Healthcare Adverse Events (FAERS)

Cypher EDA and analytics on an FDA adverse-event reporting graph, plus GDS Jaccard patient-journey similarity and Leiden sub-phenotypes — curated subgraphs only; no live database at runtime.

Source: healthcare-analytics-50.dump · 11,381 nodes · 61,453 relationships · offline GDS precompute

11381

Nodes

61453

Relationships

4307

Cases

2701

Distinct reactions

2500

Drugs

218

Top severe drug (REVLIMID)

617

Top manufacturer (PFIZER)

Nodes by label

Case4,307
Reaction2,701
Drug2,500
Therapy1,721
Manufacturer136
Outcome6
AgeGroup5
ReportSource5

Relationships by type

HAS_REACTION21,069
IS_CONCOMITANT8,393
PRESCRIBED5,441
RESULTED_IN4,461
FALLS_UNDER4,307
REGISTERED4,307
REPORTED_BY4,307
IS_PRIMARY_SUSPECT3,830
IS_SECONDARY_SUSPECT3,616
RECEIVED1,721
IS_INTERACTING1

Graph model

Cases sit at the center: manufacturers register them, drugs appear as suspects, reactions and outcomes hang off each case, and age groups provide demographics. (Therapy / ReportSource exist in the dump but are omitted from this schema sketch.)

REGISTEREDIS_PRIMARY_SUSPECTHAS_REACTIONRESULTED_INFALLS_UNDERManufacturerCaseDrugReactionOutcomeAgeGroup

Schema discovery (EDA)

Frozen Cypher from the notebook’s first pass: labels, relationship types, genders, reactions, outcomes.

Total nodes

Overall size of the restored FAERS-style graph.

1 row
MATCH (n)
RETURN count(n) AS total_nodes;
total_nodes
11381

Node labels

Eight entity types discovered via schema EDA.

8 rows
CALL db.labels() YIELD label
RETURN label
ORDER BY label;
label
AgeGroup
Case
Drug
Manufacturer
Outcome
Reaction
ReportSource
Therapy

Total relationships

Directed edge count across all relationship types.

1 row
MATCH ()-[r]->()
RETURN count(r) AS total_relationships;
total_relationships
61453

Relationship types

Eleven relationship types linking cases, drugs, reactions, and outcomes.

11 rows
MATCH ()-[r]->()
RETURN DISTINCT type(r) AS relationship_type
ORDER BY relationship_type;
relationship_type
FALLS_UNDER
HAS_REACTION
IS_CONCOMITANT
IS_INTERACTING
IS_PRIMARY_SUSPECT
IS_SECONDARY_SUSPECT
PRESCRIBED
RECEIVED
REGISTERED
REPORTED_BY
Showing 1–10 of 11 rows

Case genders

Distinct gender values on Case nodes.

3 rows
MATCH (c:Case)
WHERE c.gender IS NOT NULL
RETURN DISTINCT c.gender AS gender
ORDER BY gender;
gender
F
M
U

Distinct reactions

Unique adverse-reaction descriptions in the graph.

1 row
MATCH (r:Reaction)
WHERE r.description IS NOT NULL
RETURN count(DISTINCT r.description) AS distinct_reactions;
distinct_reactions
2701

Outcome descriptions

Seriousness categories used on Outcome nodes.

6 rows
MATCH (o:Outcome)
WHERE o.outcome IS NOT NULL
RETURN DISTINCT o.outcome AS outcome_description
ORDER BY outcome_description;
outcome_description
Congenital Anomaly
Death
Disability
Hospitalization - Initial or Prolonged
Life-Threatening
Other Serious (Important Medical Event)

Drug-safety analytics

Deeper questions: top reactions, drugs tied to severe outcomes, manufacturer footprint, and a PFIZER drill-down.

Top 20 reactions

Most frequent adverse reactions across all cases.

20 rows
MATCH (c:Case)-[:HAS_REACTION]->(r:Reaction)
WHERE r.description IS NOT NULL
RETURN r.description AS reaction, count(*) AS frequency
ORDER BY frequency DESC
LIMIT 20;
reactionfrequency
Fatigue303
Product dose omission issue285
Headache272
Nausea256
Pain253
Dyspnoea245
Pneumonia229
Diarrhoea219
Fall198
Off label use196
Showing 1–10 of 20 rows

Top drugs with severe outcomes

Drugs linked as primary/secondary suspect to cases with Death, Life-Threatening, Disability, or Hospitalization.

10 rows
MATCH (d:Drug)<-[:IS_PRIMARY_SUSPECT|IS_SECONDARY_SUSPECT]-(c:Case)
MATCH (c)-[:RESULTED_IN]->(o:Outcome)
WHERE o.outcome IN $severe
RETURN d.name AS drug_name, count(DISTINCT c) AS severe_cases
ORDER BY severe_cases DESC
LIMIT 10;
drug_namesevere_cases
REVLIMID218
NIVOLUMAB82
ATEZOLIZUMAB77
HUMAN NORMAL IMMUNOGLOBULIN; LIQUID66
POMALYST65
DEXAMETHASONE65
CYCLOPHOSPHAMIDE64
CUVITRU61
REMODULIN57
Teduglutide53

Top manufacturers by drugs with side effects

Manufacturers ranked by distinct drugs appearing on their registered cases that have reactions.

10 rows
MATCH (m:Manufacturer)-[:REGISTERED]->(c:Case)-[:HAS_REACTION]->(r:Reaction)
MATCH (d:Drug)<-[:IS_PRIMARY_SUSPECT|IS_SECONDARY_SUSPECT|IS_CONCOMITANT|IS_INTERACTING]-(c)
RETURN m.manufacturerName AS manufacturer, count(DISTINCT d) AS drugs_with_side_effects
ORDER BY drugs_with_side_effects DESC
LIMIT 10;
manufacturerdrugs_with_side_effects
PFIZER617
ROCHE595
CELGENE452
NOVARTIS386
TAKEDA356
ABBVIE352
BRISTOL MYERS SQUIBB307
JOHNSON AND JOHNSON218
GLAXOSMITHKLINE200
AMGEN191

PFIZER top drugs and side effects

Leading PFIZER drugs by case count with a sample of distinct reaction descriptions.

5 rows
MATCH (m:Manufacturer {manufacturerName: 'PFIZER'})-[:REGISTERED]->(c:Case)-[:HAS_REACTION]->(r:Reaction)
MATCH (d:Drug)<-[:IS_PRIMARY_SUSPECT|IS_SECONDARY_SUSPECT|IS_CONCOMITANT|IS_INTERACTING]-(c)
WHERE r.description IS NOT NULL
RETURN d.name AS drug_name,
       collect(DISTINCT r.description)[0..20] AS side_effects,
       count(DISTINCT c) AS case_count
ORDER BY case_count DESC
LIMIT 5;
drug_nameside_effectscase_count
LYRICAAgeusia, Pain, Hypoacusis, Temporomandibular joint syndrome, Depressed level of consciousness, Intentional product use issue, Gingival disorder, Malaise, Anosmia, Viral infection, Deafness, Dry mouth, Confusional state, Memory impairment, Toothache, Panic attack, Anxiety, Hallucination, Dyspnoea, Withdrawal syndrome120
GENOTROPINDevice breakage, Drug dose omission by device, Product prescribing error, Device use error, Device issue, Device physical property issue, Product physical issue, Wrong technique in device usage process, Device mechanical issue, Device use issue, Device leakage, Device power source issue, Blindness, Device information output issue, Incorrect dose administered by device, Poor quality device used, Device malfunction, Incorrect dose administered, Product dispensing error, Product dose omission issue77
IBRANCEProduct dose omission issue, Product prescribing error, Neoplasm progression, Neutropenia, Anaemia, Off label use, Thrombocytopenia, White blood cell count decreased, Foot fracture, Product complaint, Poor quality product administered, Diarrhoea, Blood count abnormal, Neutrophil count decreased, Dyspnoea exertional, Pulmonary embolism, Illness, Malaise, Amnesia, Cerebrovascular accident48
XELJANZ XRArthralgia, Feeling abnormal, Rheumatoid arthritis, Pain, Intentional product use issue, Peripheral swelling, COVID-19, Gout, Rotator cuff syndrome, Aphonia, Fall, Contusion, Muscle rupture, Hair growth abnormal, Musculoskeletal stiffness, Impaired healing, Confusional state, Angle closure glaucoma, Therapeutic response unexpected, Illness47
XELJANZRheumatoid arthritis, Lymphocyte count decreased, White blood cell count decreased, Pain, Off label use, Bronchitis, Hypokinesia, Autoimmune disorder, Therapeutic response unexpected, Malaise, Spinal fracture, Illness, Pain in extremity, Product dose omission issue, Arthralgia, Joint swelling, Pruritus, Nasopharyngitis, Influenza, Localised infection32

Similar patient journeys

GDS nodeSimilarity on an undirected Case–Drug–Reaction projection (Jaccard, cutoff 0.2). Top pairs share reactions and suspect drugs.

Case A Case B Similarity Shared reactions Shared drugs
1115309121585749321.00
Pneumonia
REVLIMID
1115309121249028221.00
Pneumonia
REVLIMID
1111401421472429121.00
Neutropenia
REVLIMID
1111401421949262011.00
Neutropenia
REVLIMID
1115309121649813721.00
Pneumonia
REVLIMID
1115309121249774521.00
Pneumonia
REVLIMID
1098373241259796021.00
Atrial fibrillation
REVLIMIDDEXAMETHASONE
1111401421470657821.00
Neutropenia
REVLIMID
1111401421949550721.00
Neutropenia
REVLIMID
1125504221262412821.00
Septic shock
CarfilzomibDEXAMETHASONEREVLIMID

Patient sub-phenotypes (Leiden)

Leiden communities summarize each patient sub-phenotype by case count, top genders, age groups, and adverse reactions. Strong demographic or reaction concentration suggests a clinically interpretable cluster for targeted safety surveillance. Click a card to open that community’s curated ego-network in the explorer below.

Graph explorer

The full graph (~11k nodes) is too dense to ship. Explore curated neighborhoods: a sample of cases from a Leiden community, or cases around a high-signal drug.

Pick a community or drug to load a curated neighborhood.