Prevalence of Iron Deficiency and Anemia in Yemeni Hospital Patients: A Retrospective Cross Sectional Study with Machine Learning Prediction from Complete Blood Count Parameters

Study design and settings
This retrospective cross sectional study was conducted using laboratory data from two public tertiary referral hospitals in Yemen: Al Thawra General Hospital (Ibb Center) and Al Mukha General Hospital (Taiz Hospital). These facilities were chosen to represent the central and western regions of Yemen, serving mixed urban and rural populations. The cross sectional design allowed evaluation of iron status patterns across age groups within a large routine laboratory dataset. Data were collected between January 2019 and March 2025. The study was approved by the institutional review boards of both participating hospitals, and patient identifiers were anonymized before analysis. All eligible patients were included without further sampling, as this was a record based analysis of consecutive laboratory requests.
Study population
We included all patients aged 0.1 years or older with available measurements of serum ferritin and complete blood count (CBC). Patients were excluded if they had missing age or sex, missing ferritin or key CBC parameters (hemoglobin, mean corpuscular volume [MCV], or mean corpuscular hemoglobin [MCH]), or duplicate records for the same patient identifier. Excluding patients with missing data ensured internally consistent definitions of anemia and iron deficiency across all participants. For duplicate records, only the first available measurement (earliest date) per patient was retained. After applying these criteria, a total of 1,998 patients were included in the analysis.
Data collection
Data were extracted via structured SQL queries from the electronic laboratory information system of each hospital. Variables included age (in years), sex (coded as 0 for female and 1 for male), serum ferritin (in ng/mL), hemoglobin (in g/dL), red blood cell (RBC) count (in ×10⁶/µL), packed cell volume (in %), MCV (in fL), MCH (in pg), white blood cell (WBC) count (in ×10³/µL), and platelet count (in ×10³/µL). Multiple measurements per patient were allowed before deduplication, but only the initial ferritin value was retained per patient for analysis.