Hemoglobin Risk Exposure

Risk Exposure Overview 

Hemoglobin is a protein found in red blood cells. Hemoglobin contains iron and helps to facilitate the transporation of oxygen throughout the body. Hemoglobin concentration for an individual can be assessed via a blood test.

Risk Exposures Description in GBD 

In GBD, estimation of the hemoglobin exposure distribution serves two purposes:

To estimate the prevalence envelope of the anemia impairment (and to perform anemia causal attribution). This is done for all age/sex groups and data is stored as several modelable entities, which are split by pregnancy status.
To serve as the risk exposure estimate of the hemoglobin risk factor. The hemoglobin risk factor only affects pregnancy-related outcomes in GBD, so the exposure is only combined with relative risk for the pregnant population. As a result, the exposure stored in the risk factor (REI ID 376) is from modelable entity 28869, “Mean hemoglobin for risk modelling”. The description of the GBD 2023 model marked “best” for this entity states that it is specific to the pregnant population, though it includes estimates for all ages/sexes (the non-WRA estimates are presumed to be non-pregnant exposures).

The remainder of this page is specific to the second application of hemoglobin as a risk factor exposure specifically among the pregnant population.

Starting in GBD 2023, hemoglobin exposure among the pregnant population is estimated directly. This is an update from the previous approach of estimating the hemoglobin exposure among the non-pregnant population and then applying a “pregnancy adjustment factor” to obtain an exposure estimate among the pregnant population.

Note

See this notebook that demonstrates the risk exposure data for the low hemoglobin risk factor is consistent with the MEID for hemoglobin exposure data among the pregnant population.

More details on the risk exposure model can be found in the anemia impairment portion of the GBD 2023 methods appendix.

Finally, the risk effects document for the low hemoglobin risk factor can be found here.

Vivarium Modeling Strategy 

To model hemoglobin exposure among non-pregnant populations, follow the modeling strategy found here. Details for modeling hemoglobin exposure among the pregnant population are included below.

Hemoglobin exposure will be modeled as a continuous exposure that follows an ensemble distribution parameterized by a mean and standard deviation. Simulants should be assigned a hemoglobin propensity value (random value between 0 and 1) that does not change throughout their lifetime and will be used to determine their specific hemoglobin exposure at any time by comparing it to the population distribution of hemoglobin specific to the simulant’s current age and pregnancy status. If individual simulants are tracked across pregnant and non-pregnant states in the same simulation, their hemoglobin exposure propensity should remain constant throughout these transitions, but the exposure distribution used to determine their individual exposure will change between pregnant and non-pregnant exposures accordingly. More specifically, an individual simulant’s hemoglobin exposure value will be determined by taking the percent point function (PPF) for their propensity value of the relevant population hemoglobin exposure distribution.

Todo

Determine if we want hemoglobin exposure propensity to be correlated with any other risks for the MNCNH model, such as BMI and/or hypertensive disorders

Restrictions 

GBD 2023 Risk Exposure Restrictions
Restriction Type	Value	Notes
Male only	False
Female only	True	Pregnant population ONLY
Age group start	age_group_id=7	Starts at age 10
Age group end	age_group_id=15	Ends at age 55

Assumptions and Limitations 

We assume that there are no natural variations in hemoglobin exposure throughout the duration of pregnancy. This is unrealistic as hemoglobin is typically depleted over the course of a pregnancy, especially without intervention.

Todo

Add more assumptions and limitations, especially as we receive documentation from GBD modeling

Data Description Tables 

Distribution Parameters
Parameter	Value	Note
Exposure mean	rei_id=376, source=’exposure’, `get_draws(gbd_round_id=9, gbd_id_type='rei_id', gbd_id=376, source='exposure', release_id=16)`	Expressed in grams per liter
Exposure standard deviation	rei_id=376, source=’exposure_sd’, `get_draws(gbd_round_id=9, gbd_id_type='rei_id', gbd_id=376, source='exposure_sd', release_id=16)`	Expressed in grams per liter
Ensemble distribution weights	Found at `/ihme/epi/risk/ensemble/_weights/low_hgb.csv`	40% gamma, 60% mirror gumbel
XMIN	40	Expressed in grams per liter
XMAX	150	Expressed in grams per liter

Validation Criteria 

Population mean hemoglobin among the simulated pregnant population should validate to the exposure value for REI 376 in the baseline scenario
The hemoglobin standard deviation among the simulated pregnant population should validate to the exposure SD value for REI 376 in the baseline scenario
Prior to implementation of any hemoglobin-affecting interventions, hemoglobin exposures should not change at the individual level over the course of pregnancy.