Evaluation of the HemoScreen Point-of-Care Hematology Device Compared with the Sysmex XN-1000 Analyzer and Digital Microscopy in Pediatric Inpatients

Point-of-care (POC) hematology analyzers have become increasingly relevant in recent years, mainly due to the need for rapid and reliable diagnostic information in emergency departments, acute care units, and outpatient clinics. In pediatrics, quick access to complete blood count (CBC) results is essential, as early detection of infections, cytopenias, anemia, and other hematologic abnormalities directly supports clinical decision-making and may improve patient outcomes. The limited circulating blood volume in children, combined with the discomfort and repeated sampling often required during hospitalization, highlights the value of analytical platforms that require minimal sample volume. Despite these potential advantages, evidence on the performance of POC hematology analyzers specifically in pediatric populations remains limited, and robust comparisons with central laboratory methods are needed to ensure accuracy, reliability, and appropriate integration into routine pediatric workflows. This thesis evaluated the analytical and diagnostic performance of HemoScreen, an AI-based POC hematology analyzer requiring 40 µL of blood, compared with central laboratory methods for both CBC parameters and WBC differential counts in pediatric patients. A total of 141 venous K2-EDTA samples from inpatients aged 2 months to 17 years were analyzed. CBCs were performed using the Sysmex XN-1000 and subsequently HemoScreen. Blood smears were digitized with CellaVision DC-1 and independently reviewed by experienced cytologists. Analytical agreement for WBC, red blood cells (RBC), platelets (PLT), hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), red cell distribution width (RDW), mean platelet volume (MPV), and WBC differentials (neutrophils, lymphocytes, monocytes, eosinophils, and basophils) was assessed using Spearman’s correlation, Passing-Bablok regression, Bland-Altman analysis, and proportions exceeding predefined total allowable error (TEa) limits. Diagnostic agreement was evaluated for clinically relevant categories using Cohen’s κ, and for concordance between HemoScreen flagging and cytologist review. Correlations for CBC indices were strong to excellent (ρ = 0.783-0.992), with mean bias values generally within adult European Federation of Clinical Chemistry and Laboratory Medicine (EFLM)-derived acceptable bias limits; Using the Passing-Bablok regression as the reference line, fewer than 10% of measurements exceeded TEa thresholds for WBC, RBC, PLT, and most secondary parameters. Correlations for WBC differentials were also strong to excellent (ρ = 0.625-0.977). Diagnostic agreement was excellent for leukopenia, leukocytosis, thrombocytopenia, anemia, elevated HGB, neutropenia, neutrophilia and lymphocytopenia (κ = 0.807-1.000); good for thrombocytosis, macrocytosis, microcytosis, lymphocytosis, monocytosis, and eosinophilia (κ = 0.606-0.799). HemoScreen flagged 31.9% of samples for abnormal cells, with a high negative predictive value (92.6%), consistent with a conservative flagging behavior. Overall, in pediatric venous samples, HemoScreen demonstrated high analytical and diagnostic agreement with reference methods for both CBC parameters and WBC differential counts, indicating its potential feasibility for implementation in children’s hospitals.