You are on page 1of 25

Journal of Mixed Methods Research

http://mmr.sagepub.com/

Practices for Embedding an Interpretive Qualitative Approach Within a Randomized


Clinical Trial
Vicki L. Plano Clark, Karen Schumacher, Claudia West, Janet Edrington, Laura B. Dunn, Andrea
Harzstark, Michelle Melisko, Michael W. Rabow, Patrick S. Swift and Christine Miaskowski
Journal of Mixed Methods Research published online 28 January 2013
DOI: 10.1177/1558689812474372

The online version of this article can be found at:


http://mmr.sagepub.com/content/early/2013/01/24/1558689812474372

Published by:

http://www.sagepublications.com

Additional services and information for Journal of Mixed Methods Research can be found at:

Email Alerts: http://mmr.sagepub.com/cgi/alerts

Subscriptions: http://mmr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

>> OnlineFirst Version of Record - Jan 28, 2013

What is This?

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Journal of Mixed Methods Research
XX(X) 1–24
Practices for Embedding an Ó The Author(s) 2013
Reprints and permission:
Interpretive Qualitative sagepub.com/journalsPermissions.nav
DOI: 10.1177/1558689812474372
Approach Within a http://jmmr.sagepub.com

Randomized Clinical Trial

Vicki L. Plano Clark1, Karen Schumacher2, Claudia West3,


Janet Edrington3, Laura B. Dunn3, Andrea Harzstark3,
Michelle Melisko3, Michael W. Rabow3, Patrick S. Swift4, and
Christine Miaskowski3

Abstract
The embedded approach is a mixed methods design that is most commonly used when qualita-
tive methods are embedded within intervention designs such as randomized clinical trials
(RCTs). Scholars have noted challenges associated with embedded procedures and expressed
concern that embedded designs undervalue and underutilize interpretive qualitative approaches.
This article examines these issues in the context of a study about cancer pain management
where qualitative methods were embedded within an RCT design. We describe our practices
for stating embedded research questions, designing embedded qualitative data collection within
the constraints of the RCT, and developing enriched understandings of the RCT through an
interpretive qualitative analysis. These practices provide guidance for intervention researchers
planning to embed qualitative components within RCT designs.

Keywords
embedding, nesting, randomized clinical trials, interpretive qualitative research, cancer pain
management

The mixed methods literature has focused a great deal of attention on identifying and describing
typologies of designs that are useful for conducting mixed methods research (see, e.g., Creswell
& Plano Clark, 2011, for a summary of 15 different typologies). Although some scholars ques-
tion the value of design typologies due to the unique decisions that must be made within the
context of any research study (e.g., Bazeley, 2010; Maxwell & Loomis, 2003), many writers
1
University of Cincinnati, Cincinnati, OH, USA
2
University of Nebraska Medical Center, Omaha, NE, USA
3
University of California, San Francisco, San Francisco, CA, USA
4
Alta Bates Summit Medical Center, Berkeley, CA, USA

Corresponding Author:
Vicki L. Plano Clark, 475A Dyer Hall, University of Cincinnati, Cincinnati, OH 45221-0049
Email: vicki.planoclark@uc.edu

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


2 Journal of Mixed Methods Research XX(X)

note their utility for describing and naming basic designs that represent different logics for com-
bining quantitative and qualitative approaches (e.g., Creswell & Plano Clark, 2011; Greene,
2007; Morse & Niehaus, 2009; Teddlie & Tashakkori, 2009). These basic mixed methods
designs are useful when they provide models for how to combine the quantitative and qualita-
tive components of a study, identify important issues inherent within the different models, and
advance terminology for communicating the overall approach used in a study.
The embedded (or nested) design is included in several of the available typologies. It is
defined as a design where one methodology (e.g., qualitative) is placed within the framework of
another (e.g., quantitative; Caracelli & Greene, 1997; Creswell & Plano Clark, 2011; Greene,
2007). There are many possible variants of an embedded design, such as embedding quantitative
methods within a qualitative case study framework (e.g., Christ & Makarani, 2009) or embed-
ding qualitative methods within a quantitative longitudinal framework (e.g., Turner-Cobb et al.,
2010). To date, the literature has focused on embedding qualitative methods within quantitative
intervention designs, such as randomized clinical trials (RCTs). A growing number of published
studies have used ‘‘embedded’’ or ‘‘nested’’ RCT designs across the health and social sciences
(e.g., Brady & O’Regan, 2009; Donovan et al., 2002; Kaptchuk et al., 2009; Lipman et al.,
2010). Scholars have both critiqued the limitations of traditional RCTs and discussed the bene-
fits and utility of using qualitative methods in conjunction with RCTs (e.g., Creswell, Fetters,
Plano Clark, & Morales, 2009; Flemming, Adamson, & Atkin, 2008; Miller & Crabtree, 2005;
Nastasi & Schensul, 2005; Sandelowski, 1996; Schumacher et al., 2005; Song, Sandelowski, &
Happ, 2010).
Despite the interest in embedding qualitative methods within RCTs, several issues remain to
be resolved in terms of how this design is applied in practice. Song et al. (2010) noted that qua-
litative methods within intervention studies are ‘‘still too often not planned, and tacked on with-
out much demonstrable thought’’ (p. 729). Creswell and Zhang (2009) stated that the
procedures for embedded designs are ‘‘underconceptualized’’ (p. 615). These concerns suggest
the need for more attention to the practice of embedding qualitative methods. Furthermore, sev-
eral scholars have argued that researchers using mixed methods designs guided by postpositivist
frameworks (such as embedded RCTs) do not take full advantage of the insights that might be
gained through the inclusion of interpretive qualitative approaches because they tend to under-
value and underuse interpretive approaches in practice (Giddings, 2006; Howe, 2004).
Collectively, these writings suggest that intervention researchers do not have sufficient gui-
dance for anticipating and navigating the decisions and issues inherent when embedding inter-
pretive qualitative methods within an RCT framework.
This article aims to contribute to these discussions of the embedded design by describing
our experiences with embedding interpretive qualitative methods within the context of an RCT
study that compares two doses of an intervention to help oncology outpatients manage cancer
pain. Several years into the implementation of the RCT study, we were struck (and even some-
what surprised) by the interpretive direction that our analytic process had taken. Finding that
the interpretive nature of the qualitative analysis exceeded our expectations led us to examine
the literature on embedded designs in more depth and to reflect on our specific practices in
designing the qualitative component of the RCT study. From this examination, we identified
three dimensions of embedding within the context of our study as related to the research ques-
tions, the data collection, and the data analysis. Using this framework to organize our reflec-
tions, we considered the key practices and tensions that occurred in our conceptualization and
implementation of our embedded design’s research process.
We begin by briefly describing the background of the ongoing study that provides the con-
text for this discussion. Next, we turn to the available literature on embedding, which suggested
a framework that we found useful for considering our practices and the tensions involved in

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 3

implementing this study. Using this framework, we discuss how our research questions fit an
embedded approach, how our design of the embedded qualitative data collection methods was
constrained by the parameters of the RCT study, and how our analytic process fostered an inter-
pretive qualitative analysis despite the constraints of the larger RCT study. We conclude with
the lessons learned from our experiences in this one embedded study. Our experiences can serve
as one model of the types of practices and tensions that occur for intervention researchers who
choose to embed qualitative approaches within their studies. In addition, this discussion of mul-
tiple dimensions of embedding within one study may help methodologists further conceptualize
the facets of embedding as a mixed methods design.

Context of the Current Pain Management Study


This discussion is focused on our experiences in an ongoing, 5-year empirical study titled,
‘‘Improving Cancer Pain Management through Self-Care’’ (Miaskowski, 2007-2012). We refer
to this study as the ‘‘current’’ pain management study, and we begin by providing an overview
of the current study’s background, design, and research team.

Background
The current pain management study built on prior research aimed to help oncology outpatients
learn to better manage their cancer pain and prescribed pain management regimens. In the pre-
ceding study, conducted from 1995 to 2000 and referred to as the ‘‘prior’’ pain management
study (Miaskowski et al., 2004; West et al., 2003), the research team developed and tested a 6-
week psychoeducational intervention called the PRO-SELFÓ Pain Control Program. The inter-
vention was based on theories of self-care, adult learning, academic detailing, and nurse coach-
ing (West et al., 2003). It involved a nurse interacting one-on-one with patients (and optional
family caregivers) to help them develop and apply the knowledge, beliefs, and strategies
required to maximize the effectiveness of their physician-prescribed pain management regi-
mens in the home setting. Intervention sessions were audiotaped, originally for assessment of
intervention fidelity. However, initial reviews of the audiotapes yielded important insights into
patients’ struggles with pain management. Thus, funds were obtained to transcribe a subset of
the tapes and conduct a qualitative analysis of patients’ discussions with their intervention
nurses (Schumacher et al., 2005).
The prior study produced statistically significant and clinically meaningful improvements in
patients’ pain compared with the control condition (Miaskowski et al., 2004). However, a
detailed evaluation of the effectiveness of the intervention within the intervention group
revealed that approximately half the patients did not have the desired reduction in their pain
(Miaskowski et al., 2007). Moreover, the qualitative data revealed that problem solving around
difficulties with pain management was a process that took place over time and patients were
still actively involved in problem solving at the end of the 6-week intervention. These results
raised questions as to whether patients would have benefited from a longer intervention and/or
an intervention that included more frequent interactions with a nurse. Insights from the qualita-
tive component suggested ways to enhance the intervention.

Design
Building from both the quantitative and qualitative results of the prior study, the current pain
management study was designed to compare patient outcomes for two doses of an enhanced 10-
week intervention (called PRO-SELF PlusÓ). In the low-dose condition, participants interacted

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


4 Journal of Mixed Methods Research XX(X)

with the intervention nurse once per week during the 10 weeks (10 contacts for a total of about
8 hours of interaction). In the high-dose condition, participants interacted three times per week
for the first 3 weeks and then once per week for the remaining 7 weeks (16 contacts for a total
of about 12 hours of interaction). In both conditions, the interactions occurred through in-person
visits and phone calls following a predetermined arrangement throughout the 10 weeks. Unlike
the prior study, the current study had no control condition since the prior results demonstrated
that even a 6-week intervention was significantly better than standard care.
Development of the PRO-SELF Plus intervention was based on insights gained from the
qualitative analysis in the prior study. It included an expanded assessment designed to specifi-
cally address pain management difficulties identified in the qualitative analysis completed dur-
ing the prior study. It also included an assessment of past experiences with pain that may shape
pain management strategies in the present. During the first interaction, the nurse conducted an
extensive assessment of the patient’s situation and provided structured teaching around com-
mon misconceptions about pain management as well as other related topics (e.g., scripts for dis-
cussing pain management issues with clinicians, a pillbox for organizing medications). The
nurse also identified pain management problems and initiated problem solving tailored to indi-
vidual patient’s needs. Subsequent interactions involved tailored assessments and coaching with
the patient (and optional family caregiver) on strategies that fit his or her personal needs and
whatever physician-prescribed pain management regimen he or she was trying to use in the
home setting. At the initiation of planning the current study, a qualitative component was
included to examine the experiences of patients and family caregivers engaged in the expanded
intervention.

Research Team and Reflective Process


The research team for the prior and current pain management studies included individuals with
different content and methodological expertise. The development of both the prior and current
interventions and RCT studies was led by nursing researchers (CM, CW, and JE), working with
an interdisciplinary team of clinicians (LD, AH, MM, MR, and PS), who brought extensive
expertise in conducting clinical intervention research on the topic of cancer pain management.
These individuals oversaw the design and implementation of the current RCT and all data col-
lection and preparation procedures. A nurse researcher (KS) specializing in caregiver issues
and qualitative research joined the team during the prior study when the analysis of available
text data became warranted. She led the qualitative data analysis efforts in the current study. A
methodologist (VPC) specializing in mixed methods and qualitative research with a background
in educational research joined the team after the current pain management study was under
way.
At the final writing of this article, the team has completed participant recruitment and quan-
titative and qualitative data collection. Throughout the process of analyzing the qualitative data
over the previous 4 years, the first two authors actively reflected on, memoed about, and regu-
larly discussed their emergent analytic processes. These reflections led us to examine discus-
sions of the embedded design found within the literature, which provided a framework for
organizing reflections about our processes, practices, and tensions for implementing the qualita-
tive methods embedded within this RCT study.

A Framework for the Dimensions of Embedding


When the research team first conceptualized and planned the current pain management study
around 2003, little literature existed that might have informed this study’s use of an embedded

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 5

design and the epistemological and practical issues that could be anticipated. As our qualitative
analytic processes unfolded over several years of implementing the current study, we examined
the recent mixed methods literature to help us reflect on the role of the qualitative component
within the larger RCT design of the study. Overall we found only a small amount of methodo-
logical literature specific to the embedded design in comparison with the growing use of this
approach. Even so, the definitions, issues, and differing viewpoints that we found in the litera-
ture suggested a framework useful for organizing our experiences with the design and imple-
mentation of the current pain management study.
The earliest discussion of the mixed methods embedded design was found in Caracelli and
Greene’s (1997) chapter on mixed methods evaluation designs. They introduced the concept of
the ‘‘embedded or nested’’ (p. 24) design as an example of a mixed methods approach that
incorporates a high level of integration between the quantitative and qualitative components of
a study. They defined the embedded design as one ‘‘in which the study is framed by one metho-
dology within which a different methodology is located’’ (p. 26), noting an exemplar case where
a research team conducted ethnographic methods within an experimental design framework.
They argued for the value of this approach as an example of meaningfully integrating the two
types of methodologies ‘‘in a framework of creative tension’’ (p. 24).
Since that publication, several works have built on this initial definition and delineated char-
acteristics noted as central to this design (Creswell & Plano Clark, 2007, 2011; Creswell, Plano
Clark, Gutmann, & Hanson, 2003; Greene, 2007; Plano Clark & Creswell, 2008). Collectively,
this literature has focused on the research questions and methods associated with embedded
approaches. Embedded designs are defined as having an unequal priority in terms of the relative
importance of the quantitative and qualitative components for addressing the study’s research
questions. Researchers choose an embedded approach when their research questions include pri-
mary and secondary questions, where one question (e.g., the primary question) calls for a quan-
titative approach and the other question (e.g., the secondary question) calls for a qualitative
approach. Contrasted with the primary research questions, the secondary questions are described
as having lesser priority and addressing different (but related) questions that aim to enhance the
implementation or interpretation of the larger design. In relation to the research questions, the
embedded method has less priority, is located within and constrained by the larger design, and
its role within the study has been described as supplementary, subservient, and supportive.
Greene (2007) highlighted this embedded relationship as follows:

Distinctively in [the embedded] design, the secondary method follows or adheres to key parameters of
the primary method—for example, sampling or designed controls—rather than following the para-
meters usually associated with this secondary method. (pp. 127-128, italics in original)

In embedded designs, researchers start by analyzing the quantitative and qualitative databases
separately to address the different research questions and then move to more integrative strate-
gies as used in other common mixed methods designs so that the secondary results complement
and enhance the understanding of the primary research questions and results.
Although several points of agreement about the characteristics of the embedded design are
found in the literature, variations in and disagreements about definitions and the most appropri-
ate study designs exist. For example, in her review of Plano Clark and Creswell (2008),
Bazeley (2010) noted that their definition of embedding included both concurrent and sequen-
tial approaches whereas other definitions (e.g., Greene, 2007) limit the embedded design to
approaches where the quantitative and qualitative components are implemented concurrently.
Variation in what is being embedded is demonstrated in the literature by the different ways that
researchers refer to embedding within their empirical studies. For example, Donovan et al.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


6 Journal of Mixed Methods Research XX(X)

(2002) described that their study ‘‘embedded the randomised trial within the qualitative
research’’ (p. 769). Their identification of the RCT as embedded within the qualitative research
arises because they used the qualitative results to provide insights for improving their ongoing
RCT recruitment procedures. Thus, their use of the term embedding seems to emphasize the
relationship of the qualitative results and interpretations to the overall RCT design. In contrast,
other researchers invert this relationship, such as Kaptchuk et al. (2009), who reported that they
‘‘opted to embed a qualitative study within a larger RCT’’ (p. 386). Their use of embedding
seems to emphasize the relationship of the qualitative methods to the larger RCT design. In
combination, these writings suggest that multiple dimensions of embedding are used in practice
and that a need exists to more clearly articulate these different aspects.
Epistemological questions about embedding also exist. For example, in the preface of their
text, Morse and Niehaus (2009) expressed concern about the embedded design by questioning
the logic of having one method viewed as encompassed by and submerged within the context of
another method. Writing about mixed methods in general, Giddings (2006) noted the hegemony
of postpositivist epistemologies in researchers’ use of mixed methods research. Her concern is
as relevant to embedded designs as it is to mixed methods in general. Howe (2004) raised simi-
lar concerns specific to the use of qualitative methods within experiments, arguing that such
mixed methods designs typically limit qualitative methods to auxiliary, noninterpretive roles.
Together, these authors raise important concerns about the assumptions behind the use of an
embedded design and these concerns call for examples of how researchers can thoughtfully
embed interpretive qualitative methods within RCTs.
Looking across these writings, we found that the defining characteristics of and concerns
about embedding could be organized within three stages of the overall research process: (a) the
research questions, (b) the data collection methods, and (c) the data analysis and interpretation.
Within each of these stages, important issues related to an embedded study’s assumptions, pri-
ority, and procedures were raised that pushed our reflections about our experiences in the design
and conduct of the current pain management study. In this article, we use this framework to
organize our reflections of the practices and tensions that occurred during the design and imple-
mentation of each of these stages in our study. Specifically, we considered the following:

 How does embedding occur within three stages of the research process (research question
identification, data collection, and data analysis and interpretation)?
 How was the use of an embedded approach manifested in the assumptions, decisions,
and practices that we implemented in the current pain management study?

Embedding and Research Questions


A significant theme found in the literature on embedding is the nature of the research questions
that call for an embedded approach. The key characteristics of these questions as discussed in
the prior section are summarized in Figure 1. The separation of research questions both by pri-
ority (i.e., primary and secondary) and by the methodological approach called for by the differ-
ent questions is a key characteristic that distinguishes the embedded design from other mixed
methods approaches. For example, in contrast to embedded approaches, scholars advocate that
researchers using other mixed methods concurrent approaches use research questions with equal
priority (e.g., Creswell & Plano Clark, 2011) or use a combination of methods to address each
research question of interest (e.g., Yin, 2006).
Creswell and Plano Clark (2007) further distinguished the embedded design by suggesting
that in this approach, the meaning of the secondary research questions is situated within the
context of the primary questions. It is this point that seems to have raised the concern of Morse

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 7

Figure 1. The nature of research questions in an embedded design


Source: Creswell, Plano Clark, Gutmann, and Hanson (2003); Creswell and Plano Clark (2007, 2011); Greene (2007);
Plano Clark and Creswell (2008).

and Niehaus (2009) when they wrote, ‘‘To us, embedded design makes no sense. Research
studies always contribute some knowledge. Why conduct a study that you are going to ignore,
submerge, and makes no contribution to the results?’’ (p. 11, italics in original). Although it is
unlikely that Creswell and Plano Clark (2007) were advocating that researchers ignore their
own research, this important concern compelled us to revisit how we originally stated our
research questions in the current pain management study and to engage in many discussions
about the relationship among the primary and secondary questions as the study implementation
has progressed.

Stating Secondary Research Questions Embedded in the Context of the Primary


Research Questions
In the grant application for the current pain management study, the research team stated the
research questions in the form of study aims. The primary aims of the current RCT were the
following:

 To determine if there are differences between the high-dose and low-dose intervention
groups in a variety of participant outcomes (e.g., average and worst pain intensity scores)
from baseline to the end of the 10-week intervention
 To determine if there are differences between the high-dose and low-dose intervention
groups in the sustainability of a variety of participant outcomes (e.g., average and worst
pain intensity scores) at 2, 4, and 12 weeks after the completion of the intervention

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


8 Journal of Mixed Methods Research XX(X)

These aims built on the results of the prior study that suggested that individuals who did not
improve by the end of the 6-week intervention might have benefited from either a more inten-
sive intervention or a longer intervention. The current aims were designed to test whether a
more intensive 10-week intervention produced better outcomes than a less intensive 10-week
intervention, thereby calling for an RCT design. In addition to building on prior results, these
aims were implicitly framed from postpositivist assumptions (Creswell, 2009; Neuman, 2006)
such as being able to establish cause and effect (determinism), being able to narrow a situation
to a set of variables (reductionism), being able to meaningfully measure a construct such as
pain intensity (measurement), and valuing theory testing (deductive logic).
Due to the complex nature of pain management for oncology patients and the perceived
value of the insights obtained from the small qualitative component added to the prior study,
the research team included secondary questions in the current study to describe participants’
pain management experiences within the enhanced 10-week intervention. Specifically, the sec-
ondary aims of the RCT were the following:

 To describe patterns over time in pain management difficulties from patients’ and family
caregivers’ perspectives
 To describe patterns over time in pain management strategies used by patients and family
caregivers
 To describe patterns over time in the interactions of patients and family caregivers with
their intervention nurse

These secondary aims built directly on the qualitative results of the prior study that identified
issues experienced by participants during the prior 6-week intervention. By including these sec-
ondary aims at the start of the conceptualization of the current study, the research team wanted
to expand their exploration to consider how the issues, strategies, and interactions experienced
by patients and family caregivers unfolded over the 10 weeks of the enhanced intervention.
Although there are numerous approaches and methods that could be used to ‘‘describe pat-
terns,’’ the research team viewed these aims as calling for a longitudinal qualitative approach
(e.g., Saldaña, 2003) to explore how individual participants’ experiences changed over time. At
the time of initial planning and development of the grant application, the team’s focus was on
the RCT. As such, the initial expectation was that the qualitative analysis in response to the sec-
ondary aims would likely align with a basic content analysis approach that would use open cod-
ing across data time points to expand on categories that had been identified from the prior
study’s qualitative analysis. As the study unfolded, however, the qualitative analysis completed
by team members evolved into an interpretive approach. By interpretive approach, we mean an
approach that is consistent with constructivist assumptions (Creswell, 2009; Neuman, 2006)
such as aiming to understand the subjective meanings of experiences, needing to examine the
complexity of phenomena as experienced by individuals, viewing meaning as contextualized
and socially constructed among individuals, and valuing theory generation using inductive
logic.
The characteristics of research questions calling for an embedded design (as summarized in
Figure 1) were not delineated at the time the research team developed these study aims for the
grant application for the current pain management study. It is interesting, however, to examine
in which ways and to what extent these aims are indicative of the characteristics now advanced
in the literature. In the original grant application, the team labeled the research questions as
‘‘the primary aims of the RCT’’ and ‘‘the secondary aims of this RCT’’ (italics in the original).
The use of these labels served to both convey the relative priority of the aims (primary and sec-
ondary) as well as to situate all the aims (both those that called for a quantitative approach and

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 9

those that called for a qualitative approach) within the framework of the planned RCT design.
These aims called for different methods and addressed questions that were different but related
to pain management and the intervention being tested: (a) questions about quantitative differ-
ences in outcomes from the two intervention conditions and (b) questions about qualitative
descriptions of difficulties, strategies, and interactions during the intervention.
Furthermore, the study’s funding process illustrated the notion that the secondary qualitative
aims only ‘‘make sense’’ if examined within the context of the primary RCT framework. Due
to budget constraints, only the first primary quantitative aim about differences at the end of the
10-week intervention conditions was initially funded. The team sought and successfully secured
additional funding for the secondary qualitative aims (and the aim related to quantitative assess-
ments of sustainability beyond the 10 weeks). When pursuing the additional funds, the team
included the full description of the RCT primary aims and methods in the supplemental grant
application even though these components had already been funded, explaining that the details
of the RCT provided the necessary context in which the secondary qualitative aims had mean-
ing. In justifying this strategy, the principal investigator wrote,

Because the qualitative aims, as well as the sustainability aim of this grant, are contained within the
context of the funded RCT, we have chosen to submit the original grant application with the portions
that will be funded through this grant application identified in italics. It would be extremely difficult to
re-write this application with only the sustainability and qualitative aims—because it is only within the
context of the entire grant application that these aims make sense. (Miaskowski, 2007-2012, p. 1)

Therefore, the current pain management study demonstrates how we developed and stated
study aims that called for an embedded design. In addition, serendipitously, it provides a con-
crete example of how embedded secondary questions can be considered as situated in the con-
text of the primary questions. Contrary to Morse and Niehaus’s (2009) concern that an
embedded component does not aim to add new knowledge or contribute to the results, our
embedded aims clearly intended to add new knowledge about patients’ experiences. However,
from the start we viewed this knowledge (and the methods planned to gain it) as linked to the
intervention, the primary aims, and the RCT design that framed the current pain management
study. Although these secondary aims as designed do not ‘‘make sense’’ independent of the
larger context of the study, it is exactly because of the potential to gain new understandings of
the patients’ pain management process at the heart of the RCT intervention that these aims were
included and are important in the study.

Embedding and the Design of Data Collection


The most prominent theme in the embedding literature highlighted issues related to the design
of the methods within an embedded approach. Even in the earliest writings about embedding,
scholars cautioned about the challenges in planning an embedded design. For example, in the
volume where Caracelli and Greene (1997) first introduced the concept of embedding, Chen
(1997) noted that an integrated approach such as an embedded design ‘‘requires structural
changes in the methods themselves’’ (p. 70). This notion of ‘‘structural changes’’ is consistent
with Greene’s (2007) more recent description of the embedded design as involving the design
of one method so that it adheres to the parameters of the other method. Her examples of such
design decisions include issues of sampling and level of control. This focus on design decisions
suggests that embedding includes a form of mixing that occurs at the level of the design, where
the assumptions and design requirements of one approach place constraints on the design of the

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


10 Journal of Mixed Methods Research XX(X)

other approach. That is, in an embedded design there is an interaction between the quantitative
and qualitative approaches during the design process, and we find that the design process forms
a ‘‘point of interface’’ (Morse & Niehaus, 2009, p. 25) that is different from and in addition to
the interactions or mixing that occur during data analysis and results interpretation.
The literature on embedding therefore led us to focus on the constraints placed on the design
of our embedded methods. Of course, all design decisions are shaped by constraints such as
available resources, ethical considerations, and the researchers’ assumptions. What is salient to
embedding is that in addition to these usual considerations, the decisions about the embedded
method are also constrained by the assumptions and requirements associated with the primary
design (Greene, 2007). In the case of an embedded RCT, the constraints required by the pri-
mary RCT design aim to achieve a high level of internal validity so that researchers can make
strong cause-and-effect claims in the interpretation of their study results. When the embedded
method stems from assumptions consistent with those behind the primary design, then little ten-
sion may be expected to arise. For example, researchers working exclusively from postpositivist
assumptions may plan a structured, objective qualitative approach embedded within an RCT
design and face little discord in their design decisions. However, when different assumptions
are used for the two components of an embedded design, such as those associated with postpo-
sitivism and constructivism, then points of contention and discord in the design decisions may
be expected to arise.
As a means for considering these potential tensions, we juxtapose a ‘‘typical’’ qualitative
interpretive approach that is consistent with constructivist assumptions with a ‘‘typical’’ quanti-
tative RCT deductive approach that is consistent with postpositivist assumptions in Table 1. For
each major design decision—sampling, level of control, data source, data collection procedures,
validation strategies, and relationship between researchers and participants—the middle two
columns of the table highlight the design characteristics as outlined in many research methods
textbooks (e.g., Creswell, 2008; Johnson & Christensen, 2008). Although these characteristics
are a simplification of the complexity and variety found within both categories of research, they
suggest points where we might anticipate that the assumptions behind an RCT may be at odds
with and place constraints on the design of an interpretive qualitative approach. As Table 1
highlights, RCT designs emphasize the use of control and objectivity by selecting a large num-
ber of participants using specific criteria, using standardized procedures, reducing threats of
treatment bias and from confounding variables, and maintaining distance between the research-
ers and participants. In contrast, qualitative research based on interpretive assumptions tends to
study phenomena in their natural settings using a small sample of participants purposefully
selected because of their experience with the phenomena, open-ended data sources, an emergent
process where procedures change as the researcher learns from participants, validation strategies
to ensure the trustworthiness and credibility of the findings, and a close relationship between
the researchers and participants in order to co-construct meaning about the phenomenon being
studied. That is, if one were to design a study to explore the phenomenon of managing one’s
cancer pain using an approach that falls under the umbrella of interpretive qualitative research
(e.g., interpretive case study, constructivist grounded theory, or narrative design), we would
expect that study to be consistent with many or all these characteristics.
Although the literature points to the importance of the interaction at the design level for
embedding, it provides little guidance about how an RCT framework might constrain the deci-
sions made about the embedded qualitative methods in practice. We found the key decisions
outlined in Table 1 to be useful for considering the points in the design process where we
designed the qualitative data collection methods to adhere to the parameters of the primary
RCT within the current pain management study. We now describe our design in detail and dis-
cuss the considerations that led to various design decisions.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Table 1. Ways in Which the Qualitative Data Collection Methods Adhered to the Parameters of the RCT Design

How the qualitative data collection methods


Parameters typical of Parameters adhered to the parameters of the RCT
interpretive qualitative typical of the RCT design in the current pain management
Design decisions approaches design study
Sampling  Use a small sample size to provide depth  Use a large sample size to  Qualitative sample is the same convenience
about each case provide sufficient power for sample used for the RCT study
statistical analyses
 Purposefully select participants to best  Select individuals who meet  Selection criteria include adult oncology
learn about the central phenomenon specific criteria to help control outpatients with bone metastases and
for extraneous factors uncontrolled pain who consent to
participation in the RCT study
 Qualitative data collected from all
participants enrolled in the RCT study
(target N = 308)
Level of control  Study phenomena as they occur in their  Use a high level of control to  Study pain management within the
natural settings reduce threats to internal structure of two doses of the intervention
validity
 Manipulate the conditions  Design qualitative methods to not
experienced by participants introduce bias or confounding variables
Data sources  Sources include interviews, observations,  Use data sources that are  Qualitative data collection uses unobtrusive
documents, and/or audiovisual materials uniform across participants methods
and do not alter the

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


parameters of the intervention
conditions
 Choose sources best for learning about the  Sources include audiorecordings of
central phenomenon interactions and copies of notes
 Value the use of multiple types of data  Interactions and notes are generated as
parts of the intervention conditions
(continued)

11
12
Table 1. (continued)

How the qualitative data collection methods


Parameters typical of Parameters adhered to the parameters of the RCT
interpretive qualitative typical of the RCT design in the current pain management
Design decisions approaches design study
Data collection  Gather data simultaneously and iteratively  Administer the intervention  Implement qualitative data collection
procedures with analysis and data collection methods concurrent with the quantitative
procedures in a standardized methods for each case
way across all participants
throughout the
implementation of the
intervention conditions
 Use an emergent process guided by initial  Use same fixed data collection methods for
findings each case
 Engage in prolonged engagement  Do not make changes to the data collection
protocol based on emergent findings
Validation  Prolonged engagement in the field  Time of interactions defined  Emphasize researcher-driven validation
strategies by parameters of the strategies such as memoing, keeping an
intervention audit trail, and triangulating across multiple
data sources
 Share initial findings with participants (i.e.,  Keep from sharing information
member checking) with participants and
intervention nurses that might
introduce bias into the RCT
procedures

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Relationship with  Maintain a closeness with participants to  Maintain a distance with  Qualitative data recorded by intervention
participants co-construct understandings participants to ensure nurse (audiorecordings and field notes) and
objectivity patient (notes in assessment booklets)
 Qualitative analysts kept removed and
distant from participants

Note. RCT = randomized clinical trial.


Plano Clark et al. 13

Figure 2. The data collection methods used in the current pain management study
Note. QUAN = quantitative data collection; qual = qualitative data collection. Uppercase letters indicate primary
priority; lowercase letters indicate secondary priority.

Designing Embedded Qualitative Data Collection Methods Within the Constraints of


the RCT Design
The primary design of the current pain management study is an RCT because that was deemed
the best approach to address the primary aim to determine whether there are differences in a
variety of participant outcomes (e.g., average and worst pain intensity scores) between the two
doses of the intervention. Participants recruited to the RCT were oncology outpatients with
bone metastases and poorly controlled pain. They had the option to include a family caregiver.
The research team targeted a sample size of 308 to provide sufficient power for the planned sta-
tistical tests. Once participants consented to participate, they were randomly assigned to either
the high- or low-dose condition. Figure 2 provides an overview of the data collection proce-
dures in the current pain management study. There were three stages of quantitative data collec-
tion for each participant: (a) baseline assessment, (b) repeated measures during the 10-week
intervention, and (c) follow-up assessments at Weeks 12, 14, and 22. Assessed outcome vari-
ables included pain intensity (daily average and daily worst pain), analgesic intake, knowledge,
associated symptoms and side effects, and self-efficacy.
The implementation of the intervention itself was the source of the qualitative data. As
depicted by the ovals in Figure 2, the qualitative data collection occurred concurrently with the
quantitative data collection associated with the implementation of the 10-week intervention.
The main source of qualitative data was audiorecordings of the interactions (in-person and by
phone) between participants and their intervention nurse. That is, whenever the intervention
nurse interacted with the patient (and optional family caregiver) during the intervention, she
turned on a digital recorder. Other open-ended data sources included the nurses’ field notes and

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


14 Journal of Mixed Methods Research XX(X)

the notes patients made in their assessment booklets (e.g., the daily Pain Diary). These data
sources were gathered from all participants enrolled in both intervention conditions throughout
the 10 weeks that they experienced the intervention. No formal interviews were conducted and
there was no direct contact between participants and the qualitative researchers on the research
team.
Due to the embedded nature of the study, the design of the qualitative data collection meth-
ods was directly shaped by the requirements of the RCT. The last column of Table 1 sum-
marizes how six decisions made in designing the embedded qualitative methods of the current
pain management study adhered to the parameters of the larger RCT design framework. For
each of the listed decision points, the qualitative methods were designed in ways that maintained
the integrity and rigor of the larger RCT design instead of conforming to methods typically
advocated in interpretive qualitative approaches. These structural changes led to qualitative
methods that were implemented in a context where the researchers manipulated the conditions
received by participants in terms of the two different doses of the intervention (e.g., the schedule
and duration of interactions with the nurse), used a large sample size selected to control for
extraneous factors, relied on unobtrusive data collection methods gathered in the same time
frame as the quantitative measures and held fixed throughout the study, limited validation strate-
gies to those that focus on the researchers, and kept the qualitative analysts distant and removed
from the participants.
Each of the design decisions noted in Table 1 for the current pain management study were
made to contribute important qualitative findings about the pain management intervention pro-
cess while ensuring the quality of the RCT design and its ability to determine whether differ-
ences exist between the outcomes of the two treatment groups. They illustrate specific practices
used in one study to implement qualitative methods that adhered to the parameters of a larger
quantitative design. As such, they identify specific decision points that others can consider
when designing their own embedded studies. The decisions listed in the last column of Table 1
present a view of embedding that emphasizes the secondary nature of the embedded methods
and how they are constrained and limited by the parameters of the primary design. This specific
dimension of embedding is depicted in Figure 3, which highlights the primary role of the larger
design in planning and implementing an embedded approach. Although a challenge to represent
visually, this figure attempts to convey at a conceptual level how we now picture the RCT
design impinging on the design decisions for the embedded qualitative data collection methods.
This diagram and these practices are consistent with a view of embedding that stresses the con-
straints placed on the design of the qualitative methods that are embedded within an RCT
design. We next examine the practices that contributed to the emergence of an interpretive qua-
litative analytic process in the current study despite these constraints.

Embedding and Data Analysis, Results, and Interpretation


As foreshadowed by Caracelli and Greene (1997), embedding involves ‘‘creative tensions’’ that
arise when qualitative methods are constrained by a larger quantitative framework such as
described in Table 1. Although these decisions were warranted and necessary to address the pri-
mary and secondary aims of the current pain management study, it is these types of decisions
that raise concerns with the use of embedded approaches that impose constraints on the qualita-
tive methods. Reflecting on the status of qualitative methods within mixed methods, Giddings
(2006) asked whether mixed methods research is ‘‘positivism dressed in drag?’’ (p. 195), and
Howe (2004) described mixed methods experimentalism as ‘‘methodologically retrograde’’
(p. 42) because qualitative approaches are confined to auxiliary roles in experiments. Many
scholars argue that high-quality qualitative research and methods are built from foundations of

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 15

Figure 3. The nature of constraints in designing the qualitative data collection in an embedded
randomized clinical trial (RCT) design

constructivism and interpretivism, and constraining these approaches to adhere to the para-
meters of an RCT potentially limits the value of the methods for uncovering participant mean-
ings and experiences. Although the use of interpretive qualitative approaches embedded within
RCTs has been encouraged (Creswell, Shope, Plano Clark, & Green, 2006; Howe, 2004), little
guidance exists as to how these practices can successfully occur in embedded studies. We found
it helpful to consider data analysis distinctly from the other stages of the research process in
our embedded design because the relationship between the quantitative and qualitative compo-
nents of the current study evolved as the analytic process became more and more interpretive.

Developing Enhanced Understandings of the RCT Through an Interpretive


Qualitative Analysis
To understand the interpretive qualitative analysis that has occurred, we start with a brief over-
view of the basic analysis procedures occurring in the current pain management study. For the
quantitative component of the study, the team has entered the measures gathered at the various
points of time for each patient (and optional caregiver) into an SPSS database. The planned
quantitative analyses include descriptive analyses, group comparisons, and repeated measures
linear mixed models approaches to address the RCT’s primary aims. Due to the ongoing nature
of the current study, these analyses have not been completed at the time of the writing of this
article and will be reported elsewhere.
For the qualitative component of the study, the database includes verbatim transcripts of the
audiorecordings (about 12 hours per participant in the high-dose condition and about 8 hours
per participant in the low-dose condition) and electronic scans of the field notes. The two quali-
tative analysts began working with the data immediately once transcriptions started to become

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


16 Journal of Mixed Methods Research XX(X)

available (around the end of Year 1 of the current study). Over the prior 4 years, this analysis
has focused on 42 participants (with optional family caregivers) and included open coding and
theme development, creation of in-depth longitudinal within-case summaries, identification of
cross-case themes and contexts, extensive memoing, and discussion of emergent themes, con-
texts, and reflections between the primary analysts as well as with other members of the
research team. As results become available from both databases, the research team will pursue
additional analyses to integrate and mix the two types of data, results, and interpretations to fur-
ther understand both context and outcomes. Although the mixing of the results will undoubtedly
enhance understanding of pain management within the intervention, this analytic integration
common to all concurrent mixed methods approaches is not the focus of this current discussion.
Here, we continue to focus on the practices that have occurred that specifically relate to the
implementation of the qualitative component within the embedded RCT design.
Despite the design constraints outlined in the previous section and the research team’s initial
expectation for a basic content analysis, the analysis of the qualitative database in the current
study has evolved from simple coding into an interpretive, inductive, and dynamic qualitative
analytic process in the current study that has unfolded over the past 4 years. The qualitative
findings to date provide a rich, in-depth exploration of the experiences of patients, family care-
givers, and nurses as they engage in the pain management intervention that is the focus of the
primary RCT. This RCT seeks to focus specifically on the dimensions of pain and associated
symptoms and side effects through control procedures, reliable and valid measures, and care-
fully designed intervention procedures. The interpretive qualitative findings, however, have
emerged to encompass and subsume these measured outcomes within the larger context and
experiences of each individual participant.
The expansive findings that have emerged to qualitatively describe the complexity of the par-
ticipants’ experiences during the intervention are pictured in Figure 4. Unlike the arrows depict-
ing the constraints in Figure 3, this figure uses double-headed arrows to indicate the dynamic
interaction between the results and interpretations associated with the qualitative and quantita-
tive study components. That is, the double-headed arrows depict how the patterns and themes
derived from the participants’ words provide insight into the multiple dimensions of the inter-
vention process itself as well as the contexts that shape and are shaped by the intervention expe-
rience. For the purposes of illustration, we provide a few examples of the kinds of qualitative
findings that are emerging from the analysis, noting that full results will be reported elsewhere.
The interpretive qualitative results that emerged from the current pain management study
provide detailed descriptions of the intervention process. We have learned about the treatment
parameters as they were actually implemented, such as how the nurses provided individualized
coaching in the context of participants’ real-life situations, the challenges that arose, how the
nurses adapted their interaction style, and the active ingredients for different patients. We also
learned about the participants’ reactions to the treatment parameters such as how patients actu-
ally chose to use the provided pillbox and pain diaries and the contexts for this use. We also
describe different participant reactions related to the research experience such as their questions
about the use of both quantitative and qualitative data collection strategies and their personal
hopes for the benefits of the research for others.
Although the emergent descriptions of the intervention process have been vivid, the qualita-
tive findings have gone beyond simply describing the process. They have provided insights into
the larger experience and context in which the intervention occurs and outcomes are derived.
For example, the process by which patients manage their pain medications emerged as an
important dimension of their overall cancer pain management. The management of pain medi-
cations was a demanding, ongoing process that included getting prescriptions, obtaining medi-
cations, and understanding, organizing, storing, scheduling, remembering, and taking them.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 17

Figure 4. The nature of embedded interpretations where the randomized clinical trial (RCT) process
and outcomes are embedded in the interpretive descriptions of the participants’ experiences and
contexts

This multidimensional process was shaped by system, clinical, and personal contexts and played
an important role related to the intervention as delivered by the nurses and the outcomes as
experienced by the patients. In addition, we gained new understandings of the multiple dimen-
sions and dynamic character of cancer-related pain in the context of current treatment modal-
ities. We also have described life contexts and lived experiences that have occurred before the
intervention and that take place during the intervention and how they shape the overall pain
management process and how the management of pain also shapes other aspects of individuals’
personal and social lives. These contexts and experiences are something that all participants—
patients, family caregivers, and nurses—bring to the intervention interaction and qualitatively
we have learned about the expansive and complex nature of these interactions. As we reflected
on our qualitative findings, we concluded that we had developed another dimension of embed-
ding where our qualitative results and interpretations describe how the intervention is embedded
in the lived experience of participants and how the intervention is at times constrained by these
life contexts.
Findings that provide thorough descriptions of complex phenomena are not unusual in inter-
pretive qualitative research. However, we were surprised with the richness of the understandings
that emerged from our ‘‘constrained’’ qualitative methods in contrast to our initial expectations
for the study aims. The analysis and interpretations briefly highlighted here have gone well
beyond a basic content analysis of strategies, difficulties, and interactions as initially planned
for the current pain management study. Unquestionably, our interpretive analysis is partly

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


18 Journal of Mixed Methods Research XX(X)

grounded in interpretive assumptions possessed by the team’s qualitative researchers. However,


considering the initial expectations for and constraints placed on the design of the qualitative
methods by the requirements of the RCT, simply having an interpretive epistemology would not
have been sufficient for these results to emerge. We also needed to implement specific strategies
that fostered an analysis that was consistent with an interpretive approach within our embedded
design. Reflecting on the analytic process that has unfolded, we identified six key practices that
enabled our emergent interpretive approach despite the RCT constraints on the data collection
methods. Since these practices may provide useful guidance to others considering an embedded
RCT, we delineate each of them here.
Incorporated qualitative aims and methods that built on prior research from the initial study
planning. The importance of the qualitative component for the current pain management study
was recognized and valued from the earliest planning stages, and this was critical in facilitating
the quality of its implementation throughout the study. The research team had experienced the
value of qualitative research in the prior pain management RCT study (Schumacher et al.,
2005; Schumacher, Koresawa, West, Dodd, et al., 2002; Schumacher, Koresawa, West,
Hawkins, et al., 2002a, 2002b). However, the small qualitative component in the prior study
was added on when the study was already in progress and routine quality assurance monitoring
of the audiotaped intervention sessions revealed the richness and depth of the data. The value
of the findings from this small-scale investigation in the prior study led to a commitment by the
research team to include a more substantial qualitative component in the current study from the
beginning. Although the embedded qualitative component was secondary in the current study
in terms of the priority of the research questions and methods, it was intentionally designed as
an important and valued aspect of the overall study goals that was supported by the research
team members and in the study budget.
Gathered an extensive qualitative database. Unquestionably, an interpretive qualitative analy-
sis requires an extensive qualitative database. Although the qualitative data collection proce-
dures in the current pain management study were limited to unobtrusive approaches in order
not to confound the intervention, they still produced an extensive database due to both the
quantity and quality of the gathered data as facilitated by the naturally occurring discourse of
the enhanced intervention. Qualitative data for each participant included about 8 to 12 hours of
audiorecording (depending on dose condition) spread out over 10 to 16 interactions with a
nurse during a 10-week period. Transcriptions often resulted in more than 200 pages of text for
the conversations among the patient, nurse, and family caregiver (if included) across the inter-
vention interactions for each case. These conversations covered topics ranging from an assess-
ment of current pain level and health status, to educational information provided by the nurse
on important pain management topics, pain-related problems brought to the nurses’ attention,
to problem-solving strategies to improve a patient’s pain control, to cancer treatment, other
health conditions, family issues affecting pain management, interactions with health profession-
als across multiple systems of health care, pain management and employment, and chit chat
about travels and family.
Adopted a team approach to the qualitative analysis. A team approach to the qualitative analy-
sis in the current pain management study provided more personnel time and deeper thinking
about the data and their interpretations. A second researcher with qualitative research expertise
was added to the research team as the implementation of the study was under way. Although
this addition was done largely to provide the effort needed to analyze an extensive amount of
data, the use of a team approach took the qualitative analysis far beyond simply doing more of
it. The regular (often weekly) interactions between researchers who represented different back-
grounds and prior research experiences brought different lenses and analytic approaches to the
data. The conversations resulting from these different perspectives served to push the

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 19

interpretive nature of the analysis as ongoing interpretations were continually explained and
discussed between analysts.
Focused on an in-depth analysis of a small number of cases. Recruitment is a challenge of any
research involving patients with advanced cancer. With the use of a rolling recruitment process,
the qualitative (and quantitative) data in the current pain management study came in gradually
over time. In addition, the time required to transcribe extensive amounts of audiorecordings
limited the initial availability of qualitative data. In hindsight, this challenge turned out to be a
fortuitous advantage as the qualitative analysts were provided the time to work slowly as they
analyzed the data that initially became available. Taking this time at the start to go in depth and
not rush the development of themes and analytic memos paid great dividends down the road as
it laid the foundation on which deeper interpretations and levels of analysis emerged. If the ini-
tial analyses had been rushed to simply ‘‘get through’’ a large number of cases, it is likely that
the entire analysis would have remained a more superficial content analysis.
Developed an emergent, layered approach to analysis. In response to the volumes of data and
the desire to track themes and concepts longitudinally across the 10 weeks of each case, the
qualitative analysts continued to develop new tools and analytic strategies over the years of the
analysis. These tools included data summary frameworks to facilitate identifying patterns across
time within cases and across cases and the extensive use of memos. Although the data collec-
tion procedures and study aims did not change based on the initial findings, the data analysis
process continued to evolve and unfold as new concepts and questions arose from the data.
This dynamic, layered approach permitted us to incorporate an emergent analytic process that
is a strength of interpretive qualitative research while at the same time, maintaining a fixed
approach to data collection within the RCT design.
Kept the quantitative and qualitative strands separate during initial analyses. The ongoing possi-
bilities and opportunities for mixing data and results in the current pain management study are
practically limitless. Early in the analysis, there was interest in bringing quantitative data into
the qualitative analysis process. Again, because of logistical constraints in data availability,
quantitative data and results were not readily accessible to the qualitative analysts. In hindsight,
this initial separation for practical reasons became an advantage for interpretive reasons.
Because the qualitative data include many topics and experiences well beyond the quantitative
measures, the qualitative analysis would have been limited and overly structured had we
attempted to align the two databases early in the analysis process. By not letting the qualitative
analysis be constrained by the quantitative variables, it facilitated a more interpretive approach
to analyzing the qualitative data.

Lessons Learned and Conclusions


We have described the practices that occurred in the context of embedding qualitative methods
within the RCT design of our current pain management study. These practices included stating
secondary research questions that were embedded within the context and design of the primary
research questions, designing embedded qualitative data collection methods that satisfied the
constraints of the RCT, and using strategies that facilitated the development of enriched under-
standings through an interpretive qualitative analysis. This collection of practices across differ-
ent stages of our research process highlights the creative tensions that can emerge when using
an embedded design. Our reflections suggest that care needs to go into the design of an
embedded RCT due to the attention and focus that needs to be placed on the assumptions and
design requirements of the RCT during the initial planning. However, if planned well, an
embedded RCT can achieve a qualitative component that is not only interpretive but also does
not impinge on the RCT’s internal validity or rigor.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


20 Journal of Mixed Methods Research XX(X)

As our qualitative analytic methods have become more interpretive, we find that new ten-
sions are emerging within the larger embedded design. For one, more interpretive strategies
directly translate into more time required to analyze each case. Although we see this time com-
mitment as a worthwhile trade-off, it creates a tension when considering the initial plan to com-
plete a content analysis across the more than 200 recruited cases. Possible strategies for dealing
with this tension include moving the qualitative analysis to a more deductive strategy once
saturation is reached (an approach more consistent with the postpositivist assumptions of the
RCT) and/or not analyzing the qualitative data from all cases (a strategy more consistent with
interpretive assumptions). Another logistical challenge is the attempt to include new researchers
in the qualitative analysis process. Due to the high level of interpretation developing over sev-
eral years, it is challenging to bring in new individuals who do not have the long-term continuity
of working with the analysis and emergent findings. Additional tensions related to the possibili-
ties of mixing the quantitative and qualitative data and results will continue to emerge as the
quantitative results become available and the combined analyses go forward. For example, as
the qualitative results have become more expansive, it may be harder to specify dimensions for
directly comparing the quantitative and qualitative results.
Although embedding qualitative methods within an RCT framework can lead to points of
tension, it also provides numerous opportunities. In the context of the ongoing cancer pain man-
agement study, we are already able to identify several advantages to the use of an RCT design
that included an embedded interpretive qualitative analysis. These advantages include the
following:

 Increased knowledge gains arise from being able to address different types of aims
within the same study using approaches informed by different assumptions. This appli-
cation of an embedded design permits the research team to quantitatively address ques-
tions of impact and outcomes and qualitatively address questions of process and context
within one substantive area.
 Enhanced understanding can be obtained through the synergistic integration of quantita-
tive results and qualitative interpretive findings. This integration can occur to understand
a single case, to address a single topic, to examine differences across groups, to under-
stand patterns over time, and to explain or expand on one type of results with the other,
to name only a few of the possibilities for mixing data and results.
 This approach is cost-effective in the sense that increased knowledge (i.e., outcomes and
context) is obtained within the management of a single project. The costs would have
been significantly higher had this work been completed as a separate quantitative study
and separate qualitative study, each with their own team management and participant
recruitment. In addition, if the studies were done separately, there would not be the pos-
sibility of enhanced understandings achieved through meaningful integration.
 This approach provided participants with multiple modes to describe their experiences.
Some participants expressed comfort and appreciation for their experiences being
assessed in different ways. Although many noted the importance of the quantitative
measures and initiated discussions with their nurse to ensure their adherence with and
accurate completion of the information, others commented on ways that they felt that
numbers might not adequately convey their experiences with pain management. By
incorporating qualitative data into the clinical trial, participants were often assured that
their experiences were valued in addition to their outcomes.
 Finally, embedded designs have the potential to make substantial contributions to trans-
lational research and dissemination and implementation science. Embedding an interpre-
tive qualitative component in the RCT provides insights into the messiness of real-world

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 21

application of specific intervention procedures. We are developing new understandings


about the interface between the RCT design and the real-world contexts of study partici-
pants’ lives. These insights may provide a crucial link in the process of translating
research-based results into real-world practice.

In describing the practices and tensions in this one embedded study, we found it helpful to
consider different dimensions of embedding across the research process. The delineation of
these dimensions adds to literature that is developing to help researchers consider reasons for
embedding qualitative methods into intervention designs and the design considerations required
to do this well. Although each of these dimensions of embedding emerged from the mixed
methods literature, we found little explicit recognition of the existence of multiple dimensions.
Researchers and methodologists alike may benefit from an awareness of these multiple aspects
of embedding to more clearly consider the different decisions required in the design of an
embedded study. We encourage researchers who report their use of an embedded design to
articulate what is meant by the use of this term within the context of their particular study and
to describe how embedding is reflected in the design and implementation of their study.
This discussion of embedding was limited to the experiences that occurred in one embedded
mixed methods study on the topic of cancer pain management. As with all research, the current
pain management study has limitations. Specific to this discussion, the use of unobtrusive quali-
tative methods of data collection could be considered a limitation because they did not allow
for the opportunity to ask follow-up questions based on the emerging analysis. We may have
learned more of different aspects had a qualitative researcher conducted interviews with the par-
ticipants about their experiences. However, the use of unobtrusive methods made for a good
choice allowing for an extensive database that adhered to the parameters of the RCT design and
provided a record of the interactions that touched on aspects of participants’ lives that might
have never emerged if other methods had been used. Researchers designing embedded RCT
studies need to carefully consider the best qualitative methods for their topic, research ques-
tions, and intervention parameters.
Much work still needs to be done to address the methodological, epistemological, and practi-
cal issues associated with the embedded design. Future research could weigh different qualita-
tive approaches (e.g., focus groups, interviews, observations, journals) in terms of their
adherence to RCT parameters and potential to support interpretive analytic approaches as well
as consider design issues for interventions that are not so discursive in nature. Further work is
needed to map out the decisions and assumptions involved for other applications of embedding
such as embedding quantitative methods within a narrative study. By reflecting on the practice
of mixed methods and relating that practice directly to the methodological literature, mixed
methods scholars will continue to advance our understanding of the decisions that are important
to consider, the options that are possible, and the challenges and opportunities to be anticipated.

Authors’ Note
This article is based on a paper presented at the 2011 meeting of the American Educational Research
Association, New Orleans.

Acknowledgments
The authors gratefully acknowledge the patients and family caregivers who participated in this research
and our many colleagues who assisted with participant recruitment. We thank our colleagues in the Office
of Qualitative and Mixed Methods Research at the University of Nebraska–Lincoln, Carolyn S. Ridenour,
and three anonymous reviewers for their constructive feedback on previous drafts of this article.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


22 Journal of Mixed Methods Research XX(X)

Declaration of Conflicting Interests


The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or pub-
lication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publi-
cation of this article: This work was supported by a grant (CA 116423) from the National Cancer Institute
and the National Institute of Nursing Research.

References
Bazeley, P. (2010). Review of book The mixed methods reader, V. L. Plano Clark and J. W. Creswell
(Eds.). Journal of Mixed Methods Research, 4(1), 79-81.
Brady, B., & O’Regan, C. (2009). Meeting the challenge of doing an RCT evaluation of youth mentoring
in Ireland: A journey in mixed methods. Journal of Mixed Methods Research, 3(3), 265-280.
Caracelli, V. J., & Greene, J. C. (1997). Crafting mixed-method evaluation designs. In J. C. Greene & V.
J. Caracelli (Eds.), Advances in mixed-method evaluation: The challenges and benefits of integrating
diverse paradigms (pp. 19-32). San Francisco, CA: Jossey-Bass.
Chen, H. (1997). Applying mixed methods under the framework of theory-driven evaluations. In J. C.
Greene & V. J. Caracelli (Eds.), Advances in mixed-method evaluation: The challenges and benefits of
integrating diverse paradigms (pp. 61-72). San Francisco, CA: Jossey-Bass.
Christ, T. W., & Makarani, S. A. (2009). Teachers’ attitudes about teaching English in India: An
embedded mixed methods study. International Journal of Multiple Research Approaches, 3, 73-87.
Creswell, J. W. (2008). Educational research: Planning, conducting, and evaluating quantitative and
qualitative research (3rd ed.). Upper Saddle River, NJ: Pearson Education.
Creswell, J. W. (2009). Research design: Qualitative, quantitative, and mixed methods approaches (3rd
ed.). Thousand Oaks, CA: Sage.
Creswell, J. W., Fetters, M. D., Plano Clark, V. L., & Morales, A. (2009). Mixed methods intervention
trials. In S. Andrew & L. Halcomb (Eds.), Mixed methods research for nursing and the health sciences
(pp. 161-180). Oxford, England: Blackwell.
Creswell, J. W., & Plano Clark, V. L. (2007). Designing and conducting mixed methods research.
Thousand Oaks, CA: Sage.
Creswell, J. W., & Plano Clark, V. L. (2011). Designing and conducting mixed methods research (2nd
ed.). Thousand Oaks, CA: Sage.
Creswell, J. W., Plano Clark, V. L., Gutmann, M., & Hanson, W. (2003). Advanced mixed methods
research designs. In A. Tashakkori & C. Teddlie (Eds.), Handbook of mixed methods in social &
behavioral research (pp. 209-240). Thousand Oaks, CA: Sage.
Creswell, J. W., Shope, R., Plano Clark, V. L., & Green, D. O. (2006). How interpretive qualitative
research extends mixed methods research. Research in the Schools, 13(1), 1-11.
Creswell, J. W., & Zhang, W. (2009). The application of mixed methods designs to trauma research.
Journal of Traumatic Stress, 22, 612-621.
Donovan, J., Mills, N., Smith, M., Brindle, L., Jacoby, A., Peters, T., . . . Hamdy, F. (2002). Improving
design and conduct of randomised trials by embedding them in qualitative research: ProtecT (Prostate
Testing for Cancer and Treatment) study. British Medical Journal, 325, 766-769.
Flemming, K., Adamson, J., & Atkin, K. (2008). Improving the effectiveness of interventions in palliative
care: The potential role of qualitative research in enhancing evidence from randomized controlled
trials. Palliative Medicine, 22, 123-131.
Giddings, L. S. (2006). Mixed-methods research: Positivism dressed in drag? Journal of Research in
Nursing, 11, 195-203.
Greene, J. C. (2007). Mixed methods in social inquiry. San Francisco, CA: Jossey-Bass.
Howe, K. R. (2004). A critique of experimentalism. Qualitative Inquiry, 10, 42-61.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


Plano Clark et al. 23

Johnson, B., & Christensen, L. (2008). Educational research: Quantitative, qualitative, and mixed methods
(3rd ed.). Thousand Oaks, CA: Sage.
Kaptchuk, T. J., Shaw, J., Kerr, C. E., Conboy, L. A., Kelley, J. M., Csordas, T. J., . . . Jacobson, E. E.
(2009). ‘‘Maybe I made up the whole thing’’: Placebos and patients’ experiences in a randomized
controlled trial. Culture, Medicine, and Psychiatry, 33, 382-411.
Lipman, E. L., Kenny, M., Jack, S., Cameron, R., Secord, M., & Byrne, C. (2010). Understanding how
education/support groups help lone mothers. BMC Public Health, 10(4), 4-12.
Maxwell, J. A., & Loomis, D. M. (2003). Mixed methods design: An alternative approach. In A.
Tashakkori & C. Teddlie (Eds.), Handbook of mixed methods in social & behavioral research (pp.
241-271). Thousand Oaks, CA: Sage.
Miaskowski, C. (2007-2012). Improving cancer pain management through self-care. Unpublished
proposal for a study funded by the National Cancer Institute and the National Institute of Nursing
Research (5R01CA116423).
Miaskowski, C., Dodd, M., West, C., Paul, S. M., Schumacher, K., Tripathy, D., & Koo, P. (2007). The
use of a responder analysis to identify differences in patient outcomes following a self-care intervention
to improve cancer pain management. Pain, 129, 55-63.
Miaskowski, C., Dodd, M. J., West, C., Schumacher, K., Paul, S. M., Tripathy, D., & Koo, P. (2004).
Randomized clinical trial of the effectiveness of a self-care intervention to improve cancer pain
management. Journal of Clinical Oncology, 22, 1713-1720.
Miller, W. L., & Crabtree, B. F. (2005). Clinical research. In N. Denzin & Y. Lincoln (Eds.), The Sage
handbook of qualitative research (3rd ed., pp. 605-639). Thousand Oaks, CA: Sage.
Morse, J. M., & Niehaus, L. (2009). Mixed methods design: Principles and procedures. Walnut Creek,
CA: Left Coast Press.
Nastasi, B. K., & Schensul, S. L. (2005). Contributions of qualitative research to the validity of
intervention research. Journal of School Psychology, 43, 177-195.
Neuman, W. L. (2006). Social research methods: Qualitative and quantitative approaches (6th ed.).
Boston, MA: Pearson Education.
Plano Clark, V. L., & Creswell, J. W. (2008). The mixed methods reader. Thousand Oaks, CA: Sage.
Saldaña, J. (2003). Longitudinal qualitative research: Analyzing change through time. Walnut Creek, CA:
AltaMira Press.
Sandelowski, M. (1996). Using qualitative methods in intervention studies. Research in Nursing & Health,
19, 359-364.
Schumacher, K. L., Koresawa, S., West, C., Dodd, M. J., Paul, S. M., Tripathy, D., . . . Miaskowski, C.
(2002). The usefulness of a daily pain management diary for oncology outpatients with cancer related
pain. Oncology Nursing Forum, 29, 1304-1313.
Schumacher, K. L., Koresawa, S., West, C., Dodd, M., Paul, S. M., Tripathy, D., . . . Miaskowski, C.
(2005). Qualitative research contribution to a randomized clinical trial. Research in Nursing & Health,
28, 268-280.
Schumacher, K. L., Koresawa, S., West, C., Hawkins, C., Johnson, C., Wais, E., . . . Miaskowski, C.
(2002a). Pain management autobiographies and reluctance to use opioids for cancer pain management.
Cancer Nursing, 25, 125-133.
Schumacher, K. L., Koresawa, S., West, C., Hawkins, C., Johnson, C., Wais, E., . . . Miaskowski, C.
(2002b). Putting cancer pain management regimens into practice at home. Journal of Pain and
Symptom Management, 23, 369-382.
Song, M., Sandelowski, M., & Happ, M. B. (2010). Current practices and emerging trends in conducting
mixed methods intervention studies in the health sciences. In A. Tashakkori & C. Teddlie (Eds.),
Handbook of mixed methods in social & behavioral research (2nd ed., pp. 725-747). Thousand Oaks,
CA: Sage.
Teddlie, C., & Tashakkori, A. (2009). Foundations of mixed methods research: Integrating quantitative
and qualitative approaches in the social and behavioral sciences. Thousand Oaks, CA: Sage.
Turner-Cobb, J. M., Palmer, J., Aronson, D., Russell, L., Purnell, S., Osborn, M., & Jessop, D. S. (2010).
Diurnal cortisol and coping responses in close relatives of persons with acquired brain injury: A
longitudinal mixed methods study. Brain Injury, 24, 893-903.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013


24 Journal of Mixed Methods Research XX(X)

West, C. M., Dodd, M. J., Paul, S. M., Schumacher, K., Tripathy, D., Koo, P., & Miaskowski, C. (2003).
The PRO-SELF Pain Control Program—An effective approach for cancer pain management. Oncology
Nursing Forum, 30, 65-73.
Yin, R. K. (2006). Mixed methods research: Are the methods genuinely integrated or merely parallel?
Research in the Schools, 13(1), 41-47.

Downloaded from mmr.sagepub.com at PORTLAND STATE UNIV on January 29, 2013

You might also like