Collecting and Analyzing Qualitative Data

Qualitative research methods are a key component of field epidemiologic investigations because they can provide insight into the perceptions, values, opinions, and community norms where investigations are being conducted (1,2). Open-ended inquiry methods, the mainstay of qualitative interview techniques, are essential in formative research for exploring contextual factors and rationales for risk behaviors that do not fit neatly into predefined categories. For example, during the 2014–2015 Ebola virus disease outbreaks in parts of West Africa, understanding the cultural implications of burial practices within different communities was crucial to designing and monitoring interventions for safe burials (Box 10.1). In program evaluations, qualitative methods can assist the investigator in diagnosing what went right or wrong as part of a process evaluation or in troubleshooting why a program might not be working as well as expected. When designing an intervention, qualitative methods can be useful in exploring dimensions of acceptability to increase the chances of intervention acceptance and success. When performed in conjunction with quantitative studies, qualitative methods can help the investigator confirm, challenge, or deepen the validity of conclusions than either component might have yielded alone (1,2).

Similar to quantitative approaches, qualitative research seeks answers to specific questions by using rigorous approaches to collecting and compiling information and producing findings that can be applicable beyond the study population. The fundamental difference in approaches lies in how they translate real-life complexities of initial observations into units of analysis. Data collected in qualitative studies typically are in the form of text or visual images, which provide rich sources of insight but also tend to be bulky and time-consuming to code and analyze. Practically speaking, qualitative study designs tend to favor small, purposively selected samples ideal for case studies or in-depth analysis (1). The combination of purposive sampling and open-ended question formats deprive qualitative study designs of the power to quantify and generalize conclusions, one of the key limitations of this approach.

Qualitative scientists might argue, however, that the generalizability and precision possible through probabilistic sampling and categorical outcomes are achieved at the cost of enhanced validity, nuance, and naturalism that less structured approaches offer (3). Open-ended techniques are particularly useful for understanding subjective meanings and motivations underlying behavior. They enable investigators to be equally adept at exploring factors observed and unobserved, intentions as well as actions, internal meanings as well as external consequences, options considered but not taken, and unmeasurable as well as measurable outcomes. These methods are important when the source of or solution to a public health problem is rooted in local perceptions rather than objectively measurable characteristics selected by outside observers (3). Ultimately, such approaches have the ability to go beyond quantifying questions of how much or how many to take on questions of how or why from the perspective and in the words of the study subjects themselves (1,2).

Another key advantage of qualitative methods for field investigations is their flexibility (4). Qualitative designs not only enable but also encourage flexibility in the content and flow of questions to challenge and probe for deeper meanings or follow new leads if they lead to deeper understanding of an issue (5). It is not uncommon for topic guides to be adjusted in the course of fieldwork to investigate emerging themes relevant to answering the original study question. As discussed herein, qualitative study designs allow flexibility in sample size to accommodate the need for more or fewer interviews among particular groups to determine the root cause of an issue (see the section on Sampling and Recruitment in Qualitative Research). In the context of field investigations, such methods can be extremely useful for investigating complex or fast-moving situations where the dimensions of analysis cannot be fully anticipated.

Ultimately, the decision whether to include qualitative research in a particular field investigation depends mainly on the nature of the research question itself. Certain types of research topics lend themselves more naturally to qualitative rather than other approaches (Table 10.1). These include exploratory investigations when not enough is known about a problem to formulate a hypothesis or develop a fixed set of questions and answer codes. They include research questions where intentions matter as much as actions and “why?” or “why not?” questions matter as much as precise estimation of measured outcomes. Qualitative approaches also work well when contextual influences, subjective meanings, stigma, or strong social desirability biases lower faith in the validity of responses coming from a relatively impersonal survey questionnaire interview.

The availability of personnel with training and experience in qualitative interviewing or observation is critical for obtaining the best quality data but is not absolutely required for rapid assessment in field settings. Qualitative interviewing requires a broader set of skills than survey interviewing. It is not enough to follow a topic guide like a questionnaire, in order, from top to bottom. A qualitative interviewer must exercise judgment to decide when to probe and when to move on, when to encourage, challenge, or follow relevant leads even if they are not written in the topic guide. Ability to engage with informants, connect ideas during the interview, and think on one’s feet are common characteristics of good qualitative interviewers. By far the most important qualification in conducting qualitative fieldwork is a firm grasp of the research objectives; with this qualification, a member of the research team armed with curiosity and a topic guide can learn on the job with successful results.

Semi-Structured Interviews

Semi-structured interviews can be conducted with single participants (in-depth or individual key informants) or with groups (focus group discussions [FGDs] or key informant groups). These interviews follow a suggested topic guide rather than a fixed questionnaire format. Topic guides typically consist of a limited number (10– 15) of broad, open-ended questions followed by bulleted points to facilitate optional probing. The conversational back-and-forth nature of a semi-structured format puts the researcher and researched (the interview participants) on more equal footing than allowed by more structured formats. Respondents, the term used in the case of quantitative questionnaire interviews, become informants in the case of individual semi-structured in-depth interviews (IDIs) or participants in the case of FGDs. Freedom to probe beyond initial responses enables interviewers to actively engage with the interviewee to seek clarity, openness, and depth by challenging informants to reach below layers of self-presentation and social desirability. In this respect, interviewing is sometimes compared with peeling an onion, with the first version of events accessible to the public, including survey interviewers, and deeper inner layers accessible to those who invest the time and effort to build rapport and gain trust. (The theory of the active interview suggests that all interviews involve staged social encounters where the interviewee is constantly assessing interviewer intentions and adjusting his or her responses accordingly [1]. Consequently good rapport is important for any type of interview. Survey formats give interviewers less freedom to divert from the preset script of questions and formal probes.)

Individual In-Depth Interviews and Key-Informant Interviews

The most common forms of individual semi-structured interviews are IDIs and key informant interviews (KIIs). IDIs are conducted among informants typically selected for first-hand experience (e.g., service users, participants, survivors) relevant to the research topic. These are typically conducted as one-on-one face-to-face interviews (two-on-one if translators are needed) to maximize rapport-building and confidentiality. KIIs are similar to IDIs but focus on individual persons with special knowledge or influence (e.g., community leaders or health authorities) that give them broader perspective or deeper insight into the topic area (Box 10.2). Whereas IDIs tend to focus on personal experiences, context, meaning, and implications for informants, KIIs tend to steer away from personal questions in favor of expert insights or community perspectives. IDIs enable flexible sampling strategies and represent the interviewing reference standard for confidentiality, rapport, richness, and contextual detail. However, IDIs are time-and labor-intensive to collect and analyze. Because confidentiality is not a concern in KIIs, these interviews might be conducted as individual or group interviews, as required for the topic area.

Focus Group Discussions and Group Key Informant Interviews

FGDs are semi-structured group interviews in which six to eight participants, homogeneous with respect to a shared experience, behavior, or demographic characteristic, are guided through a topic guide by a trained moderator (6). (Advice on ideal group interview size varies. The principle is to convene a group large enough to foster an open, lively discussion of the topic, and small enough to ensure all participants stay fully engaged in the process.) Over the course of discussion, the moderator is expected to pose questions, foster group participation, and probe for clarity and depth. Long a staple of market research, focus groups have become a widely used social science technique with broad applications in public health, and they are especially popular as a rapid method for assessing community norms and shared perceptions.

Focus groups have certain useful advantages during field investigations. They are highly adaptable, inexpensive to arrange and conduct, and often enjoyable for participants. Group dynamics effectively tap into collective knowledge and experience to serve as a proxy informant for the community as a whole. They are also capable of recreating a microcosm of social norms where social, moral, and emotional dimensions of topics are allowed to emerge. Skilled moderators can also exploit the tendency of small groups to seek consensus to bring out disagreements that the participants will work to resolve in a way that can lead to deeper understanding. There are also limitations on focus group methods. Lack of confidentiality during group interviews means they should not be used to explore personal experiences of a sensitive nature on ethical grounds. Participants may take it on themselves to volunteer such information, but moderators are generally encouraged to steer the conversation back to general observations to avoid putting pressure on other participants to disclose in a similar way. Similarly, FGDs are subject by design to strong social desirability biases. Qualitative study designs using focus groups sometimes add individual interviews precisely to enable participants to describe personal experiences or personal views that would be difficult or inappropriate to share in a group setting. Focus groups run the risk of producing broad but shallow analyses of issues if groups reach comfortable but superficial consensus around complex topics. This weakness can be countered by training moderators to probe effectively and challenge any consensus that sounds too simplistic or contradictory with prior knowledge. However, FGDs are surprisingly robust against the influence of strongly opinionated participants, highly adaptable, and well suited to application in study designs where systematic comparisons across different groups are called for.

Like FGDs, group KIIs rely on positive chemistry and the stimulating effects of group discussion but aim to gather expert knowledge or oversight on a particular topic rather than lived experience of embedded social actors. Group KIIs have no minimum size requirements and can involve as few as two or three participants.

Visualization Methods

Visualization methods have been developed as a way to enhance participation and empower interviewees relative to researchers during group data collection (7). Visualization methods involve asking participants to engage in collective problem- solving of challenges expressed through group production of maps, diagrams, or other images. For example, participants from the community might be asked to sketch a map of their community and to highlight features of relevance to the research topic (e.g., access to health facilities or sites of risk concentrations). Body diagramming is another visualization tool in which community members are asked to depict how and where a health threat affects the human body as a way of understanding folk conceptions of health, disease, treatment, and prevention. Ensuing debate and dialogue regarding construction of images can be recorded and analyzed in conjunction with the visual image itself. Visualization exercises were initially designed to accommodate groups the size of entire communities, but they can work equally well with smaller groups corresponding to the size of FGDs or group KIIs.

Selecting a Sample of Study Participants

Fundamental differences between qualitative and quantitative approaches to research emerge most clearly in the practice of sampling and recruitment of study participants. Qualitative samples are typically small and purposive. In-depth interview informants are usually selected on the basis of unique characteristics or personal experiences that make them exemplary for the study, if not typical in other respects. Key informants are selected for their unique knowledge or influence in the study domain. Focus group mobilization often seeks participants who are typical with respect to others in the community having similar exposure or shared characteristics. Often, however, participants in qualitative studies are selected because they are exceptional rather than simply representative. Their value lies not in their generalizability but in their ability to generate insight into the key questions driving the study.

Determining Sample Size

Sample size determination for qualitative studies also follows a different logic than that used for probability sample surveys. For example, whereas some qualitative methods specify ideal ranges of participants that constitute a valid observation (e.g., focus groups), there are no rules on how many observations it takes to attain valid results. In theory, sample size in qualitative designs should be determined by the saturation principle, where interviews are conducted until additional interviews yield no additional insights into the topic of research (8). Practically speaking, designing a study with a range in number of interviews is advisable for providing a level of flexibility if additional interviews are needed to reach clear conclusions.

Recruiting Study Participants

Recruitment strategies for qualitative studies typically involve some degree of participant self-selection (e.g., advertising in public spaces for interested participants) and purposive selection (e.g., identification of key informants). Purposive selection in community settings often requires authorization from local authorities and assistance from local mobilizers before the informed consent process can begin. Clearly specifying eligibility criteria is crucial for minimizing the tendency of study mobilizers to apply their own filters regarding who reflects the community in the best light. In addition to formal eligibility criteria, character traits (e.g., articulate and interested in participating) and convenience (e.g., not too far away) are legitimate considerations for whom to include in the sample. Accommodations to personality and convenience help to ensure the small number of interviews in a typical qualitative design yields maximum value for minimum investment. This is one reason why random sampling of qualitative informants is not only unnecessary but also potentially counterproductive.

Analysis of qualitative data can be divided into four stages: data management, data condensation, data display, and drawing and verifying conclusions (9).

Managing Qualitative Data

From the outset, developing a clear organization system for qualitative data is important. Ideally, naming conventions for original data files and subsequent analysis should be recorded in a data dictionary file that includes dates, locations, defining individual or group characteristics, interviewer characteristics, and other defining features. Digital recordings of interviews or visualization products should be reviewed to ensure fidelity of analyzed data to original observations. If ethics agreements require that no names or identifying characteristics be recorded, all individual names must be removed from final transcriptions before analysis begins. If data are analyzed by using textual data analysis software, maintaining careful version control over the data files is crucial, especially when multiple coders are involved.

Condensing Qualitative Data

Condensing refers to the process of selecting, focusing, simplifying, and abstracting the data available at the time of the original observation, then transforming the condensed data into a data set that can be analyzed. In qualitative research, most of the time investment required to complete a study comes after the fieldwork is complete. A single hour of taped individual interview can take a full day to transcribe and additional time to translate if necessary. Group interviews can take even longer because of the difficulty of transcribing active group input. Each stage of data condensation involves multiple decisions that require clear rules and close supervision. A typical challenge is finding the right balance between fidelity to the rhythm and texture of original language and clarity of the translated version in the language of analysis. For example, discussions among groups with little or no education should not emerge after the transcription (and translation) process sounding like university graduates. Judgment must be exercised about which terms should be translated and which terms should be kept in vernacular because there is no appropriate term in English to capture the richness of its meaning.

Displaying Qualitative Data

After the initial condensation, qualitative analysis depends on how the data are displayed. Decisions regarding how data are summarized and laid out to facilitate comparison influence the depth and detail of the investigation’s conclusions. Displays might range from full verbatim transcripts of interviews to bulleted summaries or distilled summaries of interview notes. In a field setting, a useful and commonly used display format is an overview chart in which key themes or research questions are listed in rows in a word processer table or in a spreadsheet and individual informant or group entry characteristics are listed across columns. Overview charts are useful because they allow easy, systematic comparison of results.

Drawing and Verifying Conclusions

Analyzing qualitative data is an iterative and ideally interactive process that leads to rigorous and systematic interpretation of textual or visual data. At least four common steps are involved:

• Reading and rereading. The core of qualitative analysis is careful, systematic, and repeated reading of text to identify consistent themes and interconnections emerging from the data. The act of repeated reading inevitably yields new themes, connections, and deeper meanings from the first reading. Reading the full text of interviews multiple times before subdividing according to coded themes is key to appreciating the full context and flow of each interview before subdividing and extracting coded sections of text for separate analysis.
• Coding. A common technique in qualitative analysis involves developing codes for labeling sections of text for selective retrieval in later stages of analysis and verification. Different approaches can be used for textual coding. One approach, structural coding, follows the structure of the interview guide. Another approach, thematic coding, labels common themes that appear across interviews, whether by design of the topic guide or emerging themes assigned based on further analysis. To avoid the problem of shift and drift in codes across time or multiple coders, qualitative investigators should develop a standard codebook with written definitions and rules about when codes should start and stop. Coding is also an iterative process in which new codes that emerge from repeated reading are layered on top of existing codes. Development and refinement of the codebook is inseparably part of the analysis.
• Analyzing and writing memos. As codes are being developed and refined, answers to the original research question should begin to emerge. Coding can facilitate that process through selective text retrieval during which similarities within and between coding categories can be extracted and compared systematically. Because no p values can be derived in qualitative analyses to mark the transition from tentative to firm conclusions, standard practice is to write memos to record evolving insights and emerging patterns in the data and how they relate to the original research questions. Writing memos is intended to catalyze further thinking about the data, thus initiating new connections that can lead to further coding and deeper understanding.
• Verifying conclusions. Analysis rigor depends as much on the thoroughness of the cross-examination and attempt to find alternative conclusions as on the quality of original conclusions. Cross-examining conclusions can occur in different ways. One way is encouraging regular interaction between analysts to challenge conclusions and pose alternative explanations for the same data. Another way is quizzing the data (i.e., retrieving coded segments by using Boolean logic to systematically compare code contents where they overlap with other codes or informant characteristics). If alternative explanations for initial conclusions are more difficult to justify, confidence in those conclusions is strengthened.

Above all, qualitative data analysis requires sufficient time and immersion in the data. Computer textual software programs can facilitate selective text retrieval and quizzing the data, but discerning patterns and arriving at conclusions can be done only by the analysts. This requirement involves intensive reading and rereading, developing codebooks and coding, discussing and debating, revising codebooks, and recoding as needed until clear patterns emerge from the data. Although quality and depth of analysis is usually proportional to the time invested, a number of techniques, including some mentioned earlier, can be used to expedite analysis under field conditions.

• Detailed notes instead of full transcriptions. Assigning one or two note-takers to an interview can be considered where the time needed for full transcription and translation is not feasible. Even if plans are in place for full transcriptions after fieldwork, asking note-takers to submit organized summary notes is a useful technique for getting real-time feedback on interview content and making adjustments to topic guides or interviewer training as needed.
• Summary overview charts for thematic coding. (See discussion under “Displaying Data.”) If there is limited time for full transcription and/or systematic coding of text interviews using textual analysis software in the field, an overview chart is a useful technique for rapid manual coding.
• Thematic extract files. This is a slightly expanded version of manual thematic coding that is useful when full transcriptions of interviews are available. With use of a word processing program, files can be sectioned according to themes, or separate files can be created for each theme. Relevant extracts from transcripts or analyst notes can be copied and pasted into files or sections of files corresponding to each theme. This is particularly useful for storing appropriate quotes that can be used to illustrate thematic conclusions in final reports or manuscripts.
• Teamwork. Qualitative analysis can be performed by a single analyst, but it is usually beneficial to involve more than one. Qualitative conclusions involve subjective judgment calls. Having more than one coder or analyst working on a project enables more interactive discussion and debate before reaching consensus on conclusions.
• Computer textual analysis software. Computer-assisted analysis has a number of advantages for qualitative analysis (10). These include
  • Systematic coding.
  • Selective retrieval of coded segments.
  • Verifying conclusions (“quizzing the data”).
  • Working on larger data sets with multiple separate files.
  • Working in teams with multiple coders to allow intercoder reliability to be measured and monitored.

The most widely used software packages (e.g., NVivo [QSR International Pty. Ltd., Melbourne, VIC, Australia] and ATLAS.ti [Scientific Software Development GmbH, Berlin, Germany]) evolved to include sophisticated analytic features covering a wide array of applications but are relatively expensive in terms of license cost and initial investment in time and training. A promising development is the advent of free or low-cost Web-based services (e.g., Dedoose [Sociocultural Research Consultants LLC, Manhattan Beach, CA]) that have many of the same analytic features on a more affordable subscription basis and that enable local research counterparts to remain engaged through the analysis phase (see Teamwork criteria). The start-up costs of computer-assisted analysis need to be weighed against their analytic benefits, which tend to decline with the volume and complexity of data to be analyzed. For rapid situational analyses or small scale qualitative studies (e.g. fewer than 30 observations as an informal rule of thumb), manual coding and analysis using word processing or spreadsheet programs is faster and sufficient to enable rigorous analysis and verification of conclusions.

Qualitative methods belong to a branch of social science inquiry that emphasizes the importance of context, subjective meanings, and motivations in understanding human behavior patterns. Qualitative approaches definitionally rely on open-ended, semistructured, non-numeric strategies for asking questions and recording responses. Conclusions are drawn from systematic visual or textual analysis involving repeated reading, coding, and organizing information into structured and emerging themes. Because textual analysis is relatively time-and skill-intensive, qualitative samples tend to be small and purposively selected to yield the maximum amount of information from the minimum amount of data collection. Although qualitative approaches cannot provide representative or generalizable findings in a statistical sense, they can offer an unparalleled level of detail, nuance, and naturalistic insight into the chosen subject of study. Qualitative methods enable investigators to “hear the voice” of the researched in a way that questionnaire methods, even with the occasional open-ended response option, cannot.

Whether or when to use qualitative methods in field epidemiology studies ultimately depends on the nature of the public health question to be answered. Qualitative approaches make sense when a study question about behavior patterns or program performance leads with why, why not, or how. Similarly, they are appropriate when the answer to the study question depends on understanding the problem from the perspective of social actors in real-life settings or when the object of study cannot be adequately captured, quantified, or categorized through a battery of closed-ended survey questions (e.g., stigma or the foundation of health beliefs). Another justification for qualitative methods occurs when the topic is especially sensitive or subject to strong social desirability biases that require developing trust with the informant and persistent probing to reach the truth. Finally, qualitative methods make sense when the study question is exploratory in nature, where this approach enables the investigator the freedom and flexibility to adjust topic guides and probe beyond the original topic guides.

Given that the conditions just described probably apply more often than not in everyday field epidemiology, it might be surprising that such approaches are not incorporated more routinely into standard epidemiologic training. Part of the answer might have to do with the subjective element in qualitative sampling and analysis that seems at odds with core scientific values of objectivity. Part of it might have to do with the skill requirements for good qualitative interviewing, which are generally more difficult to find than those required for routine survey interviewing.

For the field epidemiologist unfamiliar with qualitative study design, it is important to emphasize that obtaining important insights from applying basic approaches is possible, even without a seasoned team of qualitative researchers on hand to do the work. The flexibility of qualitative methods also tends to make them forgiving with practice and persistence. Beyond the required study approvals and ethical clearances, the basic essential requirements for collecting qualitative data in field settings start with an interviewer having a strong command of the research question, basic interactive and language skills, and a healthy sense of curiosity, armed with a simple open-ended topic guide and a tape recorder or note-taker to capture the key points of the discussion. Readily available manuals on qualitative study design, methods, and analysis can provide additional guidance to improve the quality of data collection and analysis.

1. Patton MQ. Qualitative research and evaluation methods: integrating theory and practice. 4th ed. Thousand Oaks, CA: Sage; 2015.
2. Hennink M, Hutter I, Bailey A. Qualitative research methods. Thousand Oaks, CA: Sage; 2010.
3. Lincoln YS, Guba EG. The constructivist credo. Walnut Creek, CA: Left Coast Press; 2013.
4. Mack N, Woodsong C, MacQueen KM, Guest G, Namey E. Qualitative research methods: a data collectors field guide. https://www.fhi360.org/sites/default/files/media/documents/Qualitative%20Research%20Methods%20-%20A%20Data%20Collector%27s%20Field%20Guide.pdf
5. Kvale S, Brinkmann S. Interviews: learning the craft of qualitative research. Thousand Oaks, CA: Sage; 2009:230–43.
6. Krueger RA, Casey MA. Focus groups: a practical guide for applied research. Thousand Oaks, CA: Sage; 2014.
7. Margolis E, Pauwels L. The Sage handbook of visual research methods. Thousand Oaks, CA: Sage; 2011.
8. Mason M. Sample size and saturation in PhD studies using qualitative interviews. Forum: Qualitative Social Research/Sozialforschung. 2010;11(3).
9. Miles MB, Huberman AM, Saldana J. Qualitative data analysis: a methods sourcebook. 3rd ed. Thousand Oaks, CA: Sage; 2014.
10. Silver C, Lewins A. Using software in qualitative research: a step-by-step guide. Thousand Oaks, CA; Sage: 2014.