●간호학 연구에서의 질적내용분석 : 개념, 절차, 결과의 신뢰성을 높이기 위한 방법

간호학 연구에서의 질적내용분석 : 개념, 절차, 결과의 신뢰성을 높이기 위한 방법

Qualitative content analysis in nursing research: concepts, procedures and measures to achieve trustworthiness

U.H. Graneheim*, B. Lundman

Department of Nursing, Umea University, Umea 90187, Sweden

Introduction

내용분석은 처음에는 ‘the objective, systematic and quantitative description of the manifest content of communication’으로 시작되었으나 시간이 지남에 따라서 잠재적 내용(latent content)의 분석까지 포함하기 시작했다. 많은 저자들이 내용분석에 대해 다룬 바가 있으며, 1950년대에에는 주로 정량적 측면을 다루고 있었다. 현재로서 두 가지 방법으로 주로 활용되는데 하나는 정량적 접근법이고, 다른 하나는 정성적 접근법이다. 간호학에서의 질적내용분석은 다양한 자료와 다양한 해석의 깊이로 활용되어왔다.

Initially content analysis dealt with ‘the objective, systematic and quantitative description of the manifest content of communication’ (Berelson, 1952, p. 18) but, over time, it has expanded to also include interpretations of latent content. Many authors, from a variety of research traditions, have addressed content analysis (for example, Berelson, 1952; Krippendorff, 1980; Findahl and Höijer, 1981; Woods and Catanzaro, 1988; Downe-Wamboldt, 1992; Burnard, 1991 and Burnard, 1996; Polit and Hungler, 1999). The first descriptions date from the 1950s and are predominately quantitative. Currently, two principal uses of content analysis are evident. One is a quantitative approach often used in, for example, media research, and the other is a qualitative approach often used in, for example, nursing research and education. Qualitative content analysis in nursing research and education has been applied to a variety of data and to various depths of interpretation (for example, O’Brien et al., 1997; Latter et al., 2000; Berg and Welander Hansson, 2000; Söderberg and Lundman, 2001).

질적내용분석의 의미, 개념의 활용, 절차, 해석에 있어서 아직 여러 의견과 풀리지 않은 이슈들이 있다. 이러한 다양성은 다양한 역사적 관점에서 비롯되곤 한다.

A review of literature based on common databases (Cinahl, Medline and Sociological Abstracts) as well as references from articles and books shows different opinions and unsolved issues regarding meaning and use of concepts, procedures and interpretation in qualitative content analysis. The diversities can be understood partly from a historical point of view and partly from various beliefs of the nature of reality among researchers.

이 논문은 '진실은 다양한 방식으로 해석가능하며, 진실을 어떻게 이해하는가는 주관적인 해석에 달려있다'라는 것을 전제로 한다. 질적연구는 담론과 관찰에서 얻은 자료를 바탕으로 하며, 연구자와 참가자의 이해와 협력을 필요로 한다. 왜냐하면 면담이나 관찰에서 얻은 텍스트는 상호적이며, 맥락적이고, 가치의존적이기 때문이다. 따라서 텍스트가 항상 다양한 의미를 가진다는 가정과, 텍스트에는 항상 어느 정도의 해석의 여지가 있다는 것이 우리의 기본 가정이다. 이는 질적내용분석 결과의 신뢰성을 논의할 때에 중요한 부분이다.

An assumption underlying our paper is that reality can be interpreted in various ways and the understanding is dependent on subjective interpretation. Qualitative research, based on data from narratives and observations, requires understanding and co-operation between the researcher and the participants, such that texts based on interviews and observations are mutual, contextual and value bound (Lincoln and Guba, 1985; Mishler, 1986). Thus, our presumption is that a text always involves multiple meanings and there is always some degree of interpretation when approaching a text. This is an essential issue when discussing trustworthiness of findings in qualitative content analysis.

또 다른 이슈는 양적연구의 전통이 여전히 질적내용분석에도 주요하게 작용하고 있다는 점이다. 특히 '신뢰성'을 다루는 개념에 대해서 저자와 독자 사이에 혼란과 불확실성을 야기하는 원인이 된다.

Another issue is that concepts within the quantitative research tradition still predominate when describing qualitative content analysis (for example, Krippendorff, 1980; Burnard, 1991; Downe-Wamboldt, 1992), especially the use of concepts describing trustworthiness. This causes confusion and paradigmatic uncertainty among authors and readers of scientific papers.

이 논문의 목적은 세 가지이다.

The purpose of this paper was threefold: 

first, to provide an overview of concepts of importance related to qualitative content analysis in nursing research; 

second, to illustrate the use of concepts related to the research procedure; and 

third, to address measures to achieve trustworthiness.

전반적 개념 Overview of concepts

우리는 먼저 문헌에서 다양한 개념이 어떤 식으로 활용되는가를 보고, 우리의 입장을 이후에 밝히고자 한다. 

The following provides an overview of concepts related to qualitative content analysis and is to be seen as a contribution to a debate rather than an endeavour to find consensus. First, we present various uses of concepts found in the literature, and then we give reasons for our stance. The concepts are manifest and latent content, unit of analysis, meaning unit, condensing, abstracting, content area, code, category and theme.

질적내용분석을 수행할 때 가장 기본적인 이슈는 manifest content에 집중할 것이냐 latent content에 집중할 것이냐 하는 것이다. manifest content와 latent content 모두 해석과 관련된 것이지만, 그 해석은 깊이나 추상화 정도에 있어서 차이가 있다.

A basic issue when performing qualitative content analysis is to decide whether the analysis should focus on manifest or latent content. 

• Analysis of what the text says deals with the content aspect and describes the visible, obvious components, referred to as the manifest content ( Downe-Wamboldt, 1992; Kondracki et al., 2002). 
• In contrast, analysis of what the text talks about deals with the relationship aspect and involves an interpretation of the underlying meaning of the text, referred to as the latent content ( Downe-Wamboldt, 1992; Kondracki et al., 2002). 

Both manifest and latent content deal with interpretation but the interpretations vary in depth and level of abstraction.

내용분석을 할 때 가장 기본적인 결정 중 다른 하나는 '분석의 단위'를 고르는 것이다. 문헌에서 '분석의 단위'란 연구의 대상이 될 수 있는 다양한 것들을 말한다. 어떤 저자들은 분석의 단위를 인터뷰나 일기라고 보기도 하며, 연구 또는 주제에 할당된 '공간'을 보는 사람도 있다. 텍스트의 일부는 추상화되고 코딩되며, 모든 단어와 구문은 transcript로 기록되어 분석의 단위가 될 수 있다. 우리는 가장 적절한 분석의 단위는 '인터뷰 전체' 또는 '관찰 프로토콜'로서 '분석 과정에서 전체라 부르기에 충분하면서, 의미단위라는 맥락으로서 마음에 담아둘 수 있을 만큼 작은 것'이라고 제안하고자 한다.

One of the most basic decisions when using content analysis is selecting the unit of analysis. In the literature, unit of analysis refers to a great variety of objects of study, for example, a person, a program, an organisation, a classroom or a clinic ( Mertens, 1998), or a community, state or nation ( Patton, 1987). Other authors have considered the unit of analysis as interviews or diaries in their entity, and the amount of space allocated to a topic or an interaction under study ( Downe-Wamboldt, 1992). Parts of the text that are abstracted and coded ( Weber, 1990), or every word or phrase written in the transcript ( Feeley and Gottlieb, 1998), have also been considered as units of analysis. We suggest that the most suitable unit of analysis is whole interviews or observational protocols that are large enough to be considered a whole and small enough to be possible to keep in mind as a context for the meaning unit, during the analysis process.

의미단위란, 단어 또는 문구의 무리(constellation)로서, 같은 핵심적 의미로 연관되는 것이며 content unit 또는 coding unit, idea unit, textual unit, a keyword and phrase, unit of analysis, theme 이라 불리기도 한다. 우리는 의미단위를 내용과 맥락을 통해서 다른 것들과 서로 연결될 수 있는 측면을 가지는 단어/문장/문단으로 본다.

A meaning unit, that is, the constellation of words or statements that relate to the same central meaning, has been referred to as a content unit or coding unit ( Baxter, 1991), an idea unit ( Kovach, 1991), a textual unit ( Krippendorff, 1980), a keyword and phrase ( Lichstein and Young, 1996), a unit of analysis ( Downe-Wamboldt, 1992), and a theme ( Polit and Hungler, 1991). We consider a meaning unit as words, sentences or paragraphs containing aspects related to each other through their content and context.

문헌연구에서 텍스트를 짧게 줄이고자 할 때에는 '환원', '추출', '응축' 등의 개념이 필요하다. 우리는 '응축'이라는 단어를 선호하며, 이는 핵심은 유지하면서 내용은 줄이는 것을 의미한다.

In the literature, shortening the text includes the concepts of reduction (Findahl and Höijer, 1981), distillation (Cavanagh, 1997) and condensation (Coffey and Atkinson, 1996). 

• Reduction refers to decreasing the size, but it indicates nothing about the quality of what remains. 
• Distillation deals with the abstracted quality of a text, which we see as a further step in the analysis process. 
• We prefer condensation, as it refers to a process of shortening while still preserving the core.

응축된 텍스트가 추상화되는 단계를 'aggregation' 또는 '상위 제목 아래로 그룹핑'이라고 한다. 우리는 '추상화'라는 단어를 권고하는데 왜냐하면 이 단어는 논리적 상위단계의 해석과 묘사를 강조하기 때문이다.

The process whereby condensed text is abstracted has been called aggregation (Barrosso, 1997) and ‘grouping together under higher order headings’ (Burnard, 1991, p. 462). We suggest abstraction, since it emphasises descriptions and interpretations on a higher logical level. Examples of abstraction include the creations of codes, categories and themes on varying levels.

특정 이슈만을 다루는 텍스트의 일부분을 '도메인' 또는 '개략적 구조', '클러스터', '내용영역' 이라고 한다. 우리는 content area라는 단어를 선호하는데, 이는 해석을 최소화한 내용의 특정 부분을 강조하기 때문이다. content area는 그 문헌의 이론적 가정을 토대로 하고 있는 텍스트의 한 부분, 또는 인터뷰나 관찰가이드 안에서 특정 주제를 다루는 텍스트 부분을 의미한다.

Parts of a text dealing with a specific issue have been referred to as a domain or rough structure (Patton, 1990), a cluster (Barrosso, 1997) and a content area (Baxter, 1991). We prefer content area since it sheds light on a specific explicit area of content identified with little interpretation. A content area can be parts of the text based on theoretical assumptions from the literature, or parts of the text that address a specific topic in an interview or observation guide.

의미단위에 붙이는 라벨을 'code'라고 한다. '코드'의 의미와 활용에 대해서는 일정 정도의 일치를 보인다. 

The label of a meaning unit has been referred to as a code. There seems to be agreement in the literature about the use and the meaning of a code. 

• According to Coffey and Atkinson (1996, p. 32) ‘codes are tools to think with’ and ‘heuristic devices’ since labelling a condensed meaning unit with a code allows the data to be thought about in new and different ways. 
• A code can be assigned to, for example, discrete objects, events and other phenomena, and should be understood in relation to the context.

카테고리를 만드는 것은 질적내용분석의 핵심 특징 중 하나이다. 카테고리는 공통성을 가지는 컨텐츠들의 그룹이라고 할 수 있다. Pattron은 카테고리를 내적으로 동질적이고 외적으로 이질적인 것 으로 묘사한 바 있다. Krippendorff는 카테고리는 완결성을 가지면서 상호베타적이어야 한다고 강조했다. 이는 연구목적과 관련된 어떤 자료도 카테고리 부족으로 배제되어서는 안된다는 뜻이다. 더 나아가서 어떤 데이터두 두 개의 카테고리에 포함되거나 하나 이상의 카테고리에 들어갈 수 있으면 안된다. 그러나 인간경험의 복잡성으로 인해서 언제나 상호배타적인 카테고리를 만드는 것이 가능한 것은 아니다. 카테고리는 '무엇?(what?)'이라는 질문에 대답할 수 있어야 하며, 코드 전체에 걸쳐 어떤 thread로서 발견되어야 한다. 카테고리는 내용의 묘사적 수준(descriptive level)을 주로 다루기 때문에, manifest content를 주로 다룬다고 할 수 있다. 카테고리는 종종 다수의 하위카테고리, 하위하위카테고리를 가지기도 한다. 하위카테고리는 분류와 추상화단계를 거쳐서 카테고리를 만들 수 있고, 카테고리가 하위카테고리로 분류될 수도있다.

Creating categories is the core feature of qualitative content analysis. A category is a group of content that shares a commonality ( Krippendorff, 1980). Patton (1987) describes categories as internally homogeneous and externally heterogeneous. Krippendorff (1980) emphasises that categories must be exhaustive and mutually exclusive. This means that no data related to the purpose should be excluded due to lack of a suitable category. Furthermore, no data should fall between two categories or fit into more than one category. However, owing to the intertwined nature of human experiences, it is not always possible to create mutually exclusive categories when a text deals with experiences. A category answers the question ‘What?’ ( Krippendorff, 1980) and can be identified as a thread throughout the codes. As we see it, a category refers mainly to a descriptive level of content and can thus be seen as an expression of the manifest content of the text. A category often includes a number of sub-categories or sub-subcategories at varying levels of abstraction. The sub-categories can be sorted and abstracted into a category or a category can be divided into sub-categories.

'주제'라는 개념은 다양한 의미를 가지며, 주제를 형성하는 것은 여러 카테고리에 깔린 의미들을 연결시키는 방법이기도 하다. Polt와 Hungler는 '주제'를 카테고리내 혹은 여러 카테고리 사이에서 규칙적으로 반복되는 것이라고 정의하였다. Baxter는 '주제'를 여러 도메인에 걸쳐 나타나는 의미의 가닥(thread)이라고 했다. '주제'의 개념은 다른 질적연구방법에서도 드러난다. van Manen은 주제를 '어떻게?'라는 질문에 답을 하는 것이라고 했다. 우리는 '주제'를 의미단위, 코드와 카테고리를 해석적 수준에서 응축하여 만들어낸 기저에 깔린 의미의 가닥이라고 본다. 주제를 텍스트 아래 함축된 latent content의 표현이라고 볼 수도 있다. 모든 데이터가 다양한 의미를 가질 수 있으므로 주제는 반드시 상호베타적일 필요는 없다. 함축된 의미단위, 코드, 카테고리는 하나 이상의 주제에 들어맞을 수 있다. 주제는 하위주제를 통해서 구성될 수도 있고, 주제가 하위주제로 나뉘어질 수도 있다.

The concept of theme has multiple meanings and creating themes is a way to link the underlying meanings together in categories. Polit and Hungler (1999) describe a theme as a recurring regularity developed within categories or cutting across categories. Baxter (1991) defines themes as threads of meaning that recur in domain after domain. The concept of theme is also used in literature in other qualitative methods. van Manen (1990, p. 87) considers a theme to ‘describe an aspect of the structure of experience’ and emphasises that a theme can not be an object or a thing. A theme answers the question ‘How?’ We consider a theme to be a thread of an underlying meaning through, condensed meaning units, codes or categories, on an interpretative level. A theme can be seen as an expression of the latent content of the text. Since all data have multiple meanings ( Krippendorff, 1980; Downe-Wamboldt, 1992), themes are not necessarily mutually exclusive. A condensed meaning unit, a code or a category can fit into more than one theme. A theme can be constructed by sub-themes or divided into sub-themes.

개념들의 사용예시 Illustrations of the use of concepts

In the following we illustrate the use of concepts and analysis procedures for two texts based on interviews and observations respectively. One rationale behind giving two examples is to show various ways to develop themes. The processes of analysis are described and shown in Figure 1, Figure 2 and Figure 3. Even if these descriptions point to a linear process, it is important to bear in mind that the process of analysis involves a back and forth movement between the whole and parts of the text.

Qualitative content analysis of an interview text

The unit of analysis in this example is interview text about experiences of having hypoglycaemia. The context consists of a larger study aimed at describing coping strategies related to the everyday strains of living with diabetes ( Lundman and Norberg, 1993). Twenty adults with Type 1-diabetes, aged 25–59 years, participated in the study. Interviews were performed addressing various aspects of living with Type 1-diabetes. The interview text was sorted into seven content areas: experiences related to the onset of the disease; management of the disease in daily living; experiences related to hypoglycaemia; experiences related to hyperglycaemia; self-monitoring of blood glucose; and ideas about complications and the future. Experiences related to hypoglycaemia were evoked by asking: ‘Please tell me about your experiences of having hypoglycaemia.’

The interviews were read through several times to obtain a sense of the whole. Then the text about the participants’ experiences of having hypoglycaemia was extracted and brought together into one text, which constituted the unit of analysis. The text was divided into meaning units that were condensed. The condensed meaning units were abstracted and labelled with a code. Examples of meaning units, condensed meaning units and codes are shown in Fig. 1. The whole context was considered when condensing and labelling meaning units with codes. The various codes were compared based on differences and similarities and sorted into six sub-categories and three categories, which constitute the manifest content. The tentative categories were discussed by two researchers and revised. What differed between the two researchers was their judgement about what comprised familiar and unfamiliar sensations and actions. A process of reflection and discussion resulted in agreement about how to sort the codes. Finally, the underlying meaning, that is, the latent content, of the categories, was formulated into a theme. Examples of codes, sub-categories, categories and a theme are given in Fig. 2.

Qualitative content analysis of a text based on observations

The unit of analysis in this example is text based on 14 observational notes and six reflective dialogues. The context was a study aiming to illuminate how one woman with dementia and ‘behavioural disturbances’ acted in relation to her care providers, and how the care providers acted in relation to her ( Graneheim et al., 2001). The study was performed at a residential home for people with dementia and so called ‘behavioural disturbances’. The care providers were asked to select a person whose ‘behavioural disturbances’ caused severe difficulties in daily care. Two observers participated on six occasions during morning toilet and breakfast. One observer was familiar to the setting and the participants and represented an insider perspective. The other observer was unfamiliar with these conditions and represented an outsider perspective. The participant observations focused on the interaction going on between the woman with dementia and her care providers. To further illuminate various aspects of the ongoing interaction, a reflective dialogue between the observers and the care providers followed each observation occasion. The observational notes and reflective dialogues were tape-recorded and transcribed verbatim.

The text was read through several times to obtain a sense of the whole. Six observational notes, one from each occasion, were divided into meaning units. Considering the context, the meaning units were condensed into a description close to the text, the manifest content, and, where possible, into an interpretation of the underlying meaning, the latent content. Since parts of the text were much more concentrated than an interview text, further condensation was difficult. The condensed meaning units were seen as a whole and abstracted into sub-themes. Examples of meaning units, condensed meaning units, sub-themes and theme are shown in Fig. 3. Sub-themes were threads of meaning running through the condensed text. The sub-themes were presented to the care providers and revised with consideration to their opinion. The remaining eight observational notes were analysed. A co-researcher read one-third of the observational notes and the thematisation. A process of reflection and discussion resulted in agreement on a set of sub-themes. Lastly, reflection on the sub-themes and a review of literature related to the sub-themes provided phenomena that seemed to serve as relevant headings to unify the sub-themes into themes. To reveal meaning units that rejected interpretations of the observational text the reflective dialogues were analysed and nothing that contradicted the themes could be found.

신뢰성을 높이기 위한 방법 Measures for achieving trustworthiness

Research findings should be as trustworthy as possible and every research study must be evaluated in relation to the procedures used to generate the findings. The use of concepts for describing trustworthiness differs between the qualitative and the quantitative research traditions. Within the tradition of qualitative content analysis, use of concepts related to the quantitative tradition, such as validity, reliability and generalisability, is still common (for example, Downe-Wamboldt, 1992; Olson et al., 1998; Shields and King, 2001). In qualitative research the concepts credibility, dependability and transferability have been used to describe various aspects of trustworthiness (for example, Guba, 1981; Lincoln and Guba, 1985; Patton, 1987; Polit and Hungler, 1999; Berg and Welander Hansson, 2000). However, Long and Johnson (2000, p. 31) propose that validity and reliability have ‘the same essential meaning’ irrespective of research tradition and nothing is gained by changing labels. In our paper, we suggest application of concepts linked to the qualitative tradition when reporting findings of studies using qualitative content analysis. Even though we separate the aspects of trustworthiness, they should be viewed as intertwined and interrelated.

Credibility는 연구의 초점에 대한 것이며, 얼마나 데이터와 분석과정이 목표한 초점을 잘 달성하고 있는가에 대한 자신감이라고 할 수 있다. 처음 연구의 초점이 무엇인지 결정할 때, 컨텍스트를 정할 때, 참가자와 데이터수집법을 정할 때 credibility에 대한 질문을 처음 하게 된다. 다양한 경험을 가진 참가자를 고르는 것은 다양한 관점에서 연구질문을 보게 할 가능성을 높여준다. 피면접자가 성별, 연령 등이 다양하고 관찰자의 관점이 다양하면 연구하고자 하는 현상을 더욱 풍요롭게 해주는 장점이 이삳. 가장 적합한 데이터수집 방법을 선택하고, 적절한 양의 데이터를 얻는 것도 credibility에 중요하다.

Credibility deals with the focus of the research and refers to confidence in how well data and processes of analysis address the intended focus ( Polit and Hungler, 1999). The first question concerning credibility arises when making a decision about the focus of the study, selection of context, participants and approach to gathering data. Choosing participants with various experiences increases the possibility of shedding light on the research question from a variety of aspects ( Patton, 1987; Adler and Adler, 1988). In our illustrations, interviewees’ various genders and ages, and observers with various perspectives, contributed to a richer variation of the phenomena under study. Selecting the most appropriate method for data collection and the amount of data are also important in establishing credibility. The amount of data necessary to answer a research question in a credible way varies depending on the complexity of the phenomena under study and the data quality.

Credibility와 관련하여 또 다른 중요한 이슈는 가장 적절한 '의미단위'를 선택하는 것이다. '의미단위'가 너무 포괄적이라면, 예컨대 문단 수준이라면, 너무 여러가지 의미를 포함하기 때문에 다루기가 어렵다. 너무 의미단위가 좁으면 (한 단어 수준이라면) 결과가 분절화될 수 있다. 이것의 예외적 상황은 하나 또는 여러개의 단어가 심볼이나 메타포를 이룰 때이다. 두 가지 모두 응축과 추상화 단계에서 텍스트의 의미를 상실해버릴 가능성이 있다. 어떻게 의미단위, 응축, 추상화가 되었는지를 설명하는 것이 credibility를 높여준다.

Another critical issue for achieving credibility is to select the most suitable meaning unit. Meaning units that are too broad, for example, several paragraphs, will be difficult to manage since they are likely to contain various meanings. Too narrow meaning units, for example, a single word, may result in fragmentation. An exception to this is when one or several words represent a symbol or metaphor. In both cases there is a risk of losing meaning of the text during the condensation and abstraction process. Illustrating how meaning units, condensations and abstractions are made facilitates judging credibility of the findings (see Figure 1, Figure 2 and Figure 3).

연구의 Credibility는 카테고리와 주제가 얼마나 데이터를 잘 포괄하느냐에 달렸으며, 이는 관련된 데이터가 부주의하여 혹은 시스템적인 결함으로 제외되거나, 무관한 데이터가 포함되면 안된다는 의미이다. Credibility는 어떻게 카테고리간의 차이점과 카테고리내의 유사성을 판단할 것인가에 대한 질문이기도 하다. 이에 대한 한 가지 접근법은 transcribed된 텍스트로부터 대표적인 인용문구를 보여주는 것이고, 다른 방법은 공동연구자와의 일치를 보는 것이다.

Credibility of research findings also deals with how well categories and themes cover data, that is, no relevant data have been inadvertently or systematically excluded or irrelevant data included. Credibility is also a question of how to judge the similarities within and differences between categories. One way to approach this is to show representative quotations from the transcribed text. Another way is to seek agreement among co-researchers, experts and participants.

일치도를 확인하는 것이 적절한가에 대해서는 다양한 의견이 있다. 어떤 연구자는 해석의 주관적인 측면에 의해서 다양한 현실이 존재할 수 있기 때문에 공동연구자간의 validation에 대한 의문을 갖는 사람도 있다. 비록 우리가 현실이 다면적이고 주관적이라고 하더라도 우리는 공동연구자들간의 대화의 가치를 중요시한다. 여기서의 목적은 단순히 라벨되고 분류된 데이터가 같은 방식으로 되었느냐를 확인하는 것이 아니라, 다양한 전문가가 데이터가 라벨되고 분류된 방식에 동의하는가에 대한 것이다. 관찰결과에 대한 참가자의 인식도 credibility 의 한 측면이지만 verification에 대한 것이라기보다는 confirmability에 대한 것이라고 할 수 있다.

There are various opinions about the appropriateness of seeking agreement. Sandelowski, 1993 and Sandelowski, 1998 argues that, since multiple realities exist that are dependent on subjective interpretations, validation among co-researchers, experts and participants is questionable. Even though we agree that reality is multiple and subjective, we defend the value of dialogue among co-researchers. The intent here is not merely to verify that data are labelled and sorted in exactly the same way, but to determine whether or not various researchers and experts would agree with the way those data were labelled and sorted (Woods and Catanzaro, 1988). Participants’ recognition of the findings can also be an aspect of credibility. It is not, however, a question of verification but rather a question of confirmability.

신뢰성의 다른 측면은 dependability이다. dependability는 분석과정에 걸쳐서 시간에 따라 데이터가 변하는 정도, 또는 연구자의 결정이 바뀌는 것을 의미한다.

Another aspect of trustworthiness is dependability. According to Lincoln and Guba (1985, p. 299), dependability ‘seeks means for taking into account both factors of instability and factors of phenomenal or design induced changes’, that is, the degree to which data change over time and alterations made in the researcher’s decisions during the analysis process.

데이터가 방대하고 아주 오랜 기간에 걸쳐서 수집되었다면, 데이터 수집과정에서 일관성을 상실할 가능성이 있다. 한편 모든 참가자에 대해서 같은 영역에 대한 질문을 하는 것이 중요하다. 그러나 면접과 관찰은 면접자와 관찰자가 연구를 진행하는 동안 현상에 대한 새로운 통찰력을 얻는 과정에서 진화할 수 있고, 이것 때문에 이후 면담이나 관찰에 영향을 줄 수 있다. 내용의 유사성과 차이점에 대한 판단이 긴 시간에 걸쳐서 얼마나 일관적인지는 연구팀 내에서 개방적 대화를 통해 확인될 수 있다.

When data are extensive and the collection extends over time, there is a risk of inconsistency during data collection. On one hand, it is important to question the same areas for all the participants. On the other hand, interviewing and observing is an evolving process during which interviewers and observers acquire new insights into the phenomenon of study that can subsequently influence follow-up questions or narrow the focus for observation. The extent to which judgements about similarities and differences of content are consistent over time can, as in our illustrations, be addressed by an open dialogue within the research team.

또한 transferability에 대한 질문도 있다. 이는 관찰결과가 다른 세팅이나 다른 그룹에 적용가능한 정도를 말한다. 저자는 transferability에 대한 의견을 기술할 수 있지만, 이는 독자의 판단에 따라 달라지기도 한다.

Trustworthiness also includes the question of transferability, which refers to ‘the extent to which the findings can be transferred to other settings or groups’ ( Polit and Hungler, 1999, p. 717). The authors can give suggestions about transferability, but it is the reader’s decision whether or not the findings are transferable to another context.

transferability를 향상시키기 위해서는 문화와 맥락, 참가자의 선택과정과 특성, 데이터수집과 분석과정을 분명하게 기술하는 것이 중요하다. 관찰결과를 풍요로우면서도 철저하게 제시하는 것도 중요하다.

To facilitate transferability, it is valuable to give a clear and distinct description of culture and context, selection and characteristics of participants, data collection and process of analysis. A rich and vigorous presentation of the findings together with appropriate quotations will also enhance transferability.

관찰결과에 대해서 단 하나의 옳은 의미 또는 모든 상황에 적용가능한 것은 없다. 특정 관점에서 가장 그럴듯한 의미가 있을 뿐이다. 질적연구에서 해석의 신뢰성은 가장 그럴듯한 해석에 대한 근거를 수립하는 것에 대한 것이라고 할 수 있다. 독자가 또 다른 해석방법을 찾아내도록 하는 방식으로 관찰결과가 제시된다면 신뢰성이 더욱 높아질 것이다.

There is no single correct meaning or universal application of research findings, but only the most probable meaning from a particular perspective. In qualitative research, trustworthiness of interpretations deals with establishing arguments for the most probable interpretations. Trustworthiness will increase if the findings are presented in a way that allows the reader to look for alternative interpretations.

Reflections

어떤 연구방법이 너무 부정확하여 다른 연구분야에 잘 들어맞는다면, 방법론적 접근과 자료는 단순히 도구로 보일 수도 있다. 반면, 질적내용분석은 구체적인 특징이 있고, 그 아래 깔린 가정이 있다.

When discussing meaning and use of concepts, procedures and interpretation related to qualitative content analysis, it is valuable to consider whether qualitative content analysis is a separate method or tool used within different forms of qualitative analysis. On one hand, a method that is so inexact that it fits into different research fields, methodological approaches and data can be seen as merely a tool. On the other hand, it can be assumed that qualitative content analysis has specific characteristics and underlying theoretical assumptions which need to be further illuminated.

질적내용분석의 한 특징은 이 방법이 주제와 맥락에 초점을 맞추면서 코드와 카테고리 내의 차이와 유사성을 강조한다는 것이다. 또 다른 특징은 manifest content와 latent content를 모두 다룬다는 것이다. 

One characteristic of qualitative content analysis is that the method, to a great extent, focuses on the subject and context, and emphasises differences between and similarities within codes and categories. Another characteristic is that the method deals with manifest as well as latent content in a text. The manifest content, that is, what the text says, is often presented in categories, while themes are seen as expressions of the latent content, that is, what the text is talking about.

질적내용분석에 깔려 있는 이론적 가정을 이해는 한 가지 방법은 이 연구방법을 의사소통이론과 연관시키는 것이다. 인간의 의사소통을 바탕으로 해석에 대한 이슈를 제기한 바 있는데,

• 한 가지 axiom은 '혼자서 의사소통한다는 것은 불가능하다'이다. 인터뷰와 관찰에 기반한 텍스트는 연구자와 참가자의 상호작용속에서 만들어지며, 이것은 의사소통의 하나이다. 모든 텍스트에는 해석의 여지가 있는 메시지가 잇다. 분석과정이 시작되는 순간 연구자와 텍스트 사이의 끊임없는 의사소통이 시작되는 것이다. 
• 또 다른 axiom은 '모든 의사소통은 내용적 측면과 관계적 측면이 있어서, 관계적 측면이 내용적 측면을 분류하게 되고, 따라서 관계적 측면은 일종의 meta-communication이다' 이다. 우리가 설명한 것을 대입시켜보면 카테고리는 manifest content를 나타내는 것으로 내용적 측면이라 할 수 있고, 주제는 latent content를 나타내는 것으로 관계적 측면이라 할 수 있다.

One way to understand the theoretical assumptions underlying qualitative content analysis is to relate the method to communication theory as described by Watzlawick et al. (1967). They state axioms concerning human communication that could shed light on the issue of interpretation. One axiom is that ‘one cannot not communicate’ (Watzlawick et al., 1967, p. 51). Texts based on interviews and observations are shaped within an interaction between the researcher and the participants and can be seen as a communication act. In every text there are messages to be interpreted and described. As soon as the analysis procedure begins, ongoing communication between the researcher and the text is present. Another axiom is that ‘every communication has a content aspect and a relationship aspect such that the latter classifies the former and is therefore a meta-communication’ (Watzlawick et al., 1967, p. 54). In our illustrations, categories are seen as representing the manifest content, that is, the content aspect, and themes are representing the latent content, which can be seen as the relationship aspect.

• 또 다른 axiom은 '인간은 디지털과 아날로그 방식 둘 다를 활용하여 의사소통한다'이다. 언어적 의사소통은 주로 디지털적인 측면이 강하며, 쉽게 텍스트로 transcribe될 수 있다. 그러나 비언어적 의사소통은 주로 아날로그적(analogical)이며, transcription process를 어렵게 만든다. 하지만 의미는 메시지가 전달되는 방식에 의해서 생성되며, 여기에는 음성뿐만 아니라 거기에 포함된 감정도 들어간다. 따라서 인터뷰와 관찰을 텍스트로 옮길 때에는 침묵/한숨/웃음/자세/제스쳐 등을 같이 넣는 것이 좋다. 
• 마지막 axiom은 '관계의 성격은 의사소통을 하는 사람들간의 구두점(punctuation)에 달려있다'로서 텍스트를 의미단위로 나누는 것은 끊임없이 이어지는 의사소통을 구분(punctuating)하는 것이고, 의미단위를 시작할 때와 끝낼 때 모두에 manifest content와 latent content가 중요하다. 

‘Human beings communicate both digitally and analogically’ is another axiom of Watzlawick et al. (1967, p. 66). Verbal communication is mainly digital and easily transcribed into a text while non-verbal communication is mainly analogical and often put at a disadvantage in the transcription process. However, meaning is partly created by how a message is communicated, that is, the voice or implied feeling that emerges from the reading of the text (Downe-Wamboldt, 1992). Therefore, when transcribing interviews and observations into text, it is valuable to notice silence, sighs, laughter, posture, gestures etc., as these may influence the underlying meaning. Watzlawick et al. (1967, p. 59) have also formulated the axiom that ‘the nature of a relationship is contingent upon the punctuation of the communicational sequences between the communicants’. Dividing the text into meaning units is a way of punctuating the ongoing communication in a text and is important for both manifest and latent content when beginning and ending a meaning unit.

해석의 또 다른 측면은 텍스트는 언제나 다양한 의미를 가지며, 연구자의 해석은 본인이 어떠한 경험을 했느냐에 달려있다는 것이다. 연구자는 보통 데이터를 수집하는 사람일수도 있고, 분석하는 사람일수도 있기 때문에 연구자의 자질,훈련,경험 등이 중요하다. 질적내용분석에서는 균형을 잡는 것이 중요한데, 한편으로는 연구자가 특정한 관점을 발휘하지 못하게 하는 것이 불가능하거나 바람직하지 못할 수도 있지만 다른 한편으로는 존재하지 않는 의미를 주입하지 않고 '텍스트가 스스로 말하도록(let the text talk)' 해야 한다.

Another aspect of interpretation is that a text always involves multiple meanings and the researcher’s interpretation is influenced by his or her personal history. Since the researcher is often the one who collects the data as well as the one who performs the analysis, the question of the researcher’s qualifications, training and experiences is important (Patton, 1990). In qualitative content analysis interpretation involves a balancing act. On one hand, it is impossible and undesirable for the researcher not to add a particular perspective to the phenomena under study. On the other hand, the researcher must ‘let the text talk’ and not impute meaning that is not there.

Learning and teaching how to analyse texts is a delicate matter in nursing education. Qualitative content analysis can be a valuable method for students when attending a research class for the first time due to the opportunity to perform the analysis at various degrees of difficulty. Analysing content close to the text, that is, the manifest content, can be a suitable starting point. With increasing knowledge and ability students may advance to interpret the underlying meaning, that is, the latent content, on various levels of abstraction.

In conclusion, our paper is intended to be used in nursing research and education and to contribute to a debate on qualitative content analysis. In order to clarify the underlying assumptions of qualitative content analysis, we suggest using concepts related to qualitative research when describing the research procedure and measures to achieve trustworthiness. Moreover, we apply communication theory as a way to address the issue of interpretation and clarify the underlying assumptions of qualitative content analysis.

---

Nurse Educ Today. 2004 Feb;24(2):105-12.

Qualitative content analysis in nursing research: concepts, procedures and measures to achieve trustworthiness.

Graneheim UH1, Lundman B.

Author information

Abstract

Qualitative content analysis as described in published literature shows conflicting opinions and unsolved issues regarding meaning and use of concepts, procedures and interpretation. This paper provides an overview of important concepts (manifest and latent content, unit of analysis, meaning unit, condensation, abstraction, content area, code, category and theme) related to qualitative content analysis; illustrates the use of concepts related to the research procedure; and proposes measures to achieve trustworthiness (credibility, dependability and transferability) throughout the steps of the research procedure. Interpretation in qualitative content analysis is discussed in light of Watzlawick et al.'s [Pragmatics of Human Communication. A Study of Interactional Patterns, Pathologies and Paradoxes. W.W. Norton & Company, New York, London] theory of communication.

PMID:

 

14769454

 

[PubMed - indexed for MEDLINE]