Uniformization and bounded Taylor series in Newton-Raphson method improves computational performance for a multistate...

Area of Science:

• Biostatistics
• Health Informatics
• Computational Biology

Background:

• Multistate transition models (MSTMs) are essential for understanding disease progression.
• Estimating parameters and performing inference for MSTMs is computationally challenging with large, real-world datasets.

Purpose of the Study:

• To develop an efficient and accurate statistical estimation method for MSTMs.
• To address the computational challenges associated with complex MSTMs in large datasets.

Main Methods:

• A novel bounded Taylor series within a Newton-Raphson procedure is proposed.
• The uniformization technique is leveraged to derive maximum likelihood estimates and their covariance matrix.
• The method, termed uniformization Taylor-bounded Newton-Raphson, is validated through simulations.

Main Results:

• Simulation studies demonstrate high accuracy in parameter estimation.
• The method shows significant efficiency in computation time.
• The approach proves robust across various data scenarios.

Conclusions:

• The uniformization Taylor-bounded Newton-Raphson method provides an effective solution for MSTM statistical estimation.
• This method is applicable to large electronic medical record data, as shown in a statin side effects case study.