Design and Analysis of Clinical Nursing Research Studies

Colin R Martin, David R Thompson

• Publisher: Routledge
• Year: 2002
• eBook ISBN: 9781134589715
• Edition: 1
• Topic: Medicine
• Subtopic: Health Care Delivery

This invaluable text on the design and analysis of clinical research studies explains the basis of experimental design and statistics in a way that is sensitive to the clinical context in which nurses work. It uses data from actual studies to illustrate how to: *design the study
*use and select data
*present research findings
*use a computer for statistical analysis.The scope of the study designs and associated statistical techniques covered in the book allow both the beginning nurse researcher and the more seasoned professional nurse investigator to approach a research study with confidence and optimism. The authors show how qualitative data can be approached quantitatively, what the advantages of this are from the nursing viewpoint, and how quantitative methodology can help nurses to develop a common research language with other disciplines involved in patient care.

Chapter 1

Clinical research in nursing

Introduction

Developments in nursing as a research-based profession have continued apace over recent years. A key issue in terms of developing an evidence base in any discipline is the nature of the dominant philosophy underpinning the particular profession. The philosophy underlying the discipline must inevitably inform the direction and style of knowledge seeking undertaken and the accompanying methodology to execute this pursuit of evidence. In many areas of research, this does not present any real practical difficulties. In physics research, for example, a fundamental tenet underlying the accumulation of evidence is to adopt a scientific approach to the research enterprise, the methodology for which is clearly circumscribed within established design parameters. Nursing, on the other hand, presents us with a special case. The philosophy underlying nursing in terms of developing a research identity is not at all clear. This is due, in part, to the lack of clarity about what nursing represents regarding a dominant ideology; put simply, there is no clear consensus as to whether nursing is an ‘art’ or a ‘science’ (Thompson, 1998). Inevitably, this has led to the adoption of a wide variety of research methodologies applied to nursing research, all of which are legitimised by the presence of an ambiguous nursing ideology. Methodologies used in nursing research can be grouped into two distinctive camps, these being qualitative and quantitative.

Qualitative and quantitative approaches to nursing research

The distinction between qualitative and quantitative research in nursing is representative of a complex issue which will be simplified for the purposes of clarity and brevity. Broadly speaking, qualitative research takes the view that attempting to quantify the salient aspects of human experience in relation to health and disease represents a reductionist approach that cannot encapsulate the real matter of experience. As an example, a qualitative approach to understanding the process of recovery from surgery would focus on the context within which the individuals find themselves, the personal qualities of significant contacts during the period of treatment and the individuals’ prior life experiences. It can be seen that these concepts would be difficult to measure numerically and to enter into a data analysis. This leads us then to a fundamental limitation of qualitative research in nursing, a lack of generality, replication and knowledge generation in terms of new theory.

Quantitative methodology addresses the criticisms of qualitative methodology by conducting studies that provide data in numerical form that can be entered into data analysis. Quantitative studies attempt to investigate a research question or ‘hypothesis’ by focusing on discrete and measurable aspects of an area of clinical and theoretical interest. The research question is therefore quantified. So, using the earlier example, in a quantitative study of recovery after surgery, we would be interested in measurable factors such as length of time in hospital, amount of anaesthetic given, type of anaesthetic given and number of previous procedures incurred by the patient. Further, we would want to be able to generalise the findings; so, using quantitative methodology, we may wish to compare different groups of patients defined by a particular characteristic such as diagnostic criteria and to compare the groups to investigate the differential effects of a treatment intervention. As the methodology used would be well circumscribed, replication of the study would be relatively simple.

A quantitative focus

It could be argued that an emphasis on the more qualitative aspects of nursing research in an era of evidence-based practice has led to an under-representation of quantitative methodology in present research endeavours. It is therefore crucial that nurses involved in quantitative research have available a text that covers experimental design and statistics in a context-sensitive manner. The following chapters will describe the design and statistical analysis of clinical nursing research studies using data from actual studies that have been conducted in the field. The breadth of the studies described will cover both general and mental health nursing research paradigms including data from multidisciplinary studies with a strong nursing contribution in the design and execution of the study. Although most statistical analyses are these days carried out on computer packages, these packages being discussed in Chapter 13, the emphasis within this text will be on the rationale for the tests used and their working out with reference to standard statistical tables included in the appendices. It is hoped that this approach will give the reader a more in-depth understanding of the philosophy and cerebral mechanics underpinning the tests, as well as providing the researcher with statistical independence when computer-based statistical packages are not available. The applied nature of the statistical tests described in this volume has been emphasised to indicate the ready applicability of the testing techniques to other appropriate clinical domains. The scope of the study designs and the associated statistical techniques covered has been intended to allow both the novice nurse researcher and the more seasoned professional nurse investigator to approach the design, execution and analysis of a study with confidence and optimism.

A common research language

An extremely useful by-product of nurses becoming familiar and proficient in the design, execution and analysis of quantitatively biased research projects is that it facilitates more readily collaborative ventures with other health disciplines such as medicine, clinical psychology, physiotherapy, etc. The reason for this salience is twofold. First, health disciplines other than nursing, and in particular medicine, have traditionally focused on quantitative approaches to the analysis of clinical material, this being dichotomous to the qualitative research approach which has been a traditional strength of the nursing profession. There is then a methodological issue involved in the failure to communicate between the two most influential professions involved in evidence-based patient care. Second, and more importantly for the future development of nursing as a research-based profession, the launch of the National Health Service (NHS) research and development (R&D) strategy Research for Health (Department of Health, 1991) stimulated increased interest in health-related research.

The NHS R&D strategy

This created new demands on the research community within and beyond the NHS and new opportunities for NHS staff to become involved in R&D. The importance of a sound research basis for nursing has been re-emphasised in the new strategy for nursing Making a Difference (Department of Health, 1999).

The aim of the NHS R&D strategy is to improve the health of the nation by promoting a knowledge-based health service in which clinical, managerial and policy decisions are based on sound research evidence. Although nursing needs to generate and extend its own body of knowledge through research, it is important that nursing research is seen as part of the broader scientific research community. The NHS R&D initiative offers nurses an ideal opportunity to engage in truly collaborative multidisciplinary and multiagency research, to demonstrate to others their own unique skills and distinct approach to research and to learn about the methods and approaches of others (Thompson, 1999). Further, the cross-fertilisation of skills from complementary disciplines allows freedom to develop more sophisticated research studies to address complex clinical questions previously occluded to single-discipline scientific enquiry (Martin et al., 1998).

The research question

The key to the design of a good quantitative nursing study is to ask the right research question. Although this intrinsically crucial aspect of the research process may seem intuitively simple, a review of published literature across all disciplines will demonstrate that even seasoned researchers have not consistently formulated an appropriate research question and, consequently, have produced a flawed and lacklustre piece of work. Key identifying clues are often the use of ‘esoteric’, and usually inappropriate statistical tests to attempt to save a poorly conceived and/or executed study and a discussion section that bears no line of continuity to the introduction. Time spent thinking and discussing the question that we wish to address will lead the researcher to the correct methodology to implement the study. Essentially then, all quantitative research must begin with an appraisal of two key questions: ‘I wonder if…’ and ‘I wonder why…’. These fundamental questions represent the skeleton of research design, the flesh being the hypotheses (formal research questions) and the skin being the methodology. Hopefully, this analogy illustrates the relative importance of formulating the research question; attempting to clothe a body of work with abstract statistics will in no way present a boneless and meatless study in a positive light. Time invested in considering the research question will pay dividends in getting work published and disseminated to the wider academic and clinical community.

Quantitative measurement of qualitative issues

Although much debate has taken place on the divide between qualitative and quantitative research in nursing, as a discipline nursing has come to represent a gestalt of the two approaches. This is easy to see why: patient contact focuses on the interpersonal aspects of health and ill-health in the recovery process, surely a qualitative issue. However, to influence policy, facts are required, and in terms of health care delivery and evaluation these are invariably based on numeric, and therefore quantitative, values. Considering qualitative and quantitative approaches as primary colours on the ‘research palette’, it is crucial to access all the ‘shades of colour’ between these extremes. One approach, the focus of this volume, is the quantification of qualitative data. Consider, as an example, clinical depression. There are a number of methods of assessing a person’s depression. We may have an opinion about someone being depressed or not depressed; alternatively, we could conclude that the person is a little depressed or very depressed. The opinion is a subjective qualitative judgement, which may or may not be true; certainly, there is no reliable means of telling for sure. A diagnosis of depression is much the same, essentially a professional opinion. The diagnosis, like the opinion, is a qualitative judgement, prone to error and demonstrative of a restricted range of possibilities, i.e. depressed/not depressed or a limited range of labels describing levels of depression. It is, however, possible to measure depression quantitatively using a self-report measure such as the Beck Depression Inventory (Beck et al., 1963).

This quantitative measurement of the qualitative experience of depression is enormously useful: it provides a larger range of possible levels of depression than the opinion or diagnosis; it allows comparison with a normal (i.e. non-depressed) population; it allows changes over time in levels of depression within an individual to be assessed accurately, quickly and reliably; it can inform treatment decisions and act as a measure of treatment efficacy; and it can be related to other measures of psychopathology, such as anxiety, to gain a more complete picture of the aetiology of depression within the individual. Further, comparison of defined groups of individuals on scores obtained on such quantitative measures can provide insights into the nature of depression and allow research questions to be asked that provide more answers to the causes of depression. This applies not only to psychological data, such as those obtained by the Beck Depression Inventory (BDI), but also to biological, biochemical, sociological and all behavioural data. The overarching theme between such diverse areas of speciality is that the data can be reduced to a numeric value. The advantage of obtaining a numerical value is that a statistical analysis can readily be made to tell whether there are significant differences between discrete treatment groups. This approach seeks to provide the researcher with factual evidence with which to test a hypothesis.

Hypothesis testing

If we wanted to compare the efficacy of two different treatments of any condition, it is desirable that the choice of the two treatments is based on a review of the available literature. This review would then inform us that we could reasonably expect one treatment to be more effective than the other based on this prior evidence. This expectation, based on the prior evidence that has been reviewed from the literature search, forms the basis of the experimental design and methodology that we would propose to use. This expectation regarding an outcome is common to all scientific exploratory studies and is known as the hypothesis. The hypothesis provides the formal focus from which the structure of the experimental method is defined. From the hypothesis, it is clear that predictions can be made that two treatments will differ in efficacy. The context within which health service research is constrained usually means that one new treatment or intervention is ’tested’ against the established treatment/intervention. Traditionally, the epitome of this approach is embedded within the methodology of the randomised controlled trial.

The randomised controlled trial

The randomised controlled trial (RCT) has been the cornerstone of medical research, clinical trials in particular, for many years and has more recently become popular in testing hypotheses relating to nursing care. The basic premise of the RCT is that patients in one cohort are randomly allocated to two groups. One group, known as the control group, will receive the standard treatment; the second group, known as the treatment group, will receive the novel treatment. The efficacy of the new treatment will then be evaluated by a comparative analysis with the standard established treatment.

There are extreme limitations to the RCT approach to clinical research in nursing. It is likely that most nurse-led investigations will not be focusing on the effect of a drug treatment but will be examining an aspect of clinical nursing care. We may be interested in differences in recovery rates between two types of patient. For example, we may feel that mixed-sex wards have a greater impact on female patients’ perceived level of anxiety than on male patients’ perceived level of anxiety. It would not be appropriate, or possible, to conduct an RCT in this instance because it is sex that is the issue of interest within this particular environment (a mixed-sex ward), and it is impossible, after all, to randomise sex! The experimental designs that are applicable to these type of studies can be much more sophisticated than those of the RCT, the pay-off to these more elaborate studies being that the results can lay the foundations to a more theoretically meaningful examination of the issues central to the hypothesis.

Summary

This chapter has introduced the merits of the quantitative approach to clinical nursing research. Issues such as using quantitative methodology to develop a common research language with other disciplines involved in patient care have been discussed, as has the issue of quantifying qualitative data. The notions of the research question and the research hypothesis have been introduced. Limitations of the traditional medical approach to clinical research, in the form of the RCT, have been raised.

Chapter 2

Designing a clinical study

Introduction

Having decided upon a hypothesis to be tested, the next crucial phase is the design of the clinical study. To test the hypothesis that two distinct patient groups or two alternative treatments will be different, two vital factors need to be considered. First, it is crucial to consider the groups/treatments that are being compared, and, second, what is to be measured; essentially, the numbers which will be entered into the statistical analysis. Because the researcher is exercising experimental control over the content of the patient groups or the types of treatment being compared, it is logical to conclude that the researcher is manipulating this particular variable. The variable that is being manipulated, such as group type or treatment type, is known as the independent variable (IV).

Considering the example of the mixed-sex ward in Chapter 1, the two groups of patients that are being compared are male patients and female patients. The IV in this instance is therefore sex type. Using the same example, having defined the two groups, it is then necessary to measure the effect of this manipulation; in this case, a measure of anxiety. However anxiety is measured, for example by a self-report questionnaire or by a biochemical measure such as serum cortisol, the score of anxiety on the questionnaire or the absolute amount of cortisol that is measured is known as the dependent variable (DV). This is obviously quite logical as the DV is literally dependent on the IV. The measure of anxiety expressed by each group is then compared by the use of a statistical test to find out whether male and female patients are significantly different in the terms of this measure of anxiety. It could be concluded that manipulating the IV caused the difference (if any were observed) in anxiety between the two groups. This would of course be exactly the same in the case of the IV being treatment type and the DV being a measure of outcome. Once the researcher has decided on a hypothesis to be tested and the choice of IV and DV, the methodology can be formulated within the context of the research design.

Confounding variables

Having decided on the choice of IV and DV, the researcher will need to conduct a study that has a rigorous methodology. This rigour relates to the need to show that whatever differences are seen in the DV are purely a result of the manipulation of the IV and are not the result of some other uncontrolled or confounding variable. Confounding variables are the death knell of many a promising rese...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Dedication
• Table of Contents
• List of illustrations
• Acknowledgements
• Foreword
• 1 Clinical research in nursing
• 2 Designing a clinical study
• 3 The qualities and power of data
• 4 Study sample size calculation
• 5 Choice of statistical tests
• 6 Between-subjects design and tests
• 7 Within-subjects design and tests
• 8 Mixed-group design and tests
• 9 Correlation and association between variables
• 10 Single-sample studies and analysis
• 11 Confidence intervals
• 12 Graphical presentation of data
• 13 The use of computers for statistical analysis
• 14 Considerations when designing a clinical study
• Appendix: Statistical tables
• Index