JMIR Publications

ABSTRACT

Background:

Distributed data networks (DDNs) combined with distributed regression analysis (DRA) can reduce the risk of disclosing sensitive individual and institutional information in multi-center studies. However, software and analytical packages that facilitate large-scale and efficient implementation of DRA are limited.

Objective:

To assess the precision and operational performance of a new SAS-based DRA application and a file transfer workflow within PopMedNetTM, an open-source distributed file-sharing software, within a horizontally partitioned DDN.

Methods:

We executed the application to perform distributed linear, logistic, and Cox proportional hazards regression analysis on a real-world test case with three data partners. We used PopMedNet to iteratively and automatically transfer highly summarized information between the data partners and the analysis center. We compared the DRA regression parameters, standard errors, model fit statistics, and model fit curves to the results obtained from standard SAS procedures executed on the pooled individual-level dataset. We computed the execution time of each step in the workflow to evaluate the operational performance of the application.

Results:

All DRA regression results were precise (< 10-12) and DRA model fit curves were identical or similar to those obtained from the corresponding pooled individual-level data analysis. All regression models required less than 20 minutes for full end-to-end execution.

Conclusions:

We integrated a SAS-based DRA application with an open-source file-sharing software and successfully tested the new capability within an active DDN. The study demonstrated the validity and feasibility of using DRA to enable privacy-protecting analysis in multi-center studies.