While quantitative analysis may provide a mile-high or bird’s eye view of the population being studied, qualitative analysis starts from the ground and moves upward. Qualitative analysis is ideally suited for examining processes or interactions between people within a specific context, for example doctors, nurses, and technicians in an operating room [1] and can be particularly helpful in identifying subtle and critical distinctions that are not appreciable using quantitative analysis. Qualitative analysis is useful for describing social constructs for example the (now outdated) taboo against disclosure of medical errors to patients [2], and can be instrumental for outcomes research given its power to identify latent or non-obvious processes or issues at high or low performing institutions [3]. Furthermore, qualitative analysis is a critical tool for policy creation and evaluation as it enables investigators to examine perspectives and interactions among different stakeholders [4].

Qualitative analysis can be hypothesis generating. It is a good method to start with when your question is “What is going on here?” as it allows the investigator to be open to theories or constructs that arise from the data as opposed to using the data to test a preexisting theory or hypothesis. The flexibility of the methodology helps the researcher avoid the problem of seeing only what he is looking for [5]. Finally, many investigators will use qualitative methods as a first step for survey design, both to identify important questions to ask respondents and to insure the internal validity of survey questions through the use of cognitive interviews [6, 7].

Developing a concise, important and feasible research question is a challenge for all investigations and is particularly important for a qualitative study. To start, the investigator must acknowledge his theoretical assumptions and use these assumptions to focus the boundaries around the case to be sampled. In order to study a population or phenomenon in depth, the sample size for a qualitative study is, by necessity, typically small. As such the investigator is confined to study only some actors, in some contexts dealing with some decisions [8]. To define these boundaries, the investigator posits his theoretical assumptions outright in order to determine the case to be studied, aspects in which variability is desired and dimensions where homogeneity is important.

In a quantitative investigation, it is critical that the sample is representative of the population studied in order to generalize the results. As such, the sampling mechanism for a diverse population should ensure inclusion of a range of ages, socioeconomic status, racial and cultural backgrounds (if this diversity is present in the population). For a qualitative investigation, the goal is not to achieve generalizability but rather to capture the phenomenon as it exists at a certain point in time for a particular group (further investigation may or may not demonstrate variability between groups). As such, the investigator needs to explicitly state his theoretical assumptions up front. For example he might state either that he does not believe age, race, or socioeconomic status will have an appreciable effect on the results (and give supporting evidence) or state that these effects are unknown but, due to study constraints are not the focus of the investigation at this time. In essence, the investigator is not ignoring a specific segment of the population, but is explicitly acknowledging the choices made as well as the theoretical assumptions behind these choices in order to develop an appropriate study design to answer a specific and discrete research question.

Although boundaries are described from the start, because much of the research strategy and sampling methods (often called theoretical sampling) are grounded in the data, these boundaries need to be flexible or emerging [8]. This has led to the pejorative characterization of qualitative research as “make it up as you go along research” because this strategy is distinctly different than those used for quantitative investigations. However, the iterative process involved allows the investigator to shift the sampling frame in order to follow and expand important findings as they emerge from the data. Some examples of questions that are ideally suited for qualitative investigation include, “What processes are used to decrease mortality from gastric bypass in high performing centers?”, “What are the drivers of robotic surgery?” and “How do policies to improve access to general surgery in underserved populations impact different stakeholder groups?”

Once the research question is determined, the next step is to select a sampling strategy that reflects the theoretical assumptions and enables the desired analysis. Typically this is called purposeful sampling where the selected respondents or observation units (hospital wards, operating rooms, texts [9], critical documents [10]) are specifically chosen to reflect the case that you desire to study. The sample is usually small so that it can be studied in depth and typically, the investigator does not set a predetermined sample size. Instead, respondents are sampled until the investigation reaches theoretical saturation, a point in the analysis where the themes and trends encountered occur with a degree of regularity. This can pose problems with the IRB, grant solicitation and for study staff. To avoid this, investigators often generate an estimate of their sample size but should offer a large enough range to allow sampling beyond initial estimates if needed. In addition, researchers more familiar with quantitative methods may bristle when, in a qualitative study, it is necessary for the analysis to proceed before all of the data has been collected. This step is required to determine whether additional respondents are needed, to allow for interrogation of unanticipated results emerging in the data and to determine whether theoretical saturation has been achieved.

Because the sampling strategy is theory driven, it is important to state clearly the reasoning behind the selection of the sampling method used as the rigor of the study will be judged on whether there is alignment of the data collection process with the study purpose. There is a large number of sampling strategies, well described in chapter 2 of “Qualitative Analysis” by Miles and Huberman [8]. Some examples include maximum variation sampling where respondents are selected to include high variability in order to identify common patterns, contrasting case sampling where respondents or units are analyzed against each other in order to demonstrate differences [3] and snowball sampling where respondents of a unique or distinct trait are used to identify subsequent respondents for in depth investigation of an atypical point of view or phenomenon [11].

In addition to selecting a sampling strategy, the investigator will need to select an approach for data collection. The approach can range from a highly structured instrument using open-ended interviewing to a completely unstructured method, for example participant observation.