Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context
Summary: Background: Alzheimer’s disease (AD) is a leading cause of dementia and an escalating public health concern. Although recent research has identified multiple AD risk factors, most studies examine isolated comorbidities rather than complex, sequential progressions. In this study, we sought...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-08-01
|
| Series: | EBioMedicine |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2352396425002750 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849237528429723648 |
|---|---|
| author | Mingzhou Fu Sriram Sankararaman Bogdan Pasaniuc Keith Vossel Timothy S. Chang |
| author_facet | Mingzhou Fu Sriram Sankararaman Bogdan Pasaniuc Keith Vossel Timothy S. Chang |
| author_sort | Mingzhou Fu |
| collection | DOAJ |
| description | Summary: Background: Alzheimer’s disease (AD) is a leading cause of dementia and an escalating public health concern. Although recent research has identified multiple AD risk factors, most studies examine isolated comorbidities rather than complex, sequential progressions. In this study, we sought to identify multi-step trajectories culminating in AD by analysing longitudinal electronic health records (EHRs). Methods: We analysed data from 24,473 patients in the University of California Health Data Warehouse (UCHDW). A Fine–Gray subdistribution hazard model identified temporally associated diagnoses, from which we constructed diagnostic trajectories. We employed dynamic time warping and k-means clustering to group similar trajectories, and network analyses to characterize their common structures. Causal inferences were explored using the Greedy Equivalence Search algorithm. Validation in the UCHDW included association tests and comparison to control groups. We further validated our findings in the All of Us Research Program, a diverse, nationally representative cohort. Findings: After filtering, 5762 patients contributed 6794 unique AD progression trajectories, revealing four major trajectory clusters: mental health, encephalopathy, mild cognitive impairment, and vascular disease. These clusters differed significantly in demographic and clinical features. Approximately 26% of edges showed consistent directional ordering (e.g., hypertension → depressive episode → AD). In an independent population, these multi-step trajectories conferred greater AD risk than single diagnoses alone. Our validation in the All of Us cohort confirmed the reproducibility of these trajectory patterns in a more diverse population. Interpretation: Our findings demonstrate the value of examining sequential diagnostic patterns in AD pathogenesis. Multi-step progressions reveal potential latent contributors to AD, offering pathways for risk stratification, early detection, and targeted interventions. Funding: This study was supported by the National Institutes of Health, National Institute on Aging, the National Science Foundation, the Hillblom and Fineberg Foundations, and the California Department of Public Health. |
| format | Article |
| id | doaj-art-232bddf9f4b54b8b964874eabc157784 |
| institution | Kabale University |
| issn | 2352-3964 |
| language | English |
| publishDate | 2025-08-01 |
| publisher | Elsevier |
| record_format | Article |
| series | EBioMedicine |
| spelling | doaj-art-232bddf9f4b54b8b964874eabc1577842025-08-20T04:01:57ZengElsevierEBioMedicine2352-39642025-08-0111810583110.1016/j.ebiom.2025.105831Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in contextMingzhou Fu0Sriram Sankararaman1Bogdan Pasaniuc2Keith Vossel3Timothy S. Chang4Department of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, 90095, USA; Medical Informatics Home Area, Department of Bioinformatics, University of California, Los Angeles, Los Angeles, CA, 90024, USAComputational Medicine, University of California, Los Angeles, Los Angeles, CA, 90024, USAGenomics and Computational Biology, University of Pennsylvania, PA, 19104, USADepartment of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, 90095, USADepartment of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, CA, 90095, USA; Corresponding author. 710 Westwood Plaza, Room 3149, Los Angeles, CA 90073, USA.Summary: Background: Alzheimer’s disease (AD) is a leading cause of dementia and an escalating public health concern. Although recent research has identified multiple AD risk factors, most studies examine isolated comorbidities rather than complex, sequential progressions. In this study, we sought to identify multi-step trajectories culminating in AD by analysing longitudinal electronic health records (EHRs). Methods: We analysed data from 24,473 patients in the University of California Health Data Warehouse (UCHDW). A Fine–Gray subdistribution hazard model identified temporally associated diagnoses, from which we constructed diagnostic trajectories. We employed dynamic time warping and k-means clustering to group similar trajectories, and network analyses to characterize their common structures. Causal inferences were explored using the Greedy Equivalence Search algorithm. Validation in the UCHDW included association tests and comparison to control groups. We further validated our findings in the All of Us Research Program, a diverse, nationally representative cohort. Findings: After filtering, 5762 patients contributed 6794 unique AD progression trajectories, revealing four major trajectory clusters: mental health, encephalopathy, mild cognitive impairment, and vascular disease. These clusters differed significantly in demographic and clinical features. Approximately 26% of edges showed consistent directional ordering (e.g., hypertension → depressive episode → AD). In an independent population, these multi-step trajectories conferred greater AD risk than single diagnoses alone. Our validation in the All of Us cohort confirmed the reproducibility of these trajectory patterns in a more diverse population. Interpretation: Our findings demonstrate the value of examining sequential diagnostic patterns in AD pathogenesis. Multi-step progressions reveal potential latent contributors to AD, offering pathways for risk stratification, early detection, and targeted interventions. Funding: This study was supported by the National Institutes of Health, National Institute on Aging, the National Science Foundation, the Hillblom and Fineberg Foundations, and the California Department of Public Health.http://www.sciencedirect.com/science/article/pii/S2352396425002750Alzheimer’s diseaseDisease trajectoryElectronic health recordsMachine learningNetwork analysis |
| spellingShingle | Mingzhou Fu Sriram Sankararaman Bogdan Pasaniuc Keith Vossel Timothy S. Chang Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context EBioMedicine Alzheimer’s disease Disease trajectory Electronic health records Machine learning Network analysis |
| title | Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context |
| title_full | Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context |
| title_fullStr | Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context |
| title_full_unstemmed | Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context |
| title_short | Identifying common disease trajectories of Alzheimer’s disease with electronic health recordsResearch in context |
| title_sort | identifying common disease trajectories of alzheimer s disease with electronic health recordsresearch in context |
| topic | Alzheimer’s disease Disease trajectory Electronic health records Machine learning Network analysis |
| url | http://www.sciencedirect.com/science/article/pii/S2352396425002750 |
| work_keys_str_mv | AT mingzhoufu identifyingcommondiseasetrajectoriesofalzheimersdiseasewithelectronichealthrecordsresearchincontext AT sriramsankararaman identifyingcommondiseasetrajectoriesofalzheimersdiseasewithelectronichealthrecordsresearchincontext AT bogdanpasaniuc identifyingcommondiseasetrajectoriesofalzheimersdiseasewithelectronichealthrecordsresearchincontext AT keithvossel identifyingcommondiseasetrajectoriesofalzheimersdiseasewithelectronichealthrecordsresearchincontext AT timothyschang identifyingcommondiseasetrajectoriesofalzheimersdiseasewithelectronichealthrecordsresearchincontext |