Transfer Learning via Latent Factor Modeling to Improve Prediction of Surgical Complications
This work addresses the challenge of improving surgical risk assessment for patients by enabling more effective use of diverse healthcare datasets, though it appears incremental as an extension of existing latent factor methods.
The authors tackled the problem of predicting surgical complications by developing a transfer learning framework using latent factor modeling to leverage both institutional and national surgical outcomes data (NSQIP with 4 million patients from 700+ hospitals). The result was a model that accounts for different covariance structures and complex population relationships through hierarchical priors and scale mixture formulations.
We aim to create a framework for transfer learning using latent factor models to learn the dependence structure between a larger source dataset and a target dataset. The methodology is motivated by our goal of building a risk-assessment model for surgery patients, using both institutional and national surgical outcomes data. The national surgical outcomes data is collected through NSQIP (National Surgery Quality Improvement Program), a database housing almost 4 million patients from over 700 different hospitals. We build a latent factor model with a hierarchical prior on the loadings matrix to appropriately account for the different covariance structure in our data. We extend this model to handle more complex relationships between the populations by deriving a scale mixture formulation using stick-breaking properties. Our model provides a transfer learning framework that utilizes all information from both the source and target data, while modeling the underlying inherent differences between them.