How useful are registered birth statistics for health and social policy? A global systematic assessment of the availability and quality of birth registration data

The registration and certification of births, while a near-universal practice in some parts of the world, is far less common in many low- and middle-income countries (LMICs) [1]. Birth registration has a wide array of individual and societal benefits [2], including the identification and facilitation of legal entitlements [1], citizenship and voting rights [3], social security benefits, social inclusion [4], access to health and education services [5], security benefits in times of crisis [6], and proof of age [3]. So fundamental is birth registration to legal identity that it has frequently been described as a basic human right [7‐9]. Additionally, reliable birth registration, compiled and consolidated within a national civil registration and vital statistics (CRVS) system, should be the primary data source for fertility statistics [10]. Such data are necessary to track (often rapid) changes in fertility levels and patterns, to monitor and evaluate family planning programs, to provide the denominator for an array of key maternal and child mortality indicators [11], to project future population size and structure [12], and to inform planning for future health, education, and other social services. Their fundamental and comprehensive importance for a nation’s health and social development underlies calls for universal birth registration, as reflected in the Sustainable Development Goal 16.9 that aims, by 2030, to provide legal identity for all, including birth registration [13].

As interest in universal birth registration continues to grow, it will become increasingly important for countries and development partners alike to understand the performance of birth registration systems and in particular, to have some objective basis to determine whether these systems are ‘fit-for purpose’, as described above. Yet, despite their fundamental importance, the global status of birth registration is not well understood. While multiple studies have described, assessed, and monitored the global landscape of death registration [14‐16], to our knowledge no comparable evaluations of birth registration systems exist. Some partial assessments to guide policy, however, have been undertaken. UNICEF, for example, has estimated that 71% of all children younger than 5 years have had their birth registered [17]. However, this estimate is based on self-reported survey responses that may be biased, especially as data from UNICEF show considerable discrepancies in some countries between reported birth registration and evidence of a birth certificate. For example, for only 10% of births in Rwanda that are reported to be “registered” can the family provide the birth certificate [18]. Moreover, the UNICEF approach does not include information on children who have died, especially neonatal deaths, for whom birth registration is often overlooked [10].

One reason why there has been no systematic assessment of birth registration data and systems could be the absence of a comprehensive and properly maintained and used global database. While agencies such as the World Health Organization annually aggregate and disseminate cause of death statistics based on death registration data from over 150 countries around the world, birth registration data are made public only through information provided by countries to the United Nations Statistical Division via an annual questionnaire, or through country-specific channels (e.g., national statistical offices) and other decentralized sources such as the Human Fertility Collection [19, 20].

To objectively assess the quality of birth registration data, and thus (indirectly) the performance of birth registration systems, it is first necessary to define the essential elements of data quality. While a fundamental measure of the quality of birth registration data is the completeness of registration, i.e., the percentage of all births that occur in a given year that are registered, there is other specific information about the newborn, the mother or the family that is, or should be, routinely collected for each birth and provided along with the birth certificate. Much of this information is likely to be of central importance for public health and demographic purposes, and hence reflects the utility of birth registration data. These characteristics include:

An objective, reliable, and descriptive low-cost approach to assessing the performance of birth registration systems would enable countries to monitor progress in developing their birth registration and reporting systems, by facilitating international goal-setting, facilitating monitoring of development goals, and assisting in the global efforts to improve birth registration that are already underway by identifying specific aspects of data quality or availability that require attention [23]. This paper advances efforts to improve the monitoring of global birth registration in a number of ways. First, we present the results of what we believe is the first systematic effort to assemble publicly available birth registration records for as many countries as possible into a global birth registration database, similar to what WHO maintains for death registration and causes of death. Second, we present a systematic assessment of birth registration data quality around the world. We do so by adapting an existing framework used to assess the quality and utility of death registration statistics, known as the Vital Statistics Performance Index (VSPI) [15], to the context of birth registration. We expect that the birth registration database, and our findings and framework for assessing its utility, will help enable the measurement and tracking of performance metrics, especially between countries, and thus will be of immediate use by both countries and development partners to facilitate monitoring of progress with global and national development goals.

We analyzed 2680 country-years of data, from 109 countries spanning 1980 to 2015. We found 51 countries with greater than or equal to 30 years of available data since 1980, 75 countries with greater than or equal to 20 years of available data, and 11 countries with less than or equal to five years. Available data came from 32 high-income countries, 29 countries from Eastern Europe or Central Asia, 20 countries from Latin America and the Caribbean, 12 countries from North Africa and the Middle East, 11 countries from East Asia and the Pacific, and four countries from Sub-Saharan Africa (Table 1). Notably, several populous countries (e.g., China, Bangladesh and Pakistan) did not have any birth registration data publicly-available for analysis.

The data we were able to gather represented approximately 27.9 million births per year on average, ranging from 16.8 million births recorded in 1981 to 55.3 million births recorded in 2011, for a total of 1.01 billion births registered since 1980. The available data represented only 20.8% of the estimated total number of birth worldwide. This figure varied from 13.2% in 1990 to 40.0% in 2011, the most recent year for which data was available for most of the reporting countries. In 2015, the most recent year for which data were available, global availability was estimated as 32.6%. The most notable change in global registration completeness occurred in 2011, when India began publicly reporting data. Figure 1 displays the global percentage of births registered based on publicly-available data over time.

Based on their most recent year with available data, 83 countries had estimated completeness greater than 90%, 19 countries had estimated completeness between 80 and 90%, five countries had estimated completeness between 50 and 80%, and two countries had estimated completeness below 50%. Completeness estimates for the most recent year for each country with available data are shown in Table 2. Additional file 2 displays the time series of completeness for each country.

Among the indicators of data quality, most country-years reported births by maternal age and the newborn’s sex (95.3 and 73.4% of country-years respectively), when data were available. Fewer countries reported births by live birth order and birthweight, with 55.6 and 51.1%, respectively, of country years containing these indicators. Among countries which did report each indicator, some missing values were observed as well. The indicator with the highest proportion missing was birth weight, with 2.6% of births with unknown birth weight. Maternal age and live birth order had fewer missing values: 1.0% each. Births without a recorded sex were very rare, occurring in only 0.05% of cases. Additional file 2 displays the level of each indicator over time by country.

Combining completeness, quality, and timeliness, Fig. 2 displays the VSPI-B scores for each country for their most recent year with available data. 26 countries had VSPI-B scores in the highest category, between 0.9 and 1. These countries include many high- and middle-income countries with high completeness, and are generally countries which report births by all four data quality indicators. 17 countries had VSPI-B scores in the range 0.8–0.9. These countries also typically included mostly high- and middle-income countries, and were characterized by high completeness but sporadic reporting of the four data quality indicators. 38 countries were in the range 0.6–0.8. Spanning high-, middle-, and lower-middle income countries, these countries’ VSPI-B scores were driven by a mixture of lower completeness and lack of reporting of one or more data quality indicator. 19 countries had VSPI-B scores in the 0.3–0.6 range, characterized by either lower completeness, erratic availability of data, and/or lack of reporting on multiple data quality indicators (i.e., only reporting births by mother’s age or newborn’s sex, but not the others). Finally, nine countries had VSPI-B scores that were lower than 0.3. These countries typically had only few years with available data, low completeness, and/or lack of reporting of multiple indicators of data quality. Additional file 2 displays the time series of VSPI-B scores for each country.

The results from the simulation indicate that, all else being equal, the completeness indicator has the highest weight in the VSPI-B. This is evidenced by Additional file 3, which displays the simulated accuracy associated with each indicator at varying levels among simulated samples. At high levels, all five indicators have generally similar accuracy (i.e., similar influence on the VSPI-B score), but at lower levels the indicators have quite different values. This is the result of different empirical simulation probabilities to inform the weighted samples.

This paper presents, for the first time to our knowledge, a systematic assessment of the availability and quality of data reported by birth registration systems worldwide. As Mikkelsen et al. 2015 [16] argue, vital registration data quality can be assumed to be an accurate reflection of the performance of the registration system itself. In assessing birth registration data quality, we demonstrate each country’s progress toward strengthening birth registration through an adaptation of the Vital Statistics Performance Index. We also present estimates of the country-level completeness of birth registration based on available data. This assessment is based on the largest database of its kind, containing records of over one billion births since 1980 by country, year, maternal age, sex, birthweight, and live birth order.

Although over 100 countries had at least one data point, the availability of data remains low in many parts of the world. Our assessment of birth registration availability and completeness, where available, demonstrates that sharing of birth registration data is surprisingly low compared with death registration, although it does appear to be increasing. Only around 33% of births worldwide in 2015 were registered with the aggregate records made publicly available. This (and even the 2011 peak of 40% availability) is considerably less than for deaths, where 55–60% of global deaths are now registered in publicly available data systems. Further, we found considerable variance between birth reporting systems in that some countries report births by maternal age, sex, birthweight, and live birth order, while others exclude some or all of this information.

This description of birth registration completeness in country-years where it is possible to assess is in stark contrast to other assessments, particularly UNICEF’s State of the World’s Children report [17]. In their most recent such report, completeness estimates are much higher than those presented here, even in country-years with data available to assess. For example, the available data from the Philippines in 2015 represent only 74% of the estimated births according to our assessment, but the UNICEF report estimates 90% completeness. Notable examples of large discrepancies in other parts of the world include Peru (85% completeness according to 2015 data, as compared with 98% according to UNICEF), and Serbia (73% as compared to 99%). Other countries, such as South Africa, are closer, but still different (83% as compared to 85%). The reasons for the discrepancies are likely twofold. First, the alternative estimates of global birth registration completeness are based on self-report in surveys, not actual records of birth certificates. Issues of recall bias, survival bias, and survey instruments that do not confirm the actual existence of the birth certificates, are likely to lead to over-estimates of completeness via this method. Second, the estimates of completeness we present reflect the accuracy of the estimated denominator data as much as they do the completeness of systems. As already noted, the model estimates of total births include uncertainty intervals within which the total births may fall. While it would have been ideal to propagate that uncertainty into our estimates of completeness, uncertainty estimates were not available to do so at the time of analysis.

Altogether, these findings imply that approximately 74 million births (53.1% of annual global births) per year occur in countries whose systems do not systematically register them and release the records publicly. Conversely, only about 12 million births per year (8.6%) occur in countries with high-performing registration systems, i.e., those with consistent data availability, high completeness and reporting by maternal age, sex, birthweight and live birth order.

The assessment of birth registration is not without limitations however. Primarily, these numbers are based on available data only. This caution is especially salient in that it renders estimates of global registration completeness impossible, as noted above. There are likely more births registered per year that do not get aggregated and reported in order for us to assess them. As such, these availability numbers should be considered as the minimum completeness, and are most useful in countries where data are public. Evidence from China, for example, suggests that about 10 million births per year are registered in the country, which would increase global birth registration completeness to close to 50% were they to be made available for analyses such as that reported here. The assessment of completeness where data are available may also be limited by the assumption that all registered births are reported when a public release is made. The assessment of data quality is also limited to the data that are available. Many countries may have low VSPI-B scores not because their registration systems are functioning poorly, but because the data aren’t released. That includes missing years, but also failure to report certain variables. For example, it is rare to fail to record the sex of a child on their birth certificates, but many countries have not made such information publicly available. Without further data to inform our assessment, it is impossible to distinguish the reasons for lack of reporting. Additionally, some of the details of the VSPI-B simulations are subject to limitations. Principal among them is the fact that the simulations and estimates of disaggregated birth counts are based, in part, on Demographic and Health Surveys and the Understanding Society Survey. With more data, these estimates may have been more accurate. Finally, it could be argued there are other means of measuring the quality of birth registration data; for example, the percentage of births that are registered late or with unspecified type of site of occurrence (e.g., hospital, home etc.). However, given the largest available source of data, the UN database, did not collect these data, we were not able to include them in our analysis.

Our findings have a number of important implications and uses. First, we highlight the gaps in available data. While national policymakers may have unpublished data at their disposal, international and multinational health and development organizations are often reliant on public information of registered births, which we demonstrate are unavailable in many country-years. These findings underscore the significance of open data practices for public policy.

Second, we present an objective and low-cost approach to assess the performance of birth registration systems, wherever data are available. This can be helpful to monitor country progress and benchmark efforts to improve birth registration against national and international goals, especially in an era with significant multilateral, bilateral, and philanthropic investments in strengthening CRVS systems [23]. In addition, we present a set of metrics for completeness and overall system performance that is consistent between countries and over time. As such, these results may be useful for international goal-setting.

An important outcome of this work should be to highlight both the importance of birth registration as a source of fertility statistics and the limitations of the available data. National and subnational governments require routine and timely birth registration data for a range of purposes, not least of which is to lessen reliance on costly sample surveys such as the Demographic and Health Surveys that produce fertility statistics with considerable uncertainty in small areas and which can be 2–5 years out of date once available. More generally, the data generated by civil registration systems are of paramount importance to global health and development efforts, as well as for critical epidemiologic and demographic research [10, 16, 29]. Some authors have even argued that the heightened ability to design and implement effective health policy afforded by greater civil registration has led to a measurable relationship with population health outcomes [30, 31]. This will hopefully encourage stakeholders to collect, consolidate, use, and release more data, release data more promptly, and ensure they maintain a centralized, standardized system for aggregating birth registration data. Considering that birth registration is seen as a fundamental human right, and given the enormous policy relevance of timely, accurate, and complete information on fertility patterns, our findings should be taken as an urgent call for immediate, coordinated, and sustained support to countries to strengthen birth as well as death registration systems, and for greater global efforts to register births and incorporate the minimum demographic and health indicators associated with each of them.