Scholarlink

<p>Iron deficiency is a common cause of anaemia, but ferritin testing may be costly or difficult to access in resource-constrained settings. We conducted a retrospective diagnostic prediction study using de-identified hospital laboratory records from a tertiary-care academic centre in Northeast India to develop and internally validate models based on routine complete blood count parameters (Haemoglobin, Haematocrit, RBC, MCV, MCH, MCHC, RDW SD, and RDW CV), age, and sex for identifying ferritin-defined iron deficiency among adults. Unique adults aged 18&#xa0;years or older with same-day complete blood count and serum ferritin measurements were included. Two prespecified outcomes were ferritin less than 30&#xa0;ng/mL and less than 15&#xa0;ng/mL. The primary model was penalised logistic regression, with simple baseline, random forest, and support vector machine models as comparators. Among 1122 adults, 391 (34.8%) had ferritin less than 30&#xa0;ng/mL and 256 (22.8%) had ferritin less than 15&#xa0;ng/mL. Penalised logistic regression achieved test-set ROC-AUCs of 0.822 and 0.899, respectively, with PR-AUCs of 0.749 and 0.736. At balanced thresholds, sensitivity and specificity were 0.701 and 0.791 for ferritin less than 30&#xa0;ng/mL, and 0.792 and 0.823 for ferritin less than 15&#xa0;ng/mL. Routine complete blood count parameters showed clinically useful discrimination and may support triage, prioritisation of ferritin testing, and empirical treatment decisions after external validation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A complete blood count–based prediction model for ferritin-defined iron deficiency

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月28日

查看全文
  • Indrani Devi Sarma,
  • Debopam Das,
  • Bipasha Kalita,
  • Indrajit Nath,
  • Prasad Dange,
  • Bhupen Barman,
  • Rituparna Chetia,
  • Dibyajyoti Saikia

摘要

Iron deficiency is a common cause of anaemia, but ferritin testing may be costly or difficult to access in resource-constrained settings. We conducted a retrospective diagnostic prediction study using de-identified hospital laboratory records from a tertiary-care academic centre in Northeast India to develop and internally validate models based on routine complete blood count parameters (Haemoglobin, Haematocrit, RBC, MCV, MCH, MCHC, RDW SD, and RDW CV), age, and sex for identifying ferritin-defined iron deficiency among adults. Unique adults aged 18 years or older with same-day complete blood count and serum ferritin measurements were included. Two prespecified outcomes were ferritin less than 30 ng/mL and less than 15 ng/mL. The primary model was penalised logistic regression, with simple baseline, random forest, and support vector machine models as comparators. Among 1122 adults, 391 (34.8%) had ferritin less than 30 ng/mL and 256 (22.8%) had ferritin less than 15 ng/mL. Penalised logistic regression achieved test-set ROC-AUCs of 0.822 and 0.899, respectively, with PR-AUCs of 0.749 and 0.736. At balanced thresholds, sensitivity and specificity were 0.701 and 0.791 for ferritin less than 30 ng/mL, and 0.792 and 0.823 for ferritin less than 15 ng/mL. Routine complete blood count parameters showed clinically useful discrimination and may support triage, prioritisation of ferritin testing, and empirical treatment decisions after external validation.