Tasas de jubilación e incentivos al retiro

Validity of an instrument determines the extent to which it actually reflects the construct under study (Burns and Grove, 2005) or the extent to which it “measures what it intended to measure” (Carmines and Zeller, 1979, p. 17). Three primary types

of validity generally discussed are content validity, predictive validity and construct validity. In relation to this study content validity and face validity of the research instrument were established. These processes will now be described.

Content Validity

According to Burns and Grove (2005) content validity is ascertained by the literature review together with the “representativeness” of the study population and validation by a panel of research experts. The content validity of the questionnaire package for the current study was determined by the literature review which identified the meaning of the concept “help seeking behaviour”, the influencing variables, their key dimensions and how they are measured (Figure 5.1). In addition, content validity of the questionnaire package was established through its review by a panel of experts (n=8) including three breast care nurse specialists, a specialist breast care physician, one general practitioner, two experienced nurse researchers and one woman who had previously experienced a self discovered breast symptoms. This panel was based on the recommendation of a minimum of five experts (Burns and Grove, 2005; Polit and Beck, 2008).

Content Validity Index

A content validity index (CVI) instrument was developed by the researcher to provide a numerical value to determine the content validity of the questionnaire package (Burns and Grove, 2005; Polit and Beck 2008). This instrument enabled the experts to rate the content relevance of each item using a four point rating scale reading: 1(‘not relevant’); 2 (‘unable to assess relevance/item in need of revision’) 3 (‘relevant but needs minor alteration’) and 4 (‘very relevant and succinct’), as suggested by Lynn 1986 cited in Burns and Grove, 2005 (p. 378). The standard method for calculating a

CVI at the item level (I-CVI) is based on the number of raters giving a rating of either 3 or 4 on the 4 point ‘relevant’ scale, divided by the total number of raters on the panel, with a recommended I-CVI of 0.78 or higher (Polit, Beck and Owen, 2007). A content validity index was also determined for each scale termed ‘S-CVI/Average’ (S- CVI/Ave). This involved calculating an average score across I-CVI’s for each individual scale (Polit and Beck, 2008). ‘Excellent’ content validity is indicated by ‘I- CVI’ of 0.78 or higher and ‘S-CVI’ of 0.90 or higher (Polit and Beck, 2006; Polit, Beck and Owen 2007). Content validity of the items and scales used in the questionnaire package are outlined below and detailed in Appendix 3a.

Results of I-CV1’s in the questionnaire package ranged from 0.63-1.00. The item concerning nationality scored 0.63 as three of the experts did not consider this issue to be relevant to the study. However, the researcher was confident that following review of the literature, piloting of the questionnaire package and in view of the current multicultural nature of Irish society, this question warranted inclusion in the questionnaire package.

Average S-CVI results ranged from 0.78-1.00. A score of 0.78 was computed for the ‘Alternative HSB’ 8 itemed scale (question 11) and a score of 0.85 for the 15 itemed ‘Breast Cancer Knowledge’ (question 20) scale. These consisted of both a series of questions with dichotomous (‘Yes/ NO’) responses. The ‘Health Service System Utilisation’ scale had an S/CVI-Ave of 0.87 which could have resulted from low I- CVI for two items (12 and 15), addressing issues of perceived prejudice, oftentimes not seen as a problem in Ireland. However, they were considered worthy of further study by the researcher, in relation to women’s HSB for self discovered breast symptoms. Finally the S-CVI’s for the three itemed duration subscale of the

‘Symptom Perception Questionnaire’(question 19) was computed at 0.80. The remaining of scales had S-CVI averages of 0.9-1.00, indicating good content validity overall (Appendix 3a).

Face Validity

Polit and Beck (2008) also refer to “face validity” being achieved when an instrument “looks as though it is measuring the appropriate construct” (p. 458). Face validity of the questionnaire package was determined by also asking the group of experts to comment on the overall relevance (Appendix 3b) of the questionnaire package to the aim and objectives of the study (Parahoo, 1997). Overall, responses indicated that the questionnaire package addressed the aim and objectives of the study. However, some minor changes were made to the questionnaire as suggested by members of the expert group in the open commentary included in the validity documentation given to them.

These included addition of a specific question (number 6) relating to having ‘health insurance’ or a ‘medical card’. The options of ‘husband’ or ‘partner’ and “general practitioner” were included in question number 12 concerning disclosure of the symptom to another person. Item number 11 (question 16) was edited to focus on possession of a medical card (as opposed to heatlh insurance), as it was suggested that this would impact more on access to health services. In addition, item number 26 (question 19) was edited from ‘benign’ breast problem to ‘a non-threatening/ less serious/ benign’ breast problem, in order to clarify the meaning of the medical term ‘benign’. These changes further enhanced the validity of the final questionnaire package (Appendix 1b).

Reliability of the Research Instrument

Reliability of an instrument denotes the consistency of measures obtained and indicates the extent of “random error” in a measurement (Burns and Grove, 2005). The power of a study in detecting significant differences and relationships that actually occur in the population under study is enhanced by the use of reliable instruments (Burns and Grove, 2005). Reliability is usually expressed as a “correlation coefficient” (Burns and Grove, 2005, p. 374), measured by the ‘coefficient alpha’ (Cronbach’s alpha) (Polit and Beck, 2008). A result of 1.00 indicates perfect reliability and 0.00 indicates no reliability. The lowest acceptable reliability coefficient for a well established psychosocial measurement is taken to be 0.80 while a coefficient of 0.70 is considered acceptable for a newly developed instrument (Burns and Grove, 2005). The reliability of the ordinal scales used in the study questionnaire will now be addressed in order of sequencing in the questionnaire package (Appendix 1b). These are further detailed in Table 5.1.

In the current study, low internal consistency results were obtained for some of the scales. However, according to Pallant (2007), Cronbach alpha values are sensitive to the number of items and with short scales (of less than ten items) it is not uncommon to find low Cronbach values e.g. 0.5. In such cases it is suggested that reporting the “mean inter-item correlation” for the items may be more appropriate with optimal scores ranging from 0.2-0.4 (Pallant, 2007). The number of items in most of the scales used in the current study ranged from three to seven and only one of the scales used had more than ten items (PAHS, eleven itemed scale) which may account for the low Cronbach results for some of the scales (Pallant, 2007). The mean inter-item

correlation (0.24 for the PEP scale and 0.17 for the cure/control scale) were within and close to the recommended 0.2-0.4 range.). However, since all of the scales used had previously reported coefficient values of 0.70 or over (Table 5.1), they were deemed acceptable for use in the current study. In addition, the adapted “Symptom Emotional Distress Scale” (Meechan et al., 2003) had a Cronbach’s alpha of 0.89. The adapted symptom duration and symptom consequences sub scales yielded Cronbach’s alpha results of 0.75 and 0.79 respectively, both of which are within the acceptable range for scales with less than ten items (Pallant, 2007) (Table 5.1).