Antenatal care attendance module

1.0 Overview 

This module determines a simulant’s antenatal care attendance exposure according to their ANC visit propensity value and the distribution of ANC attendance exposure categories specific to their location and broad pregnancy outcome. ANC attendance exposure categories include whether a simulant attends an ANC visit during their first trimester and/or later during their pregnancy.

ANC visit timing is particularly relevant to both the hemoglobin and facility choice components of this simulation, so for more information on how outputs from this module will be used, refer to our hemoglobin module documentation and our facility choice module documentation.

2.0 Module Input and Output Data 

2.1 Module Inputs 

Module required inputs
Input	Source module	Application	Note
Broad pregnancy outcome	Initial attributes module	Determines which set of probabilities to use for the ANC exposure categories.	Live birth or stillbirth pregnancies can be assigned any of the values A, B, C, or D, whereas we assume abortion/miscarriage/ectopic pregnancies can only attend ANC in the first trimester, so only categories B and D have nonzero probability. See exposure probability tables below.
ANC propensity	Initial attributes module	Use the ANC propensity together with the ANC attendance exposure probabilities below to select one of the ANC exposure categories A, B, C, or D, as described in the special ordering of the categories section of the facility choice model document. When sampling, the categories should be ordered D < C < B < A (highest risk to lowest risk in terms of ultrasound timing), so that higher propensity values correspond to seeking more ANC care.	ANC propensity is correlated with LBWSG category propensity and IFD propensity as described in the the correlated propensities section of the facility choice model document. Currently we assume that there is no correlation of ANC with other factors.

Additional input data
Input	Source	Definition
ANCfirst	ANCfirst is processed by the Health Systems team at IHME and available on the J drive here: `/snfs1/Project/simulation_science/mnch_grant/MNCNH portfolio/anc1_first3months_st-gpr_results_aggregates_scaled2025-05-27.csv`. We have implemented this so that we will use data specific to the `year_start` value in the artifact (2021 when using GBD 2021 data and 2023 when using GBD 2023 data). Relevant code is shown here)	Proportion of pregnant people attending ANC in the first trimester
ANC1	GBD covariate ID 7: `get_covariate_estimates(location_id=location_id, release_id=16, year_id=2023, covariate_id=7)`	Proportion of pregnant people receiving any antenatal care from a skilled provider
ANC4	GBD covariate ID 8: `get_covariate_estimates(location_id=location_id, release_id=16, year_id=2023, covariate_id=8)`	Proportion of pregnant people receiving 4 or more antenatal care visits including 1 or more from a skilled provider

2.2 Module Description 

ANC attendance will be modeled as a single variable with 4 possible exposure options:

Attends ANC during first trimester AND later pregnancy
Attends ANC during first trimester but NOT later pregnancy
Attends ANC during later pregnancy but NOT first trimester
Does not attend ANC at all during pregnancy

ANC exposure options
	Visit during late pregnancy	No visit during late pregnancy
Visit during first trimester	A	B
No visit during first trimester	C	D

The below table describes what probability values to use for each exposure option outlined above, for pregnancies resulting in live birth or stillbirth.

ANC exposure probability values for pregnancies resulting in live birth or stillbirth
ANC exposure option	Description	Probability value	Notes
A	Attends ANC during first trimester AND later pregnancy	`min(ANCfirst, ANC4)`	Assume that attending ANC in first trimester reflects “active care seeking behavior” and that it is unlikely for someone who attends first trimester ANC to attend no subsequent visits.
B	Attends ANC during first trimester but NOT later pregnancy	`ANCfirst - min(ANCfirst, ANC4)`	Prevalence of first trimester visit ONLY (and no late pregnancy visit) only occurs if ANCfirst > ANC4
C	Attends ANC during later pregnancy but NOT first trimester	`ANC1 - ANCfirst`
D	Does not attend ANC at all during pregnancy	`1 - ANC1`

Note

As of pull request #1690 we updated how we assign our ANC exposures to include the ANCfirst variable that the HS team processed and shared with us. Please see these slides for more information on this strategy update.

The above probabilities are to be implemented pregnancies resulting in live birth or stillbirth only. Abortion/miscarriage/ectopic pregnancies are assigned probabilities differently because we assume their pregnancies end before they can attend later pregnancy ANC visits. The below table describes what probabilities to use for each exposure option for abortion/miscarriage/ectopic pregnancies:

ANC exposure probabilities for abortion/miscarriage/ectopic pregnancies
ANC exposure option	Description	Probability value	Notes
A	Attends ANC during first trimester AND later pregnancy	0	Assumption
B	Attends ANC during first trimester but NOT later pregnancy	`ANCfirst`
C	Attends ANC during later pregnancy but NOT first trimester	0	Assumption
D	Does not attend ANC at all during pregnancy	Probability equal to `1 – ANCfirst`

2.3 Module Outputs 

As mentioned earlier, ANC attendance impacts hemoglobin exposure and facility choice in our model, and in order for the outputs of this component to be compatible with the data needs of these two downstream components, we will need two different outputs, one being dichotomous for the hemoglobin component and the other being polychotomous for the facility choice component.

Module outputs
Output	Value	Dependencies
First trimester ANC attendance	True for groups A and B False for groups C and D	Used as an input for the hemoglobin module. This variable is dichotomous for each pregnancy.
Late pregnancy ANC attendance	True for groups A and C False for groups B and D	Used as an input for the hemoglobin module. This variable is dichotomous for each pregnancy.
ANC attendance category	`none` for group D `later_pregnancy_only` for group C `first_trimester_only` for group B `first_trimester_and_later_pregnancy` for group A The categories of this polytomous variable are listed from highest risk (1) to lowest risk (4) in terms of ultrasound timing, in accordance with the special ordering of the categories section of the delivery facility choice model document: The categories need to be ordered D < C < B < A when sampling the ANC attendance variable using the correlated ANC propensity in order to induce the correct correlations for the facility choice model.	Used as an input for the AI Ultrasound module.

3.0 Assumptions and limitations 

We assume that the prevalence of attending both first trimester and later pregnancy visits is the minimum of ANCfirst (as processed by the HS team) and ANC4 (GBD covariate also processed by HS team). There is non-zero prevalence of first trimester visits only when ANC4 > ANC1 (such as in Pakistan). We are likely overestimating the correlation between first trimester ANC and later pregnancy ANC (i.e., the prevalence of a first trimester ANC visit ONLY is likely non-zero despite this assertion in our model.) We assume that the DHS data used to produce the ANCfirst, ANC1, and ANC4 covariates applies to abortion/miscarriage/ectopic pregnancies as well as pregnancies resulting in live birth or stillbirth.

Todo

If we decide to improve the estimation of timing for ANC visits in our model (see this JIRA ticket) we need to update our documentation accordingly.

4.0 Verification and Validation Criteria 

Overall ANC attendance should match expected ANC1 values among the overall population as well as stratified by broad pregnancy outcome
Confirm no later pregnancy ANC attendance among abortion/miscarriage/ectopic pregnancies
Confirm first trimester ANC and later pregnancy ANC attendance rate among live birth and stillbirth pregnancies is equal to minimum of ANCfirst and ANC4
Confirm first trimester ONLY ANC attendance rate among live birth and stillbirth pregnancies is equal to ANCfirst - min(ANCfirst, ANC4)
Confirm later pregnancy ONLY ANC attendance rate among live birth and stillbirth pregnancies is equal to ANC1 - ANCfirst