Background: Hemophagocytic lymphohistiocytosis (HLH) is a rare, acute, uncontrolled immune activation syndromecharacterized by clinical evidence of extreme inflammation. While ICD codes are available for HLH-relatedconditions, identifying and diagnosing HLH is challenging due complex diagnostic criteria, a high mortality rate,and heterogeneous etiology and presentation.

Aims: The objective of this study was to develop novel strategies to compile a retrospective cohort of HLH patients at alarge US healthcare system.

Methods: An Electronic Medical Record (EMR) system was queried 2009-May 2022 to identify potential patients. Threemethods were developed for patient identification. Method 1 identified patients with any mention of an HLH-related ICD code (ICD-9 288.4; ICD-10 D76.1, D76.2, D76.3). Method 2 identified patients with elevated ferritinvalues (>1,000 ng/mL or >10,000 ng/mL) or a sCD25 test (predictive markers of HLH). Method 3 identifiedpatients with etoposide and dexamethasone treatment, removing patients receiving this treatment as part of achemotherapy regimen. A clinician reviewed a random sample of patients identified by each method for HLHdiagnosis.

Results: Method 1 (N=327) was found to be predictive for HLH, with 17% of the random 20% sample confirmed HLH; allMethod 1 patients were included for full chart review. The following cohorts within Method 2 (N=1,057) werefound to be predictive of an HLH diagnosis and were included for full chart review: ferritin >10,000 ng/mL andany sCD25 test (N=42, 13% of patients in the 20% random sample confirmed HLH) and patients with ferritin1,000-10,000 ng/mL and any sCD25 test (N=94, 17% of patients in the 20% random sample confirmed HLH).Random 10% sampling of the cohorts of patients with 1) any sCD25 test and no ferritin >1,000 ng/mL (N=374)and 2) ferritin >10,000 ng/mL and no sCD25 test (N=547) were found to have few patients with HLH (N=1 or 3%and N=1 or 2%, respectively). Since full chart review of these large cohorts was not logistically feasible, dischargesummaries within these cohorts were screened for mentions of HLH or MAS resulting in 64 unique patientsidentified for full chart review. Chart review of a random 20% sample of patients within Method 3 (N=86) foundno cases of HLH. One unique patient was identified from searching discharge summary notes of this patientpopulation. After de-duplication, 528 unique patients were identified for full chart review to determine an HLHdiagnosis. Results of this chart review will be presented.

Summary/Conclusion: Our expanded algorithm captured a substantial number of patients that would have been missed withidentification using ICD codes alone, indicating a potential underestimation of HLH incidence.