This study involved two basic steps of designing and psychometrically assessing the questionnaire of factors influencing flood death according to systematic review and qualitative studies. These studies were conducted prior to the study design and psychometric evaluation of the questionnaire employed to identify the factors affecting flood deaths [2,7].
The literature used for this study had already been collected during a systematic review in which most relevant literature published from 1990 to 2017 on factors affecting flood deaths were extracted and analyzed . International databases of PubMed, Scopus, Web of Science, and Google Scholar and Iranian electronic databases IranMedex, Irandoc, Magiran, and Scientific Information Database (SID) were searched for articles related to flood deaths. we used appropriate search strategy and key terms including Flood, factor, characteristic, vulnerable, cause, element, agent, variable, mortality, fatal, death, loss of life, kill, lethal, dead and die. These key words were searched in different combinations using either OR/AND. The studies were selected regardless of their methods but based on their inclusion and exclusion criteria. The data were extracted, coded and analyzed. A descriptive and thematic analysis was performed. A total of 114 risk factors were identified in systematic review and were classified into 5 categories of hazard characteristics, individual, environmental, socio-economic, and managerial or capacities (Figure 1).
In the qualitative study using the content analysis method via a conventional approach, the underlying factors influencing flood deaths in different groups were identified including: ordinary flood-affected people, academic groups, and disaster management authorities in different parts of Iran who have experienced flood death . In this study, 7 categories of factors were identified including: hazard-related features, cultural, economic, social, demographic, management, and physical factors. These seven categories of factors had 27 sub-categories and 167 factors (Figure 1).
Consequently, to design the questionnaire, based on systematic and qualitative studies, a complete set of objective and subjective factors affecting flood deaths was collected and categorized. After editing and finalizing items that measured objective factors, the items of subjective factors were assessed and their validity and reliability were assessed. Finally, a reliable questionnaire was developed to measure these factors.
Then, the design and psychometric evaluation of the questionnaire was performed as follows:
Designing a Questionnaire to Measure the Causes of Flood Death
The design of the questionnaire consisted of three stages (Figure 1):
Item Organization (Generation, Reduction, and Classification): At this stage, first the pools of the items were formed using the items extracted in the systematic review and qualitative study. Next, after investigating each item by the research team, duplicate items were eliminated or one or more items with the same meaning were merged, with the number of items in the pool decreasing while non-repetitive items were used. Then, the items that were measuring the same scope or concept were merged together. Eventually, the sub-domains and domains were formed with each measuring a particular sub-domains and domain. Indeed, the items were classified in such a way that the related items could fit into their appropriate sub-domains and domain.
The production and finalization of questions: At this stage, the questions related to each item were designed based on the items produced and categorized in the previous step. Then, the duplicate questions were removed again and similar questions merged. Once the questions were merged, each question was finalized in terms of the writing structure, simplicity, and comprehensibility.
Categorization of questions: Once the questions were finalized, the questions were divided into two general categories of objective and subjective questions where each category was subdivided into other categories: Objective Questions: Once the objective questions were made, they were reviewed, edited, and categorized several times by the research team. Objective questions were edited and finalized three times by the research team. Subjective Questions: Once the subjective questions were made, they were reviewed, edited, and categorized by the research team. Finally, subjective questions were reviewed and edited five times by the research team.
Psychometric Assessment of the Subjective Factors Affecting Flood Death Questionnaire (SFAFDQ):
Psychometric validation of subjective factors of the questionnaire was essential and performed whereas there was no need for psychometric validation of objective factors. Subjective factors were categorized as 7 categories of factors and 55 questions. The psychometric evaluation of Subjective Factors Affecting Flood Death Questionnaire (SFAFDQ) included two main stages of validity and reliability (Figure 2):
Validity Assessment of SFAFDQ: Validity of the questionnaire consisted of three stages:
Assessment of Face Validity: Face validity is the simplest and weakest form of validation which shows the appropriateness of the questionnaire for the purpose of the study and the content of the domain . The face validity of the questionnaire was measured qualitatively and quantitatively. Face-to-face interviews are conducted with some of the target group to assess the face validity qualitatively . Accordingly, qualitative face-to-face interviews were conducted with 10 people affected by floods (a community with flood deaths) in one of Iran's flood-hit provinces on April 14, 2017. At this point, the participants were asked to give a comment about the apparent suitability of the items, readability, and difficulty level of items, the clarity of the wording, ambiguity and misinterpretations of items, the layout, style, acceptability and feeling comfortable with the items. In order to quantify the face validity quantitatively, the items’ impact scores were measured. In this step, 10 flood crisis and disaster managers and those with academic background in health and disasters who had sufficient experience or knowledge of flood were employed. The 10 participants were asked to rate the importance of each item on a 5-point Likert scale from 1 (not at all important) to 5 (very important). Impact scores of the items were measured using the following formula: The percentage of participants who gave each item scores as 4 or 5 times mean of importance for each item. For the formal acceptance of each item, an impact score ≥1.5 was considered .
Assessment of Content Validity: Assessment of the content validity of the tool was qualitatively and quantitatively assisted by 15 disaster management professionals with sufficient knowledge and experience of flood death or at least knowledge and experience of flood disaster. A total of 46% of the participants in this phase had a master's or doctoral degree while the rest had at least a bachelor's degree. Also, more than 72% of the participants in this phase had more than 10 years of experience in disaster management. In the qualitative method, the instrument was emailed to the participants to express their views on grammar, use of appropriate words, item placement and their ordering from simple to complex, categorizing and matching the items with relevant content domains, adequacy of items for relevant dimensions and appropriate scoring. The questionnaire was edited based on participants' comments.
Content Validity Index and Content Validity Ratio (CVI and CVR) were used to examine the content validity of the instrument. In order to measure the content validity ratio (CVR), experts' opinions were used, explaining the objectives of the questionnaire and asking them to rate each item on a 3-point range of " 1: It is not necessary ", "2: useful but not necessary" and "3: essential" to classify. The questions were removed from the questionnaire. The CVR index was then calculated using the following formula, followed by the Lawshe table, which scored less than 0.49 .
nE: The number of professionals choosing the "essential" option.
N: Total number of specialists
In this research, four criteria of simplicity, specificity, clarity, and relevance were used to calculate the content validity index. These criteria were individually rated by 15 experts on a four-part Likert scale (1: Irrelevant, 2: Somehow relevant, 3: Acceptably relevant, and 4: Completely relevant). CVI score was calculated by summing the percentages of agree scores for each item scored as 3 and 4 (highest score) according to the following formula :
Questions with a score above 0.79 have a good content validity index. Thus, the questions with a content validity index (CVI) score of less than 0.79 were removed from the questionnaire.
Assessment of construct validity: Based on the observations, data, and results of systematic and qualitative review studies, there were predetermined hypotheses regarding the structure and relationships between questions and categories of questions. At this stage, the confirmatory factor analysis was used to investigate these assumptions. In other words, confirmatory factor analysis was performed to evaluate the validity of questionnaire constructs. In order to conduct this analysis and check construct validity, a case‑control study was performed in 2018. The participants included 369 people in a flood affected community in different parts of Iran, who experienced flood death. The sample size required for analytical factor analysis varies from three to ten samples per item [46,47]. In this study, 10 samples were considered for each item. Since the construct validity was measured for the items of subjective factor assessment and the number of items in this section was 33 items, the sample size was estimated as 330 subjects. The research team completed a questionnaire on 369 people in the flood affected community who experienced flood death through attending the doorstep of the participants. Then, the collected data were introduced into SPSS software and confirmatory factor analysis was performed by separating the structures and the entire structure in the last step of the instrument psychometric evaluation. This step was performed after reviewing other validity and reliability steps.
Reliability Assessment of the SFAFDQ
reliability assessment of the questionnaire consisted of two stages:
Internal consistency test: this test was conducted through participation of 15 persons in the flood affected communities in one of the provinces of Iran (in Chenar village and Azarshahr city in East Azerbaijan) in 2017 flood. To measure internal consistency, Cronbach's alpha coefficient was calculated for the total questionnaire and its constructs. Based on the direct relationship between the correlation coefficient and the reliability of the instrument [45,47], the Cronbach's alpha coefficient above 2 was considered as the desired internal consistency [48,49].
Stability evaluation: For stability evolution via test-retest method, SFAFDQ was completed twice with a two-week interval in one of the floods affected communities in Iran. Then, the Pearson Correlation Coefficient (PCC) was measured. PCC of 0.5 or greater indicated sufficient stability . This coefficient was calculated separately for all questions.
Additional Statistical Analyses: in the construct validity stage, before performing confirmatory analysis, first, Kaiser-Meyer-Olkin (KMO) coefficient and Bartlett's Test of Sphericity were measured. These two tests were performed to determine the adequacy of the sampling for conducting the factor analysis. The KMO index is especially recommended when the ratio between items to variables is less than 5: 1. The value of this index is within 0 and 1, while if it is above 0.5 it is suitable for factor analysis. Also, to check the suitability for factor analysis, it is necessary to have a significant Bartlett test (p<.05) . Also, in order to identify the questions that played the most important role in each structure, the initial factor loadings of the questions and the factor loadings of the questions after being placed in the structures were measured.