•  
  •  
 

Retrieving Medical Records Within FDA’s Sentinel Distributed Network: Lessons Learned During a Protocol-Based Assessment Involving 13 Data Partners

Publication Date

8-10-2017

Keywords

observational studies, research administration, contracting, operations

Abstract

Background: The U.S. Food and Drug Administration’s Sentinel system has developed the capability to conduct active safety surveillance of marketed medical products in a large distributed network. Medical record retrieval in such an environment is logistically challenging and resource-intensive. We describe lessons learned during a protocol-based assessment of thromboembolic events (TEE) following intravenous immunoglobulin (IGIV) administration, a project that necessitated chart retrieval requests for a large number of patients from 13 Sentinel data partners.

Methods: The Sentinel IGIV-TEE workgroup requested medical records pertaining to 442 potential post-IGIV-TEE cases identified from the Sentinel Distributed Database. Charts were requested from 13 different data partners (including 4 Health Care Systems Research Network members). Six partners were claims-based health plans (CBHPs), and 7 were integrated care delivery systems (ICDSs). Dates of medical records spanned March 2006 to February 2013. Records were received and tracked by the Sentinel Operations Center at Harvard Pilgrim Health Care Institute. We report on the retrieval process, chart retrieval rates and reasons identified by data partners for nonretrieval by data partner type.

Results: The vast majority of cases identified for record retrieval were from CBHPs (88%). Retrieval processes varied by data partner type: 4 of the 7 CBHPs contracted with vendors to conduct record retrieval across their facilities, whereas the ICDSs relied on internal personnel to either electronically pull charts or seek paper records. These differing processes had implications for troubleshooting missing records and information, resulting in diverse strategies for data partner-facility consultation. Rates of retrieval varied by data partner type (64% in CBHPs vs 94% in ICDSs) as did speed of retrieval and resources. Reasons records were unobtainable fell into three major categories: inability to map Sentinel Distributed Database records to patient and provider identifiers needed for retrieval; provider refusal to participate due to legal/compliance/HIPAA concerns; and missing information (eg, no record of patient at facility, charts for the requested service dates were unavailable).

Conclusion: We will describe key lessons learned, broadly applicable to any chart retrieval project in a distributed environment, and include a discussion focused on the pros and cons of working with multiple collaborators with different data environments.

Share

COinS
 

Submitted

June 26th, 2017

Accepted

August 10th, 2017