Disability evolution in multiple sclerosis: how to deal with missing transition times in the Markov model?

Area of Science:

• Neurology
• Biostatistics
• Medical Informatics

Background:

• Markov modeling is crucial for understanding multiple sclerosis (MS) disability progression.
• Accurate transition times between disability levels are essential but often missing in MS patient data.
• Missing data present methodological challenges in analyzing MS progression.

Purpose of the Study:

• To evaluate methods for handling partly missing transition times in MS disability progression.
• To estimate the impact of prognostic factors on the risk of transitions between consecutive disability levels in MS.
• To compare the performance of multiple imputation versus listwise deletion in MS progression analysis.

Main Methods:

• Employed listwise deletion, analyzing only complete datasets.
• Utilized multiple imputation, assuming a missing at random (MAR) mechanism and imputing times using a Weibull model.
• Compared imputation results against a full dataset derived from chart review.

Main Results:

• Multiple imputation estimates were consistently closer to the full dataset results than listwise deletion estimates.
• The study confirmed that data were missing at random (MAR).
• Multiple imputation proved more effective in capturing the true transition dynamics.

Conclusions:

• Multiple imputation is a superior method for addressing missing transition times in Markov modeling of MS disability.
• Accurate estimation of prognostic factors in MS progression is enhanced by using multiple imputation.
• This methodology improves the reliability of disability progression models in multiple sclerosis research.