Qualitative Methods: What Are They
and Why Use Them?
Objective. To provide an overview of reasons why qualitative methods have been
used and canbe used in health services and health policy research, to describe arange
ofspecific methods, and to give examples oftheir application.
Data Sources. Classic and contemporary descriptions of the underpinnings and
applications ofqualitative research methods and studies that have used such methods
to examine important health services and health policy issues.
Principal Findings Qualitative research methods are valuable in providing rich de-
scriptionsofcomplexphenomena; trackingunique orunexpected events; illuminating
the experience and interpretation of events by actors with widely differing stakes
and roles; giving voice to those whose views are rarely heard; conducting initial
explorations to develop theoriesand togenerate andeven testhypotheses;andmoving
toward explanations. Qualitative and quantitative methods can be complementary,
used insequence orin tandem. Thebestqualitative research issystematic andrigorous,
and it seeks to reduce bias and error and to identify evidence that disconfirms initial
or emergent hypotheses.
Condlusions. Qualitative methods have much to contribute to health services and
health policy research, especially as such research deals with rapid change and devel-
ops amore fully integrated theory base and research agenda. However, the field must
build on the best traditions and techniques ofqualitative methods and must recognize
that special training and experience are essential to the application ofthese methods.
Key Words. Qualitative research methods, case study research, key informant inter-
views,focusgroup research, participantobservation
Behind every quantity there must lie a quality.
GertudeJaeger Selznick, Ph.D.
The field ofhealth services research is young. Its multi-disciplinary practition-
ers have borrowed conceptual frameworks and data collection and analytic
methods from a wide range of social and behavioral sciences, as well as
HSR Health Services Research 34:5 Part II (December 1999)
from public health and medicine. Many health services and health policy
researchers have used qualitative methods, either alone or in combination
with quantitative approaches. As the epigram (from a lecture by a revered
epistemologist, sociologist, and survey researcher) indicates, quantification,
which facilitates access to powerful statistical tools, must reston priorwork on
conceptualization and operationalization, and on valid and reliable measure-
ment (e.g., see Babbie 1998). Qualitative methods have the potential to con-
tribute significantly to the development ofmeaningful "quantities"; however,
theyhave inherent aswell asinstrumental value. Somephenomena, including
historical events, are so unusual that by the time one has a way to quantify
them, theyhave eitherchanged ordisappeared.' Quantification is an excellent
way to "slice up" phenomena so that they become manageable and discrete
elements ofan overall conceptual framework or analytic plan. Everythinghas
the defects of its qualities, however, and the defect of quantification is that
it does not always support, as well as qualitative work, the understanding of
complex, dynamic, and multi-dimensional "wholes" (Patton 1975).
In addition, the state-of-the-art of quantification is mixed across poten-
tially important independent and dependent variables. If we focus research
onlyonwhatwe alreadyknowhowto quantify, indeed onlyon thatwhich can
ultimately be reliably quantified, we risk ignoring factors that are more signif-
icant in explainingimportant realities and relationships. Qualitative methods
help provide rich descriptions ofphenomena. They enhance understanding
of the context of events as well as the events themselves. The use of these
methods tends to enhance peripheral vision, which is especially important
at the early stages of inquiry, as noted further on. In addition, qualitative
methods can indeed help to identify patterns and configurations among
variables and to make distinctions. Thus, qualitative research not only serves
the desire to describe; it also helps move inquiry toward more meaningful
For all of these reasons, the use of qualitative methods is growing in
health services and health policy research (Bowling 1997). Unfortunately,
however, we need to appreciate more fully the differences between sys-
tematic, rigorous, well-designed qualitative research and well-intentioned
but poorly implemented attempts to supplement quantification with more
Address correspondence to Shoshanna Sofaer, Dr.P.H., Robert P. Luciano Professor of Health
Care Policy, School ofPublic Affairs, Baruch College, 17 Lexington Ave., Box C4011, New York
NY 10010. This article, submitted to Health Services Research on February 3, 1999, was revised
and accepted for publication onJune 11, 1999.
Qualitative Methods: What and Why?
open-ended interviews. Paradoxically, many decry qualitative methods as
inherently biased because these methods depend so much on the perspective
and skills of the researcher, while others assume that qualitative research
can be carried out by anyone regardless of their training, knowledge, and
This article explores the reasons why and the situations where it is
appropriate to use qualitative methods, and it discusses some common uses
of these methods in health services and health policy research. It presents a
spectrum of qualitative methods that reflect the various ways in which they
can be used. It closes with a few examples of qualitative research to illustrate
WHY USE QUALITATIVE METHODS?
It can be useful to conceptualize research as the process of reducing our
uncertainty about important phenomena or questions. This implies that the
development ofknowledge involves the gradual reduction of uncertainty. If
we adopt such a developmental approach to inquiry, itbecomes evident that,
at the outset, there is uncertainty not only about answers, but about what the
right questions might be; abouthow they should be framed to getmeingful
answers; and about where and to whom questions should be addressed. As
understanding increases, some ofthe right questions emerge, butuncertainty
remains aboutwhether all ofthe right questions havebeen identified. Further
along, confidence grows that almost all ofthe important questions have been
identified and perhaps framed in more specffic terms, but uncertainty still
exists about the range of possible answers to those questions. Eventually, a
high level ofcertainty is reached about the range ofalmost all ofthe possible
answers. In many (butprobably not all) cases, therefore, as inquiry proceeds,
itcan move from being quite unstructured and probably largely qualitative in
nature, to being quite structured and probably largely quantitative in nature.
We can, asAbrahamKaplanputs it (1964), movefromthe context ofdiscovery
to the context ofjustification.
One way to make this more concrete is to put it in terms ofhow to ask
questions. In this developmental framework, initial questions are very open-
ended; often the researcher just asks what appear to be relevant individuals
to describe, in their own way, their experiences and responses concerning a
given situation or issue. Over time, questions remain open-ended butbecome
more specific, either in their initial wording or in accompanying "probes."
HSR: Health ServicesResearch 34:5Part II (December 1999)
As we move to more quantitative research, questions become closed-ended,
that is, they have specified response options; in addition, the sequence of
questions has become predetermined rather than left to the discretion of the
researcher. As Patton has noted (1990: ch. 7), there is a discipline, as well as
an art, to asking truly open-ended questions: genuine awareness of whether
you are asking an open-ended question can be viewed as one of the core
disciplines of the qualitative researcher.
Another way to shed light on reasons to use qualitative methods is to
address theissue oftheory. Health servicesresearchers tend toborrow theories
from social science disciplines (e.g., economics, psychology, organizational
behavior, political science, sociology, and anthropology). They have done
little to develop full-scale, independent theories. It can be argued that the
field has reached the point at which it has become most valuable to pay
greater attention to the construction oftheories that integrate understandings
from our multiple disciplinary roots. A shared set of concepts or constructs
derivedfrommore integrative andgeneric theories couldhelp ensure thatkey
questions are addressed and that key variables are measured, and measured
consistently, across studies in widely different settings and over time. This
would contribute greatly to our ability to build a coherentbody ofknowledge
with perhaps greater staying power.
Qualitative methods are quite useful in constructing or developing
theories or conceptual frameworks or, to put it another way, in generating
hypotheses. These methods can also be used in refining theories and hy-
potheses through preliminary testing. There is an important parallel to be
drawn here between rigorous qualitative and rigorous quantitative research.
In quantitative research, building on epistemological assumptions regarding
whether it is ever possible to confirm (rather than disconfirm) a hypothesis,
researchers typically try to reject anull hypothesis. An important insightfrom
Patton is that when qualitative researchers begin to see a pattern that might
be articulated as a hypothesis, they should then search systematically for
evidence that will lead to the rejection of the hypothesis (Patton 1990: ch.
8). When qualitative researchers embrace this level of discipline, they take
important and concrete steps to protect against bias, whether intentional or
Although qualitative methods are more frequently used in theory de-
velopment and refinement, some argue that they can also be used in test-
ing theory. No less apparently quantitative a social scientist than Donald
Campbell has argued that qualitative methods can be used to collect the
data needed for experimental studies (1975). Some qualitative researchers
QualitativeMethods: What and Why?
may argue that this is an inappropriate use of their approach, because it
involves a significant shift away from the type of research that involves, as
Charles Ragin (1994) has noted, in-depth, multi-dimensional exploration of
a limited number of complex wholes. If one views the use of qualitative
methods as a continuous rather than a dichotomous variable, however, it is
easier to conclude that qualitative methods can inform and be informed by
typical practice in quantitative research, and that it can provide ways to make
reliable observations that would not otherwise be possible.
USING QUALITATIVE METHODS IN
HEALTH SERVICES AND POLICY
In the field of health services research, qualitative methods have been used
to describe many kinds ofcomplex settings and complex interactions. These
include interactions amongpatients, families, and clinicians; within, between,
and among professional groups and organizations; in communities; and in
markets. Classic studies by Anselm Strauss and Barney Glaser, for example,
used qualitative methods (what they called "grounded theory" ) to
elucidate relationships among professionals and between professionals and
patients. Examples of the continuing tradition include Glaser's work (1965)
on the issue of whether or not to tell patients they are dying, Strauss's work
highlightingthevirtuallycontinuousprocess ofnegotiation inthe specification
of roles and relationships in health care settings (1978), and Corbin and
Strauss's more recent work on the life experience of family members who
care for people with chronic illnesses in home settings (1988).
Qualitativeresearch also plays animportantpartin clarifyingthe values,
language, and meanings attributed to people who play different roles in
organizations and communities. They allow people to speak in their own
voice, rather than conforming to categories and terms imposed on them
by others. Often, qualitative researchers find that they are giving voice, in
particular, to thosewho are otherwise rarelyheard, such aspatients orworkers
far down in the hierarchical chain of command. This latter tradition can
be seen as the application of the techniques of anthropology. The research
methods ofanthropology were designed both to describe and to understand
distinctive cultures in far away places. They have proved to be very useful,
as well, in understanding groups ofpeople with similarly distinctive cultures
who live in our nation, who are often either almost invisible (e.g., patients
1106HSR: Health Sennces Research 34:5PartII (December 1999)
in nursing homes or mental hospitals), alienated from mainstream society
(e.g., people with tuberculosis who are resistant to participation in treatment
programs), or simply not familiar to average Americans (e.g., many immi-
grant groups). From Goffman's classic study of mental hospital patients as
"inmates" (1961) to Diamond's more recent study ofnursinghomes (1992), a
sensitivity to the ways in which realities and roles are socially constructed and
reconstructed has helped illuminate the unspoken assumptions underlying
relationships between "health care providers" andpeople with relatively litde
power in society.
Qualitative methods are also very useful in inquiries into developmental
andhistoricalprocesses within institutions, communities, andmarkets. One of
the great advantages of qualitative methods is that they enhance the capacity
not only to describe events but to understandhow andwhy the "same" events
are often interpreted in a different, sometimes even conflicting manner, by
different stakeholders. In the context ofpolicy research in particular, qualita-
tive methods have been used to document the perspectives and interactions
amongmultiple stakeholders. This is ofgreatvaluein studies ofpolicymaling,
ofpolicy implementation, and even ofpolicy consequences.
A conundrum is inherent in this extraordinary capacity of qualitative
methods to capture the relative rather than the absolute nature of"truth." Do
we learn more about our informants or do we learn more about the events on
which they are commenting? At least some readers have probably seen the
film Rashomon, the directorial debut of the late greatJapanese director Akira
Kurosawa. This remarkable work, shot in the 1950s but set centuries ago, tells
and retells the story of a robbery, rape, and murder from the perspective of
multiple participants and observers, including the ghost ofthe victim. As the
film unfolds, the audience learns a great deal not only about the meaning of
the events to different people, but about the motivations, too, that lead them
to present events as they do. Often, these motivations reflect quite human
desires to maintain what the French call amourpropre, or what others might
call simply "image." It is not clear that anyone "knows," at the end, what
really happened. Some mightargue thatany single statement ofevents would
necessarily be less than complete or accurate. Qualitative researchers often
use the term "triangulation" to describe the process of examining different
perspectives in order to identify at a minimum what all informants seem to
agree tookplace, orwhatitmeans (Miles andHuberman 1994; also seePatton
in this issue). This is almost as if we assume that the truth exists only in the
space where multiple Venn diagrams converge. Perhaps it is more honest to
Qualitative Methods: What andWhy?
admit thatsome ofthe truth maybe found in the places in the diagram where
the circles do not converge.
Qualitative methods are frequently used in health services and health
policy evaluations (Caudle 1994; Scheirer 1994). In some ways, evaluative re-
search is aspecialkindofsocialresearch, inwhichthe dependentvariables are
by definition normative. In some cases the "desired outcomes" ofa program
are well known in advance, but in other cases the full range of outcomes
(desired and perhaps not so desired) of interventions are not fully known
at the outset. This is where the enhanced "peripheral vision" of qualitative
methods can be of special use. As an evaluation researcher, however, I have
found that the greatest value of qualitative methods in evaluative research
has been in providing a far more sophisticated approach to the specification
of the "independent variable," that is, the intervention being evaluated. All
too often, the independent variable in evaluative research, in health as well
as in other fields, has been conceptualized as dichotomous: people either
got the "intervention" or they did not. To anyone who has been close to
the process of implementing programs, particularly novel programs, this is
naive at best. Over 20 years ago, in compiling a meta-evaluation ofsubstance
abuse preventionprograms, colleagues examined over 100 studies ofdifferent
interventions to determine both the rigor ofthe evaluation methods used and
differences in the outcome across categories of intervention (Schaps et al.
1981). In a substantial subset of the studies, not enough data were available
on the nature ofthe intervention to categorize it even atthe mostrudimentary
level. Those studies presented quite a bit of outcome data but almost no
process data. In many cases, in sum, it was impossible to know exactly what
intervention was being evaluated. In contrast, the "rich description" capacity
of qualitative methods can result in a far more complete, and often far more
compelling articulation of the intervention, one that can be used both in
helping to explain outcomes and in encouraging the adoption of effective
Furthermore, "there is many a slip between cup and lip," and pro-
grams are rarely implemented exactly as planned. In multi-site evaluation
studies, considerable variation typically occurs in the nature and extent of
implementation of the "same" intervention across sites. And the level and
nature of implementation can also change over time, either increasing or
decreasing-or just swinging back and forth. Qualitative methods have been
invaluable in exploring issues such as these: How did those implementing
the intervention perceive the intentions of the original planners? Did they
have the same agendas or different ones? Did they try to implement the
HSR: Health Services Research 34:5Part II (December 1999)
original plan? If not, why not, and what did they try to implement? What
problems did they face in implementation? What new opportunities arose
that they wanted to take advantage of? What problems did they solve and
how? How did these solutions affect the intervention? Did they try to change
the intervention in response to early perceptions of what was and was not
working? What changes did they succeed in making? What changes were
they unable to make?
Within the field of evaluation, distinctions are often made between
formative and summative studies (classic references include Cronbach 1963
and Scriven 1967). Formative studies are designed to provide early feedback
to program operators. But process evaluation and formative evaluation often
become confused with each other. Although most formative studies pay a
lot of attention to issues of process and program implementation, they can,
and sometimes do, use both early quantitative data and qualitative meth-
ods (such as focus groups and interviews) to explore preliminary outcomes.
Summative studies are best conducted when an intervention has stabilized,
when its implementation has become as close to complete as it is likely to
get, and when the intervention has become consistent over time and place.
Qualitative methods can make acontributionnotonlyinidentifingthe point
at which stabilization has occurred (seldom as soon as we diink), but also in
documenting in detail the "final" intervention and in providing data on both
context and outcomes.
Finally, qualitative methods have a clear role in the development of
methods and materials. In particular, focus groups and one-to-one cognitive
testing have become state-of-the-art tools for the development ofreliable and
valid survey instruments. Focus groups are used to get an initial sense of the
dimensions thatare ofparticularrelevance to atopic and setofrespondents, so
that these dimensions can be included in a survey. Cognitive testing is used
to ensure that the items and response options on a survey are interpreted
consistently across potential respondents as the survey developers intended
(see, for example, the Consumer Assessment ofHealth Plans Study described
by Hurley in this issue). Underlying both of these activities is a fundamental
strength of qualitative methods: their ability to explore meanings and, in
particular, meanings ascribed to events and circumstances by actors rather
The same methods are also used to develop the messages and materials
used in social marketing campaigns and educational interventions. Focus
groups can be of particular assistance in identfying "hot button" concerns
ofpotential audiences and in helping to craft key messages. Cognitive testing
Qualitative Methods: What and MWy?
can help determine whether people are in fact "getting the message" from
specific materials and products and whether these productshelpmovepeople
to the decisions and actions that the developers have in mind.
A SPECTRUM OF METHODS
A wide spectrum of qualitative research methods can be used, and has
been, in health services and health policy research. The "purest" form of
qualitative research, with roots in anthropologyand sociology, is oftenknown
asethnographyornaturalistic inquiry. Naturalistic inquirymostofteninvolves
along-term exposure to a settingand/or agroup ofpeople (Lincoln andGuba
1985). The investigator makes extensive use of unstructured observations
as well as conversations (they can hardly be called interviews). Usually,
detailed field notes are kept of observations and conversations, and in the
best research, discipline is used in keeping boundaries set between what is
observed and the observer's related responses and interpretations. In some
cases, investigators may also gather and analyze the content of a wide range
of "documents" (official and unofficial) or other "traces" ofphenomena and
events. This kind of inquiry is most likely to be used when the level of
uncertainty is high (when we are not sure what questions to ask of whom),
when there is little or no theory to direct our attention, and when situations
are novel or complex.
One subset of this kind of inquiry involves not just observation, but
"participant observation," in which the investigator becomes, quite explicitly,
apart ofthe settingorprocessbeingexamined. Forthose steepedin the Carte-
sian duality between observer and observed, this approach is indeed hard to
fathom. Yet in many settings, it can be impossible (or as near as makes no
difference) to get sufficient exposure to a setting or agroup withoutbecoming
a participant For example, many ethnographers assert that it is impossible
to understand the life experience ofsomeone in a "total institution," such as
a prison or a mental hospital, without notjust spending a good deal of time
in such an institution but also being subject to the same constraints as those
who find themselves forced, by circumstance or the law, to be there. But
participant observation does not have to be quite so dramatic. Over 20 years
ago,whenIwas (andlooked) alotyounger, Iparticipated inaweekly"support
group" foryoung adultswho were atriskfor serious drugor alcohol problems.
This participant observation was part ofan evaluation ofan innovative multi-
intervention substance abuse prevention program. Because of the ground
HSR: Health Services Research 34:5PartII (December 1999)
rulesofthegroup, aperson couldnotreallycome to even one meetingwithout
being initiated into membership. The young people were quite comfortable
with adults being present, and the reasons for my presence were explained
to all of the members. I must admit that I had a good time. More important,
I learned far more by attending a few group meetings (about the kinds of
young people who participated; about the dynamics of the group; about the
style used by the group leaders; and, in fact, about the mismatch between
the program designers' expectations and the actuality ofthe program) than I
could have learnedbyusing eithermore structured qualitative methods, such
as interviews, or more closed-ended methods, such as surveys.
A more common set ofqualitative methods are those used in case study
research (e.g., Yin 1984). Some studies involve a single case, or an event
history; others involve multiple cases, chosen either at random or more
often purposively to meet the requirements of predetermined typologies.
Case study research can also be fairly long-term, but typically it does not
involve continuous immersion in a setting or group. Rather, it involves from
one to perhaps five short but intensive exposures to a setting or group. In
some cases, a single researcher goes on site alone; in other cases a small
team of researchers is used. In virtually all instances, case studies use not
only purposive sampling of sites, but purposive sampling of informants and
experiences.2 By definition, this means that the investigators have at least
some idea ofwhat they are looking for and where they need to go to find it.
Theimportantvariableshave alreadybeenidentified, and adecisionhasbeen
made about whether to explore the variable or control for it in the sampling
The methods used in case study research are similar to those of natu-
ralistic inquiry, except that data collection is typically more structured. The
mostcommon methods used are key informant interviews, structured obser-
vations of events and interactions, and the collection and content analysis of
relevant documents. Interviews can vary from the unstructured to the highly
structured, but the questions remain overwhelmingly open-ended. Typically,
a list of questions is specified and there are "probes" associated with certain
questions.Just as typically, the field researcher is given considerable latitude
about the exact wording of the questions and their sequencing. It is very
common for an informant to start answering a question that has not yet been
asked; at that point, the experienced researcher exercises judgment about
whether to follow the lead of the interviewee or to return to the sequence
in the interview protocol. It is also common for case study researchers to
use quantitative data. For example, in studying a community, researchers
QualitativeMethods: What and Why?
can and often do take advantage of the available data on the community's
demographic characteristics and/or health status. A researcher studying an
organization or a marketmay want to examine data on health care utilization
and/or financial information. These data may simplyprovide background, or
they may generate specific questions to ask about the informants' interpreta-
tion(s) ofcertain data or about their perceptions ofthe impact ofthe situation
as reflected in the data.
Structured observations ofmeetings and events are probably not used
as frequently as are key informant interviews in case study research. Such
observations, however, can be invaluable in providing investigators with a
direct experience of the ways in which actors interact in a setting. Concerns
are often expressed about whether the presence ofthe researchers influences
these interactions so extensively that the interactions are no longer authentic.
Many researchers discover that experience teaches greater humility: after a
little while, a skilled and disciplined observer does not have that much ofan
effect on the interactions. Exceptions clearly occur, however, and it can be
difficult, especially for the less experienced field researcher, to assess whether
he or she has witnessed a naturally occurring event or a well-staged drama.
Content analyses ofdocuments can also vary. In some cases, documents
are being examined in order to identify the "facts" of a situation or a series
of events. In other cases, again closer to the spirit of naturalistic inquiry, the
analysis focuses more on the meanings embedded either in language or, in
some cases, in images. This kind of analysis typically involves the identifi-
cation of assumptions, values, and priorities. It can also help to illuminate
differences in the perception of similar events across different actors.
The two last methods have already been mentioned: focus groups
(Morgan 1988) and cognitive interviews. These methods are typically used
to explore highly specific issues. Nevertheless, they retain the quality that all
information-gathering efforts should have: leaving lots of room for investi-
gators to be surprised. The focus group involves bringing together a group
of individuals chosen to meet a specific profile of characteristics. Typically,
groups are intentionally homogenous along some dimensions and heteroge-
neous along other dimensions. A structured but still informal setting is used
to explore a limited number of"focus questions." Focus groups are best used
instead of individual interviews when it is clear that the interactions among
group members will be as illuminating as the statements ofany individual. As
with other qualitative methods, focus groups are often combined with more
quantitative approaches such as surveys. Briefsurveys are often administered
atthe end ofthe group discussion. In addition, recent technological advances
HSR: Health Services Research 34:5PartII (December 1999)
have made it possible for focus group members to respond anonymously
to closed-ended questions posed by the group facilitator. In some cases the
statistical distribution ofresponses can be quickly calculated and graphically
displayed on a computer screen to become grist for additional discussion.
Sometimes responses are tested at multiple points in time to determine
whether they are affected by additional information or ideas presented to
Cognitive interviews are a still more specific data collection method. In
their typical use in survey development, for example, a small number ofone-
on-one interviews are conducted with people who would meet the criteria
for completing a particular survey. In each interview, the person is asked
to complete all or a part of the survey instrument being tested. Sometimes
the persons interviewed are asked to "think aloud" as they go along, restating
questions (and sometimesresponse options) in theirownwordsandindicating
the cognitive process theyuse inchoosingan answer. Atothertimes the survey
iscompleted first (although the interviewee is sometimes able to ask clarifying
questions along the way), and the interviewees are then "de-briefed" to go
over their interpretation of the questions and the reasons why they gave the
answers they did. This is certainly not naturalistic inquiry. But itdoesreinforce
thenotion thatqualitative methods are excellent athelpingusunderstandhow
people perceive and interpret language and their own experiences.
EXAMPLES OF THE USE OF QUALITATIVE
Here are some examples of the use of qualitative methods in my experi-
ence. The first is from a study in which my role was relatively minor. The
SUPPORT study is a multi-site, and largely quantitative, study of decisions
made when a patient in an intensive care unit is, or may be, close to the
end of life. In the last phase of the study, nurses were assigned to ICUs to
provide fairly unstructured assistance to both family members and health
professionals facing difficult decisions about whether and how to intervene.
Nurses were asked to select every "nth" patient, whom they "entered" in the
study, and to keep a daily journal to record their experiences in working
with the patient, the family, and health professionals. The contents of this
dailyjournal were loosely structured into four or five issues. The goal was to
collect detailed qualitative data on the interaction of the nurses with others,
and on the dynamics among the various "players" in the intense drama of
QualitatieMethods: What and Why?
the end of life, in order to understand better the effects ofthese dynamics on
whether decisions were made and what those were; and by whom, at what
point, and with what apparent consequences.
At this stage ofinquiry into end-of-life decision making, there was some
understanding, but it was far from complete, of who might be involved,
in what ways, and when. Little was known about those issues that would
be perceived as significant by different actors. It was clear that professional
perspectives mightwell be divergent from those ofthe patient and the family.
Clearly, this study was investigating ahighly sensitive situation, and the nurse
observer was also an active participant in the drama, reporting not only on
what she was (and was not) able to do with respect to engaging people in a
dialogue, but also on their reactions to her attempts. These qualitative data
are still being fully analyzed, but even preliminary analysis highlighted the
particular importance of divergent opinions within families, and between
family members and patients, as a barrier to decision making.
A second example is a study under my direction that examined the
implementation, at the federal and state levels, of legislation known as the
SynarAmendment, after itsnow sadly deceased sponsor, Congressman Mike
Synar of Oklahoma. This amendment required that states pass (if they had
not already) a bill prohibiting the sale of tobacco products to those under
18. In addition and more significantly, the amendment required that states
document that they were enforcing the legislation. The "stick" in the legisla-
tion made it both controversial and interesting: states that did not document
meaningful enforcementrisked losing, progressively, from 10 to 40percentof
their substance abuse block grant. The goal ofthe studywas to document and
perhaps assess the extent ofimplementation, to identify factors thathelped or
hindered implementation, and to observe unexpected or at least unintended
consequences ofthe legislation. The study examined implementation both at
the federal level and in a purposive sample of states, and it was conducted
over several years in order, as originally designed, to see if the "sticks" in the
legislation were ever needed and were ever used. The study built on a mix
of theories, including interest group theory and interorganizational relations
theory, but it sought to refine these theories and extend their application
to fairly new ground. A number of potential factors helping or hindering
implementation, especially at the state level, could already be identified at
leaston apreliminarybasis. Mostofthekeystakeholderswere alreadyknown.
Because of this, both states and informants at both the federal and state level
could be sampled to address key issues and stakeholders. One particular
interest was inhow substance abuse professionals andpolicymakers, who had
HSR: Health Services Research 34:5Part II (Decernber 1999)
nothistorically interactedwith those in the tobacco controlworld, would deal
withthechallenge inherentin thisnewlegislation. Themethodsusedincluded
semi-structured key informant interviews, tracking of legislative changes,
and content analysis of key documents (including comments made on the
Notice ofProposed Rule Makingfor the amendment). Perhaps the mostnovel
method used was a content analysis of interactions among tobacco control
activists who participated in a pre-World Wide Web computerized strategy
exchange. Tracking these interactions provided a good deal of information
on the activists who weretrulyengaged in the issue and those who were not.
The information proved quite surprising.
The history of this study shows one important value of qualitative
methods: their ability to deal with unforeseen events. The implementation
ofthe Synar Amendment was significantly affected by such events, including
health care reform efforts (which dominated the focus ofmany federal health
officials); the massive shifts in political power initiated by the election of
1994, at both the federal and state level; and the introduction by the Clinton
administration of a far broader and more ambitious youth tobacco initia-
tive spearheaded quite unexpectedly by the Food and Drug Administration.
These events created delays in implementation thatkept the "outcome" in an
ever further retreat, and it required the examination of new issues and new
stakeholders. But it also provided an opportunity to generate unexpected
insights. For example, a clear difference could be identified between the level
of national leadership engagement for the Synar Amendment, a purely con-
gressional initiative, and the FDA Youth Initiative, very much an executive
branch initiative closely linked to important broad themes of the Clinton
administration. The flexibility ofour design permitted us to expand our data
collection efforts to new key informants andnew documents. Reinterviews of
the same informants over time in response to unexpected events also proved
The final example also draws from tobacco control research, but it
can be seen more broadly as an evaluation of the extent to which multi-
organizational coalitions contribute to the passage of desired state policies
and the identification of coalition attributes that are most instrumental in
their effectiveness. This study, of the Robert WoodJohnson Foundation's
SmokeLess States Initiative, was driven by a fairly detailed conceptual frame-
work thatdrew from multiple theoretical perspectives and the work ofseveral
researchers.3 The conceptual framework articulated a set ofintermediate out-
come variables for the work oftobacco control coalitions (the passage ofkey
legislation or ordinances, the institutionalization of tobacco control efforts,
QualitativeMethods: What and Why?
and a sustained commitment ofparticipants to the coalition); a set of contex-
tual variables thatwould mediate coalition effectiveness (e.g., public attitudes
toward tobacco control at baseline, community experience in collaborative
efforts); and a sequence of independent variables relating to the structure
and functioning of the coalitions themselves. After preliminary research, a
"theory ofaction" (Patton 1997) was also developed for the SmokeLess States
Initiative that articulated a series of assumptions about the ways in which
coalitions would make a difference in tobacco control policy.
The study involved multiple site visits over two to three years to 12
program sites. A multi-disciplinary team of field researchers, proximate to
the sites, was used; each site was studied by a single researcher and most
researchers were responsible for two sites. This study is an example of a
gradual shift from a largely qualitative to a mixed qualitative and quantitative
methodology, aslearningandunderstandinggrew. Itbeganwith, andretained
until the end, typical case study research methods described earlier: in-depth
interviews with key informants; structured observations ofkey meetings and
other events; and the content analysis ofdocuments.
Even at the outset, given the explicit conceptual frameworks, instru-
ments were fairly structured. Interviews were guided by a modular interview
protocol that was actually far too long to be administered at one time to
a single informant. Instead, in the course of selecting informants, the field
researchers used a reconnaissance technique to identify which persons knew
the most about different topics, to avoid wasting time asking questions on
which people did not view themselves as knowledgeable. (Used correctly,
this approach also helps to ensure that a research team has the right set of
informants overall.) Observations were also structured by detailed protocols.
Site visit reports were developed and, at the end ofthe period ofobservation,
integrated case studies were written, using a common format. This format
specified clearly a place for the interpretive analysis and comments of the
researcher as well as a place for the researcher to assess whether and how she
or he might have had an effect on the site over the course of the study.
After about two years offield research, enough qualitative material had
been gathered to permit the research team to develop and test a closed-ended
survey instrument that would be administered to all coalition members or a
substantial sample of them. Our conceptual frameworks provided excellent
guidance about the topics to include in the survey; the years offield research
made it possible to specify the likely range of responses to questions. After
multiple iterations and expert review, the survey was cognitively tested by
being administered to people similar to the expected respondents (in this
HSR: Health ServicesResearch 34:5Part II (December 1999)
case, members of local tobacco control coalitions in states not included in
our sample). The findings from initial administration ofthe survey have been
"fed back" to the sites.
The research team isnow carrying out an integrative, cross-site analysis
ofboth qualitative and quantitative data; we are also developing briefer case
studies designed to highlight key "lessons learned" from the research. The
integrative analysis uses key common variables addressed in the qualitative
research, typically but not always by creating dichotomous variables; quanti-
tative data from the administration ofthe survey; and, finally, supplementary
secondary data on environmental characteristics (e.g., items from the Area
Resource File); changes in state and, to the extent available, local tobacco
control policies; and public opinion data drawn from surveys supported by
the Foundation through another grant. We believe, however, that our ability
to interpret the results of this more quantitative analysis will be substantially
enhanced by our experiences in the field.
Asnoted at the outset, health services andpolicyresearch isyoung as afield of
inquiry. Atthis stage ofitsdevelopment, researchers andfunderscannotafford
to ignore the potential contributions ofqualitative methods inidentfyingim-
portant questions, in building the capacity to conduct and replicate research,
and in constructing useful theories. However, the contributions of qualitative
research will not be maximized unless the methods are applied with rigor
as well as creativity. The purchasers and consumers of research have every
right to demand vigilance in ensuring that those who design and conduct this
research have the right training and experience and that they aclnowledge
and, to the extent possible, protect against the investigator-dependent nature
of this research.
1. Indeed, as descibed in one of our examples, qualitative research is particularly
useful in dealingwith unanticipated events. Luft (1986) has also described amixed-
method study in which qualitative research findings not only helped point to
significant problems and anomalies in the secondary, quantitative database his
research team had planned to use, but proved to be the only truly reliable data
source to address their questions.
QualitativeMethods: What and Why?
2. Purposive sampling is often combined with "snowball" sampling, in which ad-
ditional informants or events that deserve investigation are identified as you go
along. Note that purposive sampling can also be designed to look not only for
"typical" cases, but for atypical or extreme cases or for cases that apparently do
not fit the underlying assumptions or hypotheses of the research. Thus, it can be
designed specifically to reduce the potential for bias and support, again, the need
to look for "disconfirmation" ofemergent hypotheses.
3. Ironically, the conceptual framework was originally designed to guide an as-
sessment of coalitions set up under the National Cancer Institute's quite similar
tobacco control initiative, the ASSIST project. The qualitative nature of this
research, however, contributed to the decision that such an assessment would
not be carried out.
Babbie, E. 1998. The Practice of Social Research, 8th ed. Belmont, CA: Wadsworth
Bowling, A. 1997. InvestigatingHealth andHealth Services. Briston, PA: Open University
Campbell, D. T. 1975. "Qualitative Knowingin Action Research." In TheSocial Context
ofMethods, edited by M. Brenner, P. Marsh, and M. Brenner, pp. 184-209.
London: Croom Helm.
Caudle, S. L. 1994. "Using Qualitative Approaches." In Handbook ofPracticalProgram
Evaluation, edited byJ. S. Wholey, M. P. Hatry, andK E. Newcomer, ch. 4. San
Corbin,J. M., and A. L. Strauss. 1988. Unending Workand Care:ManagingChronic Illness
atHome. San Francisco:Jossey-Bass.
Cronbach, L.J. 1963. "Course Improvement Through Evaluation." Teachers Collge
Record 64 (5): 672-83.
Diamond, T 1992. Making Gray Gold:Narrativesfrom Inside theNursingHome. Chicago:
University ofChicago Press.
Glaser, B. 1965. Awareness ofDying. Chicago: Aldine.
Glaser, B., and A. Strauss. 1967.
Goffman, E. 1961. Asylums: &ays on the Social Situation ofMental Patients and Other
Inmates. Garden City, NJ: Anchor Books.
Kaplan, A. 1964. The Conduct ofInquiry. San Francisco: Chandler Publishing Company.
Lincoln, Y. S., and E. G. Guba, E. G. 1985. Naturalistic Inquiry. Beverly Hills, CA: Sage
Luft, H. S. 1986. "Health Services Research asaScientific Process: The Metamorphosis
ofan Empirical Research Project from Grant Proposal to Final Report." Health
Services Research 21 (4): 563-84.
Miles, M. B., and A. M. Huberman. 1994. Qualitative Data Analysis: An Fxpanded
Sourcebook. Beverly Hills, CA: Sage Publications.
Morgan, D. L. 1988. Focus Groups as Qualitative Research. Newbury Park, CA: Sage
The Discovery ofGrounded Theory. Chicago: Aldine.
HSR: Health Services Research 34:5 PartII (December 1999)
Patton, M. Q 1997. Utiliation-Focused Evaluation, 3d ed. Beverly Hills, CA: Sage
. 1990. Qualitative Evaluation and Research Methods, 2d ed. Newbury Park, CA:
. 1975. Altenative Evaluation ResearchParadigm. Grand Forks, ND: University
Ragin, C. C. 1994. Constructing Social Research: The Unity and Diversity ofMethod.
Thousand Oaks, CA: Pine Forge Press.
Schaps, E., S. Churgin, and Research Team. 1981. "A Review of 127 Drug Abuse
PreventionPrograms."JournalofDrugIssues (Special Issue, winter).
Scheirer, M. A. 1994. "Designing and Using Process Evaluation." In Handbook of
Practical Program Evaluation, edited by J. S. Wholey, H. P. Hatry, and K E.
Newcomer, ch. 3. San Francisco:Jossey-Bass.
Scriven, M. 1967. "The Methodology of Evaluation." In Perspectives of Curriculum
Evaluation, edited by R.W Tyler, R. M. Gagne, and M. Scriven. Chicago: Rand
Strauss, A. L. 1978. Negotiations: Varieties, Contexts, Processes and Social Order. San
Wholey, J. S., H. P. Hatry, and K E. Newcomer, eds. 1994. Handbook ofPractical
Program Evaluation. San Francisco:Jossey-Bass.
Yin, R K 1984. CaseStudy Research: Design and Methods. Beverly Hills, CA: Sage