Modeling zero-inflated count data when exposure varies: With an application to tumor counts.

Area of Science:

• Biostatistics
• Medical Statistics
• Epidemiology

Background:

• Count data frequently exhibit excess zeros, complicating standard statistical analysis.
• Varying exposure times are common in medical studies and can bias results if not properly accounted for.
• Existing zero-inflated models may not adequately capture the complexities of excess zeros with time-varying exposure.

Purpose of the Study:

• To develop and evaluate a novel zero-inflated count data model.
• To incorporate an underlying duration model with a Weibull hazard rate to address excess zeros.
• To compare the performance of the proposed model against the standard Poisson model with logit zero inflation.

Main Methods:

• Proposed a modified zero-inflated count data model.
• Integrated a duration model with a Weibull hazard rate to estimate the probability of excess zeros.
• Applied the models to analyze the effect of thiotepa treatment on bladder tumor incidence.

Main Results:

• The modified model demonstrated improved analysis of zero-inflated count data with varying exposure.
• The application showed differences in the estimated effect of thiotepa treatment compared to the standard model.
• The Weibull hazard rate component effectively captured excess zeros influenced by exposure duration.

Conclusions:

• The proposed modified zero-inflated model offers a more accurate approach for analyzing count data with excess zeros and time-varying exposure.
• This methodology enhances the statistical rigor in medical research, particularly in oncology studies.
• The model provides a valuable tool for understanding treatment effects in the presence of complex count data structures.