A computer graphical user interface for survival mixture modelling of recurrent infections

Area of Science:

• Biostatistics
• Epidemiology
• Medical Statistics

Background:

• Recurrent infection data present unique analytical challenges due to event dependencies.
• Existing models may not adequately capture the distinct phases (acute and stable) of recurrent events.

Purpose of the Study:

• To develop and present flexible two-component survival mixture models for analyzing recurrent infection data.
• To incorporate random effects to account for the dependency in recurrent observations.
• To provide a user-friendly implementation for practical application.

Main Methods:

• Utilized two-component survival mixture models in proportional hazards and accelerated failure time settings.
• Incorporated random effects within the conditional hazard function, akin to generalized linear mixed models.
• Developed an EM algorithm for parameter estimation, assuming Weibull or log-logistic baseline hazards.
• Implemented the methodology with a graphical user interface in Microsoft Visual C++.

Main Results:

• Demonstrated the application of the survival mixture methodology to model recurrent urinary tract infections in elderly women.
• Identified significant individual variations in both acute and stable phases of recurrent infections.
• The model effectively handles correlated and heterogeneous survival data.

Conclusions:

• The developed survival mixture methodology offers a flexible approach for analyzing recurrent infection data.
• This method enables practitioners to identify key risk factors influencing recurrent event times.
• Valid conclusions can be drawn from complex, correlated, and heterogeneous survival data, improving clinical insights.