This study aimed to evaluate the validity and reliability of the ICSQ-H. The final results of the PCA suggested a five-construct model with 12 items. Afterward, the CFA confirmed a four-construct model with 10 items. The original structure of the ICSQ (23 items) and the five-construct model suggested by the PCA did not meet the goodness of fit model requirements when tested for CFA. However, the final four-construct model (10 items) showed a good model fit where all the fit indices passed the requirements.
Our χ2/df was less than three with an insignificant p-value, which indicates a good model fit. However, there are some limitations for χ2/df model use. The main limitation is having a small sample size where χ2/df lacks power and might not be able to distinguish between good fitting models and poor fitting models (30, 36). When having a large sample size, χ2/df model is exact, which is our case (36). Our results showed that GFI, CFI, and TLI values were above 0.95. Given the detrimental effect of the sample size on the GFI index, it is recommended to be used along with other indices that we took into account when conducting our study (30). For instance, CFI is one of the most used and recommended fit indices since it is among the measures least affected by sample size. Similarly, TLI is a fit index that is less affected by sample size. In this study, the values of both CFI and TLI indicated a good model fit (30). RMSEA has recently been suggested as one of the most informative fit indices since it is affected by the total count of the estimated parameters in the model. Until the early 1990s, a value between 0.05 and 1 was considered to reflect a fair model fit (30, 37); however, in the late 1990s, a value less than 0.06 was recommended (30, 38). Our model showed a much lower RMSEA, which indicates the goodness of fit of the model. Additionally, SRMR is recommended for use since it is easier to interpret than other fit indices because of its standardized nature. Values closer to zero show a better fit, which is the case of our model (30).
Convergent validity was met for the GLVS and PPE constructs, which indicates a satisfactory level of correlation of multiple items of the same construct (33). However, the AVE of the ABHR construct was slightly below 0.5, which could still be considered acceptable. The weak correlations between the four constructs proved the discriminant validity of each. This means that the measures of distinct constructs share a little common variance and support the uniqueness of the items and the construct (32). Furthermore, it indicates that the latent constructs used for measuring the causal relationships in our model are actually different from each other and do not measure the same thing that could lead to multicollinearity (33).
Concerning the interitem correlations and the corrected item-total correlations, they were acceptable for all constructs. Furthermore, the internal consistency of the ABHR construct was below 0.6; however, its interitem correlations and the corrected item-total correlations were acceptable. This could be due to the low number of items in this construct (two items) (39).
Finally, the removal of 15 items during the different stages of this study (two items during content validity assessment, 11 items during PCA, and two items during CFA) might considerably modify the original factor structure of the ICSQ, bearing in mind that they could hold valuable and important constructs in IPC. Nevertheless, these findings further suggest the existence of repetitions of similar items measuring similar factors that compromise the construct validity of the original ICSQ (40).
Few studies have been conducted to test the psychometric properties of some IPC questionnaires that are used to assess HCWs’ knowledge about IPC measures. For instance, Duarte Valim et al. (41) validated the Knowledge Questionnaire regarding Standard Precautions Measures (QCSP) for Brazilian nurses. Convergent validity was tested using known-group methods. Reliability was tested by calculating the intraclass correlation coefficient (ICC) by applying the test-retest method. The Kappa index was used for the purpose of agreement. The Portuguese QCSP showed satisfactory ICC and Kappa. However, validation by discriminant groups did not reveal a statistically significant difference between the two groups. Similarly, the infection control evaluation tool was developed by Wu et al. (2) to assess nursing students’ knowledge about standard and additional IPC precautions. The tool was a modified version derived from two previously developed tools including 15 questions. Content validity was assessed by six experts using the CVI, where an acceptable degree of validity was found, with 68% agreement. Kuder-Richardson 20 was used to test the internal consistency, which revealed a satisfactory value of 0.76. It is worth mentioning that this tool was based on two previously developed tools, mainly Chan et al. (13), who employed the concept of UPs in measuring knowledge. Another tool was developed by Chan et al. (42) in 2008 to examine nurses’ knowledge of SPs and transmission-based precautions using four multiple-choice questions. Content validity was assessed by two experts with a CVI =0.97. Construct validity was assessed using EFA. One factor was found to include four items with factor loadings ranging from 0.76 to 0.86. The scale reliability was assessed via test-retest. Cronbach’s alpha showed an acceptable value (0.79). Finally, we noticed that only one study assessed the construct validity of the scale using EFA (42), while neither study performed CFA, which suggests that further research is needed to test the construct validity of these scales using EFA and CFA.
Strengths and limitations
Our study is the first to test the psychometric properties of the ICSQ-H. Despite that the study was performed in the Southern Transdanubian region of Hungary, we included all hospital types (university, county, and city) from different counties, so we believe that our results could be generalized to reflect the situation across Hungary. However, our study has some limitations. First, using convenience sampling might have possibly introduced selection bias. Another limitation is that we could not compare our results to other existing models. Although the ICSQ was used in several countries to assess HCWs’ knowledge about IPC, its psychometric properties were not tested and reported in other languages. Thus, future studies are needed to test the psychometric properties of the ICSQ in other languages and settings. Finally, our data were collected during the COVID-19 pandemic, so we are uncertain if the awareness level of nurses was affected due to their high alertness during this period.
Relevance to practice
Given that Hungarian is the official language in Hungary, it was necessary to validate a Hungarian tool to facilitate a more comprehensive and precise measurement of knowledge about IPC among nurses in Hungary. Based on our findings, we believe that the ICSQ-H could pave the way for more research regarding nurses’ IPC knowledge to be conducted in Hungary. Nevertheless, its validation among other HCWs is important to tailor effective interventions to enhance knowledge and awareness.