Table 1.

Potential sources of bias in low birthweight data

	Likely effect on LBW prevalence estimates*
Coverage of weighing: bias in newborns weighed at birth
Many newborns in LMICs are not weighed at birth, especially if born at home. These are more likely to be socioeconomically disadvantaged and at higher risk of LBW.	↓
Extremely preterm or sick babies, those stillborn or dying soon after birth and those born around threshold of viability are the most likely to not be weighed. These babies are at high risk of being LBW.	↓
Coverage of data system: bias in newborns included in data source
Low coverage of administrative data systems in many LMICs (eg, lower coverage of birth registration for those who die shortly after birth, missing home births, and births in private facilities even if weighed). Births in private facilities are more likely to be socioeconomically advantaged and at lower biological risk of LBW; however, high prevalence of medical interventions (eg, caesarean sections both indicated and elective before 37 weeks, may increase risk of LBW).	↓ or ↑
Loss of birthweight data: biases in missing birthweight data†
In surveys, biases in card retention (eg, birthweight not available for babies who died who are more likely to have been LBW).	↓
Missing administrative birthweight data on sickest babies (frequently LBW) who are transferred immediately to (and weighed in) a newborn ward.	↓
Measurement errors: individual measurement or recording error
Heaping of recording of birthweight on 2500 g. As definition excludes babies with birthweight exactly 2500 g, those LBW newborns with birthweight near the threshold frequently heaped at 2500 g.	↓
Errors in birthweight measurement (eg, poorly calibrated scales, inappropriate devices), suboptimal weighing practices (eg, clothed or delayed weighing until days after birth).	↓ or ↑
Extremely preterm or sick babies and those born around threshold of viability who die soon after birth are more likely to be misclassified as stillbirth. These babies are at high risk of being LBW.	↓
Measurement units error
Confusion in surveys collecting data in both lbs and grams (eg, LBW baby weighing 4·0 lb recorded as 4·0 kg).	↓
Denominator calculation errors in LBW prevalence calculation
LBW prevalence calculated as: number with birthweight <2500 per all livebirths (whether weighed or not).	↓

LBW=low birthweight.

^*↓=the potential bias is likely to lead to a decreased LBW prevalence. ↑=the potential bias is likely to lead to an increased LBW prevalence.

^†For newborns who are both included in the data source and weighed at birth.