author	authors	date	title	toc-title
Fanny Sandalinas	Fanny Sandalinas	2024-10-07	Quarto documentation	Table of contents

Preface

About this manual This manual is designed for anyone involved in the analysis of micronutrient biomarkers datasets. The methods were developed by the Micronutrient Action Policy Support (MAPS) project and made available to all for use, without warranty or liability. The MAPS project is funded by the Bill and Melinda Gates Foundation.

Authors

Fanny Sandalinas, Tineka Blake

Acknowledgements

Christopher Chagumaira, Andrew Bean, Edward Joy, Louise Ander

Who is this manual for

The goal of this documentation is to help nutritionists and data scientists to analyse datasets more quickly while remaining in full control of the decisions being made, with clear notation of the decisions made at every step. It will help them to better understand the data and obtain accurate estimates of micronutrient status, in a harmonized, reproducible and transparent way.

Introduction

Welcome to the training manual on using R and Rstudio for the analysis of micronutrient biomarker data. The parts on software requirements, data types, data structures, packages and functions can be learnt through this documentation: R for Household Consumption and Expenditure Surveys, as these parts cover the bases of R.

Summary stat function

The aim of the function is to create the summary statistics that are needed when analysing micronutrient biomarker data. That includes estimates of deficiency based on cut-offs, with disaggregation by socio-demographic characteristics, mean concentration of the micronutrient biomarkers, and histogram of distribution.

The function can produce summary statistics for serum/plasma: ferritin, soluble transferrin receptors (sTfR), hemoglobin, Retinol Binding Protein (RBP), iodine, retinol, zinc, vitamin B9, vitamin B12.

The SummaryStats function has been broken down into a series of smaller sub-functions. Each sub-function performs a specific task, i.e. preprocessing, applying a specific adjustment, etc. This has been implemented to make the large body of code more digestable, interpretable and resuable over different source files.

The SummaryStats function is therefore a series of sub-function calls that can be executed with the default parameters with the following line: SummaryStats(theData = theData, biomarkerField = biomarkerField, aggregationField = "region", group_id = "WRA", thresholds = thresholds)

Loading the function

Change working directory to this (source) file's directory

(Session > Set Working Directory > To Source File Location)

source("../../R/SummaryStats.R")

You will also need to download a few files that are all available in the same folder. You do not need to do anything with these files, just make sure they are in the same folder than your data. More explanation on these files can be found here: SummaryStats-requirements.

Place data files in the same location as this script.

• The data (S3-biomarker-group_id-variable_-theData.rda) The data must be in a .rda file. You can use a function to transform your data in the right format into .rda.

• Threshold files ('biomarker-group_id-thresholds) They are available for every micronutrient by age group. The right threshold file will be loaded once the specifications of the micronutrient and age group are entered in the summary stat function parameters.

• Text file ('S3-biomarker-group_id-variable-source.txt) The text file calls the function with the right parameters.

download and load a few packages that are needed

rm(list = ls()) 
library(srvyr) 
library(jtools) 
library(survey) 
library(dplyr) 
library(BRINDA)

Preparation of the data

To be able to run the function on your data, make sure your data follow exactly the same - name - units - format than indicated in this: Sample File WRA. The variable names are case-sensitive.

Some of the variables are mandatory: Biomarker data: at least one biomarker variable should be provided If zinc data are provided, then fasting information, timing of sample collection and age group for school-age children should be provided to apply the right cutoff to the values. If vitamin B12 data are provided, exact age of the child in month should be provided so that the right cut-off is used.

Some others are optional: CRP and AGP: these variables are inflammatory markers and are used to adjust data for inflammation. if they are not present, the BRINDA package will not be applied (see later in 2.3.2) and the summary statistics will be unadjusted. Weight: if weights were applied during the survey, the original weights must be provided so that the design of the survey is taken into account in summary estimates. Survey cluster and survey strata: if the survey was using a complex survey design, then providing the survey cluster and the survey strata will allow the use of a survey package (see 2.3.3) to take into account survey design characteristics in the analysis of data. Aggregation variables. The function performs some basic level of aggregation: • Rural/urban • Wealth quintiles • Age group

Pre-processing of data

Outliers should be removed manually. We recommend to remove values (changing to NA) following these guidance: Zinc Data Remove values of serum/plasma zinc>161 μg/dL

Haemoglobin haemoglobin<4 g/dL haemoglobin>18 g/dL

Other biomarkers

There are no publications related to implausible values for the other biomarkers and therefore the approach should be conservative, but removing values that are likely to be resulting from errors in data entry or biologically implausible.

## Run SummaryStats function over one dataset and micronutrient


prefix <- "S3-"
biomarkerField <- "haemoglobin"
group_id <- "WRA"

The function can work on several population groups of interest: Pre-school children, school-age children, women of reproductive age and men.

Pregnant women should be removed from the dataset.

It is important to chose the right population of interest as specific cutoff are and the use of inflammation adjustment also depends on the population group that has been chosen.

knitr::opts_chunk$set(echo = TRUE)
aggregationField <- "region"

Currently, the function can disaggregate results by region, wealth quintiles and residence (urban or rural)

RunSurveyWeights <- TRUE

The R survey package is used in the function to be able to take survey characteristics into account. This is particularly important for the estimation of the standard deviation as it is assumed that there is less variations between individuals of the same cluster, and the same strata than in the entire sample. If these characteristics are not entered, then the results will not be representative of the total sample.

Brinda <- TRUE

The inflammation adjustment can be performed using AGP or CRP only. For more information related to the inflammation adjustment using the BRINDA package, refer to Luo et al.

HaemAltAdjust <- TRUE

Heamoglobin increases with altitude and therefore haemoglobin values should be adjusted for altitude so that they can be compared to sea-level values. The default argument is set to TRUE, if you would not like to apply this adjustment set HaemAltAdjust = FALSE The adjustement follows the equation described in Sharma et al.

HaemSmokeAdjust <- TRUE

Smoking increases haemoglobin values and therefore haemoglobin values should be adjusted for smoking status, so they can be compared to non-smoker values. This function assumes the amount of cigarettes per day is not known and assesses only if the individual is a smoker or not. The default argument is set to TRUE, if you would not like to apply this adjustment setHaemSmokeAdjust = FALSE. The smoking adjustment follows the equation described in Sharma et al.

Load data file

load(file=paste0(prefix, biomarkerField, "-", group_id, "-", aggregationField, "-theData.rda"))

Load thresholds file

load(file=paste0(biomarkerField, "-", group_id, "-thresholds.rda"))

Make sure that the arguments of the function equal the same as the loaded data and thresholds file

source("../../R/SummaryStats.R")
output <- SummaryStats(theData = theData,
                       biomarkerField = "haemoglobin",
                       aggregationField = "region",
                       group_id = "WRA",
                       thresholds = thresholds,
                       RunSurveyWeights = TRUE,
                       Brinda = TRUE,
                       HaemAltAdjust = TRUE,
                       HaemSmokeAdjust = TRUE,
                       ZincCutoff = TRUE)