C-A4-05: Design for Identification, Retrieval and Adjudication of Acute Myocardial Infarction Cases in the Mini-Sentinel Distributed Database

Background/Aims The goal of the FDA’s Mini-Sentinel pilot is to inform development of an active safety surveillance system, the Sentinel System, for monitoring medical product safety. Mini-Sentinel is a collaborative endeavor by Data and Academic Partners including the HMORN and Kaiser, designed to evaluate barriers to the creation of this surveillance system. Orchestrated via the Mini-Sentinel Coordinating Center (MSCC), the Acute Myocardial Infarction (AMI) Validation project is a test case for medical outcome validation within a distributed system designed for medical product safety surveillance.

Methods A 3-component process for validation of medical outcomes was used :

1. Case definition and identification: Development of a clinical definition of AMI, an algorithm for identifying likely AMI from Mini-Sentinel Distributed Databases (ICD-9 code-based) and a sampling strategy to select approximately 100 likely AMI cases for chart retrieval;

2. Chart retrieval: Agreement between the MSCC and Data Partners regarding:

   1. amount of de-identified medical information to be extracted ;

   2. mode of data transfer to the MSCC; and

   3. approach to maintaining the privacy and security of protected health information;

3. Abstraction and Adjudication: creation of

   1. abstraction forms for use by trained nurse abstractors; and

   2. an adjudication process for cardiologist reviewers.

Results Challenges encountered to date include:

1. the number of participating Mini-Sentinel Data Partners;

2. decision-making surrounding centralized vs. decentralized abstraction; and

3. concerns over privacy of transmitted health information and medical record de-identification for centralized review.

Ultimately, concern over quality and variability of abstraction at the Data Partners’ sites, together with small numbers of charts requested from each Data Partner, led to selection of a centralized abstraction process. Procedures needed to be established for this novel activity that addressed how to proceed with this project as a public health activity exempt from IRB review.

Conclusions The Sentinel System can yield timely data for active surveillance of medical product safety, and may be useful for other purposes, including research and quality improvement. Ongoing challenges include ensuring the minimum necessary amount of data is transmitted from Data Partners for centralized chart review and adhering to established procedures to maintain data security.