Child underweight

Risk Exposure Overview 

Child underweight exposure is measured by weight for age z-score (WAZ). Notably, this child underweight exposure, by definition, is closely related to child stunting (height for age) exposure (weight is a function of height) as well as child wasting (weight for height exposure) exposure (both underweight and wasting exposures are a function of child weight).

Risk Exposures Description in GBD 

Underweight, a sub-component indicator of child growth failure (CGF), is based on a categorical definition using the WHO 2006 growth standards for children 0-59 months. Definitions are based on z-scores from the growth standards, which were derived from an international reference population. Mild, moderate and severe categorical prevalences were estimated for each of the three indicators. Theoretical minimum risk exposure level (TMREL) for underweight was assigned to be greater than or equal to one standard deviation below the mean (-1 SD) of the WHO 2006 standard weight-for-age (WAZ) z-score curve. This has not changed since GBD 2010.

The REI ID for child underweight is 94.

Risk exposure categories are described below:

Underweight exposure categories (range -7 to 7)
Category	Severity	WAZ range
cat4	TMREL	WAZ > -1
cat3	mild	-1 > WAZ > -2
cat2	moderate	-2 > WAZ > -3
cat1	severe	WAZ < -3

Vivarium Modeling Strategy 

We will model child underweight as an ordered polytomous variable.

We will use a “propensity exposure model” for child underweight, but with population risk exposure distributions that are conditional on child stunting and child wasting exposures. With this strategy, each simulant will be initialized with a “propensity” for child underweight that does not change throughout the duration of the simulation. The propensity should be assigned according to a uniform distribution between 0 and 1. This propensity will represent the simulant’s risk exposure percentile and will be used to determine the simulant’s child underweight exposure value by comparing it to the child underweight risk exposure distribution specific to the simulant’s joint stunting and wasting exposures.

Todo

Determine if child underweight propensity should be the same as child stunting propensity (test in notebook)

The underweight exposure distribution conditional on wasting and stunting exposures to be used in vivarium simulations can be found in the links below. Details on how these distributions were obtained can be found in the subsequent section.

Note

These values are currently not specific to MAM sub-exposures (“Better” and “Worse” MAM, as described on the wasting risk exposure document). Until these values are available, assume that the underweight exposure distribution specific to the wasting cat2 (MAM) exposure state in the linked file applies to both the “Better” and “Worse” MAM sub-exposures.

Note

The late neonatal age group should be applied to the early neonatal age group as well.

Todo

Generate and link underweight exposure distributions specific to MAM subcategories.

With this strategy, child underweight will dynamically update within a single age group in tandem with child wasting exposure changes and may also update upon transition to a new age group when child stunting exposure may change. Therefore, child underweight exposure in our simulation should be reassessed upon the following events:

Child wasting exposure state transition occurs
Child stunting exposure state changes (may only occur when simulant changes age groups)

Calculation of wasting exposure distribution conditional on stunting and wasting exposures 

The child underweight exposure distribution conditional on wasting and stunting exposure is custom calculated according to the following steps:

1. Obtain location-specific wasting, stunting, and underweight continuous correlation coefficients from Demographic Health Survey (DHS) data for the population of interest

2. In a “nano simulation,” assign simulants continuous WHZ, HAZ, and WAZ exposure propensity values according to the correlation coefficients obtained from DHS data.

3. Use these propensities to assign categorical wasting, stunting, and underweight exposure values to each simulant.

4. Calculate categorical child underweight exposure distributions specific to each joint wasting and stunting exposure category.

The notebook for which these steps were performed can be found at the links below:

Restrictions 

GBD 2021 Risk Exposure Restrictions
Restriction Type	Value	Notes
Male only	False
Female only	False
Age start	early neonatal (0-6 days), age group ID = 2	Note risk effects age start is 1 month
Age end	5 years (GBD 2019 1-4 age group ID=5; GBD 2021 2-4 age group ID = 34)

## REMINDER

#GBD 2020 age-group ids

early nn = 2
late nn = 3
1m-5m = 388
6m-11m = 389
12m-23m = 238
2y-4y = 34


#GBD 2019 age-group ids

early nn = 2
late nn = 3
post nn = 4
1-5 = 5

Assumptions and Limitations 

We model child underweight as a categorical exposure despite availability of an underlying continuous exposure distribution.

Validation Criteria 

Simulated population-level child underweight exposure should verify to GBD risk exposure distribution
Simulated child underweight exposure distributions conditional on joint wasting and stunting exposure categories should verify to conditional distribution input data
Validation criteria for child wasting and child stunting exposure models should continue to be satisfied