research instruments - Textbook of Nursing Science

According to the presence of impact, exposition in the re-search, experimental (interventional) and non-experimental (intervention-free) examinations are differentiated. In experi-mental research, the researcher manipulates the changes in

phenomena, this method is described below. In health sci-ence research it may mean clinical, territorial and population level experiments.

Non-experimental examinations in the field of nursing can be carried out using different research instruments, from which, due to limitations to size, the methods of experiment, survey, observation and interview are described here.

Experiment

In experimental research, the researcher studies the cause-effect relationships between variables systematically and rig-orously, and takes steps in order to ensure the achieved results (the effect) could only be attributed to the intervention. The most important advantage of experiments is that they can provide evidence for everyday care and practice. In healthcare experiments are generally called clinical trials.

The aim of explanatory examination is to understand how and why the intervention works, that is, it is directed at acquir-ing scientific knowledge. Practical experimental research ex-amines the effectiveness of interventions in real research situ-ations. In this case, the intervention is tailored to the condition of the patient, whereas in explanatory research, patients are selected for the intervention.

r

ealexperimeNtalresearch

Intervention, control and random sampling are jointly applied in a real experiment. A controlled randomized trial (CRT) is ca-pable of detecting a causal relationship, in which there is no difference in the treated group and the control group in terms of applied treatment, or intervention, therefore the difference appearing in the target variable can be attributed to the inter-vention (or to chance).

i

NterveNtioN

There is no experiment without intervention, the re-searcher has to do something in order to achieve an effect.

In healthcare research the term “treatment” is often used instead of the word intervention. In the course of experi-ments hypotheses and theories are tested.

E.g. Robbins et.al [44] evaluated the efficiency of infor-mation booklets and home visits in teaching about chil-dren’s diseases. The aim of the research was to evaluate the effects of the intervention on

• the use of healthcare service;

• the self-confidence of parents and their knowledge about children’s diseases;

• the parents’ intentions to treat the children’s symptoms at home;

• the parents’ intentions with regard to seeking specialist advice.

c

oNtrol

In order to prove that the only variable is responsible for the result, the researcher has to control certain external vari-ables. Those variables may belong here which themselves or together with other variables can also bring about the result. E.g. if the researcher wants to test the efficiency of consulting in the treatment of depression a conclusion may be drawn, that the patients’ condition improved as a result of consultations. However, the researcher cannot be certain of this, as it may happen that their condition would have improved anyway ”over time”, or other factors e.g. medica-tion – helped. In order to establish the real value of consul-tation sessions, setting up another group is suggested, in which patients receive the usual treatment, but they do not receive consultation. The group which receives the new in-tervention is the experimental group, and the group which is used for comparison is the control group. In addition to groups set up in this way, further possibilities to create con-trol groups are going to be described.

Intersubject design or parallel groups: In the case of parallel groups, conclusions are drawn in the experiment after com-paring two groups (treated/control) or more groups. This is the most frequent design in the case of randomised control trials (RCT). As previously mentioned, the study of Robbins et.al (2003) which explored the education of parents about children’s diseases, parents who received visits and book-lets were included in the experimental group, parents who received usual service were included in the control group.

Intersubject design may as well have more than two groups.

The number of groups depends on the aim of the experi-ment.

Within-subject or crossover design: Crossover arrangements also belong to repeated measurements. It is possible that in the case of two treatments being compared, the effect is different. However, it may show such a big spread among the subjects in the experiment, that working with two par-allel groups the difference between the treatments will not be demonstrable. In this case the differences between treat-ments can be demonstrated by applying them on the same subject one after the other. In case of chronic illnesses should be used such as(asthma, high blood pressure, diabetes, rheu-matism) when healing is not only an alleviation of pain but an improvement of state can be expected due to the treat-ment. Other examples include, if the researcher wants to find out if aromatherapy helps patients with a sleeping problem to sleep, they may select a group of 20, out of which 10 patients receive aromatherapy, the other 10 patients receive the usual sedatives during a time period of 3 weeks. Then the researcher exchanges the treatments, the first group will be given the usual sedatives, the second will be given aromatherapy. By comparing each patient’s results they can assess the effects of the new intervention.

Pearson and Hutton [37] compared the efficiency of the use of foam swabs and toothbrushes, some subjects used foam swabs, the others a toothbrush for a week, and the treatment was exchanged the second week. By this it was ensured that the effect, due to the order of treatments, could be assessed. By changing the sample the difference between the groups was minimized, ”each person func-tions as control for himself”. http://www.ncbi.nlm.nih.gov/

pubmed/12175357

The main problem with the crossover arrangement is the continuing effect which happens when the effect of the first treatment carries on during the second treatment period.

Researchers must pay special attention to this problem, and leave a period between the two interventions.

In some cases it is also possible that the same patient re-ceives two interventions at the same time. This may make the control and experimental group unnecessary, furthermore it ensures that the effect of extraneous variables is reduced to the minimum, as the same patient is involved and almost the same deviance. Its disadvantage is that this kind of experi-ment can only be applied in very few cases.

Single-case design: Experiments in which people are in-volved during healthcare provision are problematic in terms of generating a sample including a large number of elements. Even if it is possible, it is difficult to create groups according to the relevant variables. The problem is even more serious when subjects, due to different reasons, drop out of the experiment. These logistical problems are minimized by single-case design as only one participant is included at a time (obviously, if the only participant drops out, the experiment is finished). In its simplest form, a pre-test is followed by a postpre-test after the intervention. This kind of design is also known as AB design, where A is the pretest, B is the posttest, and the intervention is carried out in between the two.

It is suggested that more than one intervention should be done, repeated several times, possibly on different days, and if possible under different circumstances, because of the control of impacts. This design is useful in such cases when little is known about the effectiveness of the intervention, and it is coupled with the difficulty brought about by the large-sized traditional experiment. The biggest limitation of the individual case design is the fact that the observations cannot be generalized on a similar population since these are individual cases.

Solomon four group design: The aim of creating several groups is often not having more interventions but a better control of influencing extraneous factors. The purpose of a pretest and a posttest is usually to find out whether the in-dependent variable (intervention) caused a change in the dependent variable (result). It may occur though that the test done before the intervention itself influences the out-come. In the case of a depressed patient for example, there

are patients who are more aware of their condition, and as a consequence, they motivate themselves to perform better.

In order to avoid this, the Solomon four group construction can be applied (Table 1), which consists of four groups, two experimental groups and two control groups. The pretest and the posttest are carried out on a control and an experi-mental group, and a posttest only on another control and experimental group. This type of design eliminates the ef-fect of a pretest on the outcome. If you compare the results of the groups, you may discover if there has been such an effect.

E.g. Bakotic et. al [3] the effect of information booklets on healthy sleeping among adolescents was examined, in which secondary school students of 15-18 were included in the two experimental and two control groups. http://www.ncbi.nlm.

nih.gov/pmc/articles/PMC2681062/pdf/CroatMedJ_50_0174.

pdf

Factorial design: In experiments, the effect of the indepen-dent and depenindepen-dent variables are usually examined. For ex-ample if fluid and fibre intake affect constipation, then it is examined in both the experimental and the control group.

Probably several factors interact in real life which influence the outcome. In this case the hypothesis may be that exer-cise, fluid intake and fibre intake or their combination affect constipation. In factorial design, the effect of two or more independent variables is tested upon one or more depen-dent variables within the same research. The phrase ”factor”

refers to ”variables” in this design. This form is suitable for studying the interactions between variables, and it adds a usage value which is given by the combination of interven-tions.

r

aNdomizatioN

Some forms of random sampling have already been de-scribed in earlier chapters, which are going to be comple-mented now by some further possibilities that can be used in experiments. In experiments randomization means that subjects have an equal chance of inclusion in the experi-mental and control groups. In the case of smaller numbers, a box can be used from which slips of paper are drawn and this will decide who is included in which group, if there

are large numbers a computer may be used for the draw.

However, it is common that at admission every subject receives a non-transparent envelope with a letter A or a letter B inside. Neither the researcher nor the participant knows the content of the envelope. This is called hidden distribution.

Matched subjects design: It is possible through this meth-od that subjects are included in two groups according to the relevant characteristics. If the researcher claims the distribution of the following characteristics – middle class women between 48 and 50 with a new diagnosis of breast cancer – in the experimental and the control groups, they will seek two subjects who meet the crite-ria, and who can be included in each group. This activ-ity continues until the required number of subjects is reached. The adequately matching selection technique is usually necessary when the researcher knows in advance what the control variables are. The potential number of variables is endless, that is why it may never be certain whether all the essential characteristics of the partici-pants are the same. Furthermore, it takes a long time to find all the matched subjects for the necessary number of elements.

Randomized block design: It is useful to apply the block de-sign when a factor exerts an unwanted effect on the target variable. This can be controlled if, instead of complete ran-domization, this factor is used for stratification (the strata are the blocks), in each stratum each treatment is divided, and randomization is done within the strata. Through this, the average values of individual treatments are affected by the factor to the same degree. For example, if in an experi-ment 3 treatexperi-ments are compared, the number of subjects is 5 per treatment, and due to technical reasons all the measurements have to be taken in one day. It can be an influencing factor that the value of the target variable may vary according to the periods of the day. Furthermore the procedure is time consuming, the measurements last from morning till night. In this case the block design can be ap-plied, in which e.g. 5 blocks can be created (early morning, morning, midday, afternoon, evening), three measurements by each block (1 of each treatment) randomizing the treat-ments within the block.

Zelen design: In a traditional RCT, participants who meet the criteria are selected through randomization after they have given consent to the treatment. Sometimes this may lead to withdrawal of consent, because they are not yet divided into groups (experimental and control) as the sub-jects have hoped for. Some of them are so disappointed that they do not fully meet the protocol, and this may af-fect how they respond to treatments allocated to them.

The Zelen design offers an alternative approach to this, which is randomization of all participants (who meet the

criteria of admission) before their consent is asked for. In a traditional RCT participants are informed of the details when gaining consent.

Q

uasi experimeNtalresearch

Sometimes it happens, due to several causes – ethical and/or practical –, that it is not possible to carry out the real experi-mental research in nursing or in other healthcare areas. For example if the nurse introduces a new nursing model in the department, it is not feasible that patients are included in two groups in the same department or in different departments ei-ther, due to clinical, ethical and organizational considerations.

Possibly the new model introduced in the department can be compared with an identical department (where another model is applied – or even the same) in the same hospital. Therefore in the case of quasi experiments, there is intervention but ei-ther the control group or random sampling is missing. In this case the researcher cannot control the extraneous variables adequately, due to the uncertainty that the new intervention is really responsible for the measured change, but the verification of the close connection is possible. There are several possible versions of quasi experimental research:

• Intervention is carried out, then the outcomes are measu-red. For example the relaxation technique is applied with a group of patients if the researcher wants to know whet-her the level of anxiety lessens or not. This is the weakest form of experiment as there is no pretest, just a posttest.

• The researcher measures the level of anxiety before and after the introduction of the relaxation technique. This shows the change (if there is any) with greater reliability, though it is still difficult to state if the relaxation treat-ment is the cause, because several other factors can cause the change.

• The level of anxiety is measured on an experimental and a comparative group before and after the inter-vention. The result is even more reliable. This design is called non-equivalent group design. The subjects in both groups may be similar in many respects, but the resear-cher does not have enough control in selecting them in order to ensure equivalence.

• Interrupted time series (ITS) form includes an experimen-tal group and the series of measurements before and after the intervention. In this form four measurements are carried out (01-04) before the intervention (X), then four posttests (05-08) after the intervention. It is repre-sented as: 01 02 03 04 X 05 06 07 08.

Questionnaires

The questionnaire examination is the most frequently used primary research technique, it is suitable for descriptive, ex-planatory and exploring purposes. It is also a frequently used method in the field of nursing research, especially in examina-Table 1 Overview of the Solomon-type experimental design

Group Pretest Intervention Posttest

Experimental group1 + + +

Control group 1 + Usual treatment +

Experimental group2 – + +

Control group 2 – Usual treatment +

188 Textbook of Nursing Science Chapter 7 Fundamentals of Research Methodology and Biostatistical Knowledge 189 tions of such concepts as empathy, burn-out, social support,

pain, coping, hope, stress and life quality. In addition, infor-mation can be gathered with regard to attitudes, knowledge, beliefs, opinions, expectations and experiences of the client or the carers. The advantage of this method, is that it can be implemented relatively easily, mostly it does not burden the respondents, it is adequately compiled, and the filled-in ques-tionnaires may provide relevant information for the researcher on the above topics. In certain research topics, it is the only possibility that can be applied. Its disadvantage is subjectivity of both the compiler of the questionnaire and respondents, in some cases the lack of sincerity. The purpose of the actual compilation of the questionnaire is to ask essential questions and obtain relevant information. The most important aspects of compiling a questionnaire, the different types of questions, and general rules of their use are reviewed below. What is written down here is not compulsory rules, but considering them will make you get closer to achieving your objectives.

The questionnaire survey has four differing forms, the self-completion form when the respondent completes it in, the questioner asks the questions and records the answers, by phone and by e-mail which is becoming more a common data collection. A special form of self-completion is the group self-completion, where there is a possibility during comple-tion for the respondents to ask an interviewer who is present.

G

eNeralrulesofcompiliNGtheQuestioNNaire

In general, the characteristics of the target population should be taken into account, such as their age, social status, knowl-edge, qualifications for example. These factors may signifi-cantly affect the way the questions of the questionnaire are worded and subsequently, the compilation of the question-naire. Apart from the characteristics above, further aspects that are worth considering are the following:

• Wording should be comprehensible. The questions of the questionnaire have to be clear and unambiguous.

It may happen if the researcher is deeply involved in the particular topic, complicated questions may seem simple, and the other way round, the question will not be exact enough as a result of studying it superficially.

Do not use complex sentences, avoid using foreign words and overpolite sentences.

• Do not ask two questions at the same time. If you see the word ’and’ in a question or statement, always check and make sure you are not going to ask a ”double-bar-relled” question.

• The respondent should be competent at the question.

You should continuously pay attention to whether respondents are able to answer the question reliably, whether they have enough information to answer it.

In document Textbook of Nursing Science (Pldal 97-102)