Impact of methodological "shortcuts" in conducting public health surveys: Results from a vaccination coverage survey

• Main Authors: Luman Elizabeth T, Sablan Mariana, Stokley Shannon, McCauley Mary M, Shaw Kate M
• Format: Article
• Language: English
• Published: BMC 2008-03-01
• Series: BMC Public Health
• Online Access: http://www.biomedcentral.com/1471-2458/8/99

<p>Abstract</p> <p>Background</p> <p>Lack of methodological rigor can cause survey error, leading to biased results and suboptimal public health response. This study focused on the potential impact of 3 methodological "shortcuts" pertaining to field surveys: relying on a single source for critical data, failing to repeatedly visit households to improve response rates, and excluding remote areas.</p> <p>Methods</p> <p>In a vaccination coverage survey of young children conducted in the Commonwealth of the Northern Mariana Islands in July 2005, 3 sources of vaccination information were used, multiple follow-up visits were made, and all inhabited areas were included in the sampling frame. Results are calculated with and without these strategies.</p> <p>Results</p> <p>Most children had at least 2 sources of data; vaccination coverage estimated from any single source was substantially lower than from all sources combined. Eligibility was ascertained for 79% of households after the initial visit and for 94% of households after follow-up visits; vaccination coverage rates were similar with and without follow-up. Coverage among children on remote islands differed substantially from that of their counterparts on the main island indicating a programmatic need for locality-specific information; excluding remote islands from the survey would have had little effect on overall estimates due to small populations and divergent results.</p> <p>Conclusion</p> <p>Strategies to reduce sources of survey error should be maximized in public health surveys. The impact of the 3 strategies illustrated here will vary depending on the primary outcomes of interest and local situations. Survey limitations such as potential for error should be well-documented, and the likely direction and magnitude of bias should be considered.</p>