RT Journal
A1 Livingston EH, Wislar JS
T1 MInimum response rates for survey research
JF Archives of Surgery
JO Archives of Surgery
YR 2012
FD February 1
VO 147
IS 2
SP 110
OP 110
DO 10.1001/archsurg.2011.2169
UL http://dx.doi.org/10.1001/archsurg.2011.2169
AB 
In the interest of brevity, the answer can be simplified to a matter of bias. Several types of bias and error may be at play in any survey, but the most obvious one is nonresponse bias, which occurs when nonrespondents differ in significant ways from respondents on the variables of interest. Nonresponse bias is particularly salient when the topic is considered sensitive and the respondents would prefer to not discuss such matters. In a study like the Oreskovich et al study, surgeons who drink more heavily are potentially less likely to respond, which might underestimate the prevalence of alcohol abuse.