Academic Journal
View in EDS

Multimodal artificial intelligence-based decision support for stroke classification in medical emergency calls using call transcripts and patient health records: A retrospective diagnostic accuracy study.

Bibliographic Details
Title: Multimodal artificial intelligence-based decision support for stroke classification in medical emergency calls using call transcripts and patient health records: A retrospective diagnostic accuracy study.
Authors: Iversen E; Norwegian National Advisory Unit on Emergency Medical Communication, Haukeland University Hospital, Bergen, Norway; Department of Prehospital Emergency Medicine, Institute of Nursing and Health Promotion, Faculty of Health Sciences, Oslo Metropolitan University, Oslo, Norway; Department of Clinical Medicine, University of Bergen, Bergen, Norway. Electronic address: emil.iversen@helse-bergen.no., Ihle-Hansen H; Department of Medicine, Vestre Viken Hospital Trust, Drammen, Norway., Halle KK; Department of Medical Quality Registries, St Olavs Hospital, Trondheim University Hospital, Trondheim, Norway., Lundervold AS; Department of Strategy and Management, NHH Norwegian School of Economics, Bergen, Norway; Department of Radiology, Haukeland University Hospital, Bergen, Norway., Myrmel L; Department of Anaesthesia and Intensive Care, Haukeland University Hospital, Bergen, Norway., Vestbø AS; Department of Research and Development, Haukeland University Hospital, Bergen, Norway., Fromm A; Department of Neurology, Haukeland University Hospital, Bergen, Norway., Persia CD; Team Artificial Intelligence, Helse Vest ICT, Bergen, Norway., Autenried C; Team Artificial Intelligence, Helse Vest ICT, Bergen, Norway., Jamtli B; Department of Prehospital Emergency Medicine, Institute of Nursing and Health Promotion, Faculty of Health Sciences, Oslo Metropolitan University, Oslo, Norway; Division of Prehospital Services, Oslo University Hospital, Oslo, Norway., Brattebø G; Norwegian National Advisory Unit on Emergency Medical Communication, Haukeland University Hospital, Bergen, Norway; Department of Clinical Medicine, University of Bergen, Bergen, Norway; Department of Anaesthesia and Intensive Care, Haukeland University Hospital, Bergen, Norway.
Source: International journal of medical informatics [Int J Med Inform] 2026 Dec; Vol. 222, pp. 106713. Date of Electronic Publication: 2026 Sep 11.
Publication Type: Journal Article
Language: English
Journal Info: Publisher: Elsevier Science Ireland Ltd Country of Publication: Ireland NLM ID: 9711057 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1872-8243 (Electronic) Linking ISSN: 13865056 NLM ISO Abbreviation: Int J Med Inform Subsets: MEDLINE
Imprint Name(s): Original Publication: Shannon, Co. Clare, Ireland : Elsevier Science Ireland Ltd., c1997-
MeSH Terms: Stroke*/diagnosis , Stroke*/classification , Artificial Intelligence* , Decision Support Systems, Clinical* , Emergency Medical Service Communication Systems*, Humans ; Retrospective Studies ; Norway ; Electronic Health Records
Abstract: Background: Identifying stroke during medical emergency calls is challenging because of unstructured information, communication barriers, and heterogeneous symptoms. Artificial intelligence (AI) may improve early stroke recognition;however,evidence from emergency call settings remains limited, and the value of combining call transcripts with structured clinical data is unclear.We developed and internally validated a multimodal AI-based decision-support system for stroke classification during medicalemergency calls and compared its performance with the Emergency Medical Communication Centre (EMCC) operator baseline.
Methods: We performed a retrospective diagnostic accuracy study using data from Bergen EMCC from 2019 and 2022. Two datasets were used: (1) 473 EMCC audio files with paired manual transcripts for fine-tuning a Norwegian Whisper-based automatic speech recognition (ASR) model (2019), and (2) 2,627 labelled emergency calls for classification model developmentand evaluation (2022). Structured clinical data included hospital data, laboratory measurements, and medications. Performance was assessed using area under the receiver operating characteristic curve (AUC), precision, recall, and F1-score. A streaming evaluation at 30, 60, and 90 ssimulated real-time use.
Results: Fine-tuning improved ASR accuracy, reducing word error rate from 8.3% to 7.2%. The multimodal classification model achieved a stroke-class F1-score of 0.77, exceeding that of both the transcript-only model and the operator baseline (0.73). The macro-averaged F1-score was 0.87 and the AUC was 0.86. At an assumed 1.2% stroke prevalence, the model was estimated to generate approximately 26 false alerts per 1,000 calls, compared with 36 for the operator baseline. In the streaming evaluation, stroke F1-score increased from 0.68 at 30 s to 0.85 at 90 s, stabilising performance during the first minute.
Conclusions: A multimodal AI-based decision-support system integrating emergency call transcripts with patient health recordsimproved stroke classification performance and may support EMCC operators during stroketriage. External validation in larger datasets is required before clinical implementation.
(Copyright © 2026 The Author(s). Published by Elsevier B.V. All rights reserved.)
Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Contributed Indexing: Keywords: Artificial intelligence; Decision support system; Emergency medical communicationcentre; Medical emergency calls; Stroke; Triage
Entry Date(s): Date Created: 20260913 Date Completed: 20261006 Latest Revision: 20261006
Update Code: 20261007
DOI: 10.1016/j.ijmedinf.2026.106713
PMID: 42732713
Database: MEDLINE
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://www.doi.org/10.1016/j.ijmedinf.2026.106713?
    Name: ScienceDirect (all content) (s7799221)
    Category: fullText
    Text: View record from ScienceDirect
    MouseOverText: View record from ScienceDirect
  – Url: https://resolver.ebsco.com/c/fiv2js/result?sid=EBSCO:cmedm&genre=article&issn=18728243&ISBN=&volume=222&issue=&date=20261201&spage=106713&pages=106713&title=International journal of medical informatics&atitle=Multimodal%20artificial%20intelligence-based%20decision%20support%20for%20stroke%20classification%20in%20medical%20emergency%20calls%20using%20call%20transcripts%20and%20patient%20health%20records%3A%20A%20retrospective%20diagnostic%20accuracy%20study.&aulast=Iversen%20E&id=DOI:10.1016/j.ijmedinf.2026.106713
    Name: Full Text Finder (for New FTF UI) (ns324271)
    Category: fullText
    Text: Full Text Finder
    MouseOverText: Full Text Finder
Header DbId: cmedm
DbLabel: MEDLINE
An: 42732713
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Multimodal artificial intelligence-based decision support for stroke classification in medical emergency calls using call transcripts and patient health records: A retrospective diagnostic accuracy study.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AU" term="%22Iversen+E%22">Iversen E</searchLink>; Norwegian National Advisory Unit on Emergency Medical Communication, Haukeland University Hospital, Bergen, Norway; Department of Prehospital Emergency Medicine, Institute of Nursing and Health Promotion, Faculty of Health Sciences, Oslo Metropolitan University, Oslo, Norway; Department of Clinical Medicine, University of Bergen, Bergen, Norway. Electronic address: emil.iversen@helse-bergen.no.<br /><searchLink fieldCode="AU" term="%22Ihle-Hansen+H%22">Ihle-Hansen H</searchLink>; Department of Medicine, Vestre Viken Hospital Trust, Drammen, Norway.<br /><searchLink fieldCode="AU" term="%22Halle+KK%22">Halle KK</searchLink>; Department of Medical Quality Registries, St Olavs Hospital, Trondheim University Hospital, Trondheim, Norway.<br /><searchLink fieldCode="AU" term="%22Lundervold+AS%22">Lundervold AS</searchLink>; Department of Strategy and Management, NHH Norwegian School of Economics, Bergen, Norway; Department of Radiology, Haukeland University Hospital, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Myrmel+L%22">Myrmel L</searchLink>; Department of Anaesthesia and Intensive Care, Haukeland University Hospital, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Vestbø+AS%22">Vestbø AS</searchLink>; Department of Research and Development, Haukeland University Hospital, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Fromm+A%22">Fromm A</searchLink>; Department of Neurology, Haukeland University Hospital, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Persia+CD%22">Persia CD</searchLink>; Team Artificial Intelligence, Helse Vest ICT, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Autenried+C%22">Autenried C</searchLink>; Team Artificial Intelligence, Helse Vest ICT, Bergen, Norway.<br /><searchLink fieldCode="AU" term="%22Jamtli+B%22">Jamtli B</searchLink>; Department of Prehospital Emergency Medicine, Institute of Nursing and Health Promotion, Faculty of Health Sciences, Oslo Metropolitan University, Oslo, Norway; Division of Prehospital Services, Oslo University Hospital, Oslo, Norway.<br /><searchLink fieldCode="AU" term="%22Brattebø+G%22">Brattebø G</searchLink>; Norwegian National Advisory Unit on Emergency Medical Communication, Haukeland University Hospital, Bergen, Norway; Department of Clinical Medicine, University of Bergen, Bergen, Norway; Department of Anaesthesia and Intensive Care, Haukeland University Hospital, Bergen, Norway.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%229711057%22">International journal of medical informatics</searchLink> [Int J Med Inform] 2026 Dec; Vol. 222, pp. 106713. <i>Date of Electronic Publication: </i>2026 Sep 11.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22Elsevier+Science+Ireland+Ltd%22">Elsevier Science Ireland Ltd </searchLink><i>Country of Publication: </i>Ireland <i>NLM ID: </i>9711057 <i>Publication Model: </i>Print-Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>1872-8243 (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2213865056%22">13865056 </searchLink><i>NLM ISO Abbreviation: </i>Int J Med Inform <i>Subsets: </i>MEDLINE
– Name: PublisherInfo
  Label: Imprint Name(s)
  Group: PubInfo
  Data: <i>Original Publication</i>: Shannon, Co. Clare, Ireland : Elsevier Science Ireland Ltd., c1997-
– Name: SubjectMESH
  Label: MeSH Terms
  Group: Su
  Data: <searchLink fieldCode="MM" term="%22Stroke%22">Stroke*</searchLink>/<searchLink fieldCode="MM" term="%22Stroke+diagnosis%22">diagnosis</searchLink> <br /><searchLink fieldCode="MM" term="%22Stroke%22">Stroke*</searchLink>/<searchLink fieldCode="MM" term="%22Stroke+classification%22">classification</searchLink> <br /><searchLink fieldCode="MM" term="%22Artificial+Intelligence%22">Artificial Intelligence*</searchLink> <br /><searchLink fieldCode="MM" term="%22Decision+Support+Systems%2C+Clinical%22">Decision Support Systems, Clinical*</searchLink> <br /><searchLink fieldCode="MM" term="%22Emergency+Medical+Service+Communication+Systems%22">Emergency Medical Service Communication Systems*</searchLink><br /><searchLink fieldCode="MH" term="%22Humans%22">Humans</searchLink> ; <searchLink fieldCode="MH" term="%22Retrospective+Studies%22">Retrospective Studies</searchLink> ; <searchLink fieldCode="MH" term="%22Norway%22">Norway</searchLink> ; <searchLink fieldCode="MH" term="%22Electronic+Health+Records%22">Electronic Health Records</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Background: Identifying stroke during medical emergency calls is challenging because of unstructured information, communication barriers, and heterogeneous symptoms. Artificial intelligence (AI) may improve early stroke recognition;however,evidence from emergency call settings remains limited, and the value of combining call transcripts with structured clinical data is unclear.We developed and internally validated a multimodal AI-based decision-support system for stroke classification during medicalemergency calls and compared its performance with the Emergency Medical Communication Centre (EMCC) operator baseline.<br />Methods: We performed a retrospective diagnostic accuracy study using data from Bergen EMCC from 2019 and 2022. Two datasets were used: (1) 473 EMCC audio files with paired manual transcripts for fine-tuning a Norwegian Whisper-based automatic speech recognition (ASR) model (2019), and (2) 2,627 labelled emergency calls for classification model developmentand evaluation (2022). Structured clinical data included hospital data, laboratory measurements, and medications. Performance was assessed using area under the receiver operating characteristic curve (AUC), precision, recall, and F1-score. A streaming evaluation at 30, 60, and 90 ssimulated real-time use.<br />Results: Fine-tuning improved ASR accuracy, reducing word error rate from 8.3% to 7.2%. The multimodal classification model achieved a stroke-class F1-score of 0.77, exceeding that of both the transcript-only model and the operator baseline (0.73). The macro-averaged F1-score was 0.87 and the AUC was 0.86. At an assumed 1.2% stroke prevalence, the model was estimated to generate approximately 26 false alerts per 1,000 calls, compared with 36 for the operator baseline. In the streaming evaluation, stroke F1-score increased from 0.68 at 30 s to 0.85 at 90 s, stabilising performance during the first minute.<br />Conclusions: A multimodal AI-based decision-support system integrating emergency call transcripts with patient health recordsimproved stroke classification performance and may support EMCC operators during stroketriage. External validation in larger datasets is required before clinical implementation.<br /> (Copyright © 2026 The Author(s). Published by Elsevier B.V. All rights reserved.)
– Name: Abstract
  Label: Competing Interests
  Group: Ab
  Data: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
– Name: SubjectMinor
  Label: Contributed Indexing
  Group:
  Data: <i>Keywords: </i>Artificial intelligence; Decision support system; Emergency medical communicationcentre; Medical emergency calls; Stroke; Triage
– Name: DateEntry
  Label: Entry Date(s)
  Group: Date
  Data: <i>Date Created: </i>20260913 <i>Date Completed: </i>20261006 <i>Latest Revision: </i>20261006
– Name: DateUpdate
  Label: Update Code
  Group: Date
  Data: 20261007
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1016/j.ijmedinf.2026.106713
– Name: AN
  Label: PMID
  Group: ID
  Data: 42732713
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=42732713
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1016/j.ijmedinf.2026.106713
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        StartPage: 106713
    Subjects:
      – SubjectFull: Humans
        Type: general
      – SubjectFull: Retrospective Studies
        Type: general
      – SubjectFull: Norway
        Type: general
      – SubjectFull: Electronic Health Records
        Type: general
      – SubjectFull: Stroke diagnosis
        Type: general
      – SubjectFull: Stroke classification
        Type: general
      – SubjectFull: Artificial Intelligence
        Type: general
      – SubjectFull: Decision Support Systems, Clinical
        Type: general
      – SubjectFull: Emergency Medical Service Communication Systems
        Type: general
    Titles:
      – TitleFull: Multimodal artificial intelligence-based decision support for stroke classification in medical emergency calls using call transcripts and patient health records: A retrospective diagnostic accuracy study.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Iversen E
      – PersonEntity:
          Name:
            NameFull: Ihle-Hansen H
      – PersonEntity:
          Name:
            NameFull: Halle KK
      – PersonEntity:
          Name:
            NameFull: Lundervold AS
      – PersonEntity:
          Name:
            NameFull: Myrmel L
      – PersonEntity:
          Name:
            NameFull: Vestbø AS
      – PersonEntity:
          Name:
            NameFull: Fromm A
      – PersonEntity:
          Name:
            NameFull: Persia CD
      – PersonEntity:
          Name:
            NameFull: Autenried C
      – PersonEntity:
          Name:
            NameFull: Jamtli B
      – PersonEntity:
          Name:
            NameFull: Brattebø G
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 12
              Text: 2026 Dec
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1872-8243
          Numbering:
            – Type: volume
              Value: 222
          Titles:
            – TitleFull: International journal of medical informatics
              Type: main
ResultId 1