Subjects and study design
This study is a cross-sectional descriptive-analytic study. It was carried out from January to June 2019. The general population in this study is Iranian women, and the target population consists of all housewives in 2019. Also, the accessible population is Iranian housewives who have children and are referred to Tabriz city health centers to perform their own or their children’s health activities [11]. The sampling was multi-cluster type that in the first stage, 5 centers were randomly selected from 20 health centers in Tabriz. In the nex stage, Samples were randomly selected from within these clusters relative to the population of each cluster (in a health center with more housewives, more samples were selected). The questionary was pen-and-paper. Inclusion criteria were: having children, willing to participate in research, having at least a high school degree, and being over 18 years old. Exclusion criteria were evident mental illness (all target populations were previously registered with the health center, and the research team was able to identify people with mental health) and unwillingness to participate in the study. The purpose of the study and the importance of the study for home safety was explained to the participants participating in the study. The sample size needed to perform factor analysis to determine construct validity varies among researchers. Some literature considers a minimum of 200 for the sample size [12,13,14,15]. Klein also argues that exploratory factor analysis requires 10 or 20 samples per variable, but the minimum of 200 for sample size is also defensible [15]. Thirty-nine items were identified for the initial questionnaire using a literature review. Therefore, ten samples per tool item were selected (390 samples). The final sample size was 645, considering a 10% samples loss and a design effect coefficient of 1.5.
Instrument
In the first step, terms or concepts equivalent to home safety was extracted through a literature review. Electronic databases, including Pro-Quest, Scopus, Google Scholar, Science direct, and SID and Magiran’s Persian-language databases from 2000 to 2019, searched for related studies. Studies in Persian and English were selected. The search string comprised of (home OR house) AND (checklist OR questionnaire OR instrument) AND (safety). Two experts collected the concepts from the literature. In total, 45 terms and expressions of the questionnaire were collected. Then, these concepts were merged and finalized by a third person. After reviewing, removing, and merging common phrases, 39 phrases remained. Finally, these items were turned into questions by the research team. All statements were designed on a 5-point Likert scale (strongly agree = 1, agree = 2, no idea = 3, disagree = 4, completely disagree = 5).
Questionnaire validity
To determine the instrument’s validity, face validity, content validity, and construct validity of the made instrument were evaluated [16,17,18]. The designed questionnaire was presented to a panel of experts, including 9 Occupational Safety and Health PhD experts (8 of these experts were with a PhD and one of them was a PhD student, but all of them had at least 2 years of experience in this filed).
Face validity
Qualitative and quantitative methods were used to perform face validity. To determine qualitative face validity, ten experts were asked to comment on the level of difficulty, relevancy, and ambiguity of items in writing. Then, if the research team approved these comments, we would apply them to the item. For the quantitative approach, nine experts were asked to rate the importance of each item on a 5-point Likert scale: Completely important, important, moderately important, slightly important and not important. Then impact scores were calculated using formula 1, and items with scores less than 1.5 were excluded [19, 20]. Since all obtained impact scores were greater than 1.5 (mean score: 4.35), all items were retained for subsequent analysis.
$$\mathrm{Impact}\kern0.5em \mathrm{Score}\kern0.5em =\kern0.5em \mathrm{Frequency}\left(\%\right)\kern0.5em \times \kern0.5em \mathrm{Important}$$
(1)
In formula 1, frequency refers to the percentage of people who gave the item a score of 4 and 5, and importance is the average total score of people on importance based on the Likert scale.
Content validity
To determine content validity, both qualitative and quantitative methods were used based on the assessment of experts. The content validity ratio (formula 2) and content validity index (formula 3) were measured to assess quantitative content validity. In order to determine the content validity ratio, nine experts (different from prior step experts) were asked to examine each phrase on a three-part scale (necessary, useful but not necessary, not necessary). Due to the number of experts and considering the Lowsheh table, items whose content validity ratio was equal to or greater than 0.78 were retained [21, 22].
$$\mathrm{CVR}=\frac{\mathrm n{\displaystyle\frac{\mathrm N}2}}{\displaystyle\frac{\mathrm N}2}$$
(2)
In formula 2, “n” represents the number of experts who have chosen the necessary option and “N” represents the total number of experts. According to the content validity ratio decision table, if the panel of experts is nine people, the minimum acceptable validity for each item will be 0.78, which means that the relevant item with a significant level of reliability (P < 0.050) is essential in this tool [21, 23]. In the qualitative study of the questionnaire’s content, ten experts were asked to comment on issues such as Persian grammar, use of the right words, correct placement of items, and scoring in writing form. Then this comment was considered by the research team. In this regard, the questionnaire’s four items and rating system were corrected, and CVI and CVR were recalculated.
Waltz and Bausell method was used for Content validity [24]. For this purpose, a designed questionnaire was provided to the experts. They were asked to determine the relevance of each item in the questionnaire based on Waltz and Bausell content validity index. Thus, the degree of relevance, simplicity, and clarity was separately assessed for each item by experts using the 4-point Likert Scale. The content validity index was determined by formula 3. According to the Waltz and Bausell method, items with a score higher than 0.79 are appropriate, between 0.70–0.79 need revision, and less than 0.70 are unacceptable.
$$CVI\kern0.5em =\kern0.5em \frac{Number\kern0.5em of\kern0.5em reporters\kern0.5em who\kern0.5em have\kern0.5em selected\kern0.5em options\kern0.5em 3\kern0.5em and\kern0.5em 4}{Total\kern0.5em numbers\kern0.5em of\kern0.5em experts}$$
(3)
At this point, none of the items had scores below 0.7. Also, after revising, the CVI of one of the items ranged from 0.70 to 0.79. Consequently, the content validity index of the item was re-evaluated for the second time by experts. The mean content validity index of the questionnaire was 0.94 and is considered appropriate based on the opinion of Polit and Beck [25].
Construct validity
The two-stage strategy of Muliak and Millsap model was used to determine the construct validity [26]. For doing Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) a split-half method was used. In the first step, exploratory factor analysis was used to extract factors (latent variables). Kaiser-Meyer-Olkin (KMO) was performed to check the adequacy of the samples. The Bartlett Test of Sphericity was used in the sample to ensure that the correlation matrix underlying the factor analysis is not zero. Values above 0.7 in the KMO test and p-value less than 0.050 in Bartlett’s test were considered as the criterion of suitability for factor analysis [27]. The principal component method with direct oblimin was used for data exploratory factor analysis. To determine the number of main factors of the questionnaire, three indices, including eigenvalue, fine-grained diagram, and contribution of each factor to the sum of the total variance, were used. The milestone of 0.3 was considered the minimum factor load needed to maintain each expression in the factors extracted from factor analysis [28]. After extracting the factors and expressions in each factor, the degree of consistency of these factors with the main concept and dimensions was investigated. As a result, two sentences were omitted, and the instrument’s expression count reached 37. In the second step, confirmatory factor analysis was used to evaluate the relationships between the indicator and latent variables to validate the EFA model on a sample separately from the exploratory step. Confirmatory factor analysis shows whether tool items are appropriated and fitted to relevant factors based on theoretical expectations. The estimated method was maximum likelihood [26].
Confirmatory factor analysis
Structural equation modelling with confirmatory factor analysis was used to test the relationships between variables and the instrument’s psychometric properties. Model fit was evaluated using the chi-square statistic (χ2), chi-square ratio, and degrees of freedom (χ2 / df). Goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), mean square root approximate Error (RMSEA) and CFI > 0.9, χ2 / d < 5, GFI > 0.9, and AGFI≥0.8, RMSEA < 0.08 are considered as appropriate indices and reasonable values [26].
Questionnaire reliability
The reliability of the questionnaire means to what extent the questionnaire yields the same results under the same conditions [29]. The reliability of the study questionnaire was calculated using Cronbach’s alpha coefficient and test-retest. Therefore, 20 participants were asked to complete the questionnaire in two steps with 2 weeks intervals [30]. Cronbach’s alpha coefficient was used to investigate the internal consistency of the questionnaire. An alpha coefficient greater than or equal to 0.70 was considered as a satisfactory criterion [31]. The reliability of stability and test-retest of the questionnaire were also assessed by repeated sampling and by calculating the intra-class correlation coefficient (ICC) [32]. The whole process of designing and developing the questionnaire is shown in Fig. 1.