Estimating completeness of birth registration in South Africa, 1996 – 2011

Abstract Objective To estimate the completeness of live birth registration through South Africa’s civil registration and vital statistics system between 1996 and 2011. Methods The number of births registered by the civil registration and vital statistics system was compared with independent estimates of the true number of births derived using: (i) the reverse survival method applied to 2011 census data; (ii) the application of estimated age-specific fertility rates to population estimates from censuses and surveys; and (iii) data from the public-sector district health information system. Findings In 1996, an estimated 25% of births were registered within the calendar year of birth and 33% were registered before the end of the subsequent calendar year. By 2008, 76% of registrations occurred within the calendar year of birth, 84% occurred by the end of the following year and 90% occurred before the child’s fifth birthday. These improvements were seen in all provinces and differences in completeness between provinces narrowed markedly. Improvements in the completeness of registration coincided with government efforts to strengthen the system, new legislation on vital registration and the introduction of child support grants, which required birth certificates. Interprovincial migration of children influenced the completeness of registration in affected provinces. There was some terminological confusion among government agencies on defining the timeliness of registration and the year of birth. Conclusion The completeness of birth registration in South Africa increased rapidly between 1996 and 2004. To allow international comparison, the method for measuring the completeness of birth registration needs to be standardized.


Introduction
Actions taken globally to achieve the millennium development goals highlighted the failure of civil registration and vital statistics systems in some low-and middle-income countries to record vital events accurately and to provide data essential for monitoring key child health indicators. The United Nations Secretary-General's Global Strategy for Women's and Children's Health 2016-2030 further emphasizes the importance of these systems for monitoring maternal and child health. 1 Moreover, the sustainable development goals (SDGs) include the target of achieving universal birth registration by 2030. This target stems from the overarching objective of improving data systems for monitoring health-related indicators and reflects a desire to improve equity among population groups. 2 Birth registration, which is enshrined in the United Nations Convention on the Rights of the Child, 3 is fundamental to the legal recognition of human beings and, consequently, to their ability to secure a name and nationality. Statistically, it is defined as the continuous, permanent and universal recording of the occurrence and characteristics of births. 4 Birth registration is a pivotal component of civil registration and vital statistics, and is key along with death registration, to monitoring trends in child mortality. Well-maintained registration systems provide essential data for assessing priority areas of population health. Less recognized are the links between birth registration and social development in modern societies, 5,6 particularly in addressing poverty. In 2013, the United Nations Children's Fund (UNICEF) estimated that globally the births of only 65% of children younger than 5 years were registered. 4 To draw attention to the importance of birth registration, UNICEF produced two publications on inequities and trends in birth registration and a handbook for those working on birth registration. 4,7 In November 2016, the United Nations Population Division held an expert group meeting on evaluating the completeness and quality of vital statistics data. 8 In the publication Every child's birth right, 4 which detailed national and regional trends in birth registration, UNICEF used data in an annual report from South Africa's national statistical office to conclude that in 2011, the country achieved 95% completeness of birth registration within the first year of life. However, this estimate was derived by dividing the number of birth occurrences in 2012 by registrations in 2012, effectively it assessed the timing of registration rather than its completeness. Alternative estimates of completeness, based on different methods and data sources, are available for South Africa. According to national household surveys in 2008 and 2011, 11% of children younger than 3 years of age did not have a birth certificate. 9,10 However, proof of registration was not sought in either survey. The Department of Home Affairs, which monitors babies registered within the year of birth, reported that the proportion was 90% in the financial year 2011 to 2012. 11 Although the proportion is similar to that found in household surveys, like UNICEF's estimate, it is a measure of registration timing rather than completeness.
Evidently, a standard measure or method for monitoring the completeness of birth registration is lacking. The World Health Organization defines completeness as "a measure of the extent to which births and deaths in a country in a given year are registered by the civil registration system." 12 Any measure of completeness therefore requires an independent estimate of the number of births or deaths. One concern is the need for a clear definition of the time frame, or the cut-off Research Birth registration completeness in South Africa Nadine Nannan et al.
date, for registration. Current measures consider completeness within either 1, 3 or 5 years of birth.
To date, no estimate of the completeness of birth registration in South Africa has been based on estimates of the true number of births. The aims of this study therefore were to assess the completeness of birth registration in South Africa's nine provinces between 1996 and 2011, with particular reference to the SDG's focus on monitoring and reducing health inequities, and to propose a method for annual assessment.

Methods
South Africa, an upper-middle-income country with a population of over 56 million individuals in 2016, has a well-established civil registration and vital statistics system, but still faces challenges in using the data to track births and deaths, and particularly, to estimate child mortality. 13,14 The registration of births and deaths is governed by the Births and Deaths Registration Act of 1992, 15 which has been amended several times, probably the most important amendment was the introduction of new birth and death notification forms in 1998. 16 One crucial initiative for improving timely birth registration was providing new mothers with information about the process during antenatal care visits. 16 In addition, birth notification forms were made available at state facilities at the time of delivery. In South Africa, births are registered by the Department of Home Affairs using information provided on the birth notification forms. The information is captured on the National Population Register (i.e. civil registration) and these data are submitted to the national statistical office (Statistics South Africa), which compiles annual reports on birth statistics. 17 We used unit record data to create tables of births by year of occurrence and year of registration. The statistical office defines late registration as registration after the calendar year of birth (Statistics South Africa, unpublished data, 2015), with the calendar year of registration closing at the end of February of the following year. We regarded births registered before the end of February of the calendar year following the calendar year of birth as being registered in the year of birth and births registered in the following

Research
Birth registration completeness in South Africa Nadine Nannan et al.
calendar year as being registered in the year of birth + 1, and so on.

Estimating the number of births
First, estimates of the true number of births between 1996 and 2010 were derived using the reverse survival method from the enumerated South African-born population recorded in the 2011 census, allowing for interprovincial migration between 2001 and 2011. 18 Specifically, data on children, by age and province of birth, were used. The method followed age cohorts in the recorded population back to their year of birth by applying appropriate age, sex and time period-specific probabilities of surviving from birth to age x. Hence, the number of children aged 0 to 15 years recorded in the 2011 census were projected backward to estimate the number of births for each age cohort using the where L x is the number of person-years lived between age and age x + 1 (x can vary from 0 to 15) recorded in the cohort life table, of which l 0 is the radix (i.e. the number of births in the age cohort). Estimating the number of births in any year involved the following stages: (i) age cohort survival probabilities to 2011, which were applicable to births over the past 15 years, were derived using the Actuarial Society of South Africa's 2008 population projection model with the migration assumptions set to zero; 19 (ii) each survival factor was derived by dividing the number of individuals aged x last birthday in 2011 by the number of births these individuals originated from in the year in which they were born; (iii) each survival factor was then used to estimate the number of births in year y (B y ) by dividing the number in a particular age cohort recorded in the 2011 census (in the country or province) by the appropriate survival factor. Second, an alternative estimate of the number of births against which the completeness of birth registration could be evaluated was derived using estimated total fertility rates. We based these rates on national and provincial fertility research in South Africa between 1996 and 2011. 20 The researchers initially derived annual estimates of age-specific and total fertility from summary birth history data from the 1996, 2001 and 2011 censuses, as well as from a 2007 community survey and from full birth history data from the 1998 Demographic and Health Survey. We assumed that fertility rates changed linearly between data collection points. We then obtained estimates of total fertility by calculating the number of births occurring in any year by applying age-specific fertility rates to annual estimates of the population of women by age. We used age-specific fertility rates to derive the number of births occurring in different years for comparison with the number derived by projecting backward from the population recorded in the 2011 census.  Third, we estimated the number of births between 2004 and 2012 using routine health statistics collected by the District Health Information System from public sector health facilities in all nine provinces. In 1994, the introduction of free health care for mothers and children up to 6 years of age led to a substantial increase in the number of births taking place in the public sector. 21 Nevertheless, District Health Information System data had to be adjusted for births that occurred in private facilities or at home. A review of the 1998 and 2003 South African Demographic and Health Surveys and of the 2012 South African National HIV Prevalence, Incidence and Behaviour Survey revealed that insufficient data were available to provide robust estimates of the proportion of births that occurred either in the private sector or at home. [22][23][24] However, they did show that births in private facilities were more common in richer, more urban provinces (e.g. Gauteng and the Western Cape) than in poorer, more rural provinces (e.g. the Eastern Cape, Mpumalanga and Limpopo), where the proportion of home births was greater. We used the proportion of children younger than 1 year of age who were covered by medical aid schemes or health insurance, as reported in General Household Surveys, 25 as a proxy for the proportion of births in the private sector: the average was 11.9% between 2004 and 2012. We estimated the annual proportion of home births in each province from the proportion of rural residents because we found a strong correlation (i.e. R 2 = 0.74) between the proportion of rural residents reported in the 1996 census in a province and the proportion of home births recorded in the 1998 Demographic and Health Survey in that province. We used the following regression equation to estimate the percentage of home births in province i (hb i ) from the percentage of home births recorded nationally in 2010 (HB): where r i is the percentage of the population in province i that was rural, R is the percentage of the national population that was rural in 2010, and a and b are regression coefficients. As far as possible, we followed the Guidelines for Accurate and Transparent Health Estimates Reporting. 26  Census  60  64  69  75  80  83  80  81  80  82  ND  DHIS  ND  ND  ND  122  121  120  115  111  104  102  94  Free State  Census  66  74  79  80  83  86  82  86  83  86  ND  DHIS  ND  ND  ND  101  103  105  105  107  107  112  112  Gauteng  Census  84  87  85  90  89  85  83  82  79  76  ND  DHIS  ND  ND  ND  96  97  98  97  97  93  98  95  KwaZulu-Natal  Census  67  81  77  76  80  87  84  86  82  84  ND  DHIS  ND  ND  ND  100  105  110  110  108  107  109  108  Limpopo  Census  65  67  70  74  77  77  78  82  81  82  ND  DHIS  ND  ND  ND  98  97  96  96  96  95  99  99  Mpumalanga  Census  57  67  76  83  83  85  86  88  84  86  ND  DHIS  ND  ND  ND  113  110  110  111  110  110  113  113  Northern Cape  Census  72  76  79  84  88  89  90  93  90  91  ND  DHIS  ND  ND  ND  105  107  108  111  112  113  113  114  North West  Census  78  83  90  93  101  99  100  103  100  97  ND  DHIS  ND  ND  ND  126  130  132  134  132  132  137  134  Western Cape  Census  85  90  87  95  97  98  92  94  89  87  ND  DHIS  ND  ND  ND  104  111  112  111  110  111  115  114  Nationally  Census  73  79  79  82  86  86  84  86  84  83  ND  DHIS  ND  ND  ND  105  107  105  103  102  99  101

Research
Birth registration completeness in South Africa Nadine Nannan et al.
count, which included people not born in South Africa (data available from the corresponding author). Fig. 2 also shows point estimates of the annual number of births derived using fertility rates reported by women in censuses and surveys (   Table 2 shows the completeness of birth registration before the end of the year after the calendar year of birth in individual provinces derived using both the reverse survival method and routine health statistics adjusted for births in private health facilities and at home. Overall completeness improved considerably between 2001 and 2011, and differences between different geographical settings narrowed over time. Surprising findings were that completeness was: (i) greater than the national average in the Free State and the North West; (ii) lowest in Gauteng; and (iii) greater than 100% in the North West in some years, when calculated using the reverse survival method. These anomalies may have occurred because children migrated early in life from their province of birth to the province in which their birth was registered. For example, differences in the estimated and reported number of births between the Eastern Cape and the Western Cape suggest that migration between these provinces may have occurred. The estimate for the Western Cape derived using adjusted routine health statistics was higher than that derived using the reverse survival method (Fig. 4), whereas the reverse was observed in the Eastern Cape (Fig. 5). Some mothers from the Eastern Cape may have travelled to the Western Cape to deliver their babies.

Discussion
Our study illustrates the usefulness of a robust method for assessing the completeness of birth registration that benefits from considering multiple data sets, particularly less frequently used data sets. We found that the completeness of birth registration in South Africa increased rapidly between 1995 and 2004 and that by 2011, 83% of registrations occurred before the end of the year after the calendar year of birth. This improvement was seen in all provinces and the relative difference between provinces narrowed markedly. The increase in birth registration observed after 1995 coincided with the introduction of amendments to the 1992 Birth and Deaths Registration Act that extended it to include former homelands and with efforts to strengthen vital registration. In addition, the introduction of child support grants in 1998 also played a role because the primary eligibility requirement was that the child's birth had to be registered. 29 Our estimate of the completeness of birth registration differed somewhat from UNICEF's finding in 2011 that 95% of births in South Africa were registered within the first year of life. 4 The main reason for this difference was the method used to assess completeness. We compared the number Birth registration completeness in South Africa Nadine Nannan et al.
of births registered by a given age with an independent estimate of the total number of births in the relevant year. In contrast, UNICEF looked at the proportion of registered births that were registered within the first year of life. In 2013, UNICEF reported a variation globally: in some countries, measures of completeness were based on surveys, whereas in other countries they were based on the analysis of vital registration data. The method used to measure the completeness of birth registration needs to be standardized and international comparisons should use consistent and well defined assessment criteria and methods of evaluation.

Limitations
The accuracy of the figures we derived for the estimated numbers of births using the reverse survival method was highly dependent on the accuracy of the enumeration of children in the 2011 census. We believe the 2011 census count was accurate because the figures were comparable with subsequent assessments of the number of older children in school and with the number of survivors of children enumerated in previous censuses. 20 Furthermore, the numbers of births calculated by the reverse survival method using 2011 census data corresponded to the numbers estimated using fertility rates and adjusted routine health statistics.
Another limitation is that our analysis assessed the completeness of registration according to the number of complete years after the calendar year of birth because the exact date of registration was not available as part of the unit record data. Consequently, estimates of completeness according to the exact age of the child, which are arguably more desirable, were approxi-mate. Second, the migration of young children after birth made it difficult to interpret the completeness of birth registration in individual provinces. Nevertheless, completeness improved in all provinces and we could use data on births collected by the District Health Information System to monitor registration. Finally, because of the complexity of using multiple data sources to arrive at our estimates, we did not quantify uncertainty in completeness estimates or undertake a sensitivity analysis.
Despite the concerted effort of the South African government to improve the civil registration and vital statistics system, there remain obstacles that prevent approximately 20% of births being registered by parents within the year of birth. Insight into what these obstacles are and which communities are most affected are needed. In addition, our assessment of the completeness of birth registration identified some terminological confusion among government agencies on definitions of the timeliness of registration and the year of birth. Although any definition of the completeness of birth registration should compare the number of registered births with the actual number of births, definitions generally do not mention the timing of registration, which could be within 30 days or 12 months or before the age of 5 years. Indicators should therefore be clearly described and there should be a general agreement on standard measures.
The commendable improvement in birth registration in South Africa can be attributed to the government's commitment to improving the civil registration and vital statistics system, to changes in legislation around vital registration and to particular policies, such as cash transfers directed at the welfare of children. Evaluating the impact of these interventions would help other countries develop policies and strategies for improving birth registration. ■ Competing interests: None declared.