A simple approach to the estimation of incidence rate difference

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• The incidence rate difference (IRD) is vital for evaluating intervention impact, particularly in vaccine efficacy studies.
• Estimating IRD can be challenging due to quantitative covariates and variable follow-up durations.
• Accurate IRD estimation is critical for understanding public health intervention effectiveness.

Purpose of the Study:

• To propose a novel, easily implementable weighted least-squares regression approach for estimating the incidence rate difference (IRD).
• To extend the estimation method to accommodate repeatable outcome events, analyzing all events instead of only the first.
• To validate the proposed method's performance through simulation studies and real-world data application.

Main Methods:

• Weighted least-squares regression applied to transformed variables for IRD estimation.
• Ordinary least-squares regression on transformed data for simplified implementation.
• Consideration of four Huber-White robust standard error versions for statistical inference.
• Analysis of all events for repeatable outcomes, not just first events.

Main Results:

• The proposed method yields unbiased incidence rate difference (IRD) estimates across various simulated scenarios.
• Empirical coverage proportion of the 95% confidence intervals closely matches the nominal level, indicating reliability.
• The method demonstrated practical utility when applied to data from a Gambia-based vaccine trial.

Conclusions:

• The developed weighted least-squares method offers a robust and accessible tool for estimating the incidence rate difference (IRD) in medical research.
• This approach simplifies complex IRD estimation challenges, enhancing the assessment of public health interventions.
• The method's accuracy and reliability are confirmed by simulations and a real-world vaccine trial example.