DATA ANALYSIS

4.5.1. Statistical analysis

The statistical analysis of the follow-up inquiries was carried out using SPSS/Win 7.0 Software. The analysis begun after each inquiry by forming frequency distributions in order to identify possible coding errors and to examine the distributions of each variable. The validity of the sum variables in the instrument was tested with the data of the first inquiry with confirmatory factor analysis. This confirmed the relevancy of the factors (see Table 3), but also indicated three main factors which were as follows: (I) individual factors:

growth motivation (OC1), performance motivation (OC3), reflectivity (RF1), (II) team factors: atmosphere (AF1), team spirit (GF1), team’s functionality (GF2), commitment to work and organisation (OC2), (III) organisational factors: encouragement value of work (WF1), possibility to influence (WF2), participatory management style (MF1), performance oriented management style (MF2), task and goal systems (OF2). The sum variables were calculated next for all the follow-up inquiries, and the respective graphs were examined to check the normality of the distributions. In the third phase, the data files from the different inquiries were merged for final analysis. The analysis of variance for repeated measures was used for exploring the changes

in the prerequisites for professional development during the team supervision intervention. This method included the following components: the ward effect as the between factor, the time effect and the interaction between the ward and time as within factors. The level of significanse was set at ≤ .05. (Polit 1996, Burns and Grove 1997, Polit and Hungler 1997)

The methods of statistical process control and control chart were used for analysing and reporting the

‘Continuous self-monitoring of work’ and the ‘Patient satisfaction feedback’ questionnaires monthly. This statistical analysis was carried out using the Statistica software for Windows. The literature characterizes the methods of statistical process control as effective and practical tools as they provide an approach to describing and analysing the structures and processes that affect the quality, but also produce information that promotes the understanding of variation (Finison et al. 1993, Benneyan 1998). The method is based on an assumption that natural variation will always occur in any process and consequently the results or outcomes of the process vary as well (Finison et al. 1993, Benneyan 1998, see also Iberg 1991). The variation is caused by common and special causes. The purpose of statistical process control is to detect statistically and objectively the variation in process quality, to distinguish and differentiate variation due to common (random effects) or special causes (non-random effects) and provide longitudinal information of process quality and its possible changes. It is claimed that evidencing the variation, its nature and extent are the statistical starting point for quality improvement, as the causes of variation are generally difficult to determine through ordinary observation or intuitively because of minor variation. (Finison et al. 1993, Finison and Finison 1996, Benneyan 1998)

The control chart is defined as a graphic presentation of the performance in a process or outcome observation through the parameters of statistical process control (e.g. Finison et al. 1993) in a sample that is arrayed in some rational sequence (Finison and Finison 1996). The key components in control chart are the central line (describing the means of the observations), upper and lower warning (x + ±2σ) and acting (x +

±3σ) limits calculated from the standard deviation (SD) within the group of observations. The warning and acting lines are based on .05 (±1.95 SD) and .01 (±3.09 SD) statistical significance levels and the 95% and 99.8% confidence intervals. The process and its variation are examined against the statistically determined limits (Finison et al. 1993, Finison and Finison 1996, Benneyan 1998). The examination of control charts is done visually and utilising the warning and acting lines. The process is said to be ‘in control’ when the variation occurs between the acting limits and the causes for variation are assumed to be common (Benneyan 1998). The process is said to be ‘out of control’ when the observations are outside the acting limits and the variation is assumed to be due to special causes (Finison and Finison 1996, Benneyan 1998). The improvements in quality are expected to show statistically and in control chart through decreased variation of common causes and observations that fall constantly between the acting limits. It is claimed that the acting limits get narrower, thus reflecting the reduced variation and the shift of the central line to a positive and beneficial direction (Finison et al.1993).

58 4.5.2. Qualitative analysis

Content analysis

The written answers to the open-ended questions in the follow-up questionnaire, the continuous self-monitoring of work and the patient satisfaction feedback questionnaires were analysed using content analysis. The method is common in qualitative research studies, but it is also used in quantitative studies.

The method allows to describe the data (e.g. written answers) qualitatively, systematically and objectively, to manage large volumes of data, to quantify the categories created and to analyse them further with statistical methods (Morgan 1993, see Nieswiadomy 1993, Polit and Hungler 1997, Burns and Grove 1997, see also Krippendorff 1980). Some authors (e.g. Polit and Hungler 1997, Burns and Grove 1997) have even defined content analysis as a process during which the created categories are quantified. However, this definition has been criticised (see e.g. Krippendorff 1980, Tesch 1990) because the difference between qualitative and quantitative content analysis is regarded as blurred and too vague. When the focus of interest is on ‘what is this /what is happening here’, content analysis and its basis are qualitative, whereas the nature is quantitative when the interest is on finding ‘how many times this has happened’ (see also Morgan 1993, Nieswiadomy 1993, Polit and Hungler 1995, Burns and Grove 1997, Kyngäs and Vanhanen 1999).

This study used inductive content analysis. The analysis started by careful reading of the data that had been transcribed verbatim after every inquiry (i.e. monthly for the self-monitoring of work and patient satisfaction questionnaires and half-yearly for the follow-up inquiries). The unit of analysis was a theme describing, for instance, the patients’ positive and negative experiences during their hospital stay. The data were next organised into themes and after this the content of data in each theme was classified into sub-categories. Finally, the sub-categories were combined into categories and further into main sub-categories. (see Kyngäs and Vanhanen 1999). The analysis was undertaken separately for the staff’s self-assessment, patient satisfaction and follow-up inquiries. After the analysis was completed two researcher colleagues evaluated the relevance, clarity and completeness of the classification in the staff and the patient inquiries. The differences of opinions concerning the categories were discussed until an agreement was reached. The data on the staff’s inquiries were quantified by coding it according to the themes to be able to combine it with the rest of the statistical data.

Phenomenography

A noteworthy special feature of group interview is that the analytic unit is the group instead of individuals.

This is why an analysis and description of the group’s situation is always needed alongside the interview material. The results are examined from a related context, that is, from that of the group. The reason for this is that the group’s interaction and actions influence the nature of the data collected (Sulkunen 1991, Carey and Smith 1994). The researcher’s notes served this end.

The literature on group interview revealed that the methods of analysing group interview data are inadequately developed and that no mutually agreed technique exists. It has been suggested that depending on the purpose of the study, group interviews could be analysed using the phenomenological method,

grounded theory, the ethnographic method, narrative analysis or content analysis. (Carey and Smith 1994, Pötsönen and Pennanen 1998, also Henderson 1995). Critique has, however, been expressed concerning the incompatibility of the group-oriented data collection methods with the above-mentioned data analysis methods (Webb and Kevern 2001). The problems pointed out have concerned the ‘essence’ of the method, features of the methodological approach, approved procedures for ensuring the validity or rigour of the findings and the contradictions that may arise when the data have been collected from a group of people. In this study an assumption was made that after the long and intensive team supervision, the teams would be able to express their conceptions due to the ‘collective mind’ that had developed during the intervention (Marton 1988, see also Svensson 1984).

Several researchers (Carey and Smith 1994, Kitzinger 1994, Carey 1995, Krueger 1995, Morgan 1995) have indicated that the solutions used so far in relation to group-oriented data have not been fully satisfactory. Criticism has especially focused on the fact that insufficient attention has been paid to the impact of group context in the analysis of these type of data. The group context is inherently complex and this places demands on data analysis, the method and the description of the analysis process. The results and reporting strategies have come under criticism due to their alleged triviality and resemblance to those of individual interviews. The results of group interviews have been considered shallow, mainly lists of participants’ comments, although the data were produced in social interaction. (Carey 1995, Morgan 1995, Reed and Payton 1997, Smith 1995)

Phenomenography studies reality as people conceptualise it. Phenomenography is not interested in why people think the way they do, but in describing the variation of different conceptions of different things in a group of people under study. The interest is directed at what people are interested in (the ‘what’ aspect), but also at how study subjects construct their conceptions of the phenomenon under study (the ‘how’ aspect).

Categories formulated by the researcher, which describe the variation of the conceptions occurring in the data, are considered the results of phenomenographic research. The categories may be hierarchically, horizontally or vertically inter-related. This means that the method permits the description of the spectrum of different conceptions. (Marton 1988, Uljens 1989, Uljens 1993, Bowden 1995)

The group interviews were first transcribed verbatim and printed out. This produced 200 pages of text with 1.5 line spacing. The tapes were listened to once more ensuring the correspondence of the text and the tape. The analysis proceeded by condensing literal text without changing the content. Sequences of discussion were sought out first to discover what the group was actually talking about (see Reed and Payton 1997, cf. Launis 1994, Niemistö 1998). The next step was to examine how group members dealt with the topic while examining their experiences. Then a spectrum of conceptions started to emerge. The formation of categories required that data analysis actually proceeded at three different levels: at the level of the group, that of its individual members and of the context (i.e. wards) (Carey and Smith 1994). The initials of the interviewees were retained during the analysis process and these were included in the chosen quotations to illustrate the interaction between the interviewees. It was also deemed important to maintain a sense of the teams as a whole in the presentation of the findings and thus whenever possible quotations from at least two or more participants have been presented. During the interviews the teams could form a common conception

60

of the topic or it could be seen from different angles or even in contradictory ways. A sequential examination made it possible to describe the nature of the conceptions and the way they were composed in the teams.