Abstract:Background: The World Health Organization (WHO) reported that cardiovascular diseases (CVDs) are the leading cause of death worldwide. CVDs are chronic, with complex progression patterns involving episodes of comorbidities and multimorbidities. When dealing with chronic diseases, physicians often adopt a "watchful waiting" strategy, and actions are postponed until information is available. Population-level transition probabilities and progression patterns can be revealed by applying time-variant stochastic modeling methods to longitudinal patient data from cohort studies. Inputs from CVD practitioners indicate that tools to generate and visualize cohort transition patterns have many impactful clinical applications. The resultant computational model can be embedded in digital decision support tools for clinicians. However, to date, no study has attempted to accomplish this for CVDs. Objective: This study aims to apply advanced stochastic modeling methods to uncover the transition probabilities and progression patterns from longitudinal episodic data of patient cohorts with CVD and thereafter use the computational model to build a digital clinical cohort analytics artifact demonstrating the actionability of such models. Methods: Our data were sourced from 9 epidemiological cohort studies by the National Heart Lung and Blood Institute and comprised chronological records of 1274 patients associated with 4839 CVD episodes across 16 years. We then used the continuous-time Markov chain method to develop our model, which offers a robust approach to time-variant transitions between disease states in chronic diseases. Results: Our study presents time-variant transition probabilities of CVD state changes, revealing patterns of CVD progression against time. We found that the transition from myocardial infarction (MI) to stroke has the fastest transition rate (mean transition time 3, SD 0 days, because only 1 patient had a MI-to-stroke transition in the dataset), and the transition from MI to angina is the slowest (mean transition time 1457, SD 1449 days). Congestive heart failure is the most probable first episode (371/840, 44.2%), followed by stroke (216/840, 25.7%). The resultant artifact is actionable as it can act as an eHealth cohort analytics tool, helping physicians gain insights into treatment and intervention strategies. Through expert panel interviews and surveys, we found 9 application use cases of our model. Conclusions: Past research does not provide actionable cohort-level decision support tools based on a comprehensive, 10-state, continuous-time Markov chain model to unveil complex CVD progression patterns from real-world patient data and support clinical decision-making. This paper aims to address this crucial limitation. Our stochastic model–embedded artifact can help clinicians in efficient disease monitoring and intervention decisions, guided by objective data-driven insights from real patient data. Furthermore, the proposed model can unveil progression patterns of any chronic disease of interest by inputting only 3 data elements: a synthetic patient identifier, episode name, and episode time in days from a baseline date.

Modeling disease progression in longitudinal EHR data using continuous-time hidden Markov models

Continuous-time probabilistic models for longitudinal electronic health records

Time-dependent Probabilistic Generative Models for Disease Progression

Modeling Disease Progression Trajectories from Longitudinal Observational Data

Modelling and Analysing the Dynamics of Disease Progression from Cross-Sectional Studies

Learning transition times in event sequences: the Event-Based Hidden Markov Model of disease progression

Development of a Cohort Analytics Tool for Monitoring Progression Patterns in Cardiovascular Diseases: Advanced Stochastic Modeling Approach

Joint Hierarchical Gaussian Process Model with Application to Forecast in Medical Monitoring

A Multi-state Non-Markov Framework to Estimate Progression of Chronic Disease

Healthcare Predictive Analytics for Disease Progression: a Longitudinal Data Fusion Approach

Scalable Modeling of Multivariate Longitudinal Data for Prediction of Chronic Kidney Disease Progression

A Temporal Visualization of Chronic Obstructive Pulmonary Disease Progression Using Deep Learning and Unstructured Clinical Notes.

A Deep Learning Approach to Handling Temporal Variation in Chronic Obstructive Pulmonary Disease Progression

A joint model for multistate disease processes and random informative observation times, with applications to electronic medical records data

Joint modeling of zero-inflated longitudinal measurements and time-to-event outcomes with applications to dynamic prediction

DPVis: Visual Analytics With Hidden Markov Models for Disease Progression Pathways

Jointly Modelling Longitudinal and Survival Data

Markov-modulated marked Poisson processes for modelling disease dynamics based on medical claims data

Modeling Heterogeneity and Missing Data of Multiple Longitudinal Outcomes in Electronic Health Records

Updating Markov Models to Integrate Cross-Sectional and Longitudinal Studies.

Mixture of Input-Output Hidden Markov Models for Heterogeneous Disease Progression Modeling