A Framework for Individualizing Predictions of Disease Trajectories by Exploiting Multi-Resolution Structure

By Peter F. Schulam, Suchi Saria. 2015

104 citations recorded by the source corpus.

For many complex diseases, there is a wide variety of ways in which an individual can manifest the disease. The challenge of personalized medicine is to develop tools that can accurately predict the trajectory of an individual's disease, which can in turn enable clinicians to optimize treatments. We represent an individual's disease trajectory as a continuous-valued continuous-time function describing the severity of the disease over time. We propose a hierarchical latent variable model that individualizes predictions of disease trajectories. This model shares statistical strength across observations at different resolutions-the population, subpopulation and the individual level. We describe an algorithm for learning population and subpopulation parameters offline, and an online procedure for dynamically learning individual-specific parameters. Finally, we validate our model on the task of predicting the course of interstitial lung disease, a leading cause of death among patients with the autoimmune disease scleroderma. We compare our approach against state-of-the-art and demonstrate significant improvements in predictive accuracy.

Subfields: Artificial Intelligence, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Reliable Uncertainty Estimates in Deep Neural Networks using Noise Contrastive Priors (2018, background): “For example, in healthcare, reliable uncertainty estimates can prevent overconﬁdent decisions for rare or novel patient conditions (Schulam and Saria, 2015).”
Noise Contrastive Priors for Functional Uncertainty (2018, background): “For example, in healthcare, reliable uncertainty estimates can prevent overconfident decisions for rare or novel patient conditions (Schulam and Saria, 2015).”
Scalable Joint Models for Reliable Uncertainty-Aware Event Prediction (2017, methodology): “Under this assumption, we can ignore the process that caused missing data and infer parameters of the model only based on the observed data (see Rubin [32] and Appendix B of Schulam and Saria [33] for a longer discussion).”
Reliable Decision Support using Counterfactual Models (2017, methodology): “As predictive models are becoming more widely used in domains like healthcare where safety is critical (e.g. Li-wei et al. 2015, Schulam and Saria 2015, Alaa et al. 2016, Wiens et al. 2016, Cheng et al. 2017), the framework proposed here is increasingly pertinent.”
Using Causal Inference to Estimate What-if Outcomes for Targeting Treatments (2016, background): “Alternate forms for ui that incorporate long and short term structure in the baseline disease trajectories [Schulam and Saria, 2015] should be incorporated when relevant.”
Sparse Multi-Output Gaussian Processes for Medical Time Series Prediction (2017, methodology): “In an extension of PSM, the authors adapted patient-specific information to forecast specific clinical covariates (Schulam and Saria, 2015); the time series for each covariate was still modeled independently.”
A Bayesian Nonparametic Approach for Estimating Individualized Treatment-Response Curves (2016, unknown): “A different choice for both the mean and the covariance kernel can be made depending on the properties of the data; see Schulam and Saria (2015) for a different example of the baseline model for modeling progression in chronic diseases.”
Differentially Private Data Generative Models (2018, background): “As motivation, consider a hospital not allowed to release its private medical data to public for use, but wants to share the data with universities for, say, data-driven disease diagnosis studies [34, 55].”
Learning to Detect Sepsis with a Multitask Gaussian Process RNN Classifier (2017, background): “For instance, (Schulam & Saria, 2015) developed a longitudinal model for predicting progression of scleroderma, (Futoma et al., 2016) presented a joint model for predicting progression of chronic kidney disease and cardiac events, and (Liu et al., 2015) proposed a continuous-time hidden Markov…”
What-If Reasoning with Counterfactual Gaussian Processes (2017, background): “…of disease trajectory models, for instance, state-of-the-art approaches do not explicitly control for future actions (see e.g., Li-wei et al. 2015, Schulam and Saria 2015, Alaa et al. 2016, Wiens et al. 2016, Cheng et al. 2017), and are therefore sensitive to the treatment policy in the training…”
A Review of Challenges and Opportunities in Machine Learning for Health. (2018, background): “By personalizing over time, one can learn individual-specific treatment effects that address the cause of the syndrome 11,69,70 in a particular individual.”
A Bayesian Nonparametric Approach for Estimating Individualized Treatment-Response Curves (2016, methodology): “A different choice for both the mean and the covariance kernel can be made depending on the properties of the data; see Schulam and Saria (2015) for a different example of the baseline model for modeling progression in chronic diseases.”
Using recurrent neural network models for early detection of heart failure onset (2016, methodology): “…using a 2-dimensional continuous-time hidden Markov model, 39 the progression of lung disease using graphical models with the Gaussian process, 40 the progression of chronic obstructive pulmonary disease using the Markov jump process, 41 and the progression of multiple diseases using the…”