Methods/Statistics
Differences in Response Patterns by Socio-Demographic Characteristics in a Large Longitudinal Cohort Emanuela Mesesan* Emanuela Mesesan Daniela Dudas Mariah Landry Alpa V. Patel
Department of Population Science, American Cancer Society, Atlanta, GA
Background: Longitudinal prospective cohort studies are essential to understanding disease risk over time. Participant attrition may vary by factors such as age, race/ethnicity, or sex and could impact internal study validity. We examined response patterns over time in a large U.S. prospective cohort by demographic factors.
Methods: From 2006-2013, ~250,000 adults were enrolled in the American Cancer Society’s Cancer Prevention Study-3 (CPS-3) and completed a baseline survey. In 2015 and every three years thereafter, participants received follow-up surveys to update exposure and health information. We examined response patterns by race (white, Black, Hispanic, Asian, other), baseline age (<35, 35-44, 45-54, 55-64, and 65+ years), sex (male, female), and education (
Results: Overall response decreased over time from 74% in 2015 to 66% in 2018 to 62% in 2021. Women consistently had higher response than men, an average 68% vs 63% respectively. Response rates among participants with higher educational attainment were more than 20% higher than participants with the lowest educational attainment. Response rates were steadily higher among older participants; participants aged 55 and older had more than 10% higher response rates than participants under age 45 at all-time points. White participants were more likely to respond over time, followed by Asian participants; and Black participants were less likely to respond.
Conclusion: In CPS-3, overall response to follow-up surveys declined over time. Response rates were highest among participants who were female, white, more educated, and over 55 years. Strategies to increase sustained engagement, particularly in men, non-white, less educated, and younger individuals, should be explored to minimize attrition over time.