P10: An Automated, Scalable Data Monitoring Tool for Proactive Safety Quality Assurance
Director Digital Insights and Data Governance
Bristol-Myers Squibb Company United States
The objective of the present study was to develop a rule-based engine that enables automated and scalable identification of quality findings within a safety database to support prevention of, or rapid action on, quality findings
A safety quality monitoring portal/dashboard was created using a rule-based engine to scan the safety database to identify quality issues. Control was maintained by utilizing a workflow engine to document root cause analysis, impact analysis, corrective and preventative actions.
Using the rule-based engine, more than 22K cases with errors were identified and categorized across 45 defined error categories. Errors were categorized by granularity of records and assigned a risk priority. At the case level there were 13,116 error cases, and the biggest error category was “active case without delete reason” (n=7,830 cases; medium risk). At the event level there were 4,526 error cases, and the biggest error category was “serious event without serious criteria” (n=4,239 cases; medium risk). At the literature level there were 1,627 error cases, and the biggest error category was “non-literature case with literature information” (n=1,559; medium risk). At the product level there were 4,044 error cases, and the biggest error category was “code unbroken – unblinded product” (n=3,269 cases; high risk). Cumulatively, there were 22,939 error cases.
Data quality in pharmacovigilance is governed by regulations such as Guideline on Good Pharmacovigilance Practices (GVP) module 1. Currently, quality systems are designed to facilitate reactive review and correction of errors through routine quality assurance processes. A preventive approach to proactively monitor issues could facilitate identification of root causes and permit application of meaningful corrective actions to reduce operation costs and strengthen signal detection activities.
The novel tool described here provides greater insight into the quality of the Bristol Myers Squibb (BMS) safety database by isolating quality issues by product, study, region, classification, and other dimensions. This tool is being implemented as a self-service model which may allow BMS users to view trends and understand data holistically. Planning the implementation of this tool will facilitate our shift from a traditional reactive model of quality monitoring to a proactive model to improve efficiency of regulatory submission, reduce cost per case by reducing time spent for reconciliation, drive operational efficiency, and improve the accuracy of our scientific findings.