Qualitative research

Qualitative research is an increasingly popular method of enquiry in biomedical, clinical and behavioural research. Once regarded as the preserve of social scientists and psychologists, qualitative methods have entered the mainstream of epidemiology and clinical research, as evidenced by the publication of a series of papers in the British Medical Journal (Britten 1995; Mays and Pope 1995; Pope and Mays 1995; Pope et al. 2000). The qualitative methods to be described in this chapter offer a scientific approach to understanding and explaining the experiences, beliefs, and behaviour of defined groups of people. The contrasting features and the complementary roles of qualitative and quantitative methods of enquiry will be described. While the majority of chapters in the present volume are concerned with research methods designed to answer questions such as ‘how many?’ or ‘how frequently?’, qualitative methods enable us to explore the ‘why?’, ‘what?’, and ‘how?’ of human behaviour. Since the aim is to understand the meaning of the phenomena under study from the perspective of the individuals concerned, the direction of enquiry is guided more by respondent than researcher. This approach is particularly appropriate to complex phenomena such as the range of beliefs that underlie illness behaviour and the aspects of health care that matter to different service users. Qualitative enquiry would focus on identifying beliefs and describing the circumstances that surround particular behaviours, while quantitative research would focus on measurable characteristics of the sample and the frequency and outcome of their behaviour. An example of the contribution of the two methodological approaches is the study of variations in treatment of depression in older people. Epidemiological studies in the community and in primary care settings have found that the prevalence of depression in older adults far exceeds the prevalence of the disorder among those consulting their general practitioners. To identify the factors associated with this disparity, qualitative researchers would set out to explore the reasons why older people with depression do and do not present their symptoms to the GP. The aim would be to describe the range of beliefs about depression among attenders and non-attenders. The quantitative approach would involve establishing the strength of associations between personal characteristics, external factors, and behaviour of older people with depression. It is clear from this example that both approaches are complementary in identifying the nature of the disparity. Qualitative research is based on the premise that each individual's experience is unique and the beliefs that underlie illness behaviour can only be measured once identified and described from a variety of individual perspectives. When information of this type is combined with data on prevalence and variable risk, more appropriate services and outcome measures can be developed.

Genetic epidemiology 2: molecular genetics

The previous chapter has focused on methods for identifying familial clustering of disorders or traits, and on methods for distinguishing between shared genetic and environmental influences. The primary objective for this chapter is to outline techniques for identifying specific genes responsible for an observed phenotype. The theoretical basis of complex and quantitative traits was established many decades ago. However practical methods for the efficient molecular analysis of the human genome have only recently emerged. Alongside these developments, the molecular genetic analysis of human disorders has moved at a rapid pace. Molecular genetics has focused on single gene disorders with great success, whereas for complex psychiatric disorders, few genetic risk factors have been identified. However the tools used by the complex disorder geneticist have evolved rapidly in the last few years and better strategies and statistical methods continue to appear. This chapter outlines some established and novel approaches to the analysis of the genetics of complex human disorders. A basic understanding of genetical statistics will be useful.

Health economics for psychiatric epidemiology

This chapter examines the interface between psychiatric epidemiology and health economics, particularly in relation to mental health service evaluation. We discuss the issues inherent in conducting an economic evaluation and conclude with a summary of the applications of economic analyses.

Genetic epidemiology 1: behavioural genetics

Behavioural genetics is the study of the genetic basis of behavioural traits including both psychiatric disorders and ‘normal’ personality dimensions. Behavioural genetics derives its theoretical basis from population genetics. Soon after the laws of Mendelian inheritance were re-discovered in 1900, the implications of these laws on the genetic properties of populations were worked out. Such properties include segregation ratios, genotypic frequencies in random mating populations, the effect of population structure and systems of mating, the impact of selection, the partitioning of genetic variance, and the genetic correlation between relatives. Some appreciation of population genetics is necessary for a deep understanding of behavioural genetics. Because of the complexity of behavioural traits, genetic factors cannot be regarded in isolation, or as static. Instead, it is important to consider: (i) the relative contributions of genetic and environmental factors, (ii) the interplay between genetic and environmental factors, and (iii) the changing role of genetic factors in different stages of development from infancy to old age. The major study designs in behavioural genetics will be discussed in this chapter, namely family studies, twin studies, and adoption studies. Behavioural genetics, augmented by molecular genetics has the potential to identify specific genetic variants which influence behaviour. This will be considered in detail in Chapter 14. Mendelian inheritance Gregor Mendel first demonstrated the genetic basis of biological inheritance by studies of simple all-or-none traits in the garden pea. These traits were particularly revealing because they were completely determined by the genotype at a single chromosomal locus. Diseases caused by genetic mutation at a single locus are commonly called Mendelian or single-gene disorders. A dominant disorder is expressed when an individual has one or two copies of the mutant allele, whereas a recessive disorder is expressed only when both alleles at the locus are the mutant variant. Examples of Mendelian disorders of clinical significance in psychiatry are Huntington's disease and fragile X syndrome. Mendelian disorders tend to be relatively rare because they are usually subjected to severe negative selective pressure, due to their increased mortality. Most common disorders and continuous traits of interest in psychiatry have an aetiology involving multiple genetic and environmental factors. Categorical and dimensional traits Behavioural genetics is rooted in both psychiatry and psychology. Psychiatrists traditionally adopt a medical model where diseases are defined as categorical entities and diagnoses are either present or absent. Psychologists on the other hand prefer quantitative measures of cognitive ability, personality and other traits. The methodology of behavioural genetics research reflects this duality, although there is a trend to integrate the two approaches, especially for traits such as anxiety and depression where both diagnostic criteria and quantitative measures exist.

Cross-sectional surveys

This chapter considers the strengths and limitations, and the uses and abuses of cross-sectional surveys in psychiatric epidemiology. Certain basic aspects of research methodology; the concept of the base population, sampling strategies, representativeness, the problem of non-response and the practical logistics of population-based research are introduced here, although they are in practice equally relevant to other study designs. The chapter also introduces the problem of bias, arising both from non-response and misclassification. In conclusion, we review major surveys of psychiatric morbidity in a historical context, highlighting methodological developments and discussing the yield of information to be gleaned from them.

Measurement in psychiatry

The science of the measurement of mental phenomena (psychometrics) is central to quantitative research in psychiatry. Without appropriate, accurate, stable, and unbiased measures, our research is doomed from the outset. Much effort has been expended over the last 40 years in the development of a bewildering array of assessments. Most of our measurement strategies are based on eliciting symptoms, either by asking the participant to complete a self-report questionnaire, or by using an interviewer to question the participant. Some are long, detailed, and comprehensive clinical diagnostic assessments. Others are much briefer, designed either to screen for probable cases, or as scalable measures in their own right, of a trait or dimension such as depression, neuroticism or cognitive function or as measures of an exposure to a possible risk factor for a disease. Researchers in other medical disciplines sometimes criticise psychiatric measures for being vague or woolly, because they are not based on biological markers of pathology. For this very reason, psychiatry was among the first medical disciplines to develop internationally recognized operationalized diagnostic criteria. At the same time the research interview has become progressively refined, such that the processes of eliciting, recording, and distilling symptoms into diagnoses or scalable traits are now also highly standardized. These criticisms are therefore largely misplaced. Thanks to the careful construction and extensive validation of the better established measures in psychiatric research we can now afford to be slightly more confident of their appropriateness, accuracy, and stability than would be the case even for some biological measurements. This confidence is based on our understanding of the validity and reliability of our measures.

Psychiatric epidemiology— looking to the future

The daunting objective for this chapter is to summarize issues which face the emerging specialty of psychiatric epidemiology, and to suggest broad directions for future research. Some of these have already been highlighted and we are grateful to contributing authors for providing their opinions as to the ‘state of play’, both in their own contributions and in communications solicited with respect to this chapter. Although the editorial team take responsibility for what is written here, we hope that it can be taken to reflect a wider body of opinion in this field. The issues raised are not intended to be exhaustive, although we hope that any specific omissions can be reasonably included within one or other of the broad themes identified. Psychiatric epidemiology is a relatively young research specialty. This creates both problems and opportunities. A problem is that it has ‘grown up’ heavily influenced by prevailing paradigms from other older fields—principally general epidemiology (regarding methodologies) and other areas of psychiatric research (regarding systems of classification and diagnosis). These are not automatically appropriate or helpful and may instead be a source for difficulties encountered in research. An advantage however for a young specialty is that it can perhaps more easily discard the trappings of tradition as it seeks to make its way in the world. Current issues will be considered under three broad headings. First, the need for new methodologies will be considered. Next, interfaces will be summarized both between psychiatric epidemiology and other specialties/agencies and within the specialty itself. Finally possible new directions for psychiatric epidemiology will be considered.

Statistical methods in psychiatric epidemiology 2: an epidemiologist’ s perspective

We hope that these two chapters, while providing students with greater confidence in approaching the analysis of their data sets, will also have raised as many questions as they have provided answers. It should by now be evident that there is no single, set, correct way to analyse a given data set; many will argue with some of the approaches advocated in these chapters. The important thing is for students to be aware of the diversity of methods currently available, and to proceed judiciously in the analysis and inferencing of their data, constantly aware of the strengths and limitations of the techniques that they are using. Also, it is important to recognize that biostatistics is a constantly and rapidly evolving discipline. The introduction of logistic regression in the 1970s revolutionized modern epidemiology, influencing the design of our studies as well as the methods used to analyse them. The more recent development of multi-level modelling is likely to have a similarly profound effect upon the type of research questions that we formulate, as well as the designs that we use to test these new hypotheses. Statistical methods are therefore not just the tools that statisticians, working with epidemiologists use to analyse data. They also, as they develop drive the research agenda and influence all aspects of methodology. Ever increasing collaboration between biostatisticians, epidemiologists, and clinical researchers is therefore essential if the full creative potential of this momentum is to be realized.

Statistical methods in psychiatric epidemiology 1: a statistician's perspective

In this chapter we shall look at methods of statistical analysis used in psychiatric epidemiology. We shall focus on the issues which arise in trying to make sense of a small real dataset. We assume that readers are already familiar with the concepts of confidence interval, means, correlations, and odds ratios. What makes statistical analysis in psychiatric epidemiology different? We have given more space to methods dealing with measures than would be usual in a general text on epidemiology. This is quite deliberate. What makes psychiatric epidemiology different is the emphasis on measurement. By contrast most outcomes in medical statistics were historically binary (usually dead vs. alive). This is beginning to change (note for instance the increased interest in measuring quality of life almost everywhere). Of course psychiatry as a branch of medicine has used the concept of diagnosis freely, and so naturally we also include methods for handling such binary outcomes. We start by discussing methods for predicting an outcome, whether a measurement or a binary outcome. We then discuss a group of methods used for exploring the relationship between groups of variables where there is no single outcome.

Research synthesis: systematic reviews and meta-analysis

The process of synthesizing data from different studies is known as metaanalysis. The techniques were developed in the social sciences and only recently applied to medical research. There has been intense debate about the validity of the process and its potential contribution to research. In medicine the widest application of research synthesis techniques has been with intervention studies. In particular the Cochrane Collaboration1 has promoted the use of systematic reviews and meta-analysis to evaluate medical treatments. Recently there has been increasing attention paid to meta-analysis with other types of study (Altman 2001). This overview will focus on intervention studies, and, after describing the uses and limitations of research synthesis and the particular issues arising for psychiatric researchers, will illustrate the stages in conducting a systematic review or meta-analysis. Definitions For the purposes of this chapter systematic reviews will be taken as referring to reviews which aim to achieve comprehensive coverage of the relevant literature and meta-analysis refers to the statistical process of combining quantitative data from different studies. The need for research synthesis (1) The exponential increase in medical research over recent decades makes it impossible for doctors to have a comprehensive knowledge of research in every area relevant to their practice. (2) By virtue of bringing a fresh perspective to an area, systematic reviews may be able to reach a more objective view of the evidence. (3) Health economists and policy makers need an overview of research and a reliable estimate of efficacy to facilitate the process of resource allocation. (4) Collation of research in different settings is valuable in order to obtain a picture of the range of action of a particular intervention. (5) Many studies are not large enough to detect small effects that may be clinically useful. Combining data enhances the power of the analysis to detect such effects. (6) Systematic collation of evidence indicates which areas require more research.