An Algorithm to Identify Immunocompromised Patients in French Claims Data for Rapid Preventive or Therapeutic Interventions
Author(s)
Nicolas Capit, PhD1, Laureen Majed, PhD1, Aleksandra Anchim, PhD1, Cécile Artaud, MSc1, Barbara Lebas, MSc2, Marc Jouve, MSc2, Florent Malard, MD3, Fanny Vuotto, MD4, Philippe Gatault, MD5, Cécile Janssen, MD6.
1AstraZeneca, Courbevoie, France, 2Sancare, Paris, France, 3Service d'Hématologie Clinique et de Thérapie cellulaire, AP-HP Hôpital Saint Antoine, Paris, France, 4Infectious Diseases Unit, CHU de Lille, Lille, France, 5Service Néphrologie, Hypertension Artérielle, Dialyses et Transplantation Rénale, CHRU de Tours, Tours, France, 6Infectious Diseases Unit, Centre Hospitalier Annecy Genevois, Annecy, France.
1AstraZeneca, Courbevoie, France, 2Sancare, Paris, France, 3Service d'Hématologie Clinique et de Thérapie cellulaire, AP-HP Hôpital Saint Antoine, Paris, France, 4Infectious Diseases Unit, CHU de Lille, Lille, France, 5Service Néphrologie, Hypertension Artérielle, Dialyses et Transplantation Rénale, CHRU de Tours, Tours, France, 6Infectious Diseases Unit, Centre Hospitalier Annecy Genevois, Annecy, France.
OBJECTIVES: Immunocompromised (IC) patients represent a heterogeneous population with increased burden of respiratory infections. In the context of prevention, timely identification of high-risk patients enables early outreach for vaccination, preventive therapy, or physician notification, to prevent complications, hospitalization, or poor outcomes. As large administrative databases are increasingly utilized in research, standardized methods for identifying IC status are essential for reproducible studies. We evaluated an algorithm to identify four major IC patient categories in French hospital claims databases.
METHODS: An algorithm was developed based on literature review and expert opinion using medical coding systems in France (ICD-10, GHM, CCAM) to identify patients with solid tumors under active treatment, hematological malignancies, solid organ transplants, and end-stage kidney disease. For performance assessment, 563 patients hospitalized in 2022 in relevant departments across 4 hospitals were included. The algorithm classified patients using claims with a 5-year lookback period and was assessed against claims and electronic medical record (EMR) review by physicians (gold standard).
RESULTS: The algorithm demonstrated sensitivity of 94% and specificity of 100%, positive predictive value (PPV) 100%, and negative predictive value (NPV) 89% across all cohorts. Performance varied by IC category: solid tumors (sensitivity 95%, specificity 100%, PPV 100%, NPV 89%), hematological malignancies (sensitivity 94%, specificity 100%, PPV 100%, NPV 82%), end-stage kidney disease (sensitivity 91%, specificity 100%, PPV 100%, NPV 93%), and transplant recipients (sensitivity 100%, specificity 100%, PPV 100%, NPV 100%). The primary reason for false negatives was patients receiving outpatient chemotherapy, only reported as text in EMRs.
CONCLUSIONS: This algorithm is validated to identify IC patients for four common causes: solid tumors under active treatment, hematological malignancies, end-stage kidney disease, and solid organ transplant. Even with the limitations of claims data, this algorithm achieved high sensitivity and specificity. It offers a practical tool to support timely preventive or therapeutic interventions in high-risk populations.
METHODS: An algorithm was developed based on literature review and expert opinion using medical coding systems in France (ICD-10, GHM, CCAM) to identify patients with solid tumors under active treatment, hematological malignancies, solid organ transplants, and end-stage kidney disease. For performance assessment, 563 patients hospitalized in 2022 in relevant departments across 4 hospitals were included. The algorithm classified patients using claims with a 5-year lookback period and was assessed against claims and electronic medical record (EMR) review by physicians (gold standard).
RESULTS: The algorithm demonstrated sensitivity of 94% and specificity of 100%, positive predictive value (PPV) 100%, and negative predictive value (NPV) 89% across all cohorts. Performance varied by IC category: solid tumors (sensitivity 95%, specificity 100%, PPV 100%, NPV 89%), hematological malignancies (sensitivity 94%, specificity 100%, PPV 100%, NPV 82%), end-stage kidney disease (sensitivity 91%, specificity 100%, PPV 100%, NPV 93%), and transplant recipients (sensitivity 100%, specificity 100%, PPV 100%, NPV 100%). The primary reason for false negatives was patients receiving outpatient chemotherapy, only reported as text in EMRs.
CONCLUSIONS: This algorithm is validated to identify IC patients for four common causes: solid tumors under active treatment, hematological malignancies, end-stage kidney disease, and solid organ transplant. Even with the limitations of claims data, this algorithm achieved high sensitivity and specificity. It offers a practical tool to support timely preventive or therapeutic interventions in high-risk populations.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
RWD17
Topic
Medical Technologies, Methodological & Statistical Research, Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems, Reproducibility & Replicability
Disease
Infectious Disease (non-vaccine), Oncology, Systemic Disorders/Conditions (Anesthesia, Auto-Immune Disorders (n.e.c.), Hematological Disorders (non-oncologic), Pain)