A simple Cox approach to estimating risk ratios without sharing individual-level data in multisite studies.

Area of Science:

• Epidemiology
• Biostatistics
• Data Science

Background:

• Epidemiologic studies commonly use risk ratios to assess exposure-outcome associations.
• Centralized individual-level data analysis is often hindered by privacy constraints among multiple data partners.
• Existing distributed analysis methods present limitations, including extensive data transfers or approximate estimations.

Purpose of the Study:

• To develop a practical, privacy-preserving method for estimating risk ratios in distributed epidemiologic studies.
• To provide accurate risk ratio estimates and confidence intervals comparable to pooled individual-level data analysis.
• To facilitate efficient data analysis across multiple data partners without centralizing sensitive information.

Main Methods:

• Leveraging a risk-set method and software originally designed for Cox regression.
• Requiring only a single transfer of 8 summary-level quantities from each data partner.
• Utilizing modified Poisson regression principles for accurate risk ratio estimation.

Main Results:

• The proposed method yields risk ratio estimates and 95% confidence intervals identical to pooled individual-level data analysis.
• The method was theoretically justified and validated using simulated data.
• Successful implementation in a distributed analysis of COVID-19 data from the FDA Sentinel System.

Conclusions:

• This novel approach offers a practical and accurate solution for distributed risk ratio estimation.
• It effectively addresses privacy concerns by sharing only summary-level data.
• The method enhances the feasibility of large-scale epidemiologic research across data-sharing collaborations.