Nonstationary Multivariate Gaussian Processes for Electronic Health Records
This work addresses the challenge of modeling nonstationary correlations in electronic health records for improved patient risk prediction, representing an incremental advancement in domain-specific methods.
The authors tackled the problem of jointly modeling multiple clinical variables in electronic health records by proposing multivariate nonstationary Gaussian processes with time-dependent parameters, and they showed that this model provides better predictive performance over stationary models and uncovers latent correlation processes predictive of patient risk.
We propose multivariate nonstationary Gaussian processes for jointly modeling multiple clinical variables, where the key parameters, length-scales, standard deviations and the correlations between the observed output, are all time dependent. We perform posterior inference via Hamiltonian Monte Carlo (HMC). We also provide methods for obtaining computationally efficient gradient-based maximum a posteriori (MAP) estimates. We validate our model on synthetic data as well as on electronic health records (EHR) data from Kaiser Permanente (KP). We show that the proposed model provides better predictive performance over a stationary model as well as uncovers interesting latent correlation processes across vitals which are potentially predictive of patient risk.