Longitudinal cohort studies are traditionally difficult to sustain in terms of the resources necessary to manage and maintain a cohort with sufficient reliability and validity over long periods of time (das Gupta, Arby, Garenne, & Pison, 1997
). Bias in sample composition as a result of both recruitment and attrition can significantly affect the validity of findings, as well as the legitimacy of the original hypotheses across the length of the study time period. This paper has reflected on some of the ways in which these concerns have been evaluated, addressed and managed within the parameters of the Birth to Twenty longitudinal birth cohort study. The comparatively low attrition rate over the duration of the study, despite unique contextual challenges experienced in a developing country, serves as evidence of the success of the Birth to Twenty strategies in limiting attrition and managing the cohort. Of particular interest, given the otherwise successful maintenance of the cohort over the course of the study so far, was to examine where and why recruitment was incomplete and what determined the losses that occurred during the enrolment phase.
The case study analysis demonstrates that high rates of circular migration significantly influenced the ability to recruit participants initially and establish contact at follow-up attempts. These patterns of mobility are linked to the historical migrant labour policies that limited the movement of Black South Africans and that consequently propagated a nomadic labour culture. The movements of Black South Africans were controlled through passes
and permits that prohibited them from residing in urban areas unless gainfully employed there, or from leaving rural areas unless employed in urban areas (Jones, 1993
). The spatial boundaries of African households became extended and considerable variability in household roles and functions developed as a result of high individual mobility, conjugal disruption, illegitimacy, desertion and fragmentation of the traditional nuclear and extended African family unit (Murray, 1981
). The migrant labour system, together with the creation of so called ‘homelands’, policies of forced removals and influx control, exercised considerable influence in shaping population movements and the composition of urban, semi-urban and rural settlements.
The economic imperatives compelling people to migrate to metropolitan areas (Jooma, 1991
) have not changed with the political transformation and as a result, this stream of movement has continued. Recent census figures have confirmed the existence of a highly mobile population with a growth of 10% in the general South African population since 1996, and a 20% increase (the largest) in the Gauteng region (the province in which the BT20 study is located). This growth has been attributed to internal migration from rural areas (Statistics South Africa, 2001
). Gauteng, therefore, receives a considerable number of its residents from other parts of the country. Given such patterns of high mobility, the difficulty and extent of resource expenditure in tracking and managing a cohort such as that in Birth to Twenty is understandable, as are obstacles to follow-up in the initial recruitment phase.
Name variations and name changes also pose a challenge to follow-up procedures as was evident in the case study analysis. In fact, previous research conducted in the Birth to Twenty Study has documented the significant part played by differences in the spelling and variation of African names (Anderson & Richter, 1994
), as well as through duplicate entries in official database records. Name changes in South Africa are rooted both in cultural tradition (having multiple names) and the legacy of Apartheid
that ‘forced’ Black South Africans to renounce traditional names for English or Afrikaans substitutes to facilitate employment. If all names are not recorded, it is possible to lose contact with individuals or duplicate their status as participants within the study. This was exacerbated in the case of caregivers; many of them grew up during the Apartheid
era and hence, have English or Afrikaans variations to their names.
A significant body of research has confirmed the ways in which Apartheid
legislation attempted to dilute and deny the cultural identity of non-White South Africans (Martin, 2000
). Consequently, the post-transformation period has seen attempts to reclaim identity, and naming, as a significant constituent of identity, is one area in which such attempts have occurred. As a result caregivers may have chosen not to report or disclose an earlier English or Afrikaans name. This process complicates cohort management and may have hampered follow-up.
Our findings also demonstrate that reasons for non-participation amongst some of the cases were convenience-related (time, migration and privacy concerns). Previous studies have also cited respondent-perceived inconvenience as a major factor determining non-participation in research (Hayman, Taylor, Pearl, Galland, & Sayers, 2001
). In addition, Shavers, Lynch, & Burmeister (2002)
reported that abuse and manipulation by the medical fraternity resulted in non-participation of African-Americans in research studies (specifically medical research) and, in fact, engendered pervasive distrust of research endeavours. There are parallels between these findings and the South African context given that, in 1990, South Africa was still an Apartheid
state. Many of the intended recruits would have had negative personal experiences of the political regime and hence may have been distrustful of a historically White research endeavour. However, the considerable size of the residential sample recruited by BT20 (3273 cases), militates against a significant role for this factor in recruitment and continued participation. The employment of field workers from the local communities and the setting up of a Community Advisory Board, among other strategies, has decreased the potential for distrust between the community and the researchers.
Limited communication infrastructure also hampered recruitment and follow up, especially during the early years of the study. Given the inadequacies and poor quality of the address framework and postal system in South African townships in general at the time, follow-up via mail posed a real challenge. Boys et al. (2003)
reported that the use of a range of communication strategies in longitudinal studies including face-to-face contact, follow-up reminders, postcards and postal contact were useful in keeping attrition rates low. These communication methods are in place in BT20 and contribute to the low attrition discussed previously. However, the high rate of mobility within and out of the study area, coupled with postal system deficiencies, made it difficult to maintain regular contact with the cohort.
Telephonic follow-up was also initially difficult due to the small proportion of the sample who owned telephones, lack of work contact numbers due to high unemployment rates and, inconsistencies in home contact details because of high mobility (Richter et al., 2004
). To overcome these challenges and ensure the ability of the study to track the cohort, BT20 obtained contact details of relatives and friends as well as makes frequent visits to the recorded home address. The recent increasing availability of mobile telephones has tremendously improved communication with the cohort.
The case study analysis explored study perceptions and motivations for participation in a study such as BT20. Altruism, coupled with some benefits for the child's physical and psychological health was noted. Groves, Cialdini, and Couper (1992)
have reviewed a number of possible explanations for decisions relating to research participation. These include societal-level factors, design attributes, characteristics of the sample person, interviewer attributes, and interaction between interviewer and respondent, as well as psychological factors such as compliance and helping tendencies. The decision to participate in BT20 has been shown to vary by virtue of contextual dynamics on all these. A number of factors impacted on the relationship between the study, as a site of knowledge production, and the subjects of research. These included the historic-temporal context of the study at the brink of transformation after an extensive period of oppression, the politicization of race relations and the unequal distribution of educational and economic resources.