• No results found

The original timelines proposed for the study included opportunity for piloting and pre-testing of MAST (appendices 4 and 5; pp. 241 - 242). Oppenheim (2000, p. 47) suggests that both of these are necessary, in order to ensure that the processes developed ‘work as intended’. MAST had been re-designed slightly, with reference to what demographic data were collected. Additional categories related to other potential mechanisms whose effects may impact upon students’ attitudes at each data point were added. Hence the inclusion of the demographic data categories of ‘morale’, ‘preferred speciality’ and ‘preferred career choice upon qualification’.

A total of 25 students, not in the PG, but from the same nurse education programme, completed MAST, on one occasion, in class. This process revealed that the item asking about the specialism of their current placement was problematic for respondents; these were open to interpretation, meaning that students often categorised the same placement specialism in different ways, for example, ‘male medical’ was used, as well as ‘cancer’ for the same area. Based on the pilot work, the specialisms were rationalised and clarified. However, once the study had begun, it became apparent that the students in the CG also had their own interpretations of what particular placement specialisms were. Not piloting the tool with the CG was an oversight, because of the differences in interpretation, and not knowing for sure what students in the CG meant when identifying particular specialisms. This meant that no meaningful analyses concerning these could be done.

A formal process of pre-testing was used to determine the frequency that MAST would be distributed, allowing time for both completion of the questionnaire, and for follow-up of non-respondents (Bourque and Fielder, 2003). Because students in both study groups would be alternately out on placement during the study, in diverse locations, mail administration of MAST was the sensible option. A list was created of

106 all students in both study groups, in order that address labels could be quickly generated at each data point, to facilitate speedy administration. A pre-paid envelope was included for return of the questionnaire. A unique identifying code was added to MAST on every page, so that recipients at each data point could be tracked, and followed up if needed. Having the code on every page would prove useful, if the part of MAST containing responses became separated from the front introductory letter (appendix 9; p. 249) and sheet. Introductory letters are known to have a beneficial effect upon response rates in mail-administered questionnaires, justifying the use of one in this study (Bourque and Fielder, 2003). Pre-testing demonstrated that a five-week interval between data points was sufficient to send out and receive a questionnaire, giving respondents time to complete and send it back. This turn-around time also allowed for a reminder letter to be sent to non- respondents, if required, three weeks after initial administration at a particular data point (appendix 10; p. 250).

3:7 Creation of the Interrupted Time-Series

An interrupted time-series is a number of repeated measures, of attitudes in this case, which are ‘interrupted’ by an intervention; an educational programme in this study. It was created by administering MAST 15 times, giving 15 data points in total. Seven were prior to the programme, and a further eight after, a key feature of interrupted time-series designs (England, 2005). A cumulative score for all respondents at each data point was calculated using the scoring rules for MAST outlined in section 3:6 (pp. 101 – 102). This cumulative score was then divided by the number of responses received at each particular data point, and the time-series was charted, with ‘time’ across the x axis, and ‘mean of cumulative attitude score’

107 the cumulative attitude score alone is discussed within chapter four, alongside the charted time-series’, which shows when the educational programme (‘interruption’) occurred (cf. figures 4.2 and 4.3; pp. 135 and 136).

The time-series is made up of frequent measures of attitude over time. There were two main reasons for measuring and recording attitude scores so frequently. The first was to contribute to the management of identified threats to LMCV, and the second was due to the nature and changeability of attitudes (Oppenheim, 2000).

Numerous measures of attitude facilitate the discernment of other trends in the time- series, which may require further investigation (Cook and Shadish, 1994). Whilst there appears to be no consensus about the number of data points required within an interrupted time-series, it has been noted that a longer pre-intervention phase can be beneficial in increasing the strength of the research design to detect upward or downward trends that were occurring anyway, without the programme (Ramsay, et al., 2003). Hence, the interrupted time-series can directly contribute toward the establishment of LMCV, particularly by accounting for ‘history’ and ‘maturation’ threats, as discussed in section 3:3:1 (pp. 88 – 91).

In relation to this, when effects might occur must also be considered. For example, programme effects might not be instantaneous; they may be ‘delayed in their initial manifestation’ (Cook and Campbell, 1979, p. 209). A shorter time-series, with fewer data points (and so fewer questionnaires) pre and post-programme, might lead to incorrect conclusions, such as attributing an improvement in attitude to the programme mechanisms, when a marked improvement was naturally occurring anyway, or the programme having no effect, if it was not immediate (England, 2005). In other words, by collecting multiple measures, the researcher can ‘observe how the different groups might be changing spontaneously over time’, and hence, be

108 more certain about what exactly might have caused any fluctuations or changes in the time-series, allowing useful comparisons between both groups’ attitude scores to be made (Cook and Shadish, 1994, p. 570).

As stated, the second reason for distributing MAST 15 times was related to the nature and changeability of ‘attitudes’ themselves. As Oppenheim said:

Some attitudes are more enduring than others. For instance, a man’s political beliefs may be fairly stable throughout his lifetime, whereas his attitudes to tennis or gambling may undergo multiple changes. (Oppenheim, 2000, p. 176)

The stability of nursing students’ attitudes in this field are not known for certain, hence, it seems reasonable to assume that the students’ attitudes toward caring for ill older people may be subject to such changeability, particularly after exposure to clinical placement experiences (Engström and Fagerberg, 2011). The potential changeability of attitudes discussed by Oppenheim (2000) could be mapped by frequent distribution of MAST, over time. Less frequent distribution may not have captured such changes, again leading to spurious conclusions being drawn.