P1-409 Differential loss of participants does not necessarily cause selection bias

Department of Public Health, University of Otago, New Zealand.
Australian and New Zealand Journal of Public Health (Impact Factor: 1.98). 06/2012; 36(3):218-22. DOI: 10.1111/j.1753-6405.2012.00867.x
Source: PubMed


Most research is affected by differential participation, where individuals who do not participate have different characteristics to those who do. This is often assumed to induce selection bias. However, selection bias only occurs if the exposure-outcome association differs for participants compared to non-participants. We empirically demonstrate that selection bias does not necessarily occur when participation varies in a study.
We used data from three waves of the longitudinal Survey of Family, Income and Employment (SoFIE). We examined baseline associations of labour market activity and education with self-rated health using logistic regression in five participation samples: A) the original sample at year one (n=22,260); B) those remaining in the sample (n=18,360); C) those (at year 3) consenting to data linkage (n=14,350); D) drop outs over three years (n=3,895); and E) those who dropped out or did not consent (n=7,905).
Loss to follow-up was more likely among lower socioeconomic groups and those with poorer health. However, for labour market activity and education, the odds of reporting fair/poor health were similar across all samples. Comparisons of the mutually exclusive samples (C and E) showed no difference in the odds ratios after adjustment for sociodemographic (participation) variables. Thus, there was little evidence of selection bias.
Differential loss to follow-up (drop out) need not lead to selection bias in the association between exposure (labour market activity and education) and outcome (self-rated health).

Download full-text


Available from: Sarah Mckenzie, Jun 12, 2014
  • Source
    • "However, unless these dropout rates were jointly distributed by the social and economic exposure measures and drinking behavior outcome measures, the effect of attrition on the studied associations is likely to be minimal. We have also previously shown that for the association of employment and education with self-rated health there appears to be no difference in the association among those respondents leaving the SoFIE study (non-participants) compared to those respondents who stayed in the study (participants) [36]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Social and economic measures in early childhood or adolescence appear to be associated with drinking behavior in young adulthood. Yet, there has been little investigation to what extent drinking behavior of young adults changes within young adulthood when they experience changes in social and economic measures in this significant period of their life. Methods The impact of changes in living arrangement, education/employment, income, and deprivation on changes in average weekly alcohol units of consumption and frequency of hazardous drinking sessions per month in young adults was investigated. In total, 1,260 respondents of the New Zealand longitudinal Survey of Family, Income and Employment (SoFIE) aged 18-24 years at baseline were included. Results Young adults who moved from a family household into a single household experienced an increase of 2.32 (95% CI 1.02 to 3.63) standard drinks per week, whereas those young adults who became parents experienced a reduction in both average weekly units of alcohol (β = -3.84, 95% CI -5.44 to -2.23) and in the frequency of hazardous drinking sessions per month (β = -1.17, 95% CI -1.76 to -0.57). A one unit increase in individual deprivation in young adulthood was associated with a 0.48 (95% CI 0.10 to 0.86) unit increase in average alcohol consumption and a modest increase in the frequency of hazardous drinking sessions (β = 0.25, 95% CI 0.11 to 0.39). Conclusions This analysis suggests that changes in living arrangement and individual deprivation are associated with changes in young adult’s drinking behaviors. Alcohol harm-minimization interventions therefore need to take into account the social and economic context of young people’s lives to be effective.
    Full-text · Article · Sep 2014 · BMC Public Health
  • Source
    • "Limitations include absence of information on psychological factors, such as diagnosed mental health conditions or psychological distress: this limits our ability to consider such conditions as covariates or effect modifiers, but our model is robust whether or not the mental health component of QoL is included, suggesting this is not a major issue. Additionally, the initial response rate of 57%, while lower than desirable, is similar to other comparable Australian studies, [19] and a lower response rate does not mean that relationships between outcome and exposure are necessarily biased [28]. Participants who did not continue in the study were older, heavier, with pain at more sites at baseline than the remaining participants. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Pain and radiographic changes are common in persons with osteoarthritis, but their relative contributions to quality of life are unknown. Methods Prospective cohort study of 1098 men and women aged 50–80 years, randomly selected from the electoral roll. Participants were interviewed at baseline and approximately 2.6 and five years later. Participants self-reported prior diagnosis of arthritis and presence of joint pain. Joint space narrowing (JSN) and osteophytes at the hip and knee were assessed by X-ray. Quality of life (QoL) was assessed using the Assessment of QoL (AQoL) instrument. Data was analysed using linear regression and mixed modelling. Results The median AQoL score at baseline was 7.0, indicating very good QoL. Prevalence of pain ranged from 38-62%. Over five years of observation, pain in the neck, shoulders, back, hips, hands, knees and feet were all independently and negatively associated with QoL, in a dose–response relationship. Diagnosed osteoarthritis at all sites was associated with poorer QoL but after adjustment for pain, this only remained significant at the back. Radiographic OA was not associated with QoL. While AQoL scores declined over five years, there was no evidence of an interaction between pain and time. Conclusions Pain is common in older adults, is stable over time, and the strongest musculoskeletal correlate of QoL. It also mediates the association between diagnosed OA and QoL. Since the same factors were associated with quality of life over time as at baseline, this suggests that quality of life tracks over a five year period.
    Full-text · Article · Sep 2012 · BMC Musculoskeletal Disorders
  • [Show abstract] [Hide abstract]
    ABSTRACT: Background: There is evidence that smoking is associated with poorer mental health. However, the underlying mechanisms for this remain unclear. We used longitudinal data to assess whether smoking uptake, or failed quit attempts, are associated with increased psychological distress. Methods: Data were used from Waves 3 (2004/05), 5 (2006/07) and 7 (2008/09) of the longitudinal New Zealand Survey of Family, Income and Employment. Fixed-effects linear regression analyses were performed to model the impact of changes in smoking status and quit status (exposure variables) on changes in psychological distress (Kessler 10 (K10)). Results: After adjusting for time-varying demographic and socioeconomic covariates, smoking uptake was associated with an increase in psychological distress (K10: 0.22, 95% CI 0.01 to 0.43). The associations around quitting and distress were in the expected directions, but were not statistically significant. That is, smokers who successfully quit between waves had no meaningful change in psychological distress (K10: -0.05, 95% CI -0.34 to 0.23), whereas those who tried but failed to quit, experienced an increase in psychological distress (K10: 0.18, 95% CI -0.05 to 0.40). Conclusions: The findings provide some support for a modest association between smoking uptake and a subsequent increase in psychological distress, but more research is needed before such information is considered for inclusion in public health messages.
    No preview · Article · Oct 2012 · Tobacco control
Show more