Brief CommunicationThe Impact of Attrition in an 11-Year Prospective Longitudinal Study of Younger Women
Introduction
Attrition or loss of study participants may bias the results of longitudinal studies (1). For example, findings indicate that smokers are less likely to respond and that healthier people may be more likely to respond to subsequent surveys 2, 3, 4. However, whether the associations between smoking and health are affected by attrition is not known. In this study, we investigate whether (i) the characteristics of respondents to one, two, three or all waves are different and (ii) the relationships between health and other variables differ for those who respond to some or all waves of a longitudinal study.
Section snippets
Methods
Data come from the first four waves of the Australian Longitudinal Study on Women's Health cohort born between 1973 and 1978 (the 1973–1978 cohort) (5). Surveys were mailed in 1996, 2000, 2003, and 2006.
Results
Almost half the women were consistent respondents, and more than 80% responded to at least two waves (Table 1). Nonresponse was permanent (6%), because of discontinued participation or death, or temporary, because of lack of contact with the participant (28%) or other nonresponse at the wave (8%). At wave 1, consistent respondents had more education, were less likely to have children, be stressed about money, and to be smokers than intermittent respondents (Table 1). Consistent respondents had
Discussion
Consistent respondents comprised one half of the initial sample. Nonresponse was related to baseline characteristics (eg, smoking, poorer health). Despite these differences, the relationships between self-rated health and education, stress about money, smoking, and general practitioner visits were similar, indicating there was no serious bias in estimates of associations caused by attrition. Standard errors were larger in consistent respondent models than those including intermittent
References (11)
Attrition in longitudinal population studies: Does it affect the generalizability of the findings?
J Clin Epidemiol
(2002)- et al.
Patterns of non-response to a mail survey
J Clin Epidemiol
(1990) - et al.
Tracking and follow-up of 16,915 adolescents: Minimizing attrition bias
Controlled Clinical Trials
(1997) - et al.
Psychological attributes and changes in disability among low-functioning older persons: Does attrition affect the outcomes?
J Clin Epidemiol
(2002) - et al.
Survey nonresponse bias among young adults: The role of alcohol, tobacco, and drugs
Substance Use & Misuse
(2005)
Cited by (111)
Associations between endometriosis and common symptoms: findings from the Australian Longitudinal Study on Women's Health
2023, American Journal of Obstetrics and GynecologyResidential greenspace and anxiety symptoms among Australian women living in major cities: A longitudinal analysis
2023, Environment InternationalBiopsychosocial factors influencing the occurrence and recurrence of preterm singleton births among Australian women: A prospective cohort study
2022, MidwiferyCitation Excerpt :Participant characteristics were mostly comparable with national census data across time (e.g. at Survey 1 in 1996 (Brown et al., 1996), and Survey 6 in 2011 (Dobson et al., 2015), although Australian-born, married and tertiary educated participants were slightly overrepresented (Brown et al., 1996; Dobson et al., 2015). The response rates among women contacted at each of the follow-up surveys (Survey 2–7) were stable above 80% (Powers and Loxton, 2010). The response rate to each survey is comparable with other cohort studies (Mostafa, 2014) (Bethel et al., 2005).
Longitudinal changes in employment following a diagnosis of endometriosis: Findings from an Australian cohort study
2022, Annals of EpidemiologyCitation Excerpt :Further, the matched comparison group of women without endometriosis in this study were more likely to have attended university than the non–matched women. Women who were randomly selected for matching were more consistent responders to the ALSWH surveys, and consistent responders are more likely to be university educated [29]. Consequently, the differences in labor force participation for women with surgically confirmed endometriosis, and women without endometriosis may be underestimated.