AIM-AHEAD Federated-Network Information

Due to institutional policies, sites with diverse data are often unable to share patient-level data outside of their institutions. Federated data networks allow for patient level-data to stay at the local site and the ability for sites to opt-in to participate; allowing for more diverse data and research while promoting data privacy and safety. This model also allows for greater inclusion and equity as sites that have not been able to engage in collaborations involving data sharing can participate in a federated model. Participating sites will receive assistance to set up local federated infrastructure and will be provided with centralized code to run on their data locally. Sites share ONLY aggregate data results as they feel comfortable. Aggregate data will be shared through use of a private GitHub repository so that they can be combined centrally and then shared with other network participants. The private Github repository will be password protected. ONLY AGGREGATE data will be shared, patient-level data will not be shared or stored outside of the local site. Easy-to-use dockerized environments can be set up at each participating site with data science tool kits to analyze electronic health record (EHR) data.

The AIM-AHEAD Federated Network allows sites to participate in collaborative research and share data across sites without the need for patient-level data to leave the local site. The program is open to institutions conducting research at the intersection of healthcare practice and delivery, AI/ML methods, and the advancement of health equity. The AIM-AHEAD Consortium seeks projects that represent diversity across fields of study, institutions, geography, and investigators.

This RFA calls for site participation in AIM-AHEAD federated research programs. The research team from each site is expected to (i) assemble EHR data locally from their institution to be used for AIM-AHEAD research projects; (ii) provide informatics expertise to participate in EHR data harmonization across sites; and (iii) facilitate the analysis of EHR data using AI/ML algorithms developed by the AIM-AHEAD researchers. While all patient-level EHR data remain behind the firewalls of the local site, the research team is expected to facilitate the sharing of aggregated data to support AIM-AHEAD research projects.

Expected outcomes for the participating site include protocols for data curation and data sharing fine tuned to their specific institution along with lessons learned. The sites are expected to work with the core team members to assess capacity in running AI/ML models locally and sharing outputs of AI/ML models. During year 2 of the funding cycle, the participating sites are expected to generate and share aggregated data for relevant research questions to enable the research teams to perform federated analyses. 

The participating sites and the AIM-AHEAD Infrastructure Core are expected to jointly work on publications detailing both the process of data curation, data harmonization, AI/ML analyses with local data, as well as federated analyses for selected research questions.