Defining identity
Identity has been a strong preoccupation of philosophy and social science enquiry for hundreds of years, both individual identity and collective or group identity. This is because identity is thought to shape individual and social behaviour [
1-
4]. One of the most influential theories of identity comes from the work of French philosopher Foucault on whom the Library of Congress holds over 10,000 items. Foucault defined individual and collective identity as a social construct that is created in language. In Foucault’s writings, historical and social processes work to actively create these identities, operating sometimes subliminally through particular kinds of knowledge to shape what assumptions people accept about themselves and others [
5,
6]. An example of this, important in this study, is how nursing and patient care have been constructed over the last 25 years in the discipline of health services research, as captured in a particular database PubMed.
In discursive psychology, which builds on the work of Foucault, identity is also understood to be about how language is used to create and position the self in relation to the world [
7-
9]. This second aspect of identity is also important to this study. In this study, the focus is upon how nursing identity is positioned in relation to other concepts in the discipline of health sciences as reflected in PubMed [
8,
9]. Nursing is seen as positioned in the language of health sciences research in ways that create particular kinds of “story-lines” that suggest professional attributes [
7-
9]. Of particular interest in this study is how this identity may be measured, and whether and how it offers insights into the quality agenda in healthcare in which patient-centred care is perceived as critical.
Definitions of identity in the nursing identity literature
Identity has also been defined extensively in the nursing literature. The body of literature about nursing identity is a growing one, albeit one that sometimes lacks methods and data to support its hypotheses. For the period 1
st January 2008 to 5
th of September 2013, a total of 505 papers are listed in PubMed including the terms “nursing” OR “nurse” OR “nurses” AND “identity.” A recent review of nursing identity has made a distinction between the professional self-identity of nurses and the public identity or image of nursing [
10]. The public image has been described as partly created by default by nurses themselves who have not always participated in public discourse. The professional self-identity of nurses is a complex construct that includes, but is not limited to, this public image, work contexts, work values, education and social or cultural values [
10]. Specialist nurses such as mental health nurses have been found to have an identity comprised of a cluster of “identity characteristics” [
11]. The acquisition of a nursing professional identity through processes of “professional socialisation” is known to be complex and multi-faceted and is the focus of a substantial part of the nursing identity literature [
12]. Nursing identity has also been described as a measureable self-concept that can be “manipulated” to help improve nursing retention, although sound psychometrically validated instruments are needed [
13,
14].
Patient-centred care and its importance to nursing identity
One of the most important constructs in the popular and scholarly literature about nursing identity is “patient-centred care.” Yet it is also one of the most contested constructs. A recent review found that nurses, physicians and managers all used the discourse of “patient-centred care” to suggest that their own professional group was patient-centred while other professional groups were not [
22]. In the language of policy-makers, “patient-centredness” may be a lever for achieving change in professional attitudes, as in the United Kingdom’s public debate about graduate nurses being “too posh to wash” or deliver essential nursing care [
23,
24]. In such a politicised debate in the media, problems with nursing recruitment, as well as nursing care, have been seen as being caused by the increasingly academic nature of nurse education and the specialisation of the nursing professions [
25]. Yet in the nursing literature, the role of “essentialist caring values” is often described as important to the future of nursing [
26].
However, patient care as a component of professional identity—the degree to which caring is seen as a part of nursing professional identity—has had relatively less attention in the contemporary nursing identity literature. Caring in nursing is articulated as being linked to task-oriented approaches to nursing and being grounded in the relational empathy and connection between the professional nurse and the patient [
27]. A “dimensional analysis” of patient-centred care describes it as including an eclectic range of elements, from providing a comforting room design to emotional support to attention to individualised meals to supports for patient decision-making [
28]. The authors of this analysis considered 69 papers published between 2000–2006 and found that the common themes of patient-centred care involved complex and sustained attention in nurse-patient interactions to alleviating vulnerabilities, as well as therapeutic engagement based on a knowledge of, and relationship with, the patient [
28]. Thus, in the nursing literature, “patient-centred” care has a strongly individual patient-nurse meaning.
Beyond the literature on nursing identity as such, nursing theory and knowledge development have traditionally included a strong discourse of “caring science scholarship” [
29]. The construct of caring is a multi-faceted one that has been the subject of different conceptualisations and theoretical frameworks since the 1990s at least [
30-
33]. Watson has described the dominant construct of caring as involving emotional and physical labour components, or expressive and instrumental elements, and argued that such intuitive constructs can be operationalized and measured [
34,
35]. The construct has been most notably applied using instruments such as the Caring Dimensions Inventory to measure student perceptions of caring as having both psycho-social and professional/technical dimensions [
34-
37]. By 2001, 16 such instruments were reviewed in a monograph on assessing and measuring caring [
38]. Yet there is a perceived divide between qualitative and quantitative approaches to researching caring [
35] and caring research has been criticised for failing to yield to scientific advances in knowledge [
39].
Perceived threats to the patient-centredness of care and the quality of healthcare
Despite the importance of “patient-centred” care to nursing identity, it is clear that it is perceived as being under threat in ways that threaten the quality of healthcare. For example, in hospital settings, studies have suggested that only around a third of the time of different nursing professionals is spent on direct patient care, despite the growing complexity of these direct care needs: admission and assessment, hygiene and patient/family interaction, medication and IV administration and procedures [
40]. Concerns about the deregulation of direct patient care in, for example, areas of growing care complexity such as aged care, position a declining valuing of direct patient care and care workers as part of the problem of care quality [
41]. The whole evidence-based practice movement has been seen as devaluing the complex interpersonal element of the caring components of nursing [
42]. That is, the emphasis on evidence-based practice, outcomes-based practice, and quantifiable efficiencies has been seen as ultimately threatening the fundamental “caring” strengths of nursing identity and practice, and the quality and safety agenda [
43]. Industrially, “caring” in nursing identity is also seen as under threat by new regimes of economic management, characterised by widespread cuts to services and staff reductions [
44].
This is not to argue that patient-centredness has an unproblematic relationship to patient outcomes. The evidence is mixed regarding the efficacy of patient-centredness as a style of doctor-patient communication [
45]. The asymmetry between doctor and patient is thought to have deep and even possibly justified roots in the clinical authority of the doctor [
45]. However, as the foregoing discussion demonstrates, the nursing literature has defined patient-centredness far more broadly: in terms of tasks of caring and verbal and non-verbal interaction. “Patient-centeredness” has deep roots in the whole identity of the nursing profession itself in which a nurse has a pastoral authority for broader patient health and well-being. In a context in which it is known that health outcomes are shaped by wider determinants of health, it seems likely that 1) a narrow definition of patient-centredness will not achieve real changes in patient outcomes, and 2) the patient outcomes of a broader definition of patient-centredness consistent with nursing caring scholarship will continue to be difficult to measure.
Notwithstanding, it is certain that ideas about nursing identity and patient-centred care will continue to shape public debate and healthcare quality agendas. For example, both the UK’s Essence of Care [
46] and Australia’s Essentials of Care [
47] emphasise a re-establishment of the traditional direct care roles of nurses. Underlying such debates is an uneasiness about the relevance of new nursing identities, in a context in which a disconnect between the disciplinary mission of nursing and its constituent patient communities is known to lead to the failure of health programmes [
48].
The importance of health services research to constructing nursing identity in the 21stC
The current study explores the construction of nursing identity from another perspective—that of research evidence, specifically health services research. The discipline of health services research is a particularly important field in which to examine nursing identity because it delivers the evidence that potentially shapes practice in health services. Health sciences research will increasingly be under pressure to deliver service benefits, through the emphasis on translational research of such United States-based agencies as the Institute of Medicine, Agency for Healthcare Research and Quality, and National Institutes of Health. The entire translational research movement, originally about better articulation of research benefits from the science laboratory “bench” to clinical practice “bedside” [
49,
50] has developed over the last quarter of a century into new sub-disciplines of health research ensuring that health sciences research also includes health services research [
51].
Accordingly, this study aims to explore to what extent the emerging discipline of health services research has positioned nurses as patient caregivers over the last quarter of a century in which over 200,000 papers in this discipline have been published. As such, this paper responds to calls in a recent review of nursing identity for more conceptually-based mapping of nursing identity constructs, capturing the subtlety of changes over time [
13]. It also responds to the need for more information for nurses that can help them reflect on, and shape, the construction of their own identity, including patient-centredness, particularly given its importance in healthcare reform policy, nursing education and training, and care efficacy [
10,
52].
Machine-driven text analytics
The size of the corpus of abstracts in PubMed relevant to health services is prohibitive for traditional manual methods of qualitative analysis. However, automated “natural language processing” or “computational linguistics” — the methods that inform the software used in this study — has developed over the last 30 years to offer ways of handling such unstructured datasets [
53]. Natural language processing can be described as a diverse range of automated computational techniques for analysing and representing language. It underpins technological innovations such as Google and IBM’s Watson, as well as Apple’s Siri [
54]. This section does not reproduce the content of reviews of the field published elsewhere [
54], instead focussing on literature offering examples of applications in the health sciences and the justification for our choice of Leximancer software.
As of 15 April there were only 1,034 journal papers listed in the database PubMed with the terms “computational linguistics,” “natural language processing,” or “text analytics” in the title or abstract, 655 of which were published since 2008. These suggest that, in health, applications have been diverse and include: 1) exploration of theories in the light of social media [
55]; 2) analyses of clinical texts for the purposes of classification [
56], symptom description [
57] and diagnosis [
58], the analysis and prediction of patient outcomes [
59,
60] and evaluation of the extent of utilisation of evidence in health practice [
61]; 3) the analysis of healthcare experiences and behaviours using popular media such as YouTube [
62]; and 4) clinical decision support [
63]. This study takes a socio-cultural perspective, focussing on the extent to which patient-centred nursing is under-represented in the discipline of health services as represented in the PubMed database and as such may be described as a form of “evidence surveillance.” In this study, software associated with natural language processing is used to explore a theoretical construct — nursing professional identity as it relates to patient-centredness in the PubMed corpus.
It is acknowledged that those wanting to learn how to do natural language processing (NLP) can access the open source Natural Language Toolkit (version 2.0
http://www.nltk.org) using Python programming language (version 2.7,
http://www.python.org) [
64]. However, not all researchers will be able to become proficient computer programmers or have the resources to access computer programmers. Some researchers may also choose to use high quality standardised software products developed by advanced programmers, particularly for complex procedures such as concept mapping and sentiment analysis, to help ensure standardised results that are supported by detailed manuals and (ideally) technical validity studies. For these researchers we suggest consideration of “text analytics” software available on the market, designed using the methods of NLP. A summary of these “text analytics” products is available by market analysts [
64] and there are also digital libraries available at
http://dirtdirectory.org/.
The software package Leximancer [
65], originally created at the University of Queensland in Australia, includes two key functions—concept mapping and sentiment analysis—that are fairly standard in the field. Leximancer compares favourably with other products listed in the Hurwitz report [
64] when criteria such as scholarly literature modelling its application, cost, installation and technical running problems and support, as well as user friendliness, are considered. We have tried and tested other products including the costly IBM SPSS Modeller Premium software (Version 15.0,
http://www.spss.com.hk/software/modeler/). We have found that other products do not have the scholarly validity study [
66] or body of published scholarly applications supporting Leximancer’s concept analysis and sentiments analysis procedures [
67-
74]. However, there has not yet been an application of Leximancer to the corpus of health service research abstracts, although the software offers a novel technique for mapping this large language dataset. Notwithstanding, the software suffers from the limitations of other NLP approaches: a lack of “real world” and common-sense knowledge about the contextualised meanings of words.
Leximancer is Bayesian-based software that “learns” from an uploaded dataset that it reads iteratively. For concept mapping, Leximancer creates a network of concepts defined in “text blocks” of about a paragraph in size, normally three sentences, but defined by a process of machine learning (described below). In Leximancer, a text block is the unit of analysis. An abstract may contain one or more text blocks and one text block may contain one or more concepts, just as a sentence can contain more than one concept. In sentiment analysis, Leximancer maps the frequency and co-occurrence of concepts with an in-built thesaurus of sentiment terms (negative versus positive). In more technical terms, Leximancer is described as an unsupervised approach to transforming lexical co-occurrence information from language data into semantic patterns. There are two kinds of co-occurrence information extraction in Leximancer’s automated procedures—semantic and relational—each with a different algorithm [
66]. These algorithms employ nonlinear dynamics and machine learning as summarised in the following technical overview:
A unified body of text is examined to select a ranked list of important lexical terms on the basis of word frequency and co-occurrence usage. These terms then seed a bootstrapping thesaurus builder, which learns a set of classifiers from the text by iteratively extending the seed word definitions. The resulting weighted term classifiers are then referred to as concepts. Next, the text is classified using these concepts at a high resolution, which is normally every three sentences. This produces a concept index for the text and a concept co-occurrence matrix. By calculating the relative co-occurrence frequencies of the concepts, an asymmetric co-occurrence matrix is obtained. This matrix is used to produce a two-dimensional concept map via a novel emergent clustering algorithm. The connectedness of each concept in this semantic network is employed to generate a third hierarchical dimension, which displays the more general parent concepts at the higher levels [
66].
This means that, for example, for the concept of nursing, Leximancer works by merging similar terms e.g. nurses, nursing, nursing assistant. The classifications of nursing will be developed from the text: if “physician’s assistant” is used synonymously with “nurse” the term may be learned by Leximancer and integrated into a nursing concept. The analyst also scrutinises the output at different stages to identify the concept seed words and the thesaurus terms Leximancer produces. Terms that are obviously the same can be merged manually. This demonstrates an important feature of Leximancer: it is automated but also features multiple interactive data viewing windows that allow analyst scrutiny and intervention.
The current study is therefore not a study of types of research but rather of concepts in the corpus of abstracts about health services in PubMed. The advantage of a concept analysis is precisely that it does not involve a valuing or critiquing of research using familiar typologies of research such as quantitative or qualitative research or any other way of describing research methods or research disciplines. Rather, this study measures the prevalence of particular concepts that cut across these different types of research. An example of a text block classified by Leximancer under the nursing concept, and tending to emphasise nursing service structures rather than the immediate pastoral dimensions of caring in nursing, is as follows:
PMID: 3228766
409. Can J Nurs Adm. 1988 Oct;1(3):16–8.
Ambulatory care nursing. A new approach.
McMaster DC, Greer PM, Beanlands HE.
This article describes one hospital’s experience with integrating inpatient and outpatient nursing services. Nursing services integration enabled nursing management to combine the nursing resources allocated to the inpatient and outpatient components of a clinical service under the direction of one nurse manager. This new and creative approach was implemented in thirteen clinical services at the Victoria General Hospital, Halifax. This organizational structure was considered to be an effective approach for managing ambulatory care nursing services. The introduction of integration facilitated a change in the ambulatory care nursing role. It also provided for increased continuity of patient care and afforded nurses the opportunity to practice in another setting. Nursing services integration is considered a more effective approach for managing nursing resources.
Further technical details about Leximancer analysis are available in the validity study [
66].