EUVIMEDOpdag EUVIMED
BeobachtungsstudieKohortenstudieOpen access

Large Language Model–Based Identification of Subjective Cognitive Decline in Electronic Health Records Among Older Adults With Normal Cognitive Testing: Retrospective Cohort Study

Large-Language-Model-basierte Identifikation von subjektivem kognitivem Abbau in elektronischen Gesundheitsakten bei älteren Erwachsenen mit normalem kognitivem Test: retrospektive Kohortenstudie

Liqin Wang, Rebecca E Amariglio, Sheril Varghese, Jiazi Tian, Diane Seger, Li Zhou, Gad A Marshall

JMIR Aging · 2026-10-07 · 9 · e93011-e93011

DOI: 10.2196/93011 ↗

Auf einen Blick

Bei 15.750 älteren Erwachsenen mit normalem kognitivem Test fanden sich in 13,8 % dokumentierte kognitive Beschwerden im Vorjahr. Ein 2-stufiges LLM-Verfahren identifizierte diese Einträge mit hoher Genauigkeit.

Fragestellung

Lässt sich subjektiver kognitiver Abbau in EHRs bei älteren Erwachsenen mit normalem kognitivem Test mittels LLM-basierter Textanalyse erfassen, und wie häufig sind solche Dokumentationen?

Methode

EHR-basierte Beobachtungsstudie bei Patient:innen ab 65 Jahren mit erstem normalem kognitivem Test zwischen Januar 2019 und April 2024; 2-stufige LLM-NLP-Pipeline; Vergleich mit manueller Prüfung; logistische Regression.

Ergebnisse

15.750 ältere Erwachsene; 13,8 % (n=2175) mit mindestens 1 dokumentierter kognitiver Beschwerde im Vorjahr. Pipeline: Sensitivität 0,957, PPV 0,935, Spezifität 0,985, F1-Score 0,945. Mehr Dokumentation bei neurologischen/psychiatrischen Erkrankungen.

Schlussfolgerung der Autor:innen

Die 2-stufige LLM-Pipeline ermöglichte eine genaue Identifikation dokumentierter kognitiver Beschwerden, die mit subjektivem kognitivem Abbau vereinbar sind. Die Dokumentation war selten und von klinischen sowie soziodemografischen Faktoren abhängig.

Kernaussagen

  • 13,8 % hatten mindestens 1 dokumentierte kognitive Beschwerde im Vorjahr
  • LLM-Pipeline: Sensitivität 0,957, Spezifität 0,985
  • Dokumentation war mit Parkinson, TBI, Schlaganfall und Depression assoziiert

Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.

At a glance

Beobachtungsstudie in JMIR Aging, published online on 2026-10-07. Written by Liqin Wang and 6 co-authors. Institution: Department of Medicine, Mass General Brigham, 399 Revolution Drive, Somerville, MA, 02145, United States, 857-282-4089. Major MEDLINE topics: Electronic Health Records, Cognitive Dysfunction. Cited 1 times so far (Crossref); 20 references. The full text is freely available in PubMed Central. Funded by: NIA NIH HHS.

Auf einen BlickAbstractSubjectsBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Background

Subjective cognitive decline (SCD) typically refers to self- or informant-reported decline in cognition despite the absence of objective impairment on standardized testing. Older adults with documented normal cognitive test performance provide a pragmatic anchor cohort for electronic health record (EHR)–based SCD phenotyping. However, cognitive concerns are primarily recorded in unstructured notes and are inconsistently documented, making it unclear how often and in whom concerns are captured in routine care.

Objective

This study aims to operationalize EHR-based SCD phenotyping in an objectively normal-testing cohort using a large language model (LLM)–based natural language processing approach and to examine the frequency and clinical and sociodemographic correlates of cognitive concern documentation.

Methods

We conducted an EHR-based observational study of patients aged 65 years or older with a first normal cognitive test recorded in EHR flowsheets between January 2019 and April 2024 in a large health care system. We developed and iteratively refined a 2-stage LLM-based natural language processing pipeline to identify documented cognitive concerns in unstructured notes during the 12 months prior to the index date, and evaluated performance against manual review. We quantified the frequency of documented concerns within this normal-testing cohort and used multivariable logistic regression to assess associations with sociodemographic factors (age, marital status, insurance, and neighborhood Area Deprivation Index), sequentially adjusting for clinical comorbidities and relevant medications.

Results

Among 15,750 older adults with normal cognitive test scores, 13.8% (n=2175) had at least 1 documented cognitive concern in the prior year, captured in 1.2% (7394/605,177) of notes. On manual validation, the 2-stage pipeline (Med42-v2-8B screening followed by GPT-4o confirmation) achieved a sensitivity of 0.957, positive predictive value of 0.935, specificity of 0.985, and F 1 -score of 0.945 for cognitive concern identification. Documentation was more likely in older individuals, those with commercial (vs Medicare) insurance, and those with neurological and psychiatric conditions. In fully adjusted models, Parkinson disease (adjusted odds ratio [aOR] 5.29, 95% CI 3.60‐7.77), traumatic brain injury (aOR 4.63, 95% CI 3.15‐6.81), stroke or transient ischemic attack (aOR 3.46, 95% CI 3.15‐4.18), epilepsy (aOR 2.52, 95% CI 1.83‐3.49), depression (aOR 1.54, 95% CI 1.36‐1.75), and excessive alcohol use (aOR 1.48, 95% CI 1.06‐2.06) were among the strongest correlates of cognitive concern documentation. In contrast, obesity (aOR 0.73, 95% CI 0.65‐0.82), hyperlipidemia (aOR 0.59, 95% CI 0.51‐0.67), and residence in more deprived neighborhoods (higher Area Deprivation Index) were associated with lower odds of documented cognitive concerns.

Conclusions

A 2-stage LLM pipeline enabled accurate identification of documented cognitive concerns consistent with SCD among older adults with normal cognitive testing. Documentation was uncommon and selectively captured by clinical and sociodemographic factors, with implications for equity and the validity of EHR-based phenotypes.

Subjects

MeSH terms (MEDLINE)

Electronic Health RecordsCognitive DysfunctionHumansAgedLarge Language ModelsRetrospective StudiesMaleFemaleAged, 80 and over

Author keywords

cognitive concerndocumentationelectronic health recordslarge language modelsubjective cognitive decline

Authors and institutions

  • Department of Medicine, Mass General Brigham, 399 Revolution Drive, Somerville, MA, 02145, United States, 857-282-4089.
  • Harvard Medical School, 25 Shattuck Street, Boston, MA, 02115, United States.
  • Department of Neurology, Mass General Brigham, Boston, MA, United States.
  • Columbia University Irving Medical Center, New York, NY, United States.
  • Department of Medicine, Massachusetts General Hospital, Boston, MA, United States.
  • Research Information Science & Computing, Mass General Brigham, Boston, MA, United States.

Funding

  • NIA NIH HHS

Cite

APA

Wang, L., Amariglio, R. E., Varghese, S., Tian, J., Seger, D., Zhou, L., & Marshall, G. A. (2026). Large Language Model–Based Identification of Subjective Cognitive Decline in Electronic Health Records Among Older Adults With Normal Cognitive Testing: Retrospective Cohort Study. JMIR Aging, 9, e93011-e93011. https://doi.org/10.2196/93011

Vancouver

Wang L, Amariglio RE, Varghese S, Tian J, Seger D, Zhou L, et al. Large Language Model–Based Identification of Subjective Cognitive Decline in Electronic Health Records Among Older Adults With Normal Cognitive Testing: Retrospective Cohort Study. JMIR Aging. 2026;9:e93011-e93011. doi:10.2196/93011

BibTeX
@article{Wang2026,
  author = {Wang, Liqin and Amariglio, Rebecca E and Varghese, Sheril and Tian, Jiazi and Seger, Diane and Zhou, Li and Marshall, Gad A},
  title = {Large Language Model–Based Identification of Subjective Cognitive Decline in Electronic Health Records Among Older Adults With Normal Cognitive Testing: Retrospective Cohort Study},
  journal = {JMIR Aging},
  year = {2026},
  volume = {9},
  pages = {e93011--e93011},
  doi = {10.2196/93011}
}
Start for free →