Heterogeneous Clinical Trial Outcomes via Multi-Output Gaussian Processes

TL;DR

This paper introduces scalable multi-output Gaussian Process models with Kronecker structure for analyzing large, heterogeneous clinical trial data, enabling efficient inference and comparison with parametric models.

Contribution

It presents a novel scalable GP modeling approach for heterogeneous clinical data using Kronecker structure and repeated measures for efficient inference.

Findings

Scalable GP models outperform parametric models on large clinical datasets.

Exact sampling is enabled for non-conjugate inference with mixed endpoints.

Different GP specifications with random effects components are explored and characterized.

Abstract

We make use of Kronecker structure for scaling Gaussian Process models to large-scale, heterogeneous, clinical data sets. Repeated measures, commonly performed in clinical research, facilitate computational acceleration for nonlinear Bayesian nonparametric models and enable exact sampling for non-conjugate inference, when combinations of continuous and discrete endpoints are observed. Model inference is performed in Stan, and comparisons are made with brms on simulated data and two real clinical data sets, following a radiological image quality theme. Scalable Gaussian Process models compare favourably with parametric models on real data sets with 17,460 observations. Different GP model specifications are explored, with components analogous to random effects, and their theoretical properties are described.