Discrete proportional hazards models for mismeasured outcomes

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Outcome mismeasurement can introduce bias in statistical analyses.
• Previous research highlighted bias in logistic regression and specific survival analysis cases.
• Accurate estimation requires addressing imperfect outcome measures.

Purpose of the Study:

• To introduce a general and widely applicable adjusted proportional hazards (APH) method.
• To address estimation of cumulative survival and hazard ratios with mismeasured outcomes in discrete time.
• To evaluate the performance of the APH method under various conditions.

Main Methods:

• Development of the adjusted proportional hazards (APH) method for discrete time survival data.
• Simulation studies to assess bias in standard proportional hazards (PH) models with mismeasured failure status.
• Comparison of APH method performance against standard PH models.

Main Results:

• Failure to adjust for outcome mismeasurement in proportional hazards (PH) models can lead to conservative bias in hazard ratio estimation.
• Poor specificity of outcome measurement generally has a more severe impact on inference than poor sensitivity.
• The APH method, when mismeasurement rates are correctly specified, demonstrates strong performance across a range of simulation conditions.

Conclusions:

• The adjusted proportional hazards (APH) method provides a robust approach for survival analysis with mismeasured outcomes.
• Accounting for outcome mismeasurement is crucial for reliable estimation of survival and hazard ratios.
• The APH method offers a valuable tool for biostatisticians and epidemiologists dealing with imperfect data.