’s IVF live birth prediction models outperform the U.S. national registry model, potentially improving access, affordability, and clinical outcomes for IVF treatments.
The study, titled “Machine learning center-specific models show improved live birth predictions over US national registry-based model,” analyzed data from 4,645 patients across six centers. Univfy’s models showed significant improvements in model metrics, including F1 score (the harmonic mean of precision and recall) and precision-recall area-under-the-curve (PR AUC).
These metrics are particularly important as they measure a model’s ability to minimize false positives and false negatives, a critical factor in developing real-world economic solutions such as value-based care and actuarial models for fertility treatment.
According to the study, Univfy models correctly predicted that 76% of patients had first-cycle IVF live birth probabilities (LBP) of 50% or greater. Notably, 23% of patients correctly predicted to have LBP ≥ 50% by Univfy models were given lower probabilities by the U.S. national registry model. Additionally, Univfy correctly predicted that 11% of patients had LBP ≥ 75% (with an actual live birth rate of 81%), whereas the U.S. national registry model did not identify any of these patients as having LBP ≥ 75%.
“This publication is a testament to the rigorous science behind the Univfy AI/ML Platform enabling it to correctly predict the excellent IVF outcomes that are achieved by our collaborators and the broader IVF ecosystem yet are conventionally under-appreciated,” said Dr. Mylene Yao, CEO and Cofounder of Univfy. “Univfy was founded to improve patient-centric care, especially in supporting patient IVF prognostic counseling. Through our work to improve IVF cost-success transparency, we have now established a platform that also enables scaled production of validated economic solutions that are not only win-win for patients, providers and healthcare stakeholders but are urgently needed to help more women and couples to access and afford IVF to have a family.”
The publication builds upon prior research demonstrating 2-3x improvement in IVF utilization with Univfy PreIVF Report-based patient counseling and AI/ML IVF live birth prediction model validation for centers in the U.S., UK, and EU. This highlights the value of using locally validated AI-powered solutions in real-world clinical settings.
The publication marks a significant milestone in validating the science behind the Univfy AI/ML Platform and its potential to enable economic solutions such as value-based IVF care, which are key to expanding IVF access and affordability.
