Surveys have always been part of medical research. However, over the past two decades, the ease of quickly administering surveys to large numbers of people has markedly increased—first, because of the ability to deliver surveys by email, and more recently because of the widespread availability of web-based, easy-to-use, professional-appearing survey tools.
This ease of use comes with a downside, though. First, it is almost effortless for administrators to overwhelm their clinical staff with the number of surveys administered. Physicians and nurses commonly report being “surveyed to death,” and this overuse of surveys leads to a degradation in the quality of all responses. Second, surveys are deceptively hard to do well. The purpose of this chapter is to provide an overview of the reasons for surveys, of methodological steps needed to conduct and analyze a good survey, and of a few common pitfalls to avoid.
PERHAPS THE MOST IMPORTANT THING YOU’LL LEARN IN THIS CHAPTER
If there is one thing we want you to leave this chapter with, it’s this: surveys are easy to get wrong—and, for the uninitiated, bad surveys look remarkably similar to good surveys. Survey research is a highly specialized methodology. If you are making critical decisions based on a survey, consider getting specialist help.
Surveys can be much trickier than they appear. If you are making critical decisions based on a survey, consider getting specialist help.
This has been known for a long time. In 1895, Caroline Miles wrote an amazingly prescient passage. We’ll address many of her insights and concerns as we work through this chapter, because they still hold true today:
To ask questions is easy, but to make the questionnaire an instrument of precision is very far from easy. It has more ways of going wrong than the chronoscope … To say nothing of the general difficulty of selecting truly cardinal points for questioning about, and the special rhetorical difficulty of framing questions that shall be perfectly clear as to the information required without at the same time prejudicing the answers to be received … What allowance must be made for influences that might unconsciously mold the answers? … Some of these sources of error can be avoided, some must be recognized and allowed for and some must forbid the use of the method except under uncommon circumstances.1
In the 21st century, Jones and colleagues summarized the dilemma facing the modern healthcare delivery scientist:
There is a widespread perception that surveys are easier to conduct than other research methodologies. However, in reality, rigorously designed, implemented and analysed survey research requires substantial planning, time and effort. Conclusions drawn from survey research may be misleading or invalid if the design of the questionnaire is poor, the sample inadequate, or unrepresentative of the target population.2