Multi-state models for changes in a state (e.g., of health or illness) often assume that the transition rate is constant with time spent in the state (the “Markov” assumption). A new method and software package for addressing this often unrealistic assumption, msmbayes, was developed by Christopher Jackson, who presented the method in a recent CHDS seminar. Jackson is a Senior Statistician at the MRC Biostatistics Unit at the University of Cambridge, UK.
Rather than assume that transition rates remain constant regardless of how long an individual has been in a given state, msmbayes uses hidden states, technically referred to as “phases.” This approach extends Jackson’s msm package by adding support for Bayesian inference and phase-type semi-Markov models. It also improves posterior stability and scalability across a larger number of health states.
The development of msmbayes was motivated by the need for more accurate modeling in critical health areas, particularly where the standard assumption of a constant transition rate falls short. The package enables researchers to model more precisely the duration and incidence of infectious diseases, time spent in detectable cancer states, cancer incidence, and cognitive impairment in aging studies.
By capturing the complexities of disease incidence and progression more realistically, msmbayes offers the modeling community a powerful new tool for studying health outcomes.
Learn more: Explore Jackson’s related work on GitHub
Learn more: Join our CHDS mailing list to be notified about future CHDS seminars.
Related news: Propagating Ambiguity into Decision Analyses
Related news: Setting Optimal Test Thresholds