Validity and reliability 61

De Vos et al. (2005:160) argue that before commencement of the main study the researcher must ensure ‘acceptable levels of reliability and validity’ of the measurement instruments and measurement procedures that are used in a study.

3.2.7.1 Reliability

Reliability is defined by Bostwick and Kyte (1981, cited in De Vos et al., 2005:163) as the ‘extent to which independent administration of the same instrument (or similar instruments) consistently yields the same (or similar) results under comparable conditions’. The data collection tool or measurement tool (questionnaire) needs to produce similar results, when the same variables are measured under the same conditions. In a similar setting, the questionnaire will test the same variables under study if the RNs are also responsible for the same medication administration tasks than they are responsible for as in this research setting. The questionnaire ensured that questions were asked to all participants in a consistent way.

Reliability is achieved in four ways, as described by De Vos et al (2005: 163), which includes conceptualizing constructs in the questionnaire; increasing the level of measurement; using multiple indicators of a variable; and using a pre-test or pilot study. In order to determine whether ‘constructs are clearly conceptualized’, the questionnaire was exposed to RNs (n=12) during the pilot study and questions that were not clear, were corrected and clarified. The questionnaire was also given to experts in research at the university and the statistician to determine whether the questions were clear. In order to increase the level of measurement, the research experts and statistician provided input as to whether the questionnaire measured what it is supposed to measure and to enable data analysis. Corrections were made as to the accurate numbering of the different variables that were tested. Multiple indicators of a variable were used to increase the reliability of each measured variable through the repeat of a question in a negative to see whether the participants provide the same information or through the repeat question in a different way. As discussed in detail in Chapter 4, the measurement of a variable

in different ways was adequately achieved. Therefore, increased reliability of the information gained, was achieved.

3.2.7.2 Validity

According to De Vos et al. (2005:160), validity has two important aspects to it, namely ‘that the instrument actually measures the concept in question, and that the concept is measured accurately’. Validity is categorised according to content, face, construct and criterion validity (De Vos et al., 2005:160). For the purpose of this study, content validity and face validity were applicable and are discussed as follows:

 In order to ensure content validity, items and concepts in the questionnaire were based on an extensive literature review, as well as on the objectives set for the study. The questions were designed specifically to test the perceptions of RNs regarding the factors that influence IV medication safety in the practice environment. The factors, as described in detail in the literature review and the research setting’s policies (policy related to medication administration and policy related to nursing responsibilities regarding the medication system, which is known in the literature as CPOE) were linked and tested as items in the questionnaire. Safe medication administration practice is defined in chapter 2 to be the complete and correct check of all the medication rights through ‘double-check of medication rights’ by 2 RNs. As per the researcher’s conceptual framework (paragraph 2.7), these factors are categorized as environmental, human, system and latent oriented, which was utilized from previous study outcomes to be the cause of medication errors. A statistician, experts in research methodology and nursing quality management, as well as registered nurses in the NICU, PICU and CSICU reviewed the questionnaire. The research methodology, data collection, results and interpretation of results were also reviewed by an expert in research methodology. In addition, a pre-test of the questionnaire enabled the researcher to determine whether there was any ambiguity or inaccuracies with the data collection instrument.

 Face validity is the least scientific definition of validity and relates to the superficial appearance (face value) of a measurement procedure or technique (De Vos et al., 2005: 161). In order to ensure face validity, experts in nursing quality management reviewed the questionnaire to ascertain that the instrument measured the variables it was supposed to measure. Experts in nursing quality management at the research hospital confirmed that the questionnaire was adequately testing RNs knowledge based on institutional policies. RNs who participated in the study had reviewed the questionnaire,

to determine whether it appeared relevant to those participants who would complete it, as described by De Vos et al. (2005:161). The researcher had also compiled the questionnaire, based on a thorough literature review in order to ensure that the data collection tool accurately measured the attributes under consideration, as described in De Vos et al. (2005:161).

During the review of literature phase of this study, the researcher compiled a list of all factors that influence IV medication safety and result in medication errors. These items were organized under the sections of medication error influencing factors, creating a set of items that were based on study outcomes. The questions were compiled specifically, based on these items to test the perceptions of RNs regarding the factors that influence IV medication safety in the practice environment. The factors were based on study outcomes from the literature review, the research setting’s policies (as discussed) and the researcher’s conceptual framework (paragraph 2.6). The described factors were categorized as environmental, human, system and latent oriented, which is utilized from previous study outcomes to be the cause of medication errors.