Predict IVF outcomes using advanced machine learning techniques

Lay summary

The project aims to identify and quantify the impact of different patient related features, as well as the relationship between each of the IVF treatment cycles, on the live birth outcome of a patient’s IVF journey. Unlike previous research projects which have focused on the overall success of IVF, this project will focus on building and interpreting a state-of-the-art machine learning model leveraging linked-IVF cycle data since 2003 and quality predictors, aiming to provide more accurate personalised prediction of IVF live birth outcomes.

The predictive model will be published on the Gaia website and freely available to the public. Individuals considering IVF will be able to use this predictive model to better understand the chances of their treatment being successful. This is urgently required as currently prospective IVF patients are only able to view generic clinic success metrics that are both not easily comparable (as they are measured in different ways between clinics) and do not account for the individual’s personal characteristics. Having a better understanding of individual chances of success will help prospective IVF patients to both manage their expectations as they go through treatment and understand the likely financial cost of their IVF treatment.

In addition to making the predictive model available, this project will publish a report on the methodology used to construct this predictive model so that users can understand how their chances of success are estimated. The report will contribute to public understanding about what impacts patient IVF outcomes and allowing individuals to make choices that maximise their personal chances of success. The research will also use the predictive model to improve Gaia’s existing IVF insurance offering.

Public benefit statement

IVF is an emotional and physical ordeal made more difficult by not understanding how likely your treatment is to be successful. This research project will lead to the production of an up-to-date prediction model that will be publicly available for free without asking users to provide any personal data. Giving patients a realistic indication of what their IVF journey could look like will lead to improvement of patient care. It is anticipated to reduce treatment discontinuation resulting from unrealistic expectations, provide better understanding of the possible outcomes (reducing the emotional toll of their IVF journey) and help patients with financial planning for their IVF journey, leading to allow patients to maximise their personal chances of success. This research will provide documentation explaining the model’s predictions to help users interpret the results provided, and also publish a report on methodologies used in lay and technical terms on the Gaia website and publish it in a peer-reviewed journal.