A survey involves gaining an overview of a specific phenomenon or situation directly from those concerned (Polit & Hungler 1999). Thus it allows one to obtain a description of the phenomenon or situation under study. Correlations can be established as well as relationships of various kinds, although no interpretations of causal relationships can be established. Surveys can be used in cross-sectional or in longitudinal research (see Chapter 25), including the same people several times and over time to study how their views change (studying trends). They can also be repeated over time with different people, but the same population (trend study). Also, various types of data can be collected at different points in time but from the same people (panel study).

The other meaning often connected with the term survey is self-report, which refers to a way of collecting data rather than a design. Self-reports can be obtained in other designs also. Self-reports refer to information given to the researchers directly by the people belonging to the target sample. This can be done through personal interviews or in response to questionnaires distributed in various ways.

In this chapter, survey as a design for research will be addressed, as will the data collection method self-reports. It seems worthwhile to keep these two senses of the term survey separate, i.e. when it is used in the sense of research design or when it refers to the data collection method (see also Chapter 25 and Chapter 29).

Research questions suitable for a survey design are those where the researcher wants a broad overview of a certain phenomenon or a certain situation. This design can also be used as a platform for further in-depth studies of a certain aspect. Before going into depth in a study of fatigue, for instance, in patients with Parkinson’s disease, you may want to know how prevalent the problem is in such a sample compared with people of the same age who are healthy. You may also want to know more about the circumstances in which it is more or less prevalent, age-related, gender-related or related to the severity of the disease. The survey design may be useful in giving this broad overview of the problem and the conditions under which it may vary. This can, for instance, be an overview of a certain opinion as, for instance, that of patient satisfaction, or a certain attitude, for instance how to prioritise in the public health care system (Werntoft et al 2006), or a certain behaviour, for instance how people act in an emergency situation. Other areas that can be addressed by surveys include people’s knowledge in a certain situation, for example patients with diabetes and their knowledge of lifestyle factors contributing to control over the HbA1c level.

As for behaviour, the survey design may be useful to investigate practice in nursing care. An example of this may be the sedation of patients cared for using a ventilator and the techniques used to establish the depth to which the patient is sedated. Also, people’s living conditions can be investigated in this type of design. Thus, what is common to the questions used in a survey is that people’s views, experiences, attitudes, knowledge or the like (Polit & Hungler 1999) are elicited in a systematic and standardised manner so that the response from respondent A addresses the same issue as the response from respondent B. The internal validity (Kazdin 1998) depends, among other things, on the questions being phrased in a manner that pre-vents ambiguity, and it must be possible to understand the questions, free from arbitrary inter-pretations.

It is of utmost importance for the response rate that people included in a survey study view the issue as being important to them. People’s motivation to participate in surveys of various types has decreased over the years (Herzog & Kulka 1989). This calls for care when choosing this research design and the sample to be studied.

Various techniques can be used for data collection in survey studies. In general there are two different ways: personal interviews or some kind of self-administered questionnaire (Baker 1994). Personal interviews can be conducted in a face-to-face situation or as telephone interviews. The self-administered questionnaire can be sent by post or by handing over the questionnaire, for instance in a waiting room, on admission or when leaving hospital. The questionnaire can also be distributed at, for instance, a class meeting, or a meeting of an interest organisation such as a pensioners’ or asthma patients’ association or a group of people discussing, for example, living with heart disease. Also, it is becoming more common to use the internet for distributing questionnaires, either by attaching them to an electronic mail or by giving the respondents an internet address and a login code to respond to the questions directly on the web.

Surveys serve many purposes and have their pros and cons. Among the strengths is the fact that a large group of respondents can be reached relatively easily. If sampling is carefully planned and performed, sufficiently strong conclusions can be drawn. Among the disadvantages is especially that only superficial knowledge of a certain aspect or phenomenon can be obtained. Thus no in-depth understanding can be obtained. Other drawbacks of this design depend on the method of data collection; for instance, if it is done through personal interviews or through postal questionnaires. In both cases the dropout can be substantial. Conducting personal interviews is costly and may suffer from interviewer effects (see Chapter 27) stemming either from avoiding a specific subject in a face-to-face interview or from the fact that different interviewers phrase or rephrase the questions in different ways. Personal interviews may contribute to more reliable responses but they are very costly in terms of manpower and travelling. If several interviewers are involved they may approach the questions in different ways, thus biasing the results. There may also be a problem with some questions that may be more sensitive to respond to face-to-face.

The advantages of the personal interview is that, by establishing contact and working through an interview schedule with fixed response alternatives, the interviewer can make sure that there will be no internal dropout. The interviewer can also clarify questions that the respondent may have difficulties with. The interviewer can facilitate the respondents’ interest in the study by explaining its purpose and how it can be used.

Alternatively, telephone interviews may be less expensive but not as personal. They may work very well if personal contact has been established before the interview, for instance in discharge planning, or if the clinic is known to the person.

Self-administered questionnaires may be less expensive but there is a risk of a low response rate and steps need to be taken to prevent dropout. This can be done in various ways, one of which is to call the respondents and inform them about the study, asking for consent and explaining what is expected of them and how data will be treated. Sending reminders after a few weeks is another way of encouraging people to respond, as well as providing some token of gratitude, for instance a gift voucher. The benefits of a self-administered questionnaire are that it is less costly, and it allows for the inclusion of a sample large enough to respond to the research questions with satisfactory power (Kazdin 1998). It also allows for the inclusion of a larger geographic area. The respondents can reflect on the questions and take the time needed to respond in a satisfactory manner. Interviewer effect is eliminated but there is no one to correct misunderstanding of a question or response format.

Another shortcoming of using self-administered questionnaires is that it is hard to know whether the respondent has responded independently, received help from family members or if a different person has responded. Another limitation when a self-administered questionnaire is used is that people who have difficulties in reading and writing cannot be included and this may impose a systematic bias on the findings. Similarly, using modern techniques such as the internet for collecting the data will exclude those not familiar with this technology or those not having access to it.

The aim of the survey must be of importance to those being addressed. The respondents should feel motivated to contribute to the study and they will do so if the focus of the study is of concern to them. This has implications for the sample. Also, activities to motivate and establish a relationship between the investigator and the respondents will most likely improve the respondents’ willingness to participate in the study. Thus patients, current or previous, will be more likely to respond to questions related to their health condition, whilst the public in general may feel less motivated to respond about their health.

Whether using interviews or questionnaires, the risk of systematic dropout is extensive, involving especially vulnerable people: those who are very sick, those with reading or writing problems, the illiterate, poor people and other disadvantaged groups. This means that when there is dropout, this has to be explored, as to whether it is systematic, meaning that a particular group has dropped out of the study. The response rate may be regarded as satisfactory, for instance 80%, yet the 20% that did not respond may be these vulnerable people. Thus the findings will be distorted; for example, in a study on patient satisfaction it was shown that the most fragile people were excluded, either deliberately not given the form or not able to respond, or they died; thus the results reflected those in best condition and perhaps more likely to answer positively (Ehnfors & Smedby 1993). Dropout is unavoidable; therefore the investigator should carefully monitor the respondents and non-respondents in those variables that are likely to affect the variable under study (Hellström & Hallberg 2001). For instance, this can be age, gender, capacity for daily living, social network etc. In a study of patient satisfaction, factors such as length of hospital stay or disease severity may affect the patient’s view of the hospital stay. It is more informative to be able to tell which respondents the study failed to include than it is to tell the percentage of response rate only (Table 17.1).