Learning debiased graph representations from the OMOP common data model for synthetic data generation

  1. Schulz, Nicolas Alexander
  2. Carus, Jasmin
  3. Wiederhold, Alexander Johannes
  4. Johanns, Ole
  5. Peters, Frederik
  6. Rath, Natalie
  7. Rausch, Katharina
  8. Holleczek, Bernd
  9. Katalinic, Alexander
  10. the AI-CARE Working Group
  11. Nennecke, Alice
  12. Kusche, Henrik
  13. Heinrichs, Vera
  14. Eberle, Andrea
  15. Luttmann, Sabine
  16. Abnaof, Khalid
  17. Kim-Wanner, Soo-Zin
  18. Handels, Heinz
  19. Germer, Sebastian
  20. Halber, Marco
  21. Richter, Martin
  22. Pinnau, Martin
  23. Reiner, David
  24. Schaaf, Jannik
  25. Storf, Holger
  26. Hartz, Tobias
  27. Goeken, Nils
  28. Bösche, Janina
  29. Stein, Alexandra
  30. Weitmann, Kerstin
  31. Hoffmann, Wolfgang
  32. Labohm, Louisa
  33. Rudolph, Christiane
  34. Gundler, Christopher
  35. Ückert, Frank
  36. Gundler, Christopher

Download
12874_2024_Article_2257.pdf 2,00MB
WeightNameValue
1000 Titel
  • Learning debiased graph representations from the OMOP common data model for synthetic data generation
1000 Autor/in
  1. Schulz, Nicolas Alexander |
  2. Carus, Jasmin |
  3. Wiederhold, Alexander Johannes |
  4. Johanns, Ole |
  5. Peters, Frederik |
  6. Rath, Natalie |
  7. Rausch, Katharina |
  8. Holleczek, Bernd |
  9. Katalinic, Alexander |
  10. the AI-CARE Working Group |
  11. Nennecke, Alice |
  12. Kusche, Henrik |
  13. Heinrichs, Vera |
  14. Eberle, Andrea |
  15. Luttmann, Sabine |
  16. Abnaof, Khalid |
  17. Kim-Wanner, Soo-Zin |
  18. Handels, Heinz |
  19. Germer, Sebastian |
  20. Halber, Marco |
  21. Richter, Martin |
  22. Pinnau, Martin |
  23. Reiner, David |
  24. Schaaf, Jannik |
  25. Storf, Holger |
  26. Hartz, Tobias |
  27. Goeken, Nils |
  28. Bösche, Janina |
  29. Stein, Alexandra |
  30. Weitmann, Kerstin |
  31. Hoffmann, Wolfgang |
  32. Labohm, Louisa |
  33. Rudolph, Christiane |
  34. Gundler, Christopher |
  35. Ückert, Frank |
  36. Gundler, Christopher |
1000 Verlag BioMed Central
1000 Erscheinungsjahr 2024
1000 Publikationstyp
  1. Artikel |
1000 Online veröffentlicht
  • 2024-06-22
1000 Erschienen in
1000 Quellenangabe
  • 24(1):136
1000 Copyrightjahr
  • 2024
1000 Lizenz
1000 Verlagsversion
  • https://doi.org/10.1186/s12874-024-02257-8 |
  • https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11193245/ |
1000 Publikationsstatus
1000 Begutachtungsstatus
1000 Sprache der Publikation
1000 Abstract/Summary
  • <jats:title>Abstract</jats:title><jats:sec> <jats:title>Background</jats:title> <jats:p>Generating synthetic patient data is crucial for medical research, but common approaches build up on black-box models which do not allow for expert verification or intervention. We propose a highly available method which enables synthetic data generation from real patient records in a privacy preserving and compliant fashion, is interpretable and allows for expert intervention.</jats:p> </jats:sec><jats:sec> <jats:title>Methods</jats:title> <jats:p>Our approach ties together two established tools in medical informatics, namely OMOP as a data standard for electronic health records and Synthea as a data synthetization method. For this study, data pipelines were built which extract data from OMOP, convert them into time series format, learn temporal rules by 2 statistical algorithms (Markov chain, TARM) and 3 algorithms of causal discovery (DYNOTEARS, J-PCMCI+, LiNGAM) and map the outputs into Synthea graphs. The graphs are evaluated quantitatively by their individual and relative complexity and qualitatively by medical experts.</jats:p> </jats:sec><jats:sec> <jats:title>Results</jats:title> <jats:p>The algorithms were found to learn qualitatively and quantitatively different graph representations. Whereas the Markov chain results in extremely large graphs, TARM, DYNOTEARS, and J-PCMCI+ were found to reduce the data dimension during learning. The MultiGroupDirect LiNGAM algorithm was found to not be applicable to the problem statement at hand.</jats:p> </jats:sec><jats:sec> <jats:title>Conclusion</jats:title> <jats:p>Only TARM and DYNOTEARS are practical algorithms for real-world data in this use case. As causal discovery is a method to debias purely statistical relationships, the gradient-based causal discovery algorithm DYNOTEARS was found to be most suitable.</jats:p> </jats:sec>
1000 Sacherschließung
lokal Causal Discovery
lokal Algorithms [MeSH]
lokal Graphical Models
lokal DYNOTEARS
lokal Medical Informatics/methods [MeSH]
lokal Gradient-Based Causal Discovery
lokal Humans [MeSH]
lokal Standardized Electronic Health Records
lokal Structural Equation Models
lokal Constraint-based Causal Discovery
lokal Medical Informatics/statistics
lokal Temporal Association Rule Mining (TARM)
lokal Markov Chains [MeSH]
lokal Research
lokal Electronic Health Records/statistics
lokal Synthetic Data Generation
lokal Discrete Time Series
lokal Electronic Health Records/standards [MeSH]
1000 Fächerklassifikation (DDC)
1000 Liste der Beteiligten
  1. https://frl.publisso.de/adhoc/uri/U2NodWx6LCBOaWNvbGFzIEFsZXhhbmRlcg==|https://frl.publisso.de/adhoc/uri/Q2FydXMsIEphc21pbg==|https://frl.publisso.de/adhoc/uri/V2llZGVyaG9sZCwgQWxleGFuZGVyIEpvaGFubmVz|https://frl.publisso.de/adhoc/uri/Sm9oYW5ucywgT2xl|https://frl.publisso.de/adhoc/uri/UGV0ZXJzLCBGcmVkZXJpaw==|https://frl.publisso.de/adhoc/uri/UmF0aCwgTmF0YWxpZQ==|https://frl.publisso.de/adhoc/uri/UmF1c2NoLCBLYXRoYXJpbmE=|https://frl.publisso.de/adhoc/uri/SG9sbGVjemVrLCBCZXJuZA==|https://frl.publisso.de/adhoc/uri/S2F0YWxpbmljLCBBbGV4YW5kZXI=|https://frl.publisso.de/adhoc/uri/dGhlIEFJLUNBUkUgV29ya2luZyBHcm91cA==|https://frl.publisso.de/adhoc/uri/TmVubmVja2UsIEFsaWNl|https://frl.publisso.de/adhoc/uri/S3VzY2hlLCBIZW5yaWs=|https://frl.publisso.de/adhoc/uri/SGVpbnJpY2hzLCBWZXJh|https://frl.publisso.de/adhoc/uri/RWJlcmxlLCBBbmRyZWE=|https://frl.publisso.de/adhoc/uri/THV0dG1hbm4sIFNhYmluZQ==|https://frl.publisso.de/adhoc/uri/QWJuYW9mLCBLaGFsaWQ=|https://frl.publisso.de/adhoc/uri/S2ltLVdhbm5lciwgU29vLVppbg==|https://frl.publisso.de/adhoc/uri/SGFuZGVscywgSGVpbno=|https://frl.publisso.de/adhoc/uri/R2VybWVyLCBTZWJhc3RpYW4=|https://frl.publisso.de/adhoc/uri/SGFsYmVyLCBNYXJjbw==|https://frl.publisso.de/adhoc/uri/UmljaHRlciwgTWFydGlu|https://frl.publisso.de/adhoc/uri/UGlubmF1LCBNYXJ0aW4=|https://frl.publisso.de/adhoc/uri/UmVpbmVyLCBEYXZpZA==|https://frl.publisso.de/adhoc/uri/U2NoYWFmLCBKYW5uaWs=|https://frl.publisso.de/adhoc/uri/U3RvcmYsIEhvbGdlcg==|https://frl.publisso.de/adhoc/uri/SGFydHosIFRvYmlhcw==|https://frl.publisso.de/adhoc/uri/R29la2VuLCBOaWxz|https://frl.publisso.de/adhoc/uri/QsO2c2NoZSwgSmFuaW5h|https://frl.publisso.de/adhoc/uri/U3RlaW4sIEFsZXhhbmRyYQ==|https://frl.publisso.de/adhoc/uri/V2VpdG1hbm4sIEtlcnN0aW4=|https://frl.publisso.de/adhoc/uri/SG9mZm1hbm4sIFdvbGZnYW5n|https://frl.publisso.de/adhoc/uri/TGFib2htLCBMb3Vpc2E=|https://frl.publisso.de/adhoc/uri/UnVkb2xwaCwgQ2hyaXN0aWFuZQ==|https://frl.publisso.de/adhoc/uri/R3VuZGxlciwgQ2hyaXN0b3BoZXI=|https://frl.publisso.de/adhoc/uri/w5xja2VydCwgRnJhbms=|https://frl.publisso.de/adhoc/uri/R3VuZGxlciwgQ2hyaXN0b3BoZXI=
1000 Hinweis
  • DeepGreen-ID: 1ee48fd324ab4ee0bcbc4b12ec13d9dc ; metadata provieded by: DeepGreen (https://www.oa-deepgreen.de/api/v1/), LIVIVO search scope life sciences (http://z3950.zbmed.de:6210/livivo), Crossref Unified Resource API (https://api.crossref.org/swagger-ui/index.html), to.science.api (https://frl.publisso.de/), ZDB JSON-API (beta) (https://zeitschriftendatenbank.de/api/), lobid - Dateninfrastruktur für Bibliotheken (https://lobid.org/resources/search)
1000 Label
1000 Förderer
  1. Universitätsklinikum Hamburg-Eppendorf (UKE) |
1000 Fördernummer
  1. -
1000 Förderprogramm
  1. -
1000 Dateien
1000 Förderung
  1. 1000 joinedFunding-child
    1000 Förderer Universitätsklinikum Hamburg-Eppendorf (UKE) |
    1000 Förderprogramm -
    1000 Fördernummer -
1000 Objektart article
1000 Beschrieben durch
1000 @id frl:6517842.rdf
1000 Erstellt am 2025-07-05T07:07:43.115+0200
1000 Erstellt von 322
1000 beschreibt frl:6517842
1000 Zuletzt bearbeitet 2025-08-19T19:17:46.846+0200
1000 Objekt bearb. Tue Aug 19 19:17:46 CEST 2025
1000 Vgl. frl:6517842
1000 Oai Id
  1. oai:frl.publisso.de:frl:6517842 |
1000 Sichtbarkeit Metadaten public
1000 Sichtbarkeit Daten public
1000 Gegenstand von

View source