Ethnic-specific characteristics associated with longitudinal response patterns from prebirth to 12 years: evidence from Growing Up in New Zealand

Background Longitudinal studies can generate valuable scientific knowledge, but can be compromised by systematic attrition. Previous research shows that sociodemographic characteristics (eg, ethnicity, age, educational level, socioeconomic circumstances) are associated with attrition rates. However, little is known about whether these characteristics differ by ethnicity, and how this impacts cohort retention strategies. Methods Using antenatal to 12-year data from the Growing Up in New Zealand birth cohort study (N=6743), we examined transversal response rates by ethnicity (M & amacr;ori, Pacific, Asian, European), used sequence analysis and cluster analysis to identify unique longitudinal response patterns, and binary logistic regression to examine ethnic-specific sociodemographic characteristics associated with these response patterns. Results The overall response rate at 12 years was 71.0%, with ethnic differences in response rates widening over the six data collection waves. Three longitudinal response patterns were identified: frequent responders (65.2%), intermittent non-responders (29.6%) and frequent non-responders (5.2%). Sociodemographic characteristics such as younger maternal age and lower maternal education were associated with a higher likelihood of membership in the non-response clusters across all ethnic groups. However, there were also important nuances by ethnicity. Individual level factors (eg, household material deprivation and maternal general health) tended to be associated with non-response for Europeans, whereas structural level factors (eg, area-level deprivation and racial discrimination) tended to be associated with non-response for M & amacr;ori, Pacific and Asian peoples. Conclusion Ethnic differences in longitudinal response patterns are due to multiple factors of disadvantage, and therefore require targeted retention strategies. Stratifying analyses by ethnicity is important for revealing nuanced insights.