Classifying nursing organization in wards in Norwegian hospitals: self-identification versus observation

Nursing services in hospital wards are structured and organized in different ways, with corresponding variations in the organization of the workforce and care processes. The choice of organizational model is an important element of service provision and influences the day-to-day running of a ward.

The organization of nursing services is often described using three generic models: (1)Functional nursing: tasks are allocated by similar principles to those of production lines, e.g. one registered nurse (RN) is responsible for infusion therapy for all patients, whereas another dresses wounds; (2)Team nursing: a small group of nurses with different qualifications are responsible for the care of several patients, so reducing the number of interpersonal contacts and clarifying the lines of responsibility. An RN is the team leader, supervising the work of less qualified personnel; (3)Primary nursing: one RN carries out all the care needed for a few patients. Primary nursing is conducive to the idea of holistic care and is assumed to lead to a good work environment [1‐4]. In everyday practice the organization of nursing services is customized to fit local conditions such as staffing, workload, ward size, interdisciplinary cooperation, working hours and regulatory environment. Accordingly, the organization of two wards could have different theoretical bases and names but in practice the distinction may not be clear cut. It has been suggested, therefore, that self-identification (i.e. model identified by staff on ward) should be avoided as a way of describing nursing organization [3].

A valid classification of nursing organization would be useful to identify groups of wards that can be compared, or to be used as a covariate in multivariate analysis. Explicit and verifiable descriptions would also facilitate the review of studies of complex interventions, e.g. nursing service organization [5]. A recent review of studies on hospital organization recommended the development of standardized instruments for collection of primary data [6].

This paper presents a study of wards in Norwegian hospitals using a classification procedure that could be developed to have universal applicability and be useful in nursing research.

Researchers have used different approaches to classify nursing service organization at the ward level. In a study of staff nurses' perceptions of staffing adequacy, Kramer and Schmalenberg listed six organizational models for collecting staff nurses' descriptions of the organizational models in their units [7]: (1) new team, (2) total patient care, (3) modified primary, (4) old team, (5) true primary and (6) varying from day to day. Aiken and Patrician used the Revised Nursing Work Index questionnaire to study the nursing practice environments [8]. They asked staff nurses how much they agreed with the following statements: 'Team nursing is the nursing delivery system', 'Total patient care is the delivery system' and 'Primary nursing is the delivery system.' In a study of nursing shortage, Seago et al. [9] used three different organizational models: (1) primary/total, (2) team/functional and (3) modular/case management. A recent review by Kane et al. [10] of studies on nurse staffing and quality of patient care named five organizational models: (1) patient-focused care, (2) primary nursing, (3) total nursing care, (4) team nursing and (5) functional nursing. In addition, a study by Adams et al. [1] of aspects such as job satisfaction and within-ward cooperation identified three organizational models: (1) devolved, (2) two tier and (3) centralized. There is apparently a lack of consensus about the classification and nomenclature of nursing organization at the ward level [11].

Minnick et al. [3] suggested that the lack of conclusive research findings with regard to the strengths and weaknesses of different organizational models is a consequence of indistinct classifications; there is a need for knowledge on this topic to guide decisions in nursing administration re work force deployment, with the aim of improving patient outcomes and nurses' work conditions. The predicted health personnel shortage [12] could lead to modifications of existing organizational models [13] with a consequent increase in the need for knowledge. Previous research has indicated that higher numbers of RNs on the staff are associated with better patient outcomes [10] and human resource management practices in hospitals (including the extent of team working) are associated with hospital mortality [14].

Wards in Norwegian hospitals usually have the following names for their organizational model: primary nursing, patient-responsible nursing, modified primary nursing, team nursing and group nursing. Only licensed nursing personnel are employed in Norwegian hospitals, RNs with 3 years of college education being in the majority; the rest are licensed practical nurses with vocational qualifications.

For a nursing administration study [15], a simple and valid classification system was needed to categorize models of nursing service organization at the ward level. Our expectation was that self-identification of organizational models would only to some degree correspond with reported ward data describing activity patterns on the wards. Consequently, the aims of this study were: (1) to construct a variable to classify the organizational models in a sample of hospital wards, based on reported data about ward practice; and (2) to compare this classification with the self-identification reported by ward nurse managers.

Through the use of cluster analysis for data reduction, we classified internal ward organization based on how the organizations functioned in practice. In a representative sample of wards in Norwegian general hospitals there was a moderate association between the empirical classification and the self-identification of organizational models. In many wards, the self-identification indicated a team or primary nursing organization, whereas the data suggested that they functioned in hybrid or mixed modes. This is an example of cluster analysis being used to categorize cases where no previous categories existed - often the case in healthcare organizations.

The results support previous warnings against the use of self-identification for nursing organizational models in research [3]. Outside the research field, in education as well as in discussions of nursing organization, we must be aware that terminology is an imperfect description of practice.

Differences between healthcare systems and between countries limit the generalizability of the classes identified in the current study, although the procedure used for classification is still generally applicable.

The three-cluster solution was consistent with the distribution of the questionnaire's supplementary variables, which were not used in constructing the clusters. These supplementary variables did not, however, differ between TLD and HM wards, which may be a reflection of the continuous transition between these two clusters, as observed in Figure 2, and also a consequence of limitations of the supplementary data, e.g. scheduling of shifts may not be a good variable for checking cluster consistency because choices may depend on, for example, night staffing more than on what is best from a strictly organizational point of view. By their design, the variables used in the clustering procedure give a more comprehensive description of the organization at the level of the patient-nurse interface.

In the current study, cluster analysis can be used to justify both the two- and the three-cluster solutions. Primary and team nursing organization are the main principles in ward organization, with local adjustments probably resulting in intermediate solutions. Over the last two decades, RNs have replaced practical nurses in many Norwegian hospitals. This means that teams now consist mainly of RNs, reducing some of the previous differences between team nursing and primary nursing, as observed in the HM cluster. The degree of separation between the clusters, as indicated by the discriminant functions plot, supported a three-cluster solution. The more detailed description of the three-cluster solution was regarded as more in line with theory and previous research and therefore more useful in practical research. However, the three classes of nursing organization were not identical to the generic types presented in the introduction.

In the study of Adams et al. [1], a similar clustering procedure identified three organizational models: in one cluster, labelled 'devolved nursing', responsibilities were mainly assigned to individual nurses, which corresponds to some degree with the PND wards in the current study. A second cluster, labelled 'two tier', was characterized by team work and a prominent role for the nurse manager. A third cluster was labelled 'centralized nursing', with less team work and more control in the hands of the ward nurse manager. The two-tier cluster has team work in common with our TLD wards, although it differs with regard to the role of the ward nurse manager, whose involvement in clinical practice has been reduced in Norway. This lack of responsibilities of the ward nurse manager on the TLD wards, and the absence of a parallel to the centralized cluster in the current study, can be explained by differences in time, settings and cultures.

The results of cluster analysis depend to some extent on the choices of the researcher, e.g. scaling of input variables, and should be viewed as an exploratory technique, which must eventually be validated by other means. We have demonstrated that our classification is stable with respect to sample variability and supported by validation from supplementary data. It is also consistent with theory and previous findings.

There is a possibility that the supplementary data were biased by the ward nurse managers wishing to present a consistent set of information [22]. Ideally, the classification should have been validated by detailed descriptions of the factual functioning of the ward organization by external informants, e.g. patients. Data collected from the patients in the current ward sample did not provide this possibility, but should be included in future use of the classification procedure. The possibility of obtaining validation data from outside the ward nursing services should also be considered, e.g. organization and workflow at hospital level.

The lack of pilot testing of the questionnaire is a limitation in the current study, but this limitation is mitigated by the use of items from previously developed questionnaires and by involvement of representatives of the survey population in the construction. Field observation studies and staff nurse surveys are potential alternatives to the approach used here, but these methods would require more resources. We consider a well-informed individual an adequate data source regarding global ward characteristics.

The classification that we obtained has been used to study the association of organizational models with RNs' ratings of quality of patient care, learning climate, job satisfaction and relationships with doctors in the RN survey data from the current ward sample [15]. Organizational models alone were not associated with RNs' ratings. When the models of analysis were expanded with explanatory variables describing other global ward characteristics, the association of these characteristics with the RNs' ratings varied with different organizational models. This supports the importance of including organizational models in studies of various aspects of nursing service management at the ward level.

A meaningful and statistically valid classification of wards into three classes of nursing organizational models: team leader-dominated, primary nurse-dominated and hybrid/mixed models, was achieved through empirical cluster analysis. This was, however, only moderately consistent with the self-identification. The empirical classification is an objective approach to variable construction and can be generalized to the population of Norwegian hospitals, e.g. when comparing patient and quality-of-care outcomes across different organizational models. The actual classes cannot necessarily be generalized to other countries, but the method is generally applicable and could be developed to find a reliable and valid classification across both time and healthcare systems.