Anemia Screening

Abbreviations

Abbreviation

Definition

Note

GBD

Global Burden of Disease

Intervention Overview

Hemoglobin Screening Accuracy Instructions

Research background:

Anemia is defined as decreased blood concentration of hemoglobin, irrespective of underlying cause, red blood cell morphology, or red blood cell function. A noninvasive blood test (using a small, portable device such as a HemoCue test) can be provided to pregnant people in their 2nd or 3rd trimester at antenatal care clinics to quickly and accurately measure hemoglobin levels in blood. If a pregnant person is found to have hemoglobin less than 100 g/L based on this hemoglobin screening, they will also be screened for ferritin levels, which you can read more about in the following section. This page describes how an anemia screening intervention (including hemoglobin and ferritin screenings) can be implemented and calibrated for the MNCNH Portfolio model.

Modeling instructions:

For the “Hemoglobin screening value <100 g/L?” decision node in the hemoglobin module page, we will assess whether or not the result of a simulant’s noninvasive blood test for hemoglobin screening is <100 g/L, which may be different than whether a simulant’s actual hemoglobin exposure is <100 g/L. We will do this based on assumed sensitivity and specificity levels for the hemoglobin screening test as informed from the Gates Foundation and listed below:

  • Sensitivity (percent of true positives that test positive): 85%

  • Specificity (percent of true negatives that test negative): 80%

Follow the steps below to determine the answer to the decision node:

  1. Assess a simulant’s “true” low hemoglobin status based on their hemoglobin exposure at the time of screening, which should be based on their true hemoglobin exposure value after effects from oral iron received at the first trimester ANC visit and before any effects from interventions received at the second trimester have been applied. In other words, use oral iron-affected hemoglobin exposure for those who attend ANC during the first trimester and “IFA-deleted” hemoglobin exposure for those who do not attend ANC during the first trimester (but do later in pregnancy). Low hemoglobin status corresponds to values of <100 g/L and adequate hemoglobin status corresponds to values of 100+ g/L.

  2. For simulants that are truly low hemoglobin, assign tests low hemoglobin status to 85% (sensitivity value) and tests adequate hemoglobin status to 15% (100 - sensitivity value of 85)

  3. For simulants that are truly adequate hemoglobin, assign tests adequate hemoglobin status to 80% (specificity) and tests low hemoglobin status to 20% (100 - specificty value of 80)

  4. Use the test hemoglobin status to determine the answer to the decision node (answer is “yes” if they have test low hemoglobin status and “no” if they have test adequate hemoglobin status)

  5. Record true and test hemoglobin exposures at the time of screening to the relevant outputs (to be used for V&V in the interactive simulation)

Example Python code implementing these steps is available on the hemoglobin module page.

Ferritin Screening Instructions

Research background:

Ferritin is a protein that stores iron within the body and low blood ferritin levels can indicate low iron stores. Pregnancies that have hemoglobin less than 100 g/L based on the hemoglobin screen will also be screened for ferritin levels via a minimally invasive screening (finger prick). Pregnancies that have a hemoglobin level <100 g/L and a blood ferritin level below 15 ug/L (anemic AND iron deficient) are eligible for IV iron.

Notably, the GBD does not have any estimates related to ferritin exposure or ferritin screening. However, the GBD assigns specific causes to all cases of anemia. Some of these causes of anemia are considered “iron responsive,” indicating that they are iron deficiency anemia. An example of an iron deficiency anemia is anemia caused by maternal hemorrhage (caused by blood loss, decreasing systemic levels of both hemoglobin and iron). An example of a non-iron responsive anemia is sickle cell trait (low hemoglobin is due to a defect in hemoglobin protein rather than low iron levels). Notably, it is possible for non-iron-responsive anemias to also have low iron levels. See the anemia impairment document for a list of iron responsive and non iron responsive causes of anemia in the GBD.

Unlike hemoglobin screening, we do not explicitly model any test inaccuracy with e.g. a sensitivity and specificity. We assign simulants only a tested ferritin value, not a true underlying value.

In our model we will use the severity-specific fraction of iron-responsive anemia among all causes of anemia in GBD as a proxy measure for the fraction of anemia cases (of that severity) who receive a low ferritin test result when screened. This approach is limited in that we may slightly underestimate low ferritin by not considering the proportion of the population who has low hemoglobin due to an iron-non-responsive cause and also coincidentally has low ferritin. It is also limited in that we may overstate the differences between anemia severities with respect to ferritin test results due to testing inaccuracies (false positives and false negatives).

Note

Chris T. has suggested that we can use the fraction of iron deficiency anemia from the in-progress PRISMA study rather than GBD for this purpose. PRIMSA study results are expected in June or July of 2025.

Modeling instructions:

The probability of low ferritin screening is dependent on the simulant’s location, age group, and anemia status at the time of screening. Anemia status at the time of screening should be based on their true hemoglobin exposure value after effects from oral iron received at the first trimester ANC visit and before any effects from interventions received at the second trimester have been applied. In other words, use oral iron-affected hemoglobin exposure for those who attend ANC during the first trimester and “IFA-deleted” hemoglobin exposure for those who do not attend ANC during the first trimester (but do later in pregnancy). See the anemia/hemoglobin exposure table here for reference and remember to use the pregnancy-specific values.

The probability of low ferritin specific to location, age, and anemia status (termed exp_among_{SEVERITY} in the table below) can be calculated according to the parameters defined in the table below.

Ferritin parameters

Parameter

Definition

Value

Note

exp_among_non_anemic

Rate of low ferritin exposure among the population without anemia

exp_among_mild / 2

Model assumption given that it is definitionally expected to have a lower rate than mild anemia, but also should be >0 as it is possible to have low ferritin and adequate hemoglobin

exp_among_{SEVERITY}

Rate of low ferritin exposure among the population with a given anemia severity of mild, moderate, or severe

prev_{SEVERITY}_iron_responsive_anemia_sequelae / prev_{SEVERITY}_anemia_impairment

prev_{SEVERITY}_iron_responsive_anemia_sequelae

Sum of sequela-level prevalence for specified list of sequela IDs that represent anemia severity-specific iron responsive anemia

See get_draws call below this table

prev_{SEVERITY}_anemia_impairment

Severity-specific anemia impairment prevalence

from GBD: source=’como’, mild anemia REI = 205, moderate anemia REI = 206, severe anemia REI = 207

mild_ira_sids

List of sequela IDs that represent all mild iron responsive anemias

[144, 172, 177, 240, 182, 5393, 23030, 23034, 23038, 23046, 23042, 7202, 4976, 4952, 4955, 5627, 7214, 5009, 4985, 4988, 5678, 5567, 5579, 22989, 5225, 5249, 5273, 22990, 5228, 5252, 5276, 22991, 1016, 1421, 1373, 22992, 1024, 1433, 1385, 22993, 1032, 1445, 1397, 1106, 525, 23187, 23179, 23162, 23488, 206]

List generated in this notebook

moderate_ira_sids

List of sequela IDs that represent all moderate iron responsive anemias

[145, 173, 178, 241, 183, 5396, 23031, 23035, 23039, 23047, 23043, 7205, 4979, 4958, 4961, 5630, 7217, 5012, 4991, 4994, 5681, 5570, 5582, 22999, 5219, 5243, 5267, 23000, 5222, 5246, 5270, 23001, 1017, 1424, 1376, 23002, 1025, 1436, 1388, 23003, 1033, 1448, 1400, 1107, 526, 23188, 23180, 23163, 23489, 207]

List generated in this notebook

severe_ira_sids

List of sequela IDs that represent all severe iron responsive anemias

[146, 174, 179, 242, 184, 5399, 23032, 23036, 23040, 23048, 23044, 7208, 4982, 4964, 4967, 5633, 7220, 5015, 4997, 5000, 5684, 5573, 5585, 23009, 5213, 5237, 5261, 23010, 5216, 5240, 5264, 23011, 1018, 1427, 1379, 23012, 1026, 1439, 1391, 23013, 1034, 1451, 1403, 1108, 527, 23189, 23181, 23164, 23490, 208]

List generated in this notebook

 
from get_draws.api import get_draws
year_id = 2023
gbd_release_id = 16 # gbd 2023

{SEVERITY}_sequela_data = get_draws(release_id=gbd_release_id,
          year_id=year_id,
          sex_id=sex_id,
          age_group_id=age_group_id,
          source='como',
          gbd_id_type='sequela_id',
          gbd_id={SEVERITY}_ira_sids,
          measure_id=5, # prevalence
          sex_id=2, # female (only need female for the MNCNH simulation)
          # location_id = location_ids, # specify according to modeled locations
          # age_group_id = age_group_ids # specific according to modeled age groups
          )
prev_{SEVERITY}_iron_responsive_anemias = ({SEVERITY}_sequela_data.groupby(['location_id','age_group_id','sex_id'])
                                            [[x for x in {SEVERITY}_sequela_data.columns if 'draw' in x]].sum())

Baseline Coverage Data

Baseline coverage of the minimally invasive blood test for hemoglobin screening is defined by estimates processed by the Health Systems team. The country-specific estimates are available at J:\Project\simulation_science\mnch_grant\MNCNH portfolio\anc_bloodsample_prop_st-gpr_results_aggregates_scaled2025-05-29.csv.

Baseline coverage of ferritin screening is defined in the table below.

Baseline Coverage of Ferritin Screening

Location

Coverage Mean (%)

Coverage Distribution (%)

Notes

All (Ethiopia, Nigeria, Pakistan)

0

N/A

This is an assumption based on literature evidence that many ANC programs primarily focus on hemoglobin screening, and ferritin screening is not widely available at ANCs in Nigeria, Ethiopia, or Pakistan. (e.g. [Teichman-et-al-2021] assessed ferritin testing prevalence in high-resource settings in Ontario and found 59.4% of pregnant patients were ferritin tested during pregnancy but that this was significantly lower in low-income areas, with only 4.1% in the lowest wealth quintile.)

Assumptions and Limitations

  • We assume that if a pregnant person had their blood drawn at the ANC during their pregnancy, their hemoglobin concentration was assessed. We thereby assume that the coverage estimates for blood samples taken at ANC that we received from the Health Systems team are reasonable values for baseline coverage of hemoglobin screening at ANC in our locations of interest.

  • We assume that baseline coverage for ferritin screening at ANC is 0%, based on literature evidence that many ANC programs primarily focus on hemoglobin screening, and is not widely implemented in Nigeria, Ethiopia, or Pakistan. (e.g. [Teichman-et-al-2021] assessed ferritin testing prevalence in high-resource settings in Ontario and found 59.4% of pregnant patients were ferritin tested during pregnancy but that this was significantly lower in low-income areas, with only 4.1% in the lowest wealth quintile.)

  • We assume a hemoglobin screening sensitivity of 85% and specificity of 80%, as requested by the Gates Foundation

  • Our approach to modeling hemoglobin screening sensitivity and specificity does not vary by hemoglobin exposure. In other words, you are no more likely to have your hemoglobin exposure misclassified by the screening if your exposure is very close to the threshold than if you expsoure is far away from the threshold. This will likely result in more cases of individuals without any anemia (high hemoglobin) testing as low hemoglobin and those with very low hemoglobin testing as adequate hemoglobin than may happen in practice. This may cause us to understimate the impact of the IV iron intervention. Note that an alternative to this limited approach we are taking would be to model some error around hemoglobin exposure (sampling from some distribution and adding it to hemoglobin exposure to get test exposure, similar to what is done for gestational age assessment in the AI ultrasound model). However, in order to match the desired sensitivity and specificity of the screening test, we would need to solve for the uncertainty distribution, likely via optimization, at the location-specific level (as it will depend on the underlying population hemoglobin exposure distribution).

  • We use the severity-specific fraction of iron responsive anemia among all causes of anemia in GBD as a proxy measure for the fraction of anemia cases with low ferritin test results. This approach is limited in that we may slightly underestimate total eligibility by not considering the proportion of the population who has low hemoglobin due to an iron-non-responsive cause and also coincidentally has low ferritin. It is also limited in that we may overstate the differences between anemia severities with respect to ferritin test results due to testing inaccuracies (false positives and false negatives).

  • In the absence of data to directly inform otherwise, we assume that the population without anemia has half the rate of low ferritin test results as the population with mild anemia. We made this assumption given that the population without anemia is expected to have a low ferritin exposure level that is greater than zero but less than that of the population with mild anemia.

Todo

If we find more suitable baseline coverage data for ferritin screening in ANCs in our locations of interest, we will update this page accordingly.

Validation and Verification Criteria

The following V&V criteria should be met:

  • The coverage of each intervention (hemoglobin screening and ferritin screening) by scenario should match the proportions outlined in the Scenarios section of the MNCNH Portfolio concept model

  • There should be a sensitivity (% of true positives that test positive) of 85% and specificity (% of true negatives that test negative) of 80% for those that received hemoglobin screenings.

  • There should be the expected proportion of simulants with low and high ferritin status for those that received ferritin screenings.

References

[Teichman-et-al-2021] (1,2)

Teichman, J., Nisenbaum, R., Lausman, A., Shlozberg, M. Suboptimal iron deficiency screening in pregnancy and the impact of socioeconomic status in a high-resource setting. Blood Adv (2021) 5 (22): 4666–4673. https://doi.org/10.1182/bloodadvances.2021004352