LAERTES: How to Present SPLICER's Adverse Reaction Labeling Statistics

[jon_duke]

Hi all,

SPLICER captures labeled adverse reactions from current and archived drug labels. So for any given “set_id”, there may be 1…n versions of the label. In each version, the adverse reaction may be mentioned multiple times.

So how would folks like to see this aggregated into a single stat that describes the DOI-HOI relationship? e.g.,

– A binary value for presence or absence in any version of the label
– How many times the HOI mentioned in the current version of the label
– A value reflecting the combination of all set_id’s for a given ingredient, like 13/15 set_id’s for this ingredient contain the HOI. So the value would be a ratio on the 0-1 scale.

The LAERTES team would welcome your input.

Jon

[jon_duke]

I have made some edits to the suggested options:

How would folks like to see this aggregated into a single stat that describes the DOI-HOI relationship? e.g.,

– A binary value denoting presence or absence of the HOI in any version the label
– A sum of the total mentions of the HOI in the current version of the label
– A ratio reflecting the proportion of set_id’s for a given ingredient that contain the HOI, so if it was found in 13/15 set_id’s, the value would be 0.83.

[rkboyce]

I think I like the 1/0 option because I think its kind of hard to interpret what the other numbers actually mean in terms of evidence for an drug-HOI association. The architecture of the evidence base supporting LAERTES should be able to support a variety of “drill down” summaries through the WebAPI that could help users see more detailed info such as when the ADR first appeared, sections where mentions occur, and how complete across all labels current. -R