Related Experiment Video
Updated: Mar 28, 2026
07:32
Author Spotlight: Investigating Immune Cell Dynamics in the Tumor Microenvironment — Challenges and Innovations in Cancer Prognosis
Published on: April 12, 2024
2.1K
Identifying Hemophagocytic Lymphohistiocytosis and Describing Outcomes Using Computable Phenotypes: Retrospective
Suheyla Ocak1,2, Martin Yi3, Agata Wolochacz3
1Division of Haematology/Oncology, Hospital for Sick Children, 555 University Avenue, Toronto, ON, M5G 1X8, Canada, 1 4168131500.
JMIR Cancer
|March 26, 2026
Summary
Identifying patients with hemophagocytic lymphohistiocytosis (HLH) using electronic health records is challenging. Different methods capture distinct patient groups, highlighting the need for improved computable phenotypes for HLH research.
Area of Science:
- Hematology
- Clinical Informatics
- Immunology
Background:
- Hemophagocytic lymphohistiocytosis (HLH) is a critical, hyperinflammatory condition requiring prompt diagnosis and treatment.
- Electronic health records (EHRs) present challenges for systematically identifying HLH patients due to complex diagnostic criteria.
- Validated computable phenotypes are essential for accurate HLH patient identification and outcome analysis.
Purpose of the Study:
- To compare various EHR-based approaches for developing computable phenotypes for HLH.
- To evaluate characteristics and outcomes of HLH patients meeting diagnostic criteria, comparing those who received therapy versus those who did not.
Main Methods:
- Developed three computable phenotypes: HLH ICD-10 code, HLH treatment plan, and HLH-2004 clinical criteria.
- Analyzed patient characteristics and outcomes based on HLH-directed therapies (chemotherapy or specific agents).
- Compared outcomes, including in-hospital mortality, between treated and untreated HLH patients meeting criteria.
Main Results:
- Identified 388 potential HLH patients across cohorts; ICD-10 codes and clinical criteria were more common than treatment plans.
- Among patients meeting HLH-2004 criteria, 79% received HLH-directed therapy.
- In-hospital mortality was high in both treated (15%) and untreated (13.5%) groups; few patients with fever and elevated ferritin met diagnostic criteria.
Conclusions:
- Creating HLH patient cohorts from EHR data is complex, with distinct populations identified by diagnosis codes, treatment plans, and clinical criteria.
- Each computable phenotype approach captures unique patient subsets, indicating the need for multi-faceted strategies.
- Further research is needed to refine EHR-based identification methods for HLH.
