Calculating Low Birth Weight from DHS Can Mothers Help Improve Estimation?Amos Channon, Mac McDonald, Sabu PadmadasUniversity of Southampton
OutlineWhy is birth weight important?Data used in the studyCalculating the proportion with LBWProblems with the estimatesUsing mothers to improve estimationProblems with the methodHow is birth weight recalled?Conclusions & future direction
OverviewUnited Nations targeted a fall of one-third in the proportion of children with Low Birth Weight (LBW) by 2010Important to know the current proportions in order to measure fall in LBW accuratelyProblems with measuring birth weight in developing countriesAre there adjustments that can be made?
Importance of Birth WeightGreatest predictor of infant deathCan be used both as an outcome or as a predictorMany models regarding mortality at younger ages include birth weight as a proxy for health at birthLonger term health problems
Causes of Differences in WeightNatural variabilityInter-uterine growth retardationPrematurityDistinction between the two types in developing countries?Proximate causes are medical/biologicalRelated to many social and demographic factors
Low Birth WeightWhat is Low Birth Weight?WHO Definition: Less than 2500g Not an absolute marker for increased riskRisk of infants dying who weigh 2450g and 2550g is similarContinuous scale better, but issues with measurement and targeting of at-risk infants
Research QuestionsHow accurate are the estimates of LBW in different countries?Can alternative methods be used to calculate proportion with LBW to improve accuracy?
DataInitially 15 DHS surveys used Used DHS+ conducted since 1997Attempted to get some geographical groupingFinal analysis contained 13 countriesProblems with the data in 2 countries (Haiti and Peru)
Countries in the Analysis
Calculating LBWOnly single births studiedBirths within the five years before surveyExcluded births weighing over 6kg
Reported weights taken as correctPercentage LBW depends on how those weighing 2500g are treated
New Estimates of % LBW
% with LBWGreat variability in estimates depending on how those at 2500g are treated% heaped at 2500g ranges from 0.4% in Nicaragua to 18.7% in India
Does this represent the full story?What percentage of the sample gave a birth weight?
How much Birth Weight is Missing?
Does the Missing Data Matter?Can the infants with reported birth weights be used to calculate proportion with LBW?
Do those with a birth weight differ from those without a birth weight?
Logistic regression conducted on likelihood of birth weight being missing1 Available0 Missing
ResultsSignificant in nearly all countriesPlace of Delivery, Survival Status; Paternal EducationSignificant in most countriesPrenatal Care, Urban/Rural, Maternal EducationSignificant in few or in no countriesGender, Marital Status, Religion
Using Mothers PerceptionMost DHS+ surveys included the question:
When (NAME) was born, was he/she: very large, larger than average, average, smaller than average or very small?
Can the responses to this question be used to improve estimation?
Missing Data in Mothers PerceptionAmount of Missing Data is low:0.1% in Vietnam and Tanzania3.5% in MaliMost countries under 1% missing
How accurate are these reports?Do they agree with the reported birth weights?
Mean Weight in Perception Categories
- Mean WeightClear that the mean weight is closely aligned to perceptionAll countries mean weights in V Small category
Perception for those With and Without Birth WeightDifference in the distribution of perception between those who report a birth weight and those who do notDistribution is shifted towards the smaller categories in those without a birth weight To be expected as this group have attributes which make smaller infants more likely
Distribution of PerceptionCambodiaMali
Method to Combine Perception and WeightUsing those with weights, calculate the proportion of LBW infants in each perception group for each countryApply these proportions to those without a birth weightCombine the % with LBW for those with a birth weight with the estimated % with LBW for those without a birth weight
Example - Malawi% LBW from Birth weight only = 9.7%Proportion of LBW by perception categories:Very Small 49.8%Small 41.0%Average 6.8%Large 2.8%Very Large 2.7%
Example - MalawiDistribution of Mothers Perception in those without birth weight:
Very Small 4.2%Small 15.2%Average 58.7%Large 14.3% Very Large 7.6%
Example - MalawiApply LBW percentages to the corresponding categories % LBW for those without birth weight = 12.9%% LBW for those with a birth weight = 9.7%Combined % with LBW = 11.5%
Combined Estimated % LBWAll countries estimates of % LBW riseAmount of increase depends on how those weighing 2500g are treatedTreated half those with weight of 2500g as LBW
Estimates of % LBW
Problems with MethodAssumes that the relationship between perception and birth weight is the same for those with and without a recorded birth weightAssumes births with a birth weight are as likely to be LBW as those without a birth weight
Problems with MethodAssumes that the reported birth weight is correct:A few infants reported as being over 6kg were excluded (13 pounds)What are the mothers judging the size against?Large differences in the distribution of birth weight between weights recalled from memory and those recalled from a birth card
Recall MethodWeights recalled from memory are:Greatly heapedShow greater variabilityLess reliable?More likely to be LBW
Using memory recalled weights as reference further increases % with LBW
5101 - 52005101 - 5200
5501 - 56005501 - 5600
5601 - 57005601 - 5700
5801 - 59005801 - 5900
WeightMemoryCardTotalMemory RecallCard Recall
WeightMemoryCardTotalMemory RecallCard Recall
Memory vs. Card RecallLarge differences in distribution between recall methodsCard recall appears more normalBut still is heaped in some countriesMemory recall can be good Proportion with LBW should be calculated with reference to recall method
ConclusionsCalculating %LBW from available weights underestimates true proportionMothers perception of the size of the baby generally agrees with recorded birth weightUse of the mothers perception to estimate %LBW is a useful tool to obtain more accurate estimates
Conclusions (2)Need to be awar