A probabilistic disease progression modeling approach and its application to integrated Huntington’s disease observational data

Abstract

Objective Chronic diseases often have long durations with slow, nonlinear progression and complex, and multifaceted manifestation. Modeling the progression of chronic diseases based on observational studies is challenging. We developed a framework to address these challenges by building probabilistic disease progression models to enable better understanding of chronic diseases and provide insights that could lead to better disease management. Materials and Methods We developed a framework to build probabilistic disease progression models using observational medical data. The framework consists of two steps. The first step determines the number of disease states. The second step builds a probabilistic disease progression model with the determined number of states. The model discovers typical states along the trajectory of the target disease, learns the characteristics of these states, and transition probabilities between the states. We applied the framework to an integrated observational HD dataset curated from four recent observational HD studies. Results The resulting HD progression model identified nine disease states. Compared to state-of-art HD staging system, the model 1) covers wider range of HD progression; 2) is able to quantitatively describe complex changes around the time of clinical diagnosis; 3) discovers multiple potential HD progression pathways; and 4) reveals expected time durations of the identified states. Discussion and Conclusion The proposed framework addresses practical challenges in observational data and can help enhance the understanding of progression of chronic diseases. The framework could be applied to other chronic diseases with the help of clinical knowledge.

Artificial Intelligence for Clinical Trial Design

“…The CTS tool includes computational components for modeling drug, disease, and progression of mild cognitive impairment (MCI) and early AD that can be used for model-based clinical trial design [12]. Expanding on this effort, ML methods for disease progression modeling are being developed to provide increasingly accurate and nuanced understanding and characterization of complexity and heterogeneity of many diseases, particularly those such as AD where disease-modifying drugs are not yet available [13][14][15][16][17].…”

Personalized Input-Output Hidden Markov Models for Disease Progression Modeling

“…Wang et al [18] proposed continuous time HMMs to model the progression of chronic disease. Sun et al d[19] leveraged this approach to develop an integrated Huntington’s disease progression model. However none of these previous approaches model system inputs, such as medication, or account for heterogeneity in symptoms between patients in a cohort.…”

SAMGEP: A Novel Method for Prediction of Phenotype Event Times Using the Electronic Health Record

“…We considered as benchmarks five supervised methods using the labeled set alone: (i) LASSO-penalized logistic regression [16,17,34,37–39], (ii) random forest (RF) [40,41], (iii) linear discriminant analysis (LDA) [42], and (iv) LSTM-gated recurrent neural network (RNN) [24,39,43,44] trained with raw feature counts C i , t , as well as (v) LDA trained with patient-timepoint embeddings generated without weights , which we refer to as LDA embed . In addition, we considered a semi-supervised benchmark: hidden markov model (HMM) [26–29,45,46] with a multivariate gaussian emission trained with the weight-free embeddings . Only HMM and RNN leverage the longitudinal nature of the data, while all other comparator methods train models for predicting Y t based only on concurrent features ( C i,t or ) without considering the time sequence.…”