enrolment into birth to ten (btt): population and sample characteristics

12
Pnedintric and Perinntnl Epidemiology 1995,9,109-120 Study progress Enrolment into Birth to Ten (BTT): population and sample characteristics L. M. Richter*, D. Yacht, N. Cameron$, R. D. Griesels and T. de Wet7 *Department of Paediatrics and Child Health, *Department of Anatomy and VBirth to Ten, University of the Witwatersrand, Parktown, lohannesburg, tMedical Research Council, Pretoria and Slnstitute for Behavioural Sciences, University of South Africa, Pretoria, South Africa. Summary. The population under study in the South African longitudi- nal study of urban children and their families, ’Birth to Ten’ (BTT), comprised all births during a 7-week period from April to June 1990 in Soweto-Johannesburg. Specification of the population base for the cohort was hampered by a number of flaws in the notification and record- keeping systems of the local authorities. As far as could be ascertained, 5460 singleton births occurred during this time to women who gave a permanent address within the defined region. Enrolment into BTT took place over the first 15 months of the study and covered the antenatal, delivery, &month and 1-year periods. By the end of this time, and despite a major health service strike during the delivery phase, 74% of all births (4029 cases) had been enrolled into the study. There were marked vari- ations in levels of enrolment, however, by population group membership, residential area and place of delivery. In general, there was substantial under-enrolment of largely middle-class white women and their babies. Initial non-enrolment of specific segments of the population and attrition of the enrolled sample up to the end of the first year are discussed in the context of racial and social differentiation in South Africa. Introduction ‘Birth to Ten’ is a birth cohort study of determinants of growth, development and health of children in Soweto-Johannesburg, South Africa. It is a collaborative study Address for correspondence: Linda M. Richter, Centre for Epidemiological Research in South Africa, Medical Research Council, Private Bag X385, Pretoria 0001, South Africa. 109

Upload: l-m-richter

Post on 02-Oct-2016

222 views

Category:

Documents


4 download

TRANSCRIPT

Page 1: Enrolment into Birth to Ten (BTT): population and sample characteristics

Pnedintric and Perinntnl Epidemiology 1995,9,109-120

Study progress

Enrolment into Birth to Ten (BTT): population and sample characteristics

L. M. Richter*, D. Yacht, N. Cameron$, R. D. Griesels and T. de Wet7 *Department of Paediatrics and Child Health, *Department of Anatomy and VBirth to Ten, University of the Witwatersrand, Parktown, lohannesburg, tMedical Research Council, Pretoria and Slnstitute for Behavioural Sciences, University of South Africa, Pretoria, South Africa.

Summary. The population under study in the South African longitudi- nal study of urban children and their families, ’Birth to Ten’ (BTT), comprised all births during a 7-week period from April to June 1990 in Soweto-Johannesburg. Specification of the population base for the cohort was hampered by a number of flaws in the notification and record- keeping systems of the local authorities. As far as could be ascertained, 5460 singleton births occurred during this time to women who gave a permanent address within the defined region. Enrolment into BTT took place over the first 15 months of the study and covered the antenatal, delivery, &month and 1-year periods. By the end of this time, and despite a major health service strike during the delivery phase, 74% of all births (4029 cases) had been enrolled into the study. There were marked vari- ations in levels of enrolment, however, by population group membership, residential area and place of delivery. In general, there was substantial under-enrolment of largely middle-class white women and their babies. Initial non-enrolment of specific segments of the population and attrition of the enrolled sample up to the end of the first year are discussed in the context of racial and social differentiation in South Africa.

Introduction

‘Birth to Ten’ is a birth cohort study of determinants of growth, development and health of children in Soweto-Johannesburg, South Africa. It is a collaborative study

Address for correspondence: Linda M. Richter, Centre for Epidemiological Research in South Africa, Medical Research Council, Private Bag X385, Pretoria 0001, South Africa.

109

Page 2: Enrolment into Birth to Ten (BTT): population and sample characteristics

110 L. M. Richter et al.

involving more than 30 academics, researchers and service providers from two universities, the Medical Research Council and the Health Departments of the three city authorities involved in the study - Soweto, Johannesburg and Diepmeadow. The background to the study and a general outline of methods have been described previously.'a2 This paper reports on the population from which the cohort was drawn, factors affecting enrolment, the representativeness of the enrolled sample, and attrition.

The Soweto-Johannesburg metropolis comprises the largest urban area in South Africa and it is the centre of South African commerce, including the gold mining and associated industries. It has a population estimated to be between three and four million people and extends over nearly 200 square kilometres. People in this area range from the super-rich to the desperately poor; they live in circum- stances as diverse as tree-lined suburbs, decaying inner-city high-rise apartments and informal shanty or squatter settlements. Health services include private insti- tutions and practitioners of international standing, provincial and local hospitals and clinics, as well as non-medical practitioners. Despite the political changes set in motion in 1990, with the announcement of the final dismantling of apartheid, being classified as white or black in South Africa still has profound implications for access to employment, housing and health care.

The majority of white South Africans are urbanised and urbanisation among black people is occurring rapidly, with an estimated urbanisation growth of 3.5% per year. Birth to Ten yas initiated to study the effects of urban environments resulting from this rapid, unplanned urbanisation on child health and develop- ment in Soweto-Johannesburg. In South Africa, generally, the common potential benefits of urbanisation in terms of increased access to employment, education and health services have been severely challenged by the inability of local and national government to meet the demand for basic services.

BTT is a prospective study of children and their families, born during a seven-week period between April and June 1990. Topics being investigated include prenatal risk factors, mortality and morbidity, growth, psychological development, environmental and household air pollution, family composition and child care. It is the only longitudinal study of its kind and the largest study of child health in urban areas yet to be conducted in South Africa. BTT is now in its fifth year. From February 1994 we will be interviewing mothers and examining their 4-year-old BTT children.

Pilot studies for the main BTT programme were undertaken during 1989 to establish the potential for follow-up of children in Soweto and the adequacy of routinely collected health information through the services in the area.3 In sum- mary, these investigations revealed that nearly 20% of the women who deliver babies in Soweto-Johannesburg are not residents in the area, but are rural dwellers who come to the city to have their babies, for a variety of political, personal and health reasons, and then return home. The pilot studies also indicated the extent of

Page 3: Enrolment into Birth to Ten (BTT): population and sample characteristics

Enrolment info ‘Birth fo Ten’ 111

fragmentation of the health services and health service information systems. For example: antenatal maternal care is not linked to postnatal child care; each of the three city authorities maintain their own system of birth notifications; and deaths are not uniformly recorded within the health service information systems in this area. In response to this, BlT made a commitment to provide feedback to the health authorities on issues relating to service delivery and record kee~ing.4,~

Methods

The source of the population data was the official birth notifications, governed by a local ordinance, and completed by delivery staff at the time of every birth in the area. This information is subsequently recorded in the registers maintained by each of the three local health authorities comprising most of the metropolitan area of Soweto-Johannesburg. Notification forms differ from one clinic or hospital to another, even within a single health authority but, in general, notifications contain all or some of the following information: mother’s name, surname, address, tele- phone number, parity, age, results of blood tests conducted during pregnancy, marital status and employment; date, place, time and type of delivery; identity of the health professionals attendant at the delivery; infant sex, birthweight, esti- mated gestational age, Apgar scores and mortality status at birth. Notifications are completed, in writing, and are subsequently sent to the records office of the local health authority where the information is again transcribed by hand into registers. Information regarding the mother’s identity, recorded on the notification, is either obtained verbally from the mother, or copied from pre-existing records such as clinic files and identity documents. Given the many levels of transcription involved in the system and the fact that there are eleven major languages in use in South Africa, it is clear that there are several ways in which serious errors of recording can occur, including auditory perceptual distortions, spelling and transcription errors. These problems resulted in a large number of errors that had to be identified and cleaned, on a case-by-case basis, through comparisons with other available birth information, such as hospital records5

Additional problems with the notifications included the fact that a small number of births were not notified. From the computerised obstetric records maintained at Baragwanath Hospital in Soweto (at which 39% of all births during this time period occurred), it was established that at least 38 births took place which did not appear in the notification records of any one of the three health authorities concerned. Secondly, notifications were duplicated, at two levels. As a matter of chance, a small number (n = 25) of notifications were found, which had been duplicated at the level of the delivery centre. That is, two separate notification forms were filled in for the same birth. However, a very much larger number of notifications were duplicated at the level of health authority registers. Approxi- mately 1000 records (18% of all births in this area during the prescribed time) were

Page 4: Enrolment into Birth to Ten (BTT): population and sample characteristics

112

unsystematically duplicated between the records of the three health authorities. These duplications occurred most often when a mother resided in the area of one health authority (say Johannesburg) and gave birth in a health facility within the jurisdiction of another health authority (say Soweto), although duplications within a single health authority's register were also found. Thirdly, there is a rather arbitrary division of residential areas serviced by the different health authorities, which led to confusion about the eligibility of mothers in some residential areas for inclusion in the study.

Because of these problems, a considerable amount of time had to be spent compiling a comprehensive and accurate database of birth notifications for the designated area and time parameters of the study. Once this was completed, it could be ascertained that the population base for BTT consisted of 5460 singleton births.

L. M. Richter et al.

Results

In order to establish whether the births that occurred during the selected time period for BTT were biased in any way, it was necessary to make a comparison with births drawn from other time periods. As part of the B I T pilot studies, a survey was conducted of all births in the Johannesburg area over a 1-year period (1987-1988). This inquiry revealed that an average of 2680 children were born each month, with no apparent marked monthly or seasonal ~ar ia t ion .~ In the case of BTT enrolment, pragmatic considerations ruled out attempting to enrol deliveries over the long summer holiday period (November to January) when many people leave the Witwatersrand to visit other parts of the country, as well as during the cold winter months (July and August). A specific investigation into whether birth patterns in the Johannesburg area over a 10-year period (July 1979-June 1989) showed a seasonal pattern, indicated that October is most likely to be significantly different from the other months of the year when population group membership and gender are both considered.6 This study showed that random patterns pertained for all population groups in the area except for black African births. African births approximated a bimodal distribution, with a small increase in birth rates over May and June (2% above the annual mean) and a peak in September (7% above the annual mean). The period from October to February showed a reduction of 2-376 below the annual mean. Seasonality and birth rates is a complex topic and both biological and environmental factors are believed to exert an effect. While BTT births were enrolled during a period when a slight increase in births could be anticipated, the total additional number of births is likely to be less than a hundred.

It was particularly important, however, to establish whether the births which occurred during the selected time period for BTT (23 April4 June 1990) were different in any way from births drawn from other adjacent time periods, since a major, unanticipated, strike by health workers employed at Baragwanath Hospital

Page 5: Enrolment into Birth to Ten (BTT): population and sample characteristics

Enrolment into ‘Birth to Ten’ 113

+157=4029

6o t s 2 40 1976

+lo1 8=2994

+409=3872

+469=3463

- Antenatal Delivery 6-month 1 -year Other studies

Percentage of population of births (5460)

Figure 1. Cumulative enrolment of subjects into cohort by contact point.

and the Soweto clinics took place from the 1 to 10 May 1990, and work stoppages and violence continued in Soweto throughout the delivery period of BIT. No birth data exist, however, that are directly comparable with that of BTT, which combines information from three separate health authorities. As a compromise, and because Baragwanath Hospital was the site of the unrest, all singleton births from two 7-week periods preceding the BTT time period (15 January-2 March 1990 and 3 March 90-20 April 90) were selected from the Baragwanath Hospital records, and compared with births recorded for the 7-week BTT time period (23 April-8 June 1990). The variables on which comparisons between the data sets could be made are maternal age, gravidity and birthweight. Maternal age was assigned to four cat- egories (< 16, 17-19, 20-38 and > 38 years); gravidity to three categories (I, 2-4 and = > 5); and birthweight to four categories (< 1500,150C-2499,250&3999 and > 3999 g). Chi-square tests7 indicated that there were no statistically significant differences, on these measures, between births which occurred during the BTT intake period and births which occurred in the two 7-week periods prior to intake: maternal age (x2 = 7.46, df = 61, gravidity (x2= 2.28, df = 41, and birthweight (x2 = 10.33, df = 6) . Thus, it would appear that the strike did not affect deliveries at Baragwanath Hospital in a way that seriously biased the births which occurred during the BTT time period.

Page 6: Enrolment into Birth to Ten (BTT): population and sample characteristics

114 L. M. Richter et al.

80

$ 60 a 2 - ;I' 40

20

0 .

The first subject data for the study were collected through questionnaires administered to women who, it was estimated, would deliver within the time parameters of the study, when they were approximately 26 weeks into their pregnancies. Enrolment took place over 15 months and covered four major contacts with mothers and caregivers: questionnaires were administered antenatally, at birth and when the Bl'Tchildren were 6 months and 1 year of age. At the same time, a small number of subjects were also enrolled through specific studies under the umbrella of BTT (such as very low birthweight infants and women with chlamydia infection). During this period, 74% of the population (n = 4029) were enrolled into the study. The cumulative enrolment of subjects by contact point is shown in Figure 1.

While the overall enrolment was 74%, this figure varied dramatically by popu- lation group and by residential area, as shown in Figures 2 and 3. For example, while 86.5% of all 'coloured' (mixed race) births within the prescribed time and area were enrolled into the study, only 38% of all white births were enrolled. Similarly, 83% of women resident in Soweto who qualified for inclusion in the study were enrolled, in contrast to only 47% of women who lived in the (still largely white) suburbs of Johannesburg. Differential enrolment of subjects into BTT also occurred with regard to place of delivery. The highest enrolment (80%) was achieved

-

.

.

'

l o o r

38

86.5

69.5

- African White 'Coloured Indian

Percentage of population and group enrolled

Figure 2. Percentage of group enrolled denoted by E4 (4029).

Enrolment by population group. Percentage of population denoted by H (5460).

Page 7: Enrolment into Birth to Ten (BTT): population and sample characteristics

Enrolment into 'Birth to Ten' 115

amongst women who delivered their babies in public hospitals and clinics catering for mainly black (including 'coloured' and Indian) patients. The lowest enrolment (45%) occurred at the city hospital providing services to mainly white patients.

The effects of the health worker strike on enrolment was examined by compar- ing the total Soweto births during the intake period, the number of Soweto subjects enrolled at delivery and the total number of Soweto subjects enrolled during the 15-month enrolment period, across five 8 to 10-day periods, from before the strike to after the strike. This analysis indicated that although enrolments at the time of delivery dropped during the strike, there were no significant differences on the three measures across the five time periods, as examined by one-way analyses of variance7 (F= 1.62, 1.54 and 2.23, respectively).

Information recorded on notifications, which was available for all 5460 births, was used to describe characteristics of the enrolled as compared with the non- enrolled members of the population. The indices on which sufficient information was available for analysis were: maternal age, gravidity, gestational age, birth- weight and 1- and 5-minute Apgar scores. These comparisons are shown in Table 1. There were significant differences between enrolled and non-enrolled subjects with regard to maternal age ( P C O.OOOl), where there was a negative trend - the lower the mother's age the more likely she was to be enrolled into the study. In addition,

l o o r 83 80

60

40

20

0 SowetoIDiepmeadow

47

Suburban

83

'&loured' 8 Indian

50

Inner-city Johannesburg Johannesbkg

Percentage of population and area enrolled

Figure 3. Enrolment by residential area. Population denoted by W (5460) and area denoted by EB (4029).

Page 8: Enrolment into Birth to Ten (BTT): population and sample characteristics

116 L. M. Richter et al.

mothers whose babies were delivered at term were more likely to be enrolled than either the preterm (< 37 weeks) or post-term (42+), and there were, similarly, lower enrolment rates among the lowest and highest birthweight groups, but no signifi- cant differences with gravidity, or Apgar scores.

Table 1. Comparison of enrolled with non-enrolled subjects: maternal age

Base population Proportion enrolled (%) x2 (do P

Maternal age (years) < 17 17-19 20-38 39+ A11 known

Gravidity 1 2-4 > = 5 All known

Gestational age

< 37 3741 42+ All known

(weeks)

Birthweight (g) < 1500 1500-2499 2500-3999 4000+ All known

Apgar at 1 minute < 8 8-1 0 All known

Apgar at 5 minutes < 8 8-10 All known

158 (3%) 597 (11%)

4536 (83%) 169 (3%)

5460 (100%)

1746 (33%) 3071 (58.5%) 424 (8%)

5241 (100%)

695 (14%) 4218 (85.5%)

4935 (100%) 22 (0.5%)

79 (1.5%) 525 (10%)

4557 (86%) 148 (2.5%)

5309 (100%)

313 (11%) 2585 (89%) 2898 (100%)

96 (3%) 2962 (97%) 3 0 3 (100%)

79.1 (125) 70.7 (3) **** 77.4 (462) 73.3 (3325) 69.2 (117) 73.8 (4029)

75.3 (1315) 74.5 (2289) 73.1 (310) 74.7 (3914)

72.7 (505) 77.3 (3262) 72.7 (16) 76.7 (3783)

63.3 (50) 73.3 (385) 75.2 (3429) 68.9 (102) 74.7 (3966)

84.0 (263) 85.0 (2198) 84:s (2461)

0.95 (2) NS

7.47 (2)"

9.29 (3)'

0.22 (1) NS

90.6 (87) 2.89 (1) NS 84.2 (2495) 84.4 (2582)

**** r < o.0001; *P < 0.05; NS P > 0.05.

Page 9: Enrolment into Birth to Ten (BTT): population and sample characteristics

Enrolment into ‘Birth to Ten’ 117

Every effort was made, during the 15-month period of enrolment, to trace mothers who had not previously been enrolled. Tracing has continued since that time and the majority of non-enrolled subjects have been accounted for, in the sense that we have information about why they were not enrolled. However, only the information available up to the end of the first year, consistent with the time frame of the enrolment information, will be presented here. Of the 1431 non-enrolled subjects, 85 gave birth to stillborn babies, 70 had children who died during the 15-month period of enrolment after birth, and 11 mothers gave their babies up for adoption. In addition, contact was made with 58 women who refused to participate in the study; 78% of these were white mothers. These factors combined accounted for 15.5% of the non-enrolled subjects.

An additional 23% of the non-enrolled subjects (n = 334) could be accounted for in the following ways: the mother had died ( n =3), the child was sent to live with relatives in a rural area (111, both mother and child returned to a rural area (79), incorrect, false, or no forwarding address given (In), communication problems (for example, a newly-emigrated Taiwanese mother) and unavailable working mothers (64). The large number of women who gave inadequate or incorrect address information are likely to be peasant rural women who try to do everything in their power to deliver their babies in Johannesburg in the belief that a child’s birth document supports his or her future right to work in the Johannesburg area. Sadly, these women are not aware of the fact that laws governing internal migration have been repealed.

Of the 800 and more women who could not be traced during the first 15 months of the study, 63.5% were African, 28% white, 5% ‘coloured‘ and 3.5% were Indian. Eleven per cent of the untraced mothers were 19 years old or younger, 8.5% had a gravidity of five or more, and 11 % of their infants weighed 2499 g or less.

As stated earlier, 4029 subjects were enrolled into the study. However, some subject losses had already occurred by the end of the infants’ first year. Twenty- nine babies whose mothers had been enrolled at one of their antenatal visits were stillbirths and a further 74 babies died during their first 3 years of life. Of the 3463 subjects who had been enrolled up to the time their infants were 6 months old, 8% were lost to the sample by the time the BTT child reached the age of one year. The majority of these lost subjects were mother-infant pairs (n = 145) who permanently left the Soweto-Johannesburg area, most often to return to a rural ‘homeland‘; an additional 11 children were sent to live with relatives in rural areas. More than 130 women gave incorrect, false or no forwarding address at which they could be found. One enrolled mother died before the end of her child’s second year.

At each point when attempts were made to enrol and interview mothers, interviewers completed record sheets on unsuccessful traces, together with comments and observations. When the BTT children were 3 years old, a major effort was mounted to trace, find and obtain basic demographic information from all previously unenrolled subjects, as well as subjects who had not been seen since the

Page 10: Enrolment into Birth to Ten (BTT): population and sample characteristics

118 L. M. Richter et al.

antenatal or delivery contact. This information is rich in both quantitative and qualitative information about individual and family reasons for migrating within and out of the study area, untraceable addresses, and why subjects avoided or refused to participate in the study. The information is being written up in a separate report as is an analysis of mortality.

Discussion

In large-scale longitudinal studies, two types of sample loss have to be considered. These are initial non-enrolment of relevant subjects and attrition on follow-up, both of which may lead to systematic distortions in the nature of the eventual sample studied, in the results obtained and in the interpretation given to results. In addition, there is a belief, not necessarily supported by available evidence, that studies of high-risk populations, including subjects from low socio-economic status communities, are likely to be flawed by less than optimal participation rate^.^-^ Thus, it is important, in studies like BTT, to assess whether participating subjects are representative of the population from which they have been sampled, whether high-risk groups within the population are adequately represented, and to examine whether attrition varies systematically with subject characteristics in a way which may jeopardise the validity of results derived from continuously enrolled subjects. In this paper, data on enrolment was presented as well as interim data on attrition. However, as the area is one through which there is considerable, and often repeated, in- and out-migration, attrition is a more difficult phenomenon to assess and will be dealt with in a separate report.

Despite a major health worker strike during the birth period covered by BTT, no marked differences exist between births which occurred at Baragwanath Hospital during the 7 weeks of the BlT study and births which occurred during two preceding 7-week periods. Although the information available is not comprehen- sive, it does support the opinion that BTT covers a representative sample of births in the circumscribed geographical area.

As previously described, there was an initial non-enrolment of 26% of the relevant population. According to information currently available, two factors primarily accounted for this group: the low enrolment of more than 600 white mothers who gave birth to their babies during the time frame of the study (nearly 10% of the total population of births), and non-enrolment of rural dwellers who delivered their babies in Soweto and then returned to their homes (estimated, in the pilot studies, to comprise about 20% of the total population of births in the area at any one time). A comparison of some birth indices indicated that there were no significant differences between cases enrolled into BTT and non-enrolled subjects, except in terms of maternal age and gestation. There was a tendency for younger mothers to be enrolled into BIT, as compared with non-enrolled individuals. However, the distribution of maternal age amongst women enrolled into BTT

Page 11: Enrolment into Birth to Ten (BTT): population and sample characteristics

Enrolment into 'Birth to Ten' 119

closely paralleled the distribution in the base population, whereas the non-enrolled group deviated from this distribution with a larger proportion of older women not participating in the study.

In contrast to a major loss of subjects due to non-enrolment, attrition over a 15-month period, from the prenatal period to the end of the first year, was 8%. This trend was anticipated in the pilot studies, which suggested that the greatest subject losses would occur during the first year of the study and that they would result from, firstly, non-enrolment of white, middle-class individuals who use the pri- vate, rather than public health facilities. The second reason for subject losses would be that African rural women may give a false or non-existent address in the area, deliver their babies, and leave the Soweto-Johannesburg area immediately after the birth of their children. We speculated that attrition might be in the region of 10% during the first 2 years of a child's life while material and social uncertainties exert their major effects on the residential status of both mother and child, but that it would decline thereafter. In some long-term developmental studies, dropout rates range from 10 to 15% per a n n ~ m . ~ , ~ Thus, even if the sample in a longitudinal study is selected randomly or representative at the time of subjects' entry into the study, there is a constant danger that the sample may subsequently become progressively more distorted because of selective attrition. For this reason, attrition in BTT will be assessed regularly.

Significant disproportions were found to exist between the sampled cohort and the base population in terms of population group membership, residential area and place of delivery. As these measures are highly correlated with one another, together they indicate that BTT does not contain a proportional representation of subjects who are white, who reside in the middle-class suburban areas of Johannes- burg and who deliver their babies within the network of private, rather than public health facilities. The major reason for not enrolling this group of subjects is the lack of contact between BTT and private obstetricians and delivery centres. BTT has not been able to reach the greatest proportion of this unenrolled group through any of the public health services delivered to pregnant women and later to their children. In addition to this, the greatest percentage of refusals to participate came from this group of subjects, who probably forsee no particular health service benefits for themselves arising from participation in the study. Nonetheless, it is clear that BTT has not under-enrolled subjects with assumed high-risk social and biological characteristics. In fact, quite the opposite - the highest enrolment of subjects occurred amongst groups of subjects with personal, social and economic character- istics indicative of higher, not lower, risk status.

The political, economic and social divisions among black and white people in South Africa, as expressed in their health service use and participation in BTT, has introduced a manifest bias into the study. The consequences of this particular bias, however, have to be evaluated conceptually as well as statistically.

On the one hand it may be argued that non-enrolment of this segment of

Page 12: Enrolment into Birth to Ten (BTT): population and sample characteristics

120 L. M. Richter et al.

Soweto-Johannesburg residents is a serious flaw in a population-based study, the validity of which rests on proportional representation of salient population charac- teristics within the sample. In addition, the exclusion of the largely middle-class group of people from any statistical analyses may reduce the impact of effects mainly attributable to socio-economic status. On the other hand, it may be argued that, by inference from studies conducted in Europe and North America, a great deal is already known, on a general level, about the determinants of health and development among white, middle-class South African children. According to this view, the major theoretical and practical benefits of B I T are to be derived from unravelling the grid of risk and protective factors operating amongst the largely black and lower class people who make up the majority of the Soweto-Johannes- burg population. As a residue of apartheid reasoning, this large and heterogeneous group is constantly, and probably erroneously, lumped together in terms of both risk factors and child health and development outcomes. It is an explicit aim of BTT to identify and describe, not only children at risk, but also the factors operating amongst this group that support and promote child health and development.

References 1 Yach D, Padayachee N, Cameron N, Wagstaff L, Richter L. 'Birth to Ten' - a study of children of the 1990's living in the Johannesburg-Soweto area. Soritli Africnn Medical Iorirnnl 1990; 77:325-326. 2 Yach D, Cameron N, Padayachee N, Wagstaff L, Richter L, Fonn S. Birth to Ten: child health in South Africa in the 1990's. Rationale and methods of a birth cohort study. Paedintric and Perinatal Epidemiology 1991; 5211-233, 3 Fonn S, de Beer M, Kgamphe S, McIntyre J, Cameron N, Padayachee N, Wagstaff L, Zitha D. 'Birth to Ten' - pilot studies to test the feasibility of a birth cohort investigating the effects of urbanisation in South Africa. South Africnn Medicnl Journal 1991; 79449454. 4 de Wet T, Richter L (Eds). 'Birth to Ten': report back - initial findings. Urbanisntion and Health Newsletter 1993; IRPart 1, 1-64, Part 2, 1-50. 5 Anderson A, Richter L. 'Birth to Ten': error detection and correction in a longitudinal address database. South Africnn Journal of Epidemiology and lnfection 1994; 916-19. 6 Sparaco A. An lnvestigation ofthe Birth Patterns ofthe Four Populntioii Groups Resident in the Johannesburg Area: (July 2979-June 1989). Johannesburg: University of Witwatersrand. 7 Aylward G, Hatcher R, Stripp B, Gustafson N, Leavitt L. Who goes and who stays: Subject loss in a multicenter, longitudinal follow-up study. Developirieiital and Behavioral Pnediutrics 1985; 63-8. 8 Ebyle M., Offord D, Racine Y, Catlin G. Ontario Child Health Study follow-up: Evalu- ation of sample loss. Journal of the American Academy of Child and Adolescent Psychintry 1991; 30449456. 9 Simons C, Ritchie S, Mullet M, Mingarelle C . Subject loss and its implications for a high-risk population. lnfant Behavior and Development 1989; 12139-146.