Cross-sectional studies are usually the simplest observational studies to conduct and analyse. They are often used to describe (summarise) the characteristics, habits, or opinions of a single group of participants. The results might then be used to explain findings from other studies, or help to design further studies. Cross-sectional studies are also used to examine the relationship between an exposure and an outcome measure, although this is often more reliably done using cohort or case–control studies. Sometimes, a sequence of cross-sectional studies over time (with almost the same design) could be used to examine trends in lifestyle habits, disease occurrence, or mortality. Further details on cross-sectional studies can be found elsewhere [1, 2].
5.2 Design
To ensure that the study objectives can be addressed satisfactorily, an appropriate sampling frame must be chosen (see Chapter 1, page 16). The choice of sampling frame will depend on how generalisable the results need to be, and on resources available to conduct the study (i.e. staff and money). The sampling frame also helps to define the eligibility criteria.
Box 5.1shows four examples of sampling frames, based on studies summarised in Boxes 5.2–5.5 [3–6], discussed in this chapter. Key design features of these studies are shown in the boxes, which highlight similar aspects of cross-sectional studies in general.
The simplest sampling frame was the one based on children with terminal cancer [5] because it only required identifying participants from a single location (i.e. one hospital in Toronto), in which the research team worked. This is a common type of sampling frame, because it does not require significant resources (compared with studies that have regional or national sampling frames), and the conduct of the study can often be incorporated into the usual work of the research team. The VDPs sampling frame in Box 5.1is also simple [3]. VDPs are one of several groups of people who are listed on a regional or national register, making it relatively easy to identify and contact them for participating in a study. Such registers can be a good way of obtaining regional or national coverage. The other two studies in Box 5.1use sampling frames that cover a wide geographical area, so would require considerable dedicated resources [4, 6].
Cross-sectional studies are often described as being conducted at ‘one point in time’, but the phrase is at best imprecise and at worst ambiguous. Two common meanings are:
All the data collected from a participant (e.g. lifestyle habits and characteristics) pertain to a single time point (actually a short time period such as a week or month), but the participants could have been identified over a short or long time period. The study on childhood cancer (Box 5.4) is an example of this. Each parent was interviewed only once about the QoL of their child, relating to the experiences around the time of the interview or recent treatment, but it took 5 years to accrue a sufficient number of subjects for a reliable analysis (2005–2009).
The participants are selected and data collected at a single point in time (again this actually could be a week or month), but some of the data relate to lifestyle habits and characteristics that occurred in the near or distant past (e.g. several years before). An example is the study on chronic lung disease (Box 5.3), in which participants were selected at one time point (1991), but were asked about past exposures, such as parental smoking when they were children.
When the sampling frame and time period of recruitment have been determined, the process for selecting the participants, how this could be done, and who will administer the questionnaires or face-to-face interviews are decided. How participants are selected from the sampling frame depends on available resources and practicalities. In Box 5.1, two of the sampling frames were relatively small (VDPs, and children with cancer at one hospital), and the one for VDPs was also easy to access, so all were approached for participation in the study. However, in the other two examples (which have regional or national sampling frames of several thousand people), it was not possible to include them all, so a random sample was used instead. There are multiple methods for selecting random samples [7], but the ultimate purpose is to achieve a representative group from the sampling frame.
In the study of children with cancer, the clinicians and nurses identified and approached potential participants, and also collected the data. In the study of VDPs, all participants were under the guidance of a VDP advisor, to whom the questionnaire packs were posted for distribution to the VDPs. In the studies of passive smoking and salt intake, which involved a geographical spread of participants too wide to be covered by the central research team themselves, staff were employed specifically for the study, to approach participants and/or collect the data.
5.3 Measuring variables, exposures, and outcome measures
Central to any study design is the type and quality of data obtained. To ensure that the data are in an appropriate form to enable the study objectives to be addressed, the questionnaires should be carefully developed. If a questionnaire is being developed specifically for a study, it is good practice to first test the questions out on colleagues not involved in the study, or ideally potential participants. What might seem clear and obvious to the investigators might not be to others, especially the general public. The feedback often leads to questions being revised to improve clarity, making the questionnaire more ‘user-friendly’.
Alternatively it may be possible to use standard questionnaires that already exist and have therefore been tested and validated. Box 5.6lists several key general features of a good questionnaire.
The following sections illustrate some of the issues considered when developing questionnaires (or adopting standard ones), for the four example studies used in this chapter.
5.4 Collecting the data
Investigators can collect data in a variety of ways, depending on the study objectives and resources available (Box 1.9).
Face-to-face interviews and telephoning participants require members of the research team to undertake this often time-consuming activity (depending on the number of participants to be contacted), but a key advantage is that these approaches can significantly increase the response rate. Postal surveys can be a convenient way for participants to take part in a study, and to have greater coverage, but response rates can be relatively low (often <50%); using pre-stamped addressed envelopes usually improves this. Online surveys rely on participants having a computer at home and the software (database), which collects the data, being easy to use.
Random telephoning was, in the past, a common way of approaching potential participants, using listings of all home (domestic) telephone numbers in a regional or national area, or random digit dialling, which can be accessed after appropriate approvals have been obtained. However, many people (particularly young adults) now only have mobile telephones, so other methods have had to be developed to approach these as potential participants.
A balance must be achieved between trying to collect as much information as is required, and maximising the response rate. Response rate can be defined in a couple of ways: participants who complete some or all of the questions survey(s), or those who complete all of the questions. Participants faced with too many questions can be deterred from completing (or even starting) the questionnaire, or may rush through the questions and so answer some of them inaccurately (producing inconsistent responses to similar questions).
If the data collected directly from the participant are to be linked with other data (such as hospital records), the agreement and consent of the participant may be required, because the process is likely to involve the use of identifiers (such as date of birth and hospital number). Box 5.7shows how different studies can collect data, using the examples in this chapter.
