A Bayesian approach to adjust for diagnostic misclassification between two mortality causes in Poisson regression

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Counted data often suffers from response misclassification, leading to biased parameter estimates and underestimated uncertainty in Poisson regression models.
• Existing classical methods to address misclassification require asymptotic distribution results and supplemental validation data.

Purpose of the Study:

• To develop a novel Bayesian Poisson regression procedure that corrects for misclassification in two-category count variables.
• To provide an operationally effective method for accounting for misclassification effects within Poisson count regression models.

Main Methods:

• A new Bayesian Poisson regression procedure was derived to handle misclassified count data.
• The method incorporates validation data, expert opinion, or a combination thereof to correct misclassification parameters.
• Model performance was evaluated using a simulation study and analysis of two real-data examples.

Main Results:

• The proposed Bayesian method effectively corrects for misclassification bias and uncertainty in parameter estimation.
• Comparison with analyses ignoring misclassification demonstrated the superiority of the new Bayesian approach.
• The simulation study and real-data analyses confirmed the model's performance and utility.

Conclusions:

• The developed Bayesian procedure offers a robust and effective solution for addressing misclassification in Poisson count regression.
• This method improves the accuracy and reliability of parameter estimation when dealing with misclassified count data.
• The study highlights the importance of accounting for misclassification in statistical modeling for better scientific inference.