Dear LAERTES Team, No meeting today because of the holiday. As for updates, please note the following:
Data updates:
@jon_duke has provided an update of SPLICER that I am working on loading into the proof of concept (POC)
@nick 's student Joseph Romano helped me get going with loading MEDLINE into a postgres db using code developed by @schuemie (see https://github.com/OHDSI/MedlineXmlToDatabase). After some tweaking, I got this working and will be working on loading MEDLINE data and SemMedDB into LAERTES
Dear Laertes Team - I am canceling the telecon for this coming week to give me more time to work on tidying up the evidence base and WebAPI in preparation for the AMIA CRI presentation. I will follow up by email with folks working on specific items. Also, I plan to email out some slides for the presentation for feedback.
All - sorry for the last minute change. I need to cancel this week’s meeting because of travel to ISPOR other concerns. Let’s plan for the next meeting to be June 3rd. The topics to discuss then will be the UI scenarios for Laertes and use cases. Please think about and comment on the following:
From Tomas:
Would it be possible to extend the API to allow counts by grouping on something else than the SNOMED concept ids, since our source data (the spontaneous reports in VigiBase) use MedDRA we would like the statistics to correspond to that terminology. If things are counted based on SNOMED the counts will not be representative if the counts themselves have to be aggregated to a MedDRA term from several different SNOMED terms.
For another use case that we have we would be interested in knowing the aggregated statistics (drug, condition, source_counts) for all combinations of drugs and conditions in LAERTES. Such a query would only have to be used very rarely but otherwise we would face the same problem as you have experienced with extracting information from the VigiBase API.
Scenarios:
Pharmacovigilance investigator
Katrina is trained as an epidemiologist and works full time at a center that monitors the safety of newly released drugs and old drugs that have received new indications. She typically receives suspected adverse drug events from spontaneous case reports that she evaluates using a structured algorithm like the Naranjo scale (you can see one in the last page of the attached). The first question the scale asks is “Are there previous conclusive reports on this reaction?”. Her current process is to look in a local database of spontaneous reports, review the ADR sections of the product label, and do a query of the published literature in Medline for case reports. She might also look at clinicaltrials.gov to see if any studies with the drug have recorded the adverse event and see if she can contact the PIs of those studies to learn more about the context.
Katrina would like her search to be more organized and simple to execute. She heard about the evidence base in OHDSI and decided to try it out. She uses Hermes one of two ways:
to identify the concept id for the HOI and then goes to the
evidence tab to see if the specific drug she is interested in has
been reported as potentially associated with the HOI. When she looks
at the tab, she is especially interested in the following:
— finding the drug quickly in the list of all drugs for which the
AE has been reported
— Identifying if the report is at the clinical drug or ingredient
level
— Identifying from which source the association was noted. She
cares mostly about product labeling, case reports from the
literature, and spontaneous reports. She would like to see a summary
of the evidence across these sources and then be able to click on a
link and review a bit more information about each. For the labeling,
she would like to see the source SPL and the section with the
adverse event highlighted. Literature, the abstract with (possibly
only for SemMed) the sentence mentioning the AE highlighted. For
Spontaneus reports, the counts and PRR (i.e., strength of signal),
and then when the reports came in, where from, and some demographics
on the cases
alternately, she might want to identify the concept id for the
drug and then goes to the evidence tab to see if the specific HOI
she is interested in has been reported as potentially associated
with the drug. Similar use cases apply.
quality improvement director for a health system
Eileen is a nurse with some training (certificate level) in
epidemiology. She is in charge of ensuring that certain health
quality measures QM are are at optimal levels in the health system
she serves (which currently has five NH facilities). She typically
receives QM reports each quarter. The current report shows that
anemia is at a very rate and she would like to determine what sort
of intervention is needed. Her typical process is to raise the issue
with her colleagues after looking in the literature for evidence of
what can be associated with the QM and if other clinical
interventions have been tested.
She heard about the evidence base in OHDSI and decided to try it
out. She wants uses Hermes because the QM “anemia” is not very
specific. She would like to see what sub-types of anemia there are
and if there is any evidence of drug assocation. She finds “anemia”
concept and goes to the evidence tab. When she looks at the tab, she
is especially interested in the following:
---- Is there evidence for drug association with the general concept
“anemia”?
---- How does that evidence apply when looking at the sub-concepts
of “anemia”
---- What is the list of drugs for which there is evidence for the
sub-concept of anemia she cares about?
— Identifying from which source the association was noted. She
cares mostly about product labeling, case reports, and clinical
studies in the literature. She would like to see a summary of the
evidence across these sources and then be able to click on a link
and review a bit more information about each.
---- She would like to use other OHDSI tools to determine with of
the drugs are prescribed at her sites and then design and
characterize the cohort of individuals who are exposed to anemia and
the drugs at her facilities to get a rough idea if the potential for
the intervention to work (see http://www.slideshare.net/boycer/piloting-acomprehensivepharmacovigilanceknowledgebase2015v2)