Learn about calculations, scores, and variables that inform signal detection.
About Signal Detection Calculations
Safety Signal considers various scores and calculations to identify Product-Event Combinations (PECs) of interest. This allows for the detection of new and unexpected adverse events for a medicinal product and changes to an existing PEC.
For Vault to detect a signal, PECs must meet certain criteria and surpass specific thresholds. Your Admin configures the criteria and thresholds based on your organization’s business rules.
Case Counts
Vault counts the number of post-market Cases that are spontaneous, fatal, and serious for the related PEC.
For more details on Case counts, see Signal Detection Using PV Data.
Listedness
From the Case Assessment level Listedness (Core) field, Vault determines an event is Unlisted for the Product if at least one (1) Case Assessment is evaluated as Unlisted or there is no evaluation (Listedness (Core) is blank). To consider a PEC as Listed, all Case Assessments for that PEC must be evaluated as Listed.
Empirical Bayes Geometric Mean (EBGM)
The Empirical Bayes Geometric Mean (EBGM) is an advanced statistical method for measuring how frequently an adverse event occurs for a Product of interest compared to how often it occurs for all other Products in the database.
The EBGM calculation offers a high degree of confidence even among smaller sample sizes. EBGM scores are supplemented by the EB05 and EB95, the lower and upper 90% confidence limits.
EB05 and EB95 Scores
EB05 and EB95 scores are used as lower and upper confidence bounds, respectively, for the EBGM 90% confidence interval. The confidence interval outlines the uncertainty in a PEC estimate to address small sample sizes.
EB05 represents the lower bound of the EBGM score and indicates there is a 5% probability the true value of the EBGM score is lower than the calculated value.
EB95 represents the upper bound of the EBGM score, and indicates there is a 5% probability the true value of the EBGM score is greater than the calculated value.
Stratification
Stratification is a method for normalizing data to eliminate confounding variables that can influence the probability of experiencing an adverse event when taking a given product. For example, to eliminate the potential confounding effects of gender, the stratification method studies the data for each gender and then pools the scores together.
The following diagram demonstrates an example of how Safety Signal stratifies data by gender:
