close
Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Jan 4;24(1):10.
doi: 10.1186/s12911-023-02409-8.

A patient safety knowledge graph supporting vaccine product development

Andrew M Simms #  1 , Anshul Kanakia #  2 , Muhammad Sipra  2 , Bhaskar Dutta  2 , Noel Southall  2
Affiliations

A patient safety knowledge graph supporting vaccine product development

Andrew M Simms et al. BMC Med Inform Decis Mak. .

Abstract

Background: Knowledge graphs are well-suited for modeling complex, unstructured, and multi-source data and facilitating their analysis. During the COVID-19 pandemic, adverse event data were integrated into a knowledge graph to support vaccine safety surveillance and nimbly respond to urgent health authority questions. Here, we provide details of this post-marketing safety system using public data sources. In addition to challenges with varied data representations, adverse event reporting on the COVID-19 vaccines generated an unprecedented volume of data; an order of magnitude larger than adverse events for all previous vaccines. The Patient Safety Knowledge Graph (PSKG) is a robust data store to accommodate the volume of adverse event data and harmonize primary surveillance data sources.

Methods: We designed a semantic model to represent key safety concepts. We built an extract-transform-load (ETL) data pipeline to parse and import primary public data sources; align key elements such as vaccine names; integrated the Medical Dictionary for Regulatory Activities (MedDRA); and applied quality metrics. PSKG is deployed in a Neo4J graph database, and made available via a web interface and Application Programming Interfaces (APIs).

Results: We import and align adverse event data and vaccine exposure data from 250 countries on a weekly basis, producing a graph with 4,340,980 nodes and 30,544,475 edges as of July 1, 2022. PSKG is used for ad-hoc analyses and periodic reporting for several widely available COVID-19 vaccines. Analysis code using the knowledge graph is 80% shorter than an equivalent implementation written entirely in Python, and runs over 200 times faster.

Conclusions: Organizing safety data into a concise model of nodes, properties, and edge relationships has greatly simplified analysis code by removing complex parsing and transformation algorithms from individual analyses and instead managing these centrally. The adoption of the knowledge graph transformed how the team answers key scientific and medical questions. Whereas previously an analysis would involve aggregating and transforming primary datasets from scratch to answer a specific question, the team can now iterate easily and respond as quickly as requests evolve (e.g., "Produce vaccine-X safety profile for adverse event-Y by country instead of age-range").

Keywords: Datasets; Knowledge graph; Pharmacovigilance.

PubMed Disclaimer

Conflict of interest statement

AS was employed as a contractor at AstraZeneca, AK and NS are full-time employees of AstraZeneca, MS was a contractor at AstraZeneca, BD was formerly a full-time employee of AstraZeneca.

Figures

Fig. 1
Fig. 1
PSKG Ontology, depicting nodes categorized by data source, and relationship edges
Fig. 2
Fig. 2
EudraVigilance and VAERS high-level native structure, and links to MedDRA
Fig. 3
Fig. 3
PSKG Generator class actions. Each generator reads from raw source data, assigns identifiers, align outcomes, and performs type conversions (e.g. strings to datetimes)
Fig. 4
Fig. 4
Organization of generator objects for VAERS and EudraVigilance adverse event data
Fig. 5
Fig. 5
Minimal Cypher query to count all VAERS vaccine cases
Fig. 6
Fig. 6
Stratifying VAERS vaccine cases by vaccine type, and returning the results ordered by descending number of cases
Fig. 7
Fig. 7
Identifying cases that include two adverse events (as defined by multiple preferred terms)
Fig. 8
Fig. 8
Identifying cases that include one adverse event and exclude a second adverse event (as defined by multiple preferred terms)
Fig. 9
Fig. 9
Identifying the most frequently reported adverse event terms, and then extracting case counts (stratified by age and sex). An initial MATCH statement result finds the top reported adverse event terms, and is chained to an additional clause that extracts stratified case counts
Fig. 10
Fig. 10
Example investigation of concomitant medications and related conditions. Here a simplistic inclusion criteria (cases with more than one adverse event term) are extracted and matched to concomitant medications and related indications, and the number of cases are summarized by indication, medication
Fig. 11
Fig. 11
Finding available stratification for exposure data
Fig. 12
Fig. 12
Adverse events reported in VAERS over 30 years
Fig. 13
Fig. 13
Parsing and analyzing data in EudraVigilance. EudraVigilance data are produced in a tabular form, but columns can contain complex values. Simple columns can be projected directly (e.g. Report Type), other columns such as suspect and concomitant drugs must be restructured. A typical flow is illustrated here showing the projection of simple columns to the result set, along with additional processing of multi-valued fields, some of which are combined (e.g. Outcomes and Seriousness)
Fig. 14
Fig. 14
VAERS case structure illustrating how vaccination records and symptoms are linked to a case
Fig. 15
Fig. 15
EudraVigilance case structure illustrating the reaction list preferred term(s), suspect and concomitant medication(s), and optional indication preferred terms
See this image and copyright information in PMC

References

    1. AstraZeneca. Two billion doses of AstraZeneca’s COVID-19 vaccine supplied to countries across the world less than 12 months after first approval. AstraZeneca; 2021. https://www.astrazeneca.com/media-centre/press-releases/2021/two-billion.... Accessed 2 July 2022.
    1. US Department of Health and Human Services. VAERS - Report an Adverse Event. 2022. https://vaers.hhs.gov/reportevent.html. Accessed 2 July 2022.
    1. European Medicines Agency. ADR reporting - patient guideline. 2022. https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/a.... Accessed 2 July 2022.
    1. US Department of Health and Human Services. Vaccine Adverse Event Reporting System (VAERS). 2022. https://vaers.hhs.gov/index.html. Accessed 2 July 2022
    1. US Department of Health and Human Services. VAERS - Guide to Interpreting VAERS Data. 2022. https://vaers.hhs.gov/data/dataguide.html. Accessed 2 July 2022.

Publication types

MeSH terms