Survey Fatigue: Investigating the Effects on Response Rates of a 100% Non-Hispanic Black Oversample in Three Wisconsin Counties

Monday, June 20, 2016: 11:42 AM
Tubughnenq' 3, Dena'ina Convention Center
Sarah P. Blackwell , Wisconsin Department of Health and Family Services, Madison, WI
Richard Miller , Wisconsin Department of Health and Family Services, Madison, WI
BACKGROUND: In 2011, the Wisconsin Pregnancy Risk Assessment Monitoring System (PRAMS) received funding for a non-Hispanic black oversample in four University of Wisconsin-Madison Lifecourse Initiative for Healthy Families (LIHF) counties. In three of these counties, Wisconsin PRAMS samples 100% of non-Hispanic black mothers. Recent declines in response rates in this population may be due in part to survey fatigue among women receiving the PRAMS survey across multiple years of the oversample. This analysis examines the effect of repeat sampling across consecutive years of the PRAMS survey on non-response rates in the LIHF oversample population.   

METHODS: PRAMS data from 2011-2012 were joined to a dataset of linked successive deliveries to mothers across years in order to identify women repeatedly sampled by PRAMS in the first two years of the 100% oversample. The linked deliveries dataset contains a unique identifier for women across deliveries using mother’s first name, maiden name, and date of birth to determine a match. PRAMS operations variables for a mother’s second PRAMS record were renamed and linked to the mother’s first PRAMS record. Analysis was restricted to women residing in the oversample counties for either birth who were multiparous in 2012. Differences in 2012 response by repeat sample status were assessed using Fisher’s Exact tests and log-binomial regression to calculate crude and adjusted prevalence ratios.   

RESULTS: Wisconsin PRAMS sampled 440 non-Hispanic black multiparous mothers resident in Kenosha, Racine, and Rock counties in 2012. Forty-nine of these mothers (11.1%) were also sampled in 2011. In 2012, less than a quarter of repeatedly sampled mothers responded to PRAMS compared to 50% of non-repeatedly sampled mothers. In the crude log-binomial model, repeatedly sampled mothers were 51% (p<0.01) more likely to not respond to the survey than their non-repeat counterparts. In a model adjusting for maternal age, the only covariate associated with both repeat sampling and nonresponse, repeatedly sampled mothers were 27% (p=0.02) more likely than non-repeat mothers to not respond.   

CONCLUSIONS: The PRAMS census of non-Hispanic black mothers in the three LIHF counties may result in many women being sampled repeatedly over a relatively short time frame, which could negatively affect response rates without additional outreach to these mothers. More highly-powered and nuanced investigation of the effects of repeated sampling on response rates is needed and will become possible as more years of oversample data become available.