"SPARKLIS over Pegase Knowledge Graph: A New Tool for Pharmacovigilance"

This page contains the additional material for the paper "SPARKLIS over Pegase Knowledge Graph: A New Tool for Pharmacovigilance" by Carlos Bobed, Laura Douze, Sébastien Ferré, and Romaric Marcilly.

In particular, the videos included in this page have been elaborated during the evaluation of SPARKLIS over Pegase Knowledge Graph based on the evaluation of the cognitive path of the users. Such evaluation was carried out after the reengineering process performed to adapt SPARKLIS to the pharmacovigilance use case to identify remaining usability issues.

The particular queries presented in these videos (Use Cases 1-5) are use cases detected by our team of ergonomists in their study of how pharmacovigilance specialists perform their daily work, as well as the building of a SMQ (Extra Use Case) which is aligned to one of the Trifiro's list of 23 first priority pharmacovigilance safety topics¹.

Trifirò, G. , Pariente, A. , Coloma, P. M., Kors, J. A., Polimeni, G. , Miremont-Salamé, G. , Catania, M. A., Salvo, F. , David, A. , Moore, N. , Caputi, A. P., Sturkenboom, M. , Molokhia, M. , Hippisley-Cox, J. , Acedo, C. D., van der Lei, J. and Fourrier-Reglat, A., Data mining on electronic health record databases for signal detection in pharmacovigilance: which events to monitor? Pharmacoepidem. Drug Safe., 18: 1176-1184 (2009), doi:10.1002/pds.1836

Use Case 1: Looking for ADEs related to Cardiac Valvulopathy - Knowing previously the PTs


	Scenario Description The user wants to find how many adverse events related to cardiac valvulopathy there are in the database. The user knows previously the PT that she wants to use. Actions [~00:05] The user selects the main type/class of its search (Fiche Dense - cases in the database). This changes the query (give me every Fiche Dense) and the focus (highlighted in green). The focus dictates the suggestions and modification that the user can apply to the current query. [~00:10] The user navigates trough the possible properties of Fiche Dense to find effet indésirable (MedDRA) and selects it. This modifies the query adding a constraint on Fiche Dense (to have an effet indésirable) and moves the focus onto the property. Now the suggestions are related to the values that a Fiche Dense can have for effet indésirable. [~00:15] The user navigates through the possible values of effet indésirable. [~00:18] As the user knows the PT she is looking for, the user directly inputs a search string in the keyword search form (cardiaque). This filters the suggestions according to the keywords matched. [~00:21] The user navigates the filtered results, and selects Valvulopathie cardiaque. [~00:29] As the user want to count the number of Fiche Dense that hold the query, she changes the focus to Fiche Dense. [~00:31] The user applies the aggregator operator (the number of _) to the current focus (the whole query).

Scenario Description

The user wants to find how many adverse events related to cardiac valvulopathy there are in the database.
The user knows previously the PT that she wants to use.

Actions

[~00:05] The user selects the main type/class of its search (Fiche Dense - cases in the database). This changes the query (give me every Fiche Dense) and the focus (highlighted in green). The focus dictates the suggestions and modification that the user can apply to the current query.
[~00:10] The user navigates trough the possible properties of Fiche Dense to find effet indésirable (MedDRA) and selects it. This modifies the query adding a constraint on Fiche Dense (to have an effet indésirable) and moves the focus onto the property. Now the suggestions are related to the values that a Fiche Dense can have for effet indésirable.
[~00:15] The user navigates through the possible values of effet indésirable.
[~00:18] As the user knows the PT she is looking for, the user directly inputs a search string in the keyword search form (cardiaque). This filters the suggestions according to the keywords matched.
[~00:21] The user navigates the filtered results, and selects Valvulopathie cardiaque.
[~00:29] As the user want to count the number of Fiche Dense that hold the query, she changes the focus to Fiche Dense.
[~00:31] The user applies the aggregator operator (the number of _) to the current focus (the whole query).

Use Case 2: Looking for ADEs related to Cardiac Valvulopathy - Searching for the appropriate PTs


	Scenario Description The user wants to find how many adverse events related to cardiac valvulopathy there are in the database. The user does not know previously the PT that she wants to use, and she has to look for it previously. Actions [~00:01] The user unfolds the hierarchy of the types of med DRA term. [~00:03] The user selects PT to state that she wants to look for preferred terms. [~00:05] The user selects finding site property to establish a constraint on the PTs retrieved. The suggestions for the finding site are arranged in a hierarchy of SNOMED CT concepts. [~00:08] The user unfolds the different upper levels of SNOMED CT hierarchy, browsing for the particular location she is looking for. [~00:11] As SNOMED CT hierarchy is quite deep, the user uses the free text input to narrow the search. She inputs heart, which filters the results and modifies the results. [~00:14] The user navigates the filtered results, unfolding part of the the subhierarchy, and selects Heart structure (body structure). Note how after the filtering, there is still the possibility of navigating through the resulting subhierarchies. [~00:17] The user modifies the focus to show the PTs that hold the query (recall that the results and suggestions are always referring to the current focus - highlighted in green in the query). [~00:20] The user filters the suggestions (valvulo...) to find the exact preferred term she was looking for, Valvulopathie cardiaque, and she selects it. [~00:28] After selecting the term, the query has changed, and the user wants just to keep the term to start another query. To do so, she has to remove the different constraints applied in the current query. First, she moves the focus to the inner element (Heart structure (body structure)) and removes it by clicking the red cross. [~00:30] Removal of the next focus (finding site). [~00:32] Now, the user has the exact term she was looking for, and the user starts the search backwards (from the term to the associated Fiche Dense). The user selects the is the effet indésirable (MedDRA) property, which is the inverse of has effet indésirable present for Fiche Dense. [~00:41] The user establishes that she wants to retrieve Fiche Dense, and, by selecting it, the focus moves to it. [~00:44] Finally, the user applies the aggregator operator (the number of _) to the current focus (Fiche Dense) to count how many there are.

Scenario Description

The user wants to find how many adverse events related to cardiac valvulopathy there are in the database.
The user does not know previously the PT that she wants to use, and she has to look for it previously.

Actions

[~00:01] The user unfolds the hierarchy of the types of med DRA term.
[~00:03] The user selects PT to state that she wants to look for preferred terms.
[~00:05] The user selects finding site property to establish a constraint on the PTs retrieved. The suggestions for the finding site are arranged in a hierarchy of SNOMED CT concepts.
[~00:08] The user unfolds the different upper levels of SNOMED CT hierarchy, browsing for the particular location she is looking for.
[~00:11] As SNOMED CT hierarchy is quite deep, the user uses the free text input to narrow the search. She inputs heart, which filters the results and modifies the results.
[~00:14] The user navigates the filtered results, unfolding part of the the subhierarchy, and selects Heart structure (body structure). Note how after the filtering, there is still the possibility of navigating through the resulting subhierarchies.
[~00:17] The user modifies the focus to show the PTs that hold the query (recall that the results and suggestions are always referring to the current focus - highlighted in green in the query).
[~00:20] The user filters the suggestions (valvulo...) to find the exact preferred term she was looking for, Valvulopathie cardiaque, and she selects it.
[~00:28] After selecting the term, the query has changed, and the user wants just to keep the term to start another query. To do so, she has to remove the different constraints applied in the current query. First, she moves the focus to the inner element (Heart structure (body structure)) and removes it by clicking the red cross.
[~00:30] Removal of the next focus (finding site).
[~00:32] Now, the user has the exact term she was looking for, and the user starts the search backwards (from the term to the associated Fiche Dense). The user selects the is the effet indésirable (MedDRA) property, which is the inverse of has effet indésirable present for Fiche Dense.
[~00:41] The user establishes that she wants to retrieve Fiche Dense, and, by selecting it, the focus moves to it.
[~00:44] Finally, the user applies the aggregator operator (the number of _) to the current focus (Fiche Dense) to count how many there are.

Use Case 3: Looking for cases of Paresthesia related to Anafranil - Knowing previously the PTs


	Scenario Description The user wants to know if there is any case of "paresthesia" related to "Anafranil" reported in the database. The user knows previously the PT that she wants to use. Actions [~00:07] The user selects the main type/class of its search (Fiche Dense - cases in the database). [~00:11] The user navigates trough the possible properties of Fiche Dense to find effet indésirable (MedDRA) and selects it. [~00:15] As the user knows the PT she is looking for, the user directly inputs a search string in the keyword search form (paresth...). This filters the suggestions according to the keywords matched. [~00:21] The user navigates selects Parésthesie. This builds the part of the query give me every Fiche Denses whose effect indérisable is Paresthésie. [~00:24] The user changes the focus to Fiche Dense to add more constraints to the fiche denses that she wants to retrieve. [~00:27] The user selects the property a prise de médicament associée to constrain Fiche Dense. This allows the user to establish which kind of medicament she is looking for. [~00:29] The user selects the property nom to access the name of the medicament associated to the adverse effect. [~00:34] The user looks for Anafranil among the names of the medicaments (note that they are only those associated to patients that have had Paresthésie as an adverse effect). [~00:39] The user selects Anafranil and sees that there are two cases.

Scenario Description

The user wants to know if there is any case of "paresthesia" related to "Anafranil" reported in the database.
The user knows previously the PT that she wants to use.

Actions

[~00:07] The user selects the main type/class of its search (Fiche Dense - cases in the database).
[~00:11] The user navigates trough the possible properties of Fiche Dense to find effet indésirable (MedDRA) and selects it.
[~00:15] As the user knows the PT she is looking for, the user directly inputs a search string in the keyword search form (paresth...). This filters the suggestions according to the keywords matched.
[~00:21] The user navigates selects Parésthesie. This builds the part of the query give me every Fiche Denses whose effect indérisable is Paresthésie.
[~00:24] The user changes the focus to Fiche Dense to add more constraints to the fiche denses that she wants to retrieve.
[~00:27] The user selects the property a prise de médicament associée to constrain Fiche Dense. This allows the user to establish which kind of medicament she is looking for.
[~00:29] The user selects the property nom to access the name of the medicament associated to the adverse effect.
[~00:34] The user looks for Anafranil among the names of the medicaments (note that they are only those associated to patients that have had Paresthésie as an adverse effect).
[~00:39] The user selects Anafranil and sees that there are two cases.

Use Case 4: Looking for cases of Paresthesia related to Anafranil AND/OR Tercian - Knowing previously the PTs


	Scenario Description The user wants to know if there is any case of "paresthesia" related to "Anafranil AND/OR Tercian" reported in the database. The user knows previously the PT that she wants to use. Actions Until [~00:27] the user performs the same actions as in the previous query (Use Case 3). [ ~00:28] The user moves the focus to ANAFRANIL. [~00:29] The user adds another possible name to the medicament related to Paresthésie. In this way, the user extends the possible names to check whether there exists any Fiche Dense associated to Paresthésie that is related to Tercian as well. [~00:33] The user checks whether there is any entry associated to Tercian. SPARKLIS suggests nothing, but anything; which logically means that there is no possible value (there is not any reported case of Tercian causing Paresthesia). [~00:35] The user just removes the extension of the constraint to have the same query as in Use Case 3. [~00:42] The user moves the focus and counts the number of cases.

Scenario Description

The user wants to know if there is any case of "paresthesia" related to "Anafranil AND/OR Tercian" reported in the database.
The user knows previously the PT that she wants to use.

Actions

Until [~00:27] the user performs the same actions as in the previous query (Use Case 3).
[ ~00:28] The user moves the focus to ANAFRANIL.
[~00:29] The user adds another possible name to the medicament related to Paresthésie. In this way, the user extends the possible names to check whether there exists any Fiche Dense associated to Paresthésie that is related to Tercian as well.
[~00:33] The user checks whether there is any entry associated to Tercian. SPARKLIS suggests nothing, but anything; which logically means that there is no possible value (there is not any reported case of Tercian causing Paresthesia).
[~00:35] The user just removes the extension of the constraint to have the same query as in Use Case 3.
[~00:42] The user moves the focus and counts the number of cases.

Use Case 5: Looking for cases of Peripheral Demyelination related to Amiodarone - Knowing previously the PTs


	Scenario Description The user wants to know if there is any case of "peripheral demyelination" related to "Amiodarone" reported in the database. The user knows previously the PT that she wants to use. Actions [~00:04] The user looks for the PT associated to demyelination. The user selects PT, and then, filters with the free text input. [~00:12] The user selects Démyélinisation. [~00:20] The user selects is the effet indésirable (MedDRA) to navigate towards the patient records. [~00:24] The user selects Fiche Dense to be more accurate in the query construction. [~00:36] The user selects a prise de médicament associée to narrow the patient records she wants to specify. [~00:38] The user selects nom to be able to check whether Amiodarone is among the medicaments that are associated to Démyélinisation. [~00:45] The user checks whether there are any answer (any patient record) using the text matching filtering.

Scenario Description

The user wants to know if there is any case of "peripheral demyelination" related to "Amiodarone" reported in the database.
The user knows previously the PT that she wants to use.

Actions

[~00:04] The user looks for the PT associated to demyelination. The user selects PT, and then, filters with the free text input.
[~00:12] The user selects Démyélinisation.
[~00:20] The user selects is the effet indésirable (MedDRA) to navigate towards the patient records.
[~00:24] The user selects Fiche Dense to be more accurate in the query construction.
[~00:36] The user selects a prise de médicament associée to narrow the patient records she wants to specify.
[~00:38] The user selects nom to be able to check whether Amiodarone is among the medicaments that are associated to Démyélinisation.
[~00:45] The user checks whether there are any answer (any patient record) using the text matching filtering.

Extra Use Case: Building a Query Equivalent to the SMQ - HLT: Bullous conditions


	Scenario Description The user wants to retrieve all the medDRA terms associated to the formal definition of an SMQ which has been evaluated to be equivalent to the HLT "Bullous conditions". The user does not know previously the terms. The user is familiar with the SNOMED CT terms needed to define the medDRA terms she wants. Actions [~00:04] The user selects the type of elements she is looking for (medDRA term). [~00:7] The user narrows the search by constraining the medDRA terms to have a finding site by selecting the property has a finding site. This shows the hierarchy of SNOMED CT concepts in the suggestions (collapsed) [~00:12] The user filters the SNOMED CT hierarchy using the text filter (Skin...) and selects Skin AND subcutaneous tissue structure (body structure). [~00:16] The user moves the focus to med DRA term to establish new constraints. [~00:18] The user selects has associated morphology to further narrow the search of med DRA terms. [~00:23] The user filters the SNOMED CT hierarchy to select Blister (morphologic abnormality). [~00:29] The user moves the focus to Blister... to expand the query and the possible answers. [~00:32] The user selects to refine the condition using OR operator, which makes the SNOMED CT hierarchy of suggestions be shown again. [~00:35] The user filters the hierarchy of suggestions using vesic... and selects multiple elements by keeping Control key pressed while selecting them. [~00:37] The user decides to group the selected terms using OR. [~00:42] The user moves the focus to medDRA term to see the results of the query.

Scenario Description

The user wants to retrieve all the medDRA terms associated to the formal definition of an SMQ which has been evaluated to be equivalent to the HLT "Bullous conditions".
The user does not know previously the terms.
The user is familiar with the SNOMED CT terms needed to define the medDRA terms she wants.

Actions

[~00:04] The user selects the type of elements she is looking for (medDRA term).
[~00:7] The user narrows the search by constraining the medDRA terms to have a finding site by selecting the property has a finding site. This shows the hierarchy of SNOMED CT concepts in the suggestions (collapsed)
[~00:12] The user filters the SNOMED CT hierarchy using the text filter (Skin...) and selects Skin AND subcutaneous tissue structure (body structure).
[~00:16] The user moves the focus to med DRA term to establish new constraints.
[~00:18] The user selects has associated morphology to further narrow the search of med DRA terms.
[~00:23] The user filters the SNOMED CT hierarchy to select Blister (morphologic abnormality).
[~00:29] The user moves the focus to Blister... to expand the query and the possible answers.
[~00:32] The user selects to refine the condition using OR operator, which makes the SNOMED CT hierarchy of suggestions be shown again.
[~00:35] The user filters the hierarchy of suggestions using vesic... and selects multiple elements by keeping Control key pressed while selecting them.
[~00:37] The user decides to group the selected terms using OR.
[~00:42] The user moves the focus to medDRA term to see the results of the query.