Open Collections

The effect of survey design on response rates, costs, and sampling representativeness in the British Columbia Health Survey : a randomized experiment

University of British Columbia

Background: Population-based survey is an essential surveillance tool applicable to various settings, including collecting information regarding community health and public living standards. In the recent decades, there have been numerous reports of decreasing response rates in population-based data collection. There is a need to redesign surveys in a way that is both more appealing to participants and maximizes response rats. Objectives: The current study explored the effects of several survey design features on participant response rates, costs, and data representativeness in a general population health survey in British Columbia. Methods: The British Columbia Health Survey was conducted by the Arthritis Research Centre of Canada and was designed to target all non-institutionalized adults in BC. Seven variants of the survey, each contained a different combination of survey design features, were developed. Survey features under examination were survey mode of administration (paper vs. online), prepaid cash incentive ($2 vs. none), lottery (instant vs. end-of-study lottery), questionnaire length (10 min vs. 30 min), and sampling frame (Info Canada vs. Canada Post). 8000 households in BC were randomly allocated to one of the seven sample groups (Table 6.1). Results: The overall response rate was 27.9% (min-max: 17.1-43.4). The survey mode elicited the largest effect on the odds of response (OR 2.04, 95% CI 1.61-2.59), while the sampling frame showed the least effect (OR 1.14, 95% CI 0.98-1.34). With the exception of the Info Canada sampling frame, all survey features under examination led to statistically significant differences in response rate. Cost analysis for the seven groups showed a negative association between the number of survey features and the resulting cost per response. The baseline survey (no incentives attached) exhibited the lowest cost per surveys sent ($12.76), while the paper survey group (including all possible incentives) showed the highest cost per survey sent ($17.87). Data representativeness results showed significant differences between our survey and the population-weighted Canadian Community Health Survey (CCHS) in terms of socio-demographic variables, but similar distributions for health variables. Findings from this study provided further insight into ways to improve response rates as well as cost-efficiency in self-administered general population health surveys.

Text

2014-06-26

Attribution-NonCommercial-NoDerivs 2.5 Canada

http://hdl.handle.net/2429/47059

Population and Public Health

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/2.5/ca/