UNICEF-WHO LBW and Preterm birth estimates (2020) analysis

This repository contains the code and input data required to generate the WHO/UNICEF LBW and Preterm birth 2020 estimates [1] [2] [3], and the regional and global LBW and Preterm outputs.

The background, methods and results for these analyses can be found in the corresponding papers published in The Lancet [1] [3].

Procedure for Generating LBW and Preterm Modelled Estimates

Contents

1. LBW Bayesian Modelling Code

• functions: folder containing all the functions needed for the modelling code
• inputs: folder containing all of the input data needed for the LBW estimates. This includes the LBW admin and survey data, and the WPP livebirths numbers. [4]
• models: folder containing ".txt" files containing the code for the JAGS model
• output: folder containing all outputs of the analysis.
• R code which contains the order at which to run the R scripts for the modelling ("0.master.R")
• R code which contains the file names for all the input data ("0.fileNames.R")
• R code which contains the packages needed for the code ("0.loadPackages.R")
• 1-6 numbered R scripts: contatining all of the commands needed set up the project, create the input database, run the model, produce the country, regional and global estimates and create country estimates.

2. Quick Access Inputs and Outputs - LBW

• LBWfinalInputDatabase: Final input database that was used in the LBW modelling, after inclusion and exclusion criteria were applied. This is included as ".csv" and ".rds" files. This is produced in the modelling code, and will be located in the output folder with the same name.

• LBWcovariates: These are the five covariate time-series (1995-2020) used for the LBW estimates: gross national income per person purchasing power parity (GNI) (constant 2017 international $), prevalence of underweight among female adults, adult female literacy rate, modern contraception prevalence rate and percentage urban population. These are included as ".csv" and ".rds" files.

• LBWcountryEstimates - pCCFullModel_16000_Estimates: This includes the LBW estimates for 2000-2020 for 195 countries and areas.

  • est, estL and estU: the point LBW rate estimates with 95% credible intervals.
  • predicted: 1 if there was no country data so the LBW rate is estimated using country intercepts and time trends from the region, and country-level covariates.
  • wpp_lb: the WPP 2022 [4] livebirths estimates.
  • estN, estNL and estNU: estimated number of LBW births with 95% credible interval, calculated by multiplying the point estimates by the wpp_lb.

• LBWregionalEstimates - pCCFullModel_16000_regionalAndGlobalEstimates: SDG revision 1 regional and global estimates.

  • est, estL and estU: the point LBW prevalence estimates with 95% credible intervals (multiply by 100 for rates).
  • predicted: 1 if there was no country data so the LBW rate is estimated using country intercepts and time trends from the region, and country-level covariates.
  • wpp_lb: the WPP 2022 [4] livebirths estimates country estimates summed to the regional level.
  • estN, estNL and estNU: estimated number of LBW births with 95% credible interval, calculated by multiplying the point estimates by the wpp_lb.

• LBW Other Regional and Global Estimates: ".csv" files containing LBW regional and global estimates as above, for 12 different regional groupings.

3. Preterm Bayesian Modelling Codes

• functions: folder containing all the functions needed for the modelling code
• inputs: folder containing all of the input data needed for the Preterm birth estimates. This includes the LBW admin and survey data, and the WPP livebirths numbers. [4]
• models: folder containing ".txt" files containing the code for the JAGS model
• output: folder containing all outputs of the analysis.
• R code which contains the order at which to run the R scripts for the modelling ("0.masterP.R")
• R code which contains the file names for all the input data ("0.fileNames.R")
• R code which contains the packages needed for the code ("0.loadPackages.R")
• 1-6 numbered R scripts: contatining all of the commands needed set up the project, create the input database, run the model, produce the country, regional and global estimates, and create country estimates.

4. Quick Access Inputs and Outputs - Preterm

• PretermfinalInputDatabase: Final input database that was used in the Preterm modelling, after inclusion and exclusion criteria were applied. This is included as ".csv" and ".rds" files. This is produced in the modelling code, and will be located in the output folder with the same name.

• Pretermcovariate LBWcountryEstimates - pCCFullModel_16000_Estimates: This is the same file that is produced in the LBW code for the 195 country estimates, which is used as the covariate in the preterm modelling.

• PTBcountryEstimates - pCCFullModelP_dataType1_16000_Estimates_1: This includes the Preterm estimates for 2010-2020 for 195 countries and areas. Variables as described in the equivalent LBW files.

• PTBregEst - pCCFullModelP_dataType1DQC1_16000_regionalAndGlobal: SDG revision 1 regional and global estimates. Variables as described in the equivalent LBW files.

• Preterm Other Regional and Global Estimates: ".csv" files containing Preterm regional and global estimates as above, for 12 different regional groupings.

Set-up

For the chosen estimate:

1. You will need to install JAGS on your machine. This is a separate program outside of R which is what the RJags package runs through. Go to http://www.sourceforge.net/projects/mcmc-jags/files and click the green button Download latest version.
2. Open the "0.loadPackages.R" file, and ensure all packages are installed.
3. Run the "0.loadPackages.R" file and "0.fileNames.R" file.
4. Open the "0.master.R" file and run the files in the indicated order. All of the outputs are written to the "outputs" folder.

Any questions on running the code please email Ellen Bradley (ellen_bradley@outlook.com).

Model

The models used in both estimates are heirarchical bayesian regression models, incorporating country-specific interepts, covariates, penalised splines to capture non-linear time trends and bias adjustments based on a data quality categorisation and other data quality indicators.

Acknowledgements

Code

We thank Ellen Bradley (@ellenBradley18) for the development and implementation of the two model codes. We thank Gerard Lopez (@LOPEZ-Gerard) for the testing and enchancement of this code into github.

Conceptualisation of the Models

We thank Ellen Bradley, Eric Ohuma, Alexandra Lewin, Yemi Okwaraji, Hannah Blencowe and Joy Lawn (LSHTM), Julia Krasevec, Joel Conkle, Samuel Chakwera, Jennifer Requejo and Chika Hayashi (UNICEF), Gretchen Stevens, Ann-Beth Moller, Jenny A Cresswell, Emily White Johansson, Elaine Borghi, and Allisyn Morran (WHO) and Laith Hussain-Alkhateeb (University of Gothenburg).

References

[1] Okwaraji Y, Krasevec J, Bradley E et al. National, regional and global estimates of low birthweight in 2020, with trends from 2000: a systematic analysis. Lancet (in press).

[2] United Nations Children’s Fund and the World Health Organization. UNICEF-WHO low birthweight estimates: Levels and trends 2000–2020. New York: UNICEF; 2023 Licence: CC BY-NC-SA 3.0 IGO.

[3] Ohuma E, Moller A-B, Bradley E et al. National, regional and global estimates of preterm birth in 2020, with trends from 2010: a systematic analysis. Lancet (in press).

[4] United Nations Department of Economic and Social Affairs Population Division. World Population Prospects 2022, Online Edition. 2022 [Available from: https://population.un.org/wpp/Download/Standard/Population/]