BMJ Global Health Outcome The primary outcome was the percentage of children under five (0–59 months) without birth registration (also referred to as non-­registration) as reported by the caregiver. This is the inverse of the MICS and DHS definition of birth registration and described children who did not have a birth certificate, whose births were not registered with the ‘civil authorities’ or whose caregivers did not know whether the child’s birth had been registered. The denominator was the number of children under five included in the nationally representative survey sample. We defined complete birth registration as non-­ registration less than or equal to 5%, indicating that most children in the country had their births registered. The DHS questions on birth registration were consistent over survey rounds. However, changes were made to the calculation of birth registration across MICS rounds: to allow for comparability, we recalculated birth registration estimates from MICS2 and MICS3 according to the indicator definition in MICS4, and the estimates Bhatia A, et al. BMJ Global Health 2019;4:e001926. doi:10.1136/bmjgh-2019-001926 presented here may differ from estimates included in the MICS2 and MICS3 national reports. Online supplementary table S1 provides a description of survey questions. Covariates We included three sociodemographic covariates for disaggregation and estimating inequalities. The groups we hypothesised to be the ‘best off’ and have the lowest percentage of children under five without birth registration were selected as the reference category for analysis. Covariates included: sex of the child (boys (referent (ref)), girls), residential location (urban (ref), rural) and wealth which was used in the analysis both as an ordinal variable (quintiles from poorest to richest (ref)) and as a binary variable (quintiles 1 and 2, quintiles 3–5 (ref)). Consistent with the DHS and MICS methodology, wealth quintiles were calculated by the survey programme based on a household asset index constructed using principal components analysis where wealth quintile 1, for example, represented the poorest 20% of the households.41 Because relevant assets may vary in urban and rural households, separate principal component analyses were carried out in each area and then combined into a single score using a scaling procedure to allow comparability. This score was then divided into quintiles.42 Statistical analyses Cross-sectional analyses For each survey, we calculated point estimates and 95% CIs for the percentage of children under five without birth registration on average and stratified by sex, residential location and wealth. We estimated the absolute difference in non-­ registration coverage among girls compared with boys and among children living in rural compared with urban areas. To estimate wealth inequalities, we calculated the slope index of inequality (SII), which accounts for the distribution of individual children across household wealth quintiles.43 Absolute measures of inequality are more easily interpretable and less sensitive to small differences than relative measures of inequality. We conducted tests of statistical significance to examine whether estimates were significantly different from zero, the null value, which represents no inequality. Changes in birth registration All analyses were country-­specific and we estimated change in non-­registration on average and stratified by covariates. For the primary analyses, we calculated the difference in non-­registration between the first available and most recent survey in each country and divided this by the number of years between surveys to estimate annual change. In addition, we estimated change by survey wave. We considered countries which maintained non-­ registration coverage below 5% between the first and most recent survey to have achieved complete registration. 3 BMJ Global Health: first published as 10.1136/bmjgh-2019-001926 on 16 December 2019. Downloaded from https://gh.bmj.com on 21 July 2025 by guest. Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies. expect to underestimate both the average percentage of children under five without birth registration as well as inequalities in registration. Because of our interest in changes in both birth registration coverage and inequalities, only countries with two or more surveys with data on birth registration were included in the analysis. Surveys had to include relevant covariates (household wealth quintile, urban/ rural location and sex of the child) and be designed to produce nationally representative estimates that could be compared over time: surveys were excluded if birth registration questions were posed differently across years or if national boundaries changed. We used publicly available data from the World Bank to create a list of regions and income-­groups by country.40 To select the final sample of surveys, we used information on the DHS and MICS websites and the dataset repository in the International Centre for Equity in Health at the Federal University of Pelotas Brazil to identify the 100 of the 218 World Bank economies with publicly available DHS and MICS surveys in September 2018. A total of 68 countries had more than one survey that met the inclusion criteria. To create our final sample, we grouped surveys into four waves based on time intervals, which corresponded to the years of MICS and DHS survey rounds: surveys conducted prior to 2004 (wave 1), between 2004 and 2008 (wave 2), between 2009 and 2012 (wave 3) and after 2013 (wave 4). The majority of countries (n=57) had one survey per wave. For a small sample of countries (n=11) where there were two surveys per wave, we retained one survey per wave ensuring we included the oldest and most recent survey to preserve the longest time interval and otherwise selected the survey with the larger sample size. One country was excluded from the final sample after applying survey waves as it had two surveys conducted between 2009 and 2012 (wave 3). Our final sample included 67 countries and 173 surveys.