A joint logistic regression and covariate-adjusted continuous-time Markov chain model.

Area of Science:

• Biostatistics
• Medical Informatics
• Longitudinal Data Analysis

Background:

• Predicting categorical outcomes with longitudinal data is crucial in research.
• Joint models are used for correlated outcomes but limited for longitudinal predictors with categorical outcomes.
• Modeling longitudinal predictors to reflect biological mechanisms is challenging.

Purpose of the Study:

• To propose a novel joint logistic regression and Markov chain model.
• To predict binary cross-sectional responses using longitudinal predictors.
• To model unobserved transition rates of a two-state continuous-time Markov chain as covariates.

Main Methods:

• Developed a joint logistic regression and Markov chain model.
• Employed maximum likelihood estimation for parameter estimation.
• Validated the method through a simulation study.

Main Results:

• Simulation results demonstrated adequate estimation with high coverage probabilities and low bias.
• The model was applied to traumatic brain injury (TBI) patient data.
• Physiological changes over time improved the prediction of 6-month functional status in TBI patients.

Conclusions:

• The proposed joint model is effective for predicting categorical outcomes from longitudinal data.
• This approach enhances the prediction of long-term functional status in critically ill patients.
• Longitudinal physiological data offers valuable insights for outcome prediction in TBI.