Planning involves deciding details such as when the data collection will take place, who will collect the data, and where data collection will take place. Design, on the other hand, focuses on how exactly the information required is going to be collected.
If done correctly all methods of data collection are capable of producing excellent reports. Likewise, if done poorly, it is possible to collect data of no meaning or which cannot be analyzed. Therefore planning and designing the method of data collection for a research study is extremely important.
This article aims to explain the most common methods of data collection with examples of when their use may be suitable.
Quantitative or qualitative?
Quantitative studies involve collection of numbers, percentages etc, whereas qualitative studies involve collection of descriptive words and sentences (Blumberg et al, 2011). When deciding on how to collect data for research it is important to consider the type of data required to meet the research aim, whether this be quantitative, qualitative or a combination of both.
For example, a research study into the effect of different methods of maintaining body temperature post operatively would involve the collection of temperature readings and would be quantitative. However, a study regarding the reasons why clients choose a particular veterinary practice would involve the collection of data such as clients’ observations and opinions and would be qualitative. It is possible that some studies may lend themselves to a combination of quantitative and qualitative data collection. For example, collection of data such as the quantity and types of food sold within practice (quantitative) alongside discussing with clients why they purchase a particular food (qualitative).
Additionally, where appropriate, qualitative data can be converted into quantitative data, such as the use of body condition scores, pain scores, stress scores etc (Kessler and Turner, 1997; University of Glasgow Faculty of Veterinary Medicine, 2010). This technique of assigning a numerical value to observations allows for more straight forward analysis of such observations to be made during research. An example of this is a study by Kessler and Turner (1997) who used cat stress scores to compare a number of different boarding environments to ascertain which caused the least stress to the feline inhabitants.
Methods of data collection
Questionnaires
Questionnaires are one of the most commonly used forms of data collection (Blaxter et al, 2010) and can be used for both quantitative and qualitative studies.
Producing a questionnaire may seem like a simple task but extensive thought is required to ensure the questionnaire will be understood by respondents, completed correctly by respondents and will obtain the maximum number of responses. Incorrect design can easily discourage respondents or lead to questions being answered incorrectly therefore impacting on results and analysis.
Initially a decision on what questions to ask should be made (what information is required from respondents?). The process of the literature review should help with this, and organizing focus groups or interviews (see below) with a selection of the intended respondents can also help in deciding topic areas of interest.
Another important consideration when designing a questionnaire is the type of questions to be asked; should they be open-ended, closed or a combination of the two? The type of question that is appropriate will depend on the intended aim of the research and the data that are required. Closed questions present the respondent with fixed responses and are easier to analyze statistically as answers have some direction and can be counted and grouped. Open-ended questions have no fixed responses and the respondent answers the question as they see fit. They are more difficult to analyze statistically due to the variation in responses received. Some examples of closed and open-ended questions can be seen in Table 1. For quantitative studies closed questions are required to allow answers to be tallied and comparisons made although the inclusion of a small number of open-ended questions may allow the collection of opinions etc, which could be of interest in the discussion of results.
| Closed questions |
| How often do you exercise your dog? (please tick one option) |
| How old (in years) is your pet? |
| In your opinion how important is puppy socialization? (please tick one option) |
| What is important when purchasing pet food? Please rank in order of importance from 1 (most important) to 3 (least important) |
| Open question Open questions |
| What changes/improvements would you like to see made at the practice? |
| Why did you choose to bring your pet to this veterinary practice? |
There are multiple methods by which questionnaires can be administered: postal; face to face; telephone; and email or web based. All have various advantages and disadvantages some of which are shown in Table 2.
| Postal/web based/e-mail | Face to face/telephone | ||
|---|---|---|---|
| Advantages | Disadvantages | Advantages | Disadvantages |
| Can send large number easily | Potentially poor answers as researcher not present to answer any queries | Better response rate | Time consuming for interviewer |
| May get more honest answers as more anonymous | Low response rate | Allows queries to be answered by interviewer | Respondents more likely to get embarrassed about answers speaking directly to someone |
| Allows for large geographic coverage and access to hard to reach respondents | Cannot be too long or complex | Allows inclusion of illiterate respondents | Face to face requires access to respondents therefore more difficult to cover wide geographic region |
| Web based allows rapid data collection | Questions may be altered or answers coached by interviewer | ||
Layout is also an important consideration in questionnaire design, as Brace (2008) states the success of a questionnaire often reflects its appearance and user friendliness. The questionnaire must not discourage respondents from partaking in the research for example if the questionnaire is too long, if layout is poor and cramped, or if unnecessary personal questions are asked. The following points should be considered when designing a questionnaire:
Finally, as Davies (2007) states, after considering the above and when the questionnaire design is complete it is advisable to pilot the questionnaire. It is suggested that the minimum any questionnaire should be subjected to is an informal pilot with the possibility of larger scale pilots if time and funds allow. An informal pilot is when a small number (minimum 3–5) of respondents (target subjects or colleagues) are asked to complete the questionnaire and then feedback is obtained. The purpose of this is to:
This allows problems in the questionnaire, which might have affected analysis or the reliability or validity of the results, to be rectified. Moving on from this, a larger scale pilot is where 50 or more respondents of the target population are used. Statistical analysis can be conducted on the responses to ensure the data obtained are suitable for analysis, and are appropriate to meet the aim of the investigation (Brace, 2008). In large scale investigations this can be valuable to ensure appropriate data are being collected prior to implementing the full scale investigation.
Interviews and focus groups
Interviews are when respondents are directly asked questions or involved in discussion with the researcher either face to face or over the telephone. They may follow a specific structure and form part of quantitative research (delivering a set of closed questions in person), or be semi-structured or unstructured and form part of qualitative research (Blaxter et al, 2010). Semi-structured interviews begin with set questions which then lead the respondent into areas of discussion, whereas unstructured interviews have no set questions just topic areas to be discussed.
An additional form of interview is a focus group where the researcher brings together a group of people and interviews them together. This allows for interaction between the group members to gain further insight into the topic being discussed (Blaxter et al, 2010).
In general (apart from structured interviews) this form of data collection is used for qualitative research, although it may be useful prior to commencement of quantitative research to gain a better understanding into the topic area of choice.
Examples of veterinary research topics where the use of interviews and focus groups might be appropriate include:
Although the aim of interviews/focus groups is to enable the respondents to do most of the talking, planning is still important. The following points need to be considered:
Experimental research and design
As termed by Blumberg et al (2011) an experiment involves measurement of a particular variable when one or more variables are altered by the researcher. This is to determine any relationships present between variables.
Experimental research is useful within veterinary research to determine relationships between treatments and outcomes and therefore decide which treatments are best. For example the measurement of pain following two different methods of performing a surgical procedure (i.e. flank versus midline neutering in cats), allowing veterinary practitioners to determine which method leads to the least postoperative pain or if there is no difference between methods.
Experiment design
Hypothesis testing
When designing an experiment Davies (2007) explains that the starting point is the hypothesis, which is a statement of relationship between two variables. The terms null hypothesis (H0) and alternative hypothesis (H1) are often used, the first representing no relationship between two variables and the latter representing the presence of a relationship. For example:
H0 = the method of neutering has no effect on the amount of pain felt post operatively.
H1 = the method of neutering has an effect on the amount of pain felt post operatively.
The next step is to design an experiment to test the hypothesis. The statistical analysis of the results gained will determine whether the null hypothesis should be accepted (and the alternative hypothesis rejected), or rejected (and the alternative hypothesis accepted).
Controls
In order to make a comparison, something to compare with is required. It cannot be said that a new treatment is better if it has not been compared with the standard treatment or with no treatment at all. A control group is therefore a group that has received the standard treatment (positive control) or no treatment (negative control) (Petrie and Watson, 2006). In the example discussed it may be that midline neutering is the standard treatment and the flank approach is the new treatment, therefore midline approach is a positive control. An example of a negative control may be investigating the use of dressings to improve wound healing post surgery. No dressing is the negative control and with dressing is the new treatment. Using a negative control in such a case establishes if the period of time has the effect on the wound healing, or the application of the dressing.
It is only ethical to use a negative control if no positive control exists, or if subjecting an animal to no treatment will not lead to adverse effects (Petrie and Watson, 2006).
Variables
Blumberg et al (2011) state that in every experiment there are one or more independent variables and one dependant variable. The independent variable is the element of the experiment which is purposely different between groups of individuals creating different treatment groups (i.e. midline or flank surgical approach). The dependant variable is the element that is measured by the researcher (pain felt post operatively) and that is expected to alter depending on the independent variable.
Independent variables should be randomly assigned to cases to prevent bias (Petrie and Watson, 2006). This could be the case of the researcher randomly assigning treatments to cases if it is ethical to do so. In the example discussed comparing two neutering methods, it is not ethical to randomly assign neutering methods to cases for the purpose of research. In this case collecting data from a number of practices who use different surgical approaches to neutering cats would be appropriate.
When measuring dependent variables it is important to remember that experimental research is quantitative and therefore concerned with the collection of numerical data. This may be simply the collection of temperatures or heart rates, but some projects may not be so straightforward. In the example used the dependant variable to be measured is pain (non-numerical) and therefore a method of recording such variables numerically should be established i.e. pain scores such as those designed by University of Glasgow Faculty of Veterinary Medicine (2010).
Having set independent and dependant variables the next step is to make sure all other variables are controlled (as far as is reasonably possible). This is to minimize the influence of other such variables on the dependant variable. This process of controlling all other variables can be challenging within a veterinary environment and may mean that some experiments are too impractical to implement. This may be because there is no means of controlling certain variables, or the possibility of gaining an appropriate number of controlled cases within the time scale of the research is low. For example for research into the effectiveness of weight loss diets, participating patients should be of similar age, degree of obesity, breed, with no other health problems, given a set diet and nothing else, with a set exercise regimen etc. Would this be practical to implement within a veterinary environment and ensure client compliance? Additionally it may be unethical to control certain variables for the purpose of research such as medications administered, unless there was a license granted by the Home Office (see below).
Once variables are set and experimental design is complete, as with any research, a pilot study is recommended. This will help reveal problems/difficulties in the design so they can be rectified before the final experiment takes place (Petrie and Watson, 2006).
Other considerations
Analysis
It is worth considering at the design stage of a research project how the data collected is going to be analyzed. This will help avoid situations where time and money have been spent collecting data that is not suitable for analysis or to meet the research aim.
Ethical considerations
Ethics have to be considered prior to conducting any research and if the project is deemed unethical then it should not be undertaken. In the case of questionnaires, interviews and focus groups participants must be made fully aware of the research taking place, give their consent to be involved and be able to cease partaking at any point if they wish (Davies, 2007). The confidentiality or anonymity of respondents and their responses must also be taken into account (Blaxter et al, 2010). In the case of experimental research involving the use of animals further considerations are required including:
Animals (Scientific Procedures) Act 1986
In certain cases a license may be required under the Animals (Scientific Procedures) Act 1986. Procedures requiring a license are those which could cause the animal pain, suffering, distress or lasting harm. Procedures are exempt from this if they are carried out for the purpose of husbandry or veterinary practice and are of benefit to the animal. This includes procedures undertaken for the purpose of diagnosis (i.e. blood sampling), treatment (i.e. injecting medication), or husbandry practices (such as castration). In summary procedures that might have an adverse effect on an animal cannot be carried out solely for the purpose of research unless a license is gained.
Conclusion
There are many methods by which data can be collected for research purposes. This article has covered the most common methods but there are other methods available. Different types of research will lend themselves to different data collection methods although there are no definite right or wrong answers as to which is the best method for each project. Ultimately it is sufficient planning and design that will determine the success of each research project.