Chapter 6 Limitations of the study and recommendations
6.1 Study limitations
Each research design has its general strengths and limitations (Watson et al, 2004). Limitations of the study are the restrictions or problems in a study (Burns and Grove, 2005) may compromise the generalisability of the findings. Limitations can be identified as threats to the internal or external validity of the study (Abbot and Sapsford, 1999).
6.1.1 Threats to internal validity
This refers to the “extent to which the effects that are detected in the study are a true reflection of reality rather than the result of extraneous variables” (Porter and Carter, 2000, p.30). The main threats to internal validity are:
History is an example of a possible internal threat to validity, which is described by Polit et al (2001) as being the occurrence of events concurrent with the study, which may independently cause stress. The disquiet expressed by nurses regarding pay and conditions in June 2007, which has received so much media attention, as outlined by the Health Service Executive (HSE, 2007; Wall, 2007) may have a confounding effect on stress levels of both groups. Also the embargo on appointment of new healthcare posts in the month preceding the study may have had a profound effect on participants who were most likely be in temporary employment or anticipating employment in the near future (HSE, 2007). But this limitation was also critical to the study, as the effects of this embargo need to be captured at this time. This not only had a possible affect on the level of stress, but also affected the distribution of questionnaires as some newly qualified nurses were working in different areas dependent on agency employment.
Effects of Selection: A matched group design would be more robust with more equivalent groups (Burns and Grove, 2005) such as ensuring student and newly qualified nurse work in the same area such as Accident and Emergency department. However this option was not chosen as there may be a recruitment drive in one clinical area which may attract newly qualified nurses which may be disproportionate to the number of fourth-year nurses in the same setting, therefore leaving the feasibility questionable. The problem with selection bias may affect the results, as this was a convenience sample. Reason for non-response is not known. The reason for participation may be due to being stressed or even a busy workload, which could add to stress. It may be the very reason why a questionnaire was not completed or for non-response. Interestingly in the UK, Harrison (2004) noted that staff were too
stressed into partake in a study on occupational stress in a healthcare facility. Paradoxically, there is evidence to account low return rate with work overload (Barr et al, 2008). Further evaluation in a larger sample size would be more beneficial. Because the questionnaires were sent to the hospital there is a possibility of participants completed them out together and ‘group think’ may affect the responses.
Instrumentation: The Nursing Stress Scale is a well-validated tool. However, given the different numbers of items in each factor, it was difficult to compare results from each factor. Some other studies that used the Nursing Stress Scale used a range from 1-4 as opposed to 0-3. This made comparison difficult with other studies. The wording of the open-ended question ‘Do you have any comments or suggestions to make in relation to stress and stressors in the clinical environment’ may have been leading. However, this wording was changed on the advice and conditions for access approval from one of the research ethics committee at one institution. Some written responses also included wording from the Nursing Stress Scale, thus suggesting that the scale may have led to some of the responses. There was no opportunity for the researcher to clarify the questions, nor was there an opportunity on behalf of the respondents to clarify their answers in closed-ended questionnaires such as The Nursing Stress Scale. Furthermore the dissonance between the results from this scale and the results from the open–ended question, particularly regarding relationships with other nurses exemplifies the weakness of the use of a single tool and therefore prompts further studies to incorporate mixed method and multi-instrument use. However, the return rate may be compromised if the questionnaires are too lengthy.
Subjectivity: Stress, being a subjective experience is by its nature difficult to assess. Innate variables within the participants regardless of the clinical setting may affect the outcome, for example an anxious student or nurse by nature or the responses may be affected by the workload on that particular day. Despite numerous stress scales, physiological tests and qualitative methods to explore assess and measure stress; the concept is always subject to conceptualisation by each individual. This has been explored by Haslam (2004). Studies on stress yield individualistic results making comparison between professions difficult. However, comparison of two groups in the same setting attempts to control this limitation somewhat. Non-verbal communication is lost in postal questionnaire, which may reveal pertinent information (Parahoo, 1997) and there is a tendency for superficiality in the responses (Polit et al, 2001). Threats such as mortality, testing and maturation would not occur owing to the once off study (LoBiondo-Wood and Haber, 2002).
6.1.2 Threats to External Validity
This refers to “the generalisability of the research findings in relation to other settings or samples” (Polit et al, 2001, pp194). The cross sectional, but open-ended ‘once only’ test may not be sufficient for credibility. The use of a number of instruments together could enhance validity of the responses and concomitantly test the validity and reliability of each tool therein. This was deliberately excluded as often-lengthy questionnaires lead to low return rate (Edwards et al, 2002). The open-ended question was included to address this limitation. It was anticipated at the outset to randomise the sample, however owing to the constraints relating to the Data Protection Act (Department of Justice and Law Reform, 2003); no names could be used to initiate the
smaller owing to the lack of personalising the information leaflet attached to the questionnaire.
Sampling was limited to a convenience sample. However, the small sampling frame left this opportunity difficult and also the gender could not be identified to ensure anonymity. It would be more representative though not feasible to randomly select. While this method obtains the maximum number of participants, this non-probability sampling technique could increase the risk of sampling bias, which would affect the representation of the findings (LoBiondo-Wood and Haber, 2002). Also, the fact that return of questionnaires were self selected the generalisability of findings could be compromised and increase the risk of bias affecting the results (Fain, 1999). Further evaluation in a larger sample size would be more beneficial with all branches of nursing included in the study. A longitudinal follow-up study of the cohort of student nurses when newly qualified would be interesting to compare findings with their previous results. Demographic data were not sought in the questionnaire because this was a requirement of the ethics committee due to concerns about anonymity. This was a limitation in this study as it meant that correlation studies to further explore some of the findings were not possible such as age, sex or amount of previous experience. Information was excluded relating to the size of hospital, type of ward, whether employed in a permanent temporary capacity or working with a nursing agency. Information associated with sociodemographic data could form the basis for addressing issues relating to stress and stressors in the clinical environment where patterns arose.