MSMpred: interactive modelling and prediction of individual evolution via multistate models

Area of Science:

• Biostatistics
• Medical Informatics
• Epidemiology

Background:

• Multistate models (MSM) are crucial for analyzing disease progression and identifying prognostic factors.
• Analyzing diseases with increasing severity or those preceding death requires complex MSMs.
• The complexity of MSMs necessitates user-friendly tools for broader application.

Purpose of the Study:

• To introduce MSMpred, a web-based tool designed to simplify the fitting and application of multistate models.
• To enable researchers and medical personnel to easily model disease trajectories and predict clinical outcomes.
• To enhance the interpretability of complex MSM analyses through intuitive visualizations.

Main Methods:

• Developed using the shiny R package, MSMpred allows users to fit MSMs from provided data.
• Users define states, transitions, and covariates (e.g., age, gender) for model fitting.
• The tool generates plots for covariate distributions and state-specific patient data.

Main Results:

• MSMpred facilitates prediction of clinical evolution for individual subjects.
• Key predictions include 30-day mortality probability and likely state at a future time point.
• Visualizations like stacked transition probabilities enhance the understanding of predictions.

Conclusions:

• MSMpred is an intuitive and visual application that simplifies working with multistate models.
• The tool supports biostatisticians in their modeling tasks.
• MSMpred aids medical personnel in interpreting complex MSM results for clinical application.