Research Support - Sample Patient Counseling Report

Point-of-care patient counseling report

This figure shows a sample portion of the patient counseling report that shows the probability of having a baby from IVF based on machine learning, center-specific (MLCS), IVF live birth probability prediction model. This report is commercially known as Univfy® PreIVF Report*.

Note: This sample report uses fictitious name, medical record number, date of birth, report ID and clinical information for the purposes of demonstration. No real person’s data is used to generate this report.

A real report would show wording customized for each fertility center and the numbers would also be center-specific.

Research Support - Sample ML-based IVF Refund Program Cost Comparison

Sample MLCS model-enabled IVF Refund Program Cost Comparison

By using ML and a transparency-driven patient counseling and provider support system, we have been able to support providers in offering shared risk programs that qualify up to 80% of patients, resulting in a much larger impact on IVF utilization. In addition, ML enables providers to offer more equitable pricing that confers significant cost savings for patients while preserving profitability for fertility centers.

The ML-based IVF Refund Program Cost Comparison figure (to the right) illustrates how having an accurate and personalized IVF prediction model can inform IVF pricing to lower the cost of doing up to 3 IVF treatments to maximize IVF live birth outcomes. In essence, the term "shared risk" is simply value-based pricing for patients as consumers, whereas the term value-based pricing is used directly in the context of health plan or state-funded reimbursements.

An ML- and transparency-driven shared risk program can be offered to the majority of patients while enabling compliance with the transparency and patient-focused requirements outlined by the Ethics Committee of the American Society of Reproductive Medicine (ASRM, Fertil Steril 2024;121(5):783-786. doi: 10.1016/j.fertnstert.2023.12.032. Epub 2024 Jan 25. PMID:38276940.)

ML = machine learning model. MLCS model = machine learning center-specific model.
Sample machine learning, center specific (MLCS) model-enabled IVF refund program (aka "shared risk") value-based pricing options can be offered by the fertility center to eligible patients. Eligibility is determined objectively by MLCS model with parameters reviewed and approved by the fertility center. Such a cost comparison chart would be shown to patients who have been counseled by their providers regarding their personalized IVF live birth probabilities from doing 1, 2, or up to 3 IVF treatments using the Univfy® PreIVF Report.
The pricing and cost savings shown are for illustration only and are in US currency, based on commonly encountered IVF fees in the US. The actual pricing is dependent on local fertility center’s fee schedule and ML-pricing analysis based on the MLCS IVF live birth prediction model.

Research Support - Artificial Intelligence / Machine Learning (AI/ML) Platform for IVF

AI/ML Platform for IVF - Examples of Processes and Applications

This artificial intelligence (AI)/machine learning (ML) platform* uses data processing and model testing pipelines to transform raw data to deliverables including IVF prognostics patient counseling report; value-based IVF pricing to minimize cost per IVF baby for patients paying out-of-pocket, commercial and government payers; and research findings to advance racial parity and precision medicine. (This AI/ML platform is known as the Univfy AI/ML Platform for IVF.*)

Research Support - Examples of Clinical Variables used in IVF Prediction Models

Practical selection of clinical predictors for feature testing based on clinical goals & availability

This slide shows examples of clinical variables that can be tested for usage in a predictive model. The selection of predictors depends on the expected clinical usage of the prediction model. The variables used in the model should be known and easily available to the provider and patient at the time of clinical usage of the model. Many of these variables may also be outcomes that can in turn be predicted by “upstream” features.

MLCS model = machine learning, center-specific model

Research Support - Requirements of AI/ML Point-of-Care Delivery

Requirements for point-of-care delivery of artificial intelligence/machine learning (AI/ML)-based prognostics at scale

Description

Research Support - Development-to-Deployment Life Cycle

The development-to-deployment life cycle of the machine learning-based, center-specific (MLCS), prognostic model for use at point-of-care to support patient counseling

(A) The MLCS-based, PreIVF model (MLCS model) product life cycle comprises the steps of data pre-processing, model training and validation, deployment and post-deployment validation (or “live model validation”, LMV). MLCS1, MLCS2, etc. indicates that each MLCS model will be replaced by an updated MLCS model trained and tested with a more recent data set which may also become cumulatively larger. (B) Model pipeline supports feature testing, model training, validation analysis, deployment to production and quality testing.

Note: MLCS model = machine learning, center-specific model. In (B), “MLCS” is used generically to indicate the steps used for MLCS1, MLCS2 or any subsequent updates of MLCS model for a particular fertility center.