Building a Bridge or Digging a Pipeline? Clinical Data Mining in Evidence-Informed Knowledge Building

Abstract

Challenging the “bridge metaphor” theme of this conference, this article contends that current practice–research integration strategies are more like research-to-practice “pipelines.” The purpose of this article is to demonstrate the potential of clinical data-mining studies conducted by practitioners, practitioner-oriented PhD students, and methodologically pluralist researchers to promoting a truly collaborative, two-way traffic in evidence-informed knowledge building.

Keywords

clinical data mining evidence-informed practice bridge metaphor

Metaphors Matter. Some enlighten. Others mislead. Some preserve the status quo. Others promote change. A persistent metaphor for linking social work research and practice is “building a bridge” (Davis et al., 2013; Hess & Mullen, 1995) but prevailing strategies for doing so in social work research are more like “digging a pipeline.” The bridge metaphor implies two-way traffic; the pipeline, a unidirectional flow from research to practice.

The challenge of integrating research and practice is hardly unique to social work; nor is the bridge/pipeline distinction. So for example, a recent National Cancer Institute Cyber-seminar (2012, p. 1) on evidence-based medicine and oncology practice, asserts:

The translational gap between research and practice has long been discussed, often as a one-way street—get practitioners to recognize and utilize the research that is being conducted. While important, equally important is the reverse—integrating practice-based evidence and context into the research being conducted. We need a bridge between the two, not a pipeline. (author’s italics)

Similarly, in a Centers for Disease Control and Prevention workshop on public health research and practice, Green (2007) employed the same distinction with a slide entitled “The Bridge (not the Pipeline) from Research to Practice and Back.” In the title slide of that workshop, Green urged public health academics to do more to engage public health practitioners in collaborative research, noting “[i]f we want more evidence-based practice (EBP), we need more practice-based evidence.”

In clinical psychology, a 2010 American Psychological Association (APA), online publication recommends new ways “to bridge” research and practice as an alternative to a one-way, “pipeline” strategy. Challenging both sets of stakeholders, it argues that “[i]nstead of bemoaning the differences between researchers and practitioners, both parties can take concrete steps to make their work more accessible to the other …. ” (De Angelis, 2010, p. 45). Citing an inaugural article by APA past President Alan Kazdin (2008), De Angelis endorses “the Kazdin way of mending the split” (p. 45).

Kazdin’s bridging strategies involve contextualizing practice through qualitative research, giving more research attention to treatment moderators, and systematically analyzing the practice experiences of clinicians (De Angelis, 2010). The latter strategy closely parallels but does not reference the practitioner-conducted, clinical data-mining (CDM) approach that international academic colleagues and I have promoted for over a decade—both here and abroad, where we have conducted CDM studies with social workers and with allied health practitioners (Epstein, 2001, 2010; Giles, Epstein, & Vertigan, 2011; Joubert & Epstein, 2005, 2013; Lalayants et al., 2012).

Whether the distance between research and practice in social work is conceptualized as a “difference,” a “split,” a “gap,” a “divide,” or most recently a “chasm” (Wretman, 2013), I think we all agree that the need for building an effective bridge between research and practice in social work persists today as stubbornly as it did when Rubin and Rosenblatt (1979) convened their truly groundbreaking conference on social work research utilization in 1977. As someone who presented at that conference and who has worked both banks of the social work knowledge stream since 1968, what strikes me is how much and how little has changed since then.

Over the past three decades, the research-to-practice (“R2P” in the vernacular of this conference) pressure in social work has been significantly increased by a combination of EBP ideology and research productivity, “manualization” of empirically supported treatments (EST), the entrepreneurial efforts of EST program “purveyors,” translational and dissemination research, and the geometric expansion of information technology. In this heady context, it’s hard to imagine that the articles presented at the 1977 National Institute of Mental Health conference never once mentioned computers or computerization. They were written on typewriters—some electric, mostly manual. Much has changed indeed.

But, in my opinion, a tsunami of academically driven EBP and EST rhetoric and entrepreneurial activity has effectively drowned out the voices and collective experience of social work practitioners and excluded them from the knowledge-production enterprise of the profession. Once sought as coproducers and collaborators in the quest for knowledge, practitioners are now relegated to the role of research consumer, facilitator, or faithful implementer. Intervention “fidelity” to EST’s rather than clinical reflection and practice inventiveness is the standard by which the ideal practitioner is now judged. When practitioners resist, they are viewed by academic researchers as anywhere from nonadherent to subversive of efforts to integrate research and practice.

Admittedly, I have no hard data to present about negative researcher attitudes toward practitioners, but I hear what they say in public forums and at unguarded, belly-up-to-the-bar moments at research conferences. At one conference session I attended recently with one of my coauthors (she heard it too) on the topic “Conducting Research in Practice Settings,” the lead presenter advised attendees to “build a brick wall between your research and agency practitioners.” Anecdotal I know, but hardly a “bridging” metaphor.

On firmer empirical ground, recent studies of the characteristics of practitioners who do and do not employ research-based interventions (Parrish & Rubin, 2012; Patterson, Maguin, Dulmus, & Nisbet, 2013; Pignotti & Thyer, 2009) provide evidence that seamless adoption of R2P integration efforts remains a serious problem.

For the past two decades, I have been working from the practice side of the research/practice integration divide (practice-to-research [P2R]?). During that time, in practice-based research (PBR) and CDM workshops I have conducted in this country and abroad, practitioners have routinely complained about academic researchers’ denial of the value of their “practice wisdom,” their loss of discretionary authority, imposition of “cookbook” manualized intervention models, and their sense of disempowerment in R2P programs in which they have been enlisted or compelled to participate. Efforts on my part to have practitioners in EBP and EST programs systematically record their clinically motivated and effectiveness-intended treatment modifications have been routinely scuttled by fidelity-minded managers and evaluation researchers. When treatment modifications happen, many managers and researchers would seem to prefer not knowing about it.

Of course, these antiresearcher testimonies come from those who attend my workshops or with whom I consult. Consequently, generalization is problematic. Nonetheless, their voluntary and enthusiastic participation in PBR and CDM challenges the long-standing myth that practitioners have no interest in research, have no inclination to conduct it, and have no desire to utilize it. Ironically, my most quoted article is subtitled “Teaching research to the reluctants” (Epstein, 1987). Although its purpose was to demonstrate how it can be done, it is frequently cited to claim that it can’t.

In 2011, I published a “Viewpoint” article in Social Work highly critical of what I portrayed as antipractitioner bias implicit in the rhetoric of EBP and EST proponents. Not unexpectedly, researcher-colleagues were offended by my negative characterization of their attitudes toward practitioners, saying I had grossly misrepresented them and their intentions. Put simply, by publishing the piece, I had “betrayed” my research buddies.

On the other hand, Social Work’s editor eagerly published it and the practitioners that I heard from were universally supportive. They cheered me on. Here again, I’m talking about a small, self-selected sample, so generalizations are inappropriate. Still, this provocative piece struck a note—maybe two. And they were discordant. At this conference, bringing attention to the sentiments expressed in this article assured that my claims would be heard on both sides of the river, but admittedly, it was no way to “build a bridge.” And my ultimate intention is a meaningful “reconciliation” somewhere in the middle of that bridge rather than a shouting match from opposite shores.

In a small and perhaps unnoticed step in that direction, my Social Work article advocated CDM studies as one strategy to involve practitioners in researching their own practice-based evidence, in using their findings programmatically, and in presenting and publishing them externally. If acknowledged as a legitimate form of research, the latter would promote a P2R return flow of practitioner ideas and practice-based evidence into knowledge generation.

The purpose of this article is to make a more positive and substantial contribution to the bridge-building process. Still starting from the practice shore, but describing CDM studies in greater detail, I argue for the “value-added” potential that CDM studies conducted by practitioners, by practice-oriented PhD students, and potentially by methodologically pluralist academic researchers can make in enriching our empirical understanding of multiple forms of practice, in promoting practice–research collaboration, and enhancing a two-way traffic of ideas—that is, from P2R as well as R2P.

Exemplars of selected practitioner-conducted CDM and CDM doctoral dissertations will be offered. However, for a more in-depth discussion of CDM as a distinct research methodology—its strengths as well as its limitations—the reader is referred to my recent text on the subject (Epstein, 2010).

What Is CDM?

A simple definition of CDM is “the practitioner’s use of available agency data for practice-based research purposes” (Epstein, 2001, p. 3). More operationally, CDM may be defined as a practice-based research strategy by which practice researchers systematically retrieve, codify, analyze, and interpret available qualitative and/or quantitative information concerning client characteristics and needs, services and interventions received, and outcomes achieved derived from available clinical records for the purpose of reflecting upon the practice and policy implications of their findings.

Although the research use of available administrative and practice data by agency and academic researchers is hardly new, CDM is a more self-conscious and systematic P2R approach that has been successfully employed by social workers and allied health practitioner-researchers—for example, music therapists, occupational therapists, physical therapists, and speech pathologists in single and multidisciplinary teams (Epstein & Blumenfield, 2001; Giles et al., 2011; Joubert & Epstein, 2005, 2013).

By my working definition, CDM differs from secondary analysis (SA) in that the latter, generally conducted by researchers, makes use of available data originally intended for research purposes. Consequently, SA databases usually include standardized measures that have been reliability tested and normed, explicit sampling strategies, larger more representative samples, and so on (Sales, Lichtenwalter, & Fevola, 2006). What SA studies offer in research rigor and generalizability, however, they often sacrifice in practice relevance and practitioner salience. More to the point here is that SA databases are generated by researchers for researchers rather than by practitioners from their own practice.

For over a decade, I have spent much of my career “at the coalface,” working alongside practitioners in helping them extract, refine, analyze, and interpret their own practice data. The first CDM article I published was subtitled “Mining for silver while dreaming of gold” (Epstein, 2001). That metaphor was intended to convey the positive use of experimental (randomized controlled trial [RCT]) logic and the incorporation of basic research concepts and techniques into CDM studies.

CDM studies range from descriptive or preexperimental to quasi-experimental to ex post facto experiments (Sainz & Epstein, 2001). Depending on the richness of the available practice data, they can be analyzed qualitatively or converted from qualitative statements to quantitative data for quantitative analysis. In addition, they can be combined with other research designs and original data collection strategies. Naturally, these more ambitious mixed-method studies are more typical of PhD dissertations than of practitioner-initiated CDM studies.

Despite their design and topical flexibility, CDM studies are generally retrospective; often reliant on simple, nonstandardized intervention and outcome measures; and involving comparisons of nonrandomly assigned intervention groups. Occasionally, we “get lucky” and discover that standardized assessment, intervention, and/or outcome measures are routinely employed and available. Even then, CDM studies cannot rival RCT studies in making causal inferences concerning the linkage between interventions and outcomes. Random assignment and true control groups are incompatible with the norms of most practice settings. As a result, CDM studies are generally excluded from EBP meta-analyses and systematic reviews.

Nonetheless, it is my contention that CDM has an important role to play in the development of empirically based knowledge of “best practices” and in the generation of a more fully contextualized, differentiated, and collaboratively constructed “evidence-informed” practice (Epstein, 2009). Moreover, CDM study findings can even enhance the knowledge gained from RCTs and meta-analyses.

Practitioner-Initiated CDM Studies

In three decades of PBR consulting with social work practitioners at Mount Sinai hospital in New York City, my CDM prototype emerged rather late from a study of psychosocial factors associated with successful and unsuccessful liver transplantation (Epstein, Zilberfein, & Snyder, 1997). That first CDM study fully engaged practitioners in every stage in the research process—from conceptualization to dissemination (Epstein, 2010). Moreover, it produced unanticipated findings about the absence of a relationship between patient prior history of substance abuse and liver transplant mortality. That clinically significant “nonfinding” was a profound surprise to all involved and ran counter to liver transplant program theories and policies in other hospitals that denied transplants to anyone with a prior history of substance abuse.

Once I recognized the practice knowledge-generation potential of studies with practitioners using routinely available clinical information, I approached all subsequent practice–research collaborations with the same question—“What data are currently available that might inform the practice–research question you are seeking to answer?” The answers given in multiple settings revealed surprising storehouses of potential data that were rarely conceived as useful for research purposes. At times, they were even referred to as “databases” although they remained mostly unutilized. When used, they might be superficially skimmed for external accountability purposes. Mostly, they sat lifeless in agency file cabinets or computers.

Sensing the potential for practice–research, I developed a model of research consultation practice that began with routinely available practice data. These collaborative research efforts always led to formative findings that enhanced practitioners’ understanding of their own programs as well as producing more summative contributions to the research literature (Scriven, 1995).

By now, several collections of practitioner-initiated CDM studies appear in the literature (Epstein & Blumenfield, 2001; Joubert & Epstein, 2005; Peake, Epstein, & Medeiros, 2005) as well as occasional individual studies in peer-reviewed journals (see, e.g., Dobrof, Ebenstein, Dodd, & Epstein, 2006; Ross, Walther, & Epstein, 2004). The most recent collection appears as a special issue of Social Work in Health Care and is based on CDM studies conducted by health social workers and other allied health practitioners in Melbourne Australia (Joubert & Epstein, 2013).

Types of Practitioner-Initiated CDM Studies

Practitioner-initiated CDM studies tend to fall into three general categories: need studies that describe clients and their presenting problems, monitoring studies that describe clinical and program interventions, and outcome studies that document client and patient outcomes once services have been completed. Depending on data available, some of these studies consider empirical relationships within these general categories, whereas others look at bivariate and multivariate relationships across categories—for example, between client characteristics and services received or between services received and outcomes.

For example, a collection of CDM studies in adolescent health and mental health engaged various groups of Mount Sinai Adolescent Health Center (AHC) program staff in empirically studying relationships between age, sex, race, and sexual preference with adolescent strengths, presenting problems and risk taking (Peake et al., 2005). The findings tested, challenged, supported, and extended many of the stated and unstated theories that AHC practitioners brought to their work, identified unmet client needs, generated new program possibilities, and offered insights for the field about different clusters of adolescent risk taking. Because the multistudy project was based entirely on an intake questionnaire routinely administered to all service applicants, no data concerning interventions or outcomes were available but the administrative need for future systematic data collection of intervention and outcome data for future evaluation emerged from that collective practitioner effort.

Another Mount Sinai, practitioner-initiated CDM study described services provided and outcomes achieved at various stages in the treatment of adults with end-stage renal disease (ESRD; Dobrof et al., 2000, 2001). This study empirically demonstrated the complexity of social work practice with ESRD patients and how practice patterns changed over the course of dialysis, showed how social work intervention was associated with fewer rehospitalizations and emergency room visits, and yielded retrospective depression-reduction findings that were remarkably similar to the results of a prospective ESRD/RCT published contemporaneously (Beder, 1999; Sainz & Epstein, 2001). Demonstration of complexity was especially important because of the move in many hospitals to deprofessionalize ESRD social work. Because Beder’s RCT required randomization to a “minimal” or an “enriched” social work intervention, additional questions were raised about ethical aspects of her RCT and what its conduct meant for the practitioners who were expected to faithfully implement it (Epstein, 2010, pp. 120–121).

After a failed ESRD experiment in Israel by an academic colleague of hers, Auslander then applied the CDM methodology developed at Mount Sinai to an ESRD social work study in Israel. Comparison of the two CDM studies yielded anticipated similarities about ESRD social workers’ roles but unanticipated differences in the ways ESRD services were provided in the two countries and major differences in patient depression rates (Auslander, Dobrof, & Epstein, 2001).

The foregoing exemplars and many more suggest that when provided with collaborative, respectful, and appreciative research consultation and starting with their own practice data, practitioners eagerly employ CDM to research their own practice and critically reflect upon whatever they find—positive or negative. Hence, in an atmosphere of mutual trust and scientific inquiry, practitioners are open to empirically based discoveries and conceptual insights that are not always pleasing. In the process, they critically review relevant research conducted by others and also learn to appreciate the practical and ethical complexity of successfully conducting research in a practice setting. Understandably, however, practitioner-initiated CDM designs and analyses are less methodologically complex than CDM dissertation studies conducted by doctoral students and/or potentially by seasoned researchers.

The CDM Doctoral Dissertation

Shortly after realizing the potential of CDM in research with practitioners, I began considering its applicability for practice-oriented DSW and PhD students whose dissertations I was supervising. Over the past decade, I have supervised, been a dissertation committee member on or an external examiner of several CDM dissertations in the DSW and PhD programs in social welfare at the City University of New York in which I teach as well as in a number of Australian and Hong Kong universities where I have provided CDM workshops. Subject matter has ranged quite literally from cradle to grave. Units of analysis have ranged from individuals to groups, to single and multiple organizations as well as national and international programs. A few have employed qualitative CDM; most have been quantitative CDM. One combined both qualitative and quantitative CDM. Some have mixed CDM with other more traditional research methods, suggesting an expansion of what we mean by “mixed methods” to include mixing available and original data collection.

A brief discussion of only a few of these CDM doctoral dissertations is intended to demonstrate the usefulness and flexibility of CDM as an alternative or as a complementary dissertation methodology and of its potential for contributing to evidence-informed practice.

The first explicitly defined CDM dissertation I supervised concerned Intensive Family Preservation (IFP), a placement prevention program model that had already been widely researched with RCTs as well as trenchantly critiqued by Rossi (1992). In many respects, this dissertation was both an IFP inquiry and a CDM feasibility study (Hanssen, 2003). Using available case data from a single IFP agency, Hanssen applied published research instruments to establish fidelity to the IFP service model in the agency in which she conducted her study (Hanssen & Epstein, 2006). Once fidelity was established, Hanssen employed a very simple multivariate analysis and the research standard if practice questionable outcome measure (i.e., placement prevention 1-month postintervention) employed in previous IFP/RCTs to explore the differential effectiveness of IFP—that is, the social characteristics of individuals and families with whom it “worked” and those with whom it didn’t.

Although Rossi was extremely critical of the 1-month, nonplacement “success” measure routinely employed by IFC evaluators, Hanssen demonstrated that even with that measure different family configurations and client characteristics were differentially associated with 1-month placement outcomes (Hanssen & Epstein, 2007). Relying on experimental logic, available data, and correlational techniques, her study findings suggested that CDM inquiries might enhance knowledge generated by RCTs and meta-analyses that focus largely on effect size. More about this issue will be discussed in relation to our final CDM dissertation exemplar.

Another early CDM dissertation exemplar concerned family reunification in foster care. Based solely on successful reunification cases—that is, children stably returned from foster care to their biological families—Cordero conducted an entirely qualitative CDM dissertation “thickly” describing social worker interventions with children, family members, and foster parents at different stages in the foster care process and taking into account a typology of reasons for placement (Cordero, 2000). Although causal inferences would be inappropriate based on her qualitative design, a small sample and only successful cases, Cordero demonstrated the feasibility of qualitative CDM in describing foster social work (Cordero, 2004). More importantly from a practice point of view she showed how even in successful cases, foster care workers are confronted with practice issues that differ dependent on the original reason for placement—that is, parental substance abuse, domestic violence, or neglect (Cordero & Epstein, 2005). These contextual findings have important implications for the training of foster care workers.

A more recent, mixed-method child welfare CDM dissertation employed original qualitative data collection and quantitative analysis to study a multidisciplinary child protection consultation program (Lalayants, 2010). A winner of the Outstanding Dissertation of the Year Award of the American Professional Society on the Abuse of Children, Lalayants converted available qualitative records of responses to requests for consultation to quantitatively analyze whether the recommended practices of substance abuse, mental health, and domestic violence consultants were and were not consistent with agency practice principles of the Administration for Children’s Services in New York City. Another expression of “fidelity,” these practice principles favored strength-based, child-centered, family-focused, culturally sensitive interventions. However, in this very different fidelity study, the practitioners were the “clients” and consultation patterns revealed significant departures from agency principles as well as variations by consultant expertise (Lalayants, Epstein, & Adamy, 2011).

Turning from child welfare in New York City to palliative care in Hong Kong, Chan employed quantitative CDM to operationally define and study “good death” in a sample of over 900 Hong Kong Chinese terminal cancer patients (Chan, 2007; Chan & Epstein, 2012). Clearly, the sensitivity and ethical complexity of his chosen topic precluded Chan’s consideration of original data collection via patient interviews or questionnaires. An RCT was out of the question. Using available data alone, Chan was able to identify family predictors of a physically, interpersonally, spiritually, and psychologically pain-free state at death as indicated by routine nursing entries into patient records (Chan, Epstein, Reese, & Chen, 2009). By comparing patients on these indicators when they entered palliative care and their final entries, he was also able to demonstrate how palliative was associated with positive patient differences on several of these dimensions. Admittedly, this was not definitive “proof” of the value of palliative care, but given the circumstances it was as close as Chan could ethically approach the issue of causality.

Arguably the most complex CDM/PhD dissertation exemplar I have been involved with was also conducted in Hong Kong and combined a retrospective CDM and a prospective RCT (Lo, 2011). Working as a practitioner in a family service agency, Lo was interested in studying the effectiveness of cognitive behavioral group therapy (CBGT) in reducing anxiety and depression in Hong Kong Chinese adults. For CDM research purposes, it was fortuitous (aka “lucky”) that Lo’s agency routinely employed and computerized standardized quantitative measures for assessing agency clients at intake and at program completion. Unlike Chan, he did not need to convert voluminous qualitative patient chart entries into quantitative data. This made extraction and analysis relatively simple as well as unobtrusive. Because his sample exceeded 600 and all agency clients routinely completed the General Health Questionnaire-12 and PSRS self-reported health and mental health inventories, he had rich potential database to mine what was previously seen as useful only on a case-by-case basis.

Though he did know what to expect, Lo was able to show that the effective size of CBGT in his study sample was comparable to that found in prior RCTs with Asian and non-Asian populations. However, recognizing that even a relatively high effect size in the .80 range does not indicate that CBGT is effective with everyone, Lo proceeded to research numerous psychosocial factors as well as treatment variables associated with better or worse CBGT outcomes. These included but were not limited to personality types, energy levels, and whether clients were concurrently receiving individual therapy (Lo, Epstein, Ng, Chan, & Kwan, 2011).

Having found that CBGT worked best with clients who scored high on rigid and unreflective personality styles, Lo went on in his PhD dissertation to conduct an RCT employing a mindfulness intervention on the assumption that less directive and more self-reflective intervention modalities would be more effective with those who scored low on personal rigidity.

From the standpoint of our discussion of CDM’s contribution to this inquiry, Lo’s dissertation not only allowed him replicate the findings of previous CBGT/RCTs with non-Asian clients, but it went beyond them to identify potential moderators that allow for differential assessment and implementation of different forms of intervention based on client personality factors as well as other contextual variables. This empirically based contextualization is rarely possible or even sought in RCTs or meta-analyses. Depending on data available, however, it is clearly possible in CDM studies. Moreover, from a practice standpoint, it provides at least correlational support for making differential treatment decisions. From a research standpoint, these differential findings represent more refined hypotheses for further testing.

Ultimately, I would argue that Lo’s complex and highly differentiated findings are more useful from a practice standpoint, than an RCT-based, meta-analysis demonstrating the universal efficacy of group behavioral intervention with depressed adolescents. But that debate may be beyond the scope of this article.

Discussion

In this article, I have attempted to demonstrate the potential of CDM studies for engaging social work practitioners and practice-oriented doctoral dissertation students in PBR with routinely available qualitative and quantitative practice data. For the practitioner, CDM eases the transition from P2R roles and fosters empirically based reflection about one’s own practice. For doctoral dissertation researchers, it ensures practice-relevant studies that can make a significant contribution to knowledge without necessarily requiring original or intrusive forms of data collection. For practitioners, students, and methodologically pluralist researchers, CDM offers unique opportunities to study phenomena that for various reasons do not lend themselves to original data collection or experimental designs and to contextualize knowledge in ways that RCTs and meta-analyses either ignore or intentionally “strip away.” And as indicated earlier, CDM studies may be effectively combined with other designs or data collection strategies in mixed method studies.

It should be clear, however, that I am not saying that CDM is the only possible strategy promoting P2R traffic. Starting with an R2P collaboration model proposed by McKay and Paikoff (2007), Dodd and I proposed a two-way model of practitioner–research collaboration (Dodd & Epstein, 2012, p. 198). Likewise, practitioner-friendly, clinical information systems can be constructed (Grasso & Epstein, 1993) to support practitioner research and decision making as well as administrative monitoring and evaluation. Virtual communities of practice can be systematically “mined” for best practices (Cook-Craig & Sabah, 2009) and practitioner contributions to practice manuals can be more systematically collected and empirically tested (Galinsky, Fraser, Day, & Richman, 2013).

Agency-based infrastructures can be created to encourage practitioner research. Under the leadership of the late Helen Rehr, Mount Sinai has been doing this for 40 years. Influenced by Rehr’s example, Melbourne, Australia, now has five hospitals with PBR units staffed by practice-oriented PhDs with joint hospital–university appointments. Last fall, I conducted PBR and CDM training with 40 health and mental health practitioners under the auspice of the Hong Kong Hospital Authority. Despite these disparate P2R efforts and possibilities, most current practice–research integration efforts still take an R2P “pipeline” approach with too little to show in practice utilization for all the energy and money pumped into it.

Conclusion

In his review of my CDM text, Barber (2011) justifiably criticizes me for pitting EBP and RCTs against PBR and CDM studies. He comments:

The point is that Epstein’s CDM is just as much in the business of gathering evidence as any RCT is. His methods may have their limitations (of which he is well aware) but so do RCTs. And so did Charles Darwin’s. So what? You make the best decisions you can according to the best evidence you have. Ironically, this is at the very heart of evidence-based practice. So why all the arguing? Are not CDM and EBP in furious agreement on this very point? (p. 488)

Barber goes on to say that when social work research stops making invidious comparisons between practitioners and researchers, between qualitative and quantitative research, and between “gold” and “silver” standard evidence; then social work research will have truly come of age. That includes me. And isn’t it about time? But if other helping professions can do CDM and P2R, why can’t we?

Footnotes

Acknowledgment

A special thank-you to Dean Ira Colby for his ongoing support of this effort.

Author’s Note

This article was invited and accepted by the Guest Editor of this special issue, Danielle E. Parrish, PhD. This article was previously presented at the conference on Bridging the Research and Practice gap: A Symposium on Critical Considerations, Successes and Emerging Ideas, sponsored by the University of Houston Graduate College of Social Work, Houston, TX, April 5–6, 2013.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

References

Auslander

Dobrof

Epstein

(2001). Comparing social work’s role in renal dialysis in Israel and the United States: The practice-based research potential of available clinical information. Social Work in Health Care, 33, 129–152.

Barber

(2011). Book review. Clinical data mining: Integrating practice and research. Research on Social Work Practice, 21, 487–488.

Beder

(1999). Evaluation research on the effectiveness of social work intervention on dialysis patients: The first three months. Social Work in Health Care, 30, 15–30.

Chan

(2007). A clinical data-mining study of the psycho-social status of Chinese cancer patients in palliative care (Unpublished doctoral dissertation). University of Hong Kong, Hong Kong.

Chan

Epstein

(2012). Researching “good death” in a Hong Kong palliative care program: A clinical data-mining study. Omega: Journal of Death & Dying, 64, 203–222.

Chan

Epstein

Reese

Chen

(2009). Family predictors of psychosocial outcomes among Hong Kong Chinese cancer patients in palliative care: Living and dying with the “support paradox”. Social Work in Health Care, 48, 519–532.

Cook-Craig

P. G.

Sabah

(2009). The role of virtual communities of practice in supporting collaborative learning among social workers. British Journal of Social Work, 39, 725–739.

Cordero

(2000). When reunification works: A family strengths perspective (Unpublished doctoral dissertation). Graduate Faculty in Social Welfare, The City University of New York, New York, NY.

Cordero

(2004). When family reunification works: Data-mining foster care records. Families in Society, 85, 571–580.

10.

Cordero

Epstein

(2005). Refining the practice of reunification: “Mining” successful foster care case records of substance abusing families. In Mallon

G. P.

Hess

P. M.

(Eds.), Child welfare for the twenty-first century: A handbook of practices, policies and programs (pp. 392–403). New York, NY: Columbia University Press.

11.

Davis

Gervin

White

Williams

Taylor

McGriff

(2013). Bridging the gap between research, evaluation, and evidence-based practice. Journal of Social Work Education, 49, 16–29.

12.

De Angelis

(2010). Mending the split, the Kazdin way. Monitor on psychology, 41, 45.

13.

Dobrof

Dolinko

Lichtiger

Uribarri

Epstein

(2000). The complexity of social work practice with dialysis patients: Risk and resiliency factors, interventions and health-related outcomes. Journal of Nephrology Social Work, 20, 21–36.

14.

Dobrof

Dolinko

Lichtiger

Uribarri

Epstein

(2001). Dialysis patient characteristics and outcomes: The complexity of social work practice with end-stage renal disease. Social Work in Health Care, 33, 21–36.

15.

Dobrof

Ebenstein

Dodd

Epstein

(2006). Caregivers and professional partnership caregiver resource center: Assessing a hospital support program for family caregivers. Journal of Palliative Medicine, 9, 196–205.

16.

Dodd

S. J.

Epstein

(2012). Practice-based research in social work: A guide for reluctant researchers. London, England: Routledge.

17.

Epstein

(1987). Pedagogy of the perturbed: Teaching research to the reluctants. Journal of Teaching in Social Work, 1, 71–89.

18.

Epstein

(2001). Using available information in practice-based research: Mining for silver while dreaming of gold. In Epstein

Blumenfield

(Eds.), Clinical data-mining in practice-based research: Social work in hospital settings (pp. 15–32). Binghamton, NY: Haworth Press.

19.

Epstein

(2009). Promoting harmony where there is commonly conflict: Evidence-informed practice as an integrative strategy. Social Work in Health Care, 48, 216–231.

20.

Epstein

(2010). Clinical data-mining: Integrating practice and research. New York, NY: Oxford University Press.

21.

Epstein

Blumenfield

(Eds.). (2001). Clinical data-mining in practice-based research: Social work in hospital settings. Binghamton, NY: Haworth Press.

22.

Epstein

Zilberfein

Snyder

(1997). Using available information in practice-based outcomes research: A case study of psycho-social risk factors and liver transplant outcomes. In Mullen

E. J.

Magnabosco

J. L.

(Eds.), Outcome measurement in the human services: Cross-cutting issues and methods (pp. 224–233). Washington, DC: NASW Press.

23.

Galinsky

M. J.

Fraser

M. W.

Day

S. H.

Richman

J. M.

(2013). A primer for the design of practice manuals: Four stages of development. Research on Social Work Practice, 23, 219–228.

24.

Giles

Epstein

Vertigan

(Eds.). (2011). Clinical data-mining in an allied health organization: A real world experience. Sydney, Australia. University of Sydney Press.

25.

Grasso

A. J.

Epstein

(Eds.). (1993). Information systems in child, youth and family agencies: Planning, implementation and service enhancement. Binghamton, NY: Haworth Press.

26.

Green

L. W.

(2007). PRECEDE-PROCEED & RE-AIM as frameworks for practice-based planning and evaluation. CDC Oral Health Workshop, Atlanta, October. Retrieved from http://www.astdd.org/docs/LarryGreenPresentationSelectedSlides.pdf

27.

Hanssen

(2003). Looking inside the black box of intensive family preservation services (Unpublished doctoral dissertation). Graduate faculty in Social Welfare, The City University of New York, New York, NY.

28.

Hanssen

Epstein

(2006). A “black box” study of intensive family preservation services: Utilization of clinical data-mining. Family Preservation, 9, 7–22.

29.

Hanssen

Epstein

(2007). Learning what works: Demonstrating practice effectiveness with children and families through retrospective investigation. Family Preservation, 10, 24–41.

30.

Hess

P. M.

Mullen

E. J.

(1995). Bridging the gap. In Hess

P. M.

Mullen

M. J.

(Eds.), Practitioner-researcher partnerships: Building knowledge from, in and for practice (pp. 1–30). Washington, DC: NASW Press.

31.

Joubert

Epstein

(Eds.). (2005). Multi-disciplinary data-mining in allied health practice: Another perspective on Australian research and evaluation, Special Issue. Journal of Social Work Research and Evaluation, 6, 139–141.

32.

Joubert

Epstein

(Eds.). (2013). Current themes in health social work practice research: Academic, practice partnerships at the University of Melbourne. Special Issue. Social Work in Health Care, 52, 105–109.

33.

Kazdin

A. E.

(2008). Evidence-based treatment and practice: New opportunities to bridge clinical research and practice, enhance the knowledge base, and improve patient care. American Psychologist, 63, 146–159.

34.

Lalayants

(2010). Multidisciplinary clinical consultation in child protection: Contextual influences and stakeholder perceptions of best practices (Unpublished dissertation, PhD Program in Social Welfare). Graduate Center of the City University of New York, New York, NY.

35.

Lalayants

Epstein

Adamy

(2011). Multidisciplinary consultation in child protection: A clinical data mining evaluation. International Journal of Social Welfare, 20, 156–166.

36.

Lalayants

Epstein

Auslander

Chan

Fouche

Giles

… Vertigan

(2012). Clinical data-mining: Learning from practice in international settings. International Journal of Social Work, 56, 1–23.

37.

Epstein

Chan

Kwan

(2011). When cognitive behavioral group therapy works and when it doesn’t?: Clinical data-mining good and poor outcomes for depression and anxiety among Hong Kong Chinese. Social Work in Mental Health, 9, 456–472.

38.

H. H. M.

(2011). A body-mind-spirit approach to depression and anxiety: Application of mindfulness and compassion training for Hong Kong Chinese (PhD dissertation). The University of Hong Kong, Hong Kong.

39.

McKay

Paikoff

(Eds.). (2007). Community collaborative partnerships: The foundation for HIV prevention research efforts. New York, NY. Taylor & Francis.

40.

National Cancer Institute Cyber-seminar. (2012). Bridging research and reality: Practice-based evidence & evidence-based practice, September 11. Retrieved from https://researchtoreality.cancer.gov/node/939

41.

Parrish

Rubin

. (2012). Social workers’ orientations toward the evidence-based practice process: A comparison with psychologists and licensed marriage and family therapists. Social Work, 57, 201–210.

42.

Patterson

D. A.

Maguin

Dulmus

C. N.

Nisbet

(2013). Individual worker-level attitudes toward empirically supported treatments. Research on Social Work Practice, 23, 95–99.

43.

Peake

Epstein

Medeiros

(Eds.). (2005). Clinical and research uses of an adolescent intake questionnaire: What kids need to talk about. Binghamton, NY: Haworth Press.

44.

Pignotti

Thyer

B. A.

(2009). Use of novel unsupported and empirically supported therapies by licensed clinical social workers: An exploratory study. Social Work Research, 33, 5–17.

45.

Ross

Walther

Epstein

(2004). Screening risks for intimate partner violence in primary care settings. Social Work in Health Care, 38, 1–24.

46.

Rossi

(1992). Assessing family preservation programs. Children and Youth Services Review, 14, 77–97.

47.

Rubin

Rosenblatt

(Eds.). (1979). Sourcebook on research utilization. New York, NY: Council on Social work Education.

48.

Sainz

Epstein

(2001). Creating experimental analogs with available clinical information: Credible alternatives to ‘gold standard experiments’? Social Work in Health Care, 33, 163–164.

49.

Sales

Lichtenwalter

Fevola

(2006). Secondary analysis in social work research education: Past, present and future promise. Journal of Social Work Education, 42, 543–558.

50.

Scriven

(1995). The logic of evaluation and evaluation practice. In Fournier

(Ed.), Reasoning in evaluation: Inferential links and leaps (New Directions for Program Evaluation, No. 68) (pp. 49–70). San Francisco, CA: Jossey-Bass.

51.

Wretman

C. J.

(2013). Book review of the profession of social work: Guided by history, led by evidence. Research on Social Work Practice, 23, 363–364.