A literature review was performed to review available published articles, documents, and reports. First, the research question and search strategy were developed. Then, to determine the nature and format of the main social vulnerability indicators in the COVID-19 pandemic, available evidence, including articles, documents, and reports, was searched and reviewed. This literature review was performed in partnership with a research team and medical information specialists through December 2020. The Web of Science, PubMed, and Google Scholar were searched for relevant articles. For keywords, the combination of the following Medical Subject Headings (MeSH) in consultation with our research team and experts, including an experienced medical information specialist, was used: “COVID-19,” “social vulnerability”, “management”, “indicator”, “epidemic”, “pandemic”, “outbreak”, and “infectious disease”. In addition, we searched specialized databases and websites such as the World Health Organization (WHO), the Federal Emergency Management Agency (FEMA), and the Michigan Department of Health and Human Services (MDHHS) with the mentioned keywords. Then, references and bibliographies of all relevant articles were reviewed to identify additional studies. All found studies and literature were entered into the EndNote software version X7.

Articles were selected by applying the inclusion and exclusion criteria. The inclusion criteria were as follows: (1) addressing the social vulnerability indicators and infectious diseases and (2) the availability of the full text of articles for free. The exclusion criteria were as follows: (1) articles published in non-English languages; (2) studies that solely focused on concepts or theoretical frameworks and did not provide metrics or measurement indicators; and (3) news articles, abstracts, and those studies for which full texts were not available.

Duplicate articles were eliminated via the EndNote software. Articles were screened by assessing the titles and abstracts for eligibility by two reviewers independently. The full text of the retrieved articles was reviewed by the research team. After screening studies based on the inclusion criteria, the final set of studies was summarized using a data extraction form. Data were charted and presented according to the research objective. The process of doing this phase with more details had been published earlier by one of the members of the research team in the present study (24).

Expert panels in two rounds were planned from 15 January 2020 to 20 February 2021 to discuss the findings and examine the feasibility of the recommended indicators in Iran. Due to the conditions of COVID-19, a face-to-face panel was not possible, and telephone, mobile social media applications (Skype), and email were used to provide the experts' opinions. For this purpose, academics, policymakers, infectious disease specialists, and social and welfare experts in Iran were invited to develop discussions about the indicators of social vulnerability during the COVID-19 pandemic based on the results of the literature review.

A list of participants was prepared. After obtaining their permission and approval for the dates and time of the panel, an official invitation letter along with a research brief was sent to them by email or Skype. However, before participation, they were informed of the study's objective and considered a topic in the expert panel session. A discussion guide, due to the literature review (social vulnerability indicators in the COVID-19 pandemic), was developed to assess the feasibility of using the retrieved indicators in the data accessibility and context of Iran. Two members of the research team served as facilitators, presenting the questions most appropriate according to the participants' backgrounds and in relation to the study objective. Additionally, the participation was voluntary, and the participants could withdraw from the study without any consequence. The facilitators used an informed consent form for the participants and ensured that they felt free to answer any question, had the possibility of ending their participation in this study, and had the right to prevent or stop the recording. In total, eight experts participated in the study in addition to the research team. Two meeting sessions were coordinated via Skype for each round of the expert panel. Additionally, the comments of two experts were collected via email in each round of the expert panel. The Skype sessions were recorded.

Finally, the consensus between the opinions of the experts was reached by the facilitator of the meeting in such a way that at least 70% of the members agreed with including the indicators in the study (25). The indicators of social vulnerability in the COVID-19 pandemic in Iran that received less agreement than 70% were eliminated.

After finalizing the main social vulnerability indicators in the COVID-19 pandemic in Iran, a questionnaire was designed based on a pairwise comparison of all indicators, and finally, we estimated relative weights for selected indicators using the AHP technique.

The university professors who are specialists in the fields of disaster and emergency health, infectious disease, and social and welfare were asked to prioritize the importance of each of these indicators. To assess the reliability of the results, the consistency ratio (CR) was evaluated (26). In this stage of the study, 12 experts participated and estimated relative weights. The participants had to choose an indicator in the pairwise comparison questionnaire and identify the degree of more or less importance of each selected indicator. To help with the comparison, a nine-point scale of importance was created (27). The suggested numbers were used to express the degree of importance between each of the two indicators based on the nine-point scale as shown in Table 1. Then, the researchers calculated the weights for each indicator.

The most repeated domains of social vulnerability to COVID-19 in studies were family composition and disability, public health infrastructures, racial and ethnic minorities, socioeconomic status, housing status, and transportation. Additionally, the relevant indicators for the mentioned domains, such as the percentage of single-parent households, the immigrant population, population density, unemployed rate, percentage of healthcare service workers, and percentage of the population who had access to health insurance, were extracted from the literature review. Then, they were adjusted as a draft of social vulnerability, including five domains and 38 indicators for the next stage.

The expert panel was conducted in two rounds subsequently. For each round, two sessions were organized via Skype. During these two rounds, the experts discussed and reached a consensus agreement on domains and indicators that would be useful as a basis for developing a tool for assessing social vulnerability in COVID-19 in Iran. Additionally, the collected comments of two experts via email were involved in the including or excluding of domains and indicators in two rounds of an expert panel.

In the first round of the expert panel, the number of extracted domains from the literature review increased from five to six, according to the consensus of experts. In the draft version of domains from the literature review, education was accounted for as one of the indicators of the “household composition and disability” domain. The experts reached an agreement to categorize education as an individual domain of social vulnerability in COVID-19. Moreover, the five indicators that received <70% agreement were eliminated. The eliminated indicators in the first round were “no high school diploma” and “percentage of households with heads aged 60 years or older” from the socioeconomic domain, “income” and “percentage of children under poverty” from the socioeconomic status domain, and “population does not have a safe home to shelter in place” from the housing and transportation domain. Then, the following two indicators were defined for the education domain: “no high school diploma” and “percentage of the population with university education.” In total, there were 35 indicators at the end of the first round of the expert panel.

The second round of the expert panel was held by the same participants as the previous round. The domains and indicators were discussed again, and the number of domains remained unchanged but the number of indicators decreased from 35 to 31. The eliminated indicators included “percentage of households with children aging from 0 to 4 years” from the household composition and disability domain, “percentage of children with no access to the internet” from the socioeconomic status domain, “working population spend more than 60 min on commute” and “percentage of pedestrian and bike commuters” from the housing and transportation domain, and “hospital bed per capita” from the public health infrastructure domain. Additionally, the indicator of “place of living (urban or rural areas)” was added to the housing and transportation domain. Consequently, the final version of domains and indicators of social vulnerability in COVID-19 included 6 and 31 items, respectively (Table 2).

The indicators were assessed due to the availability of data during the COVID-19 pandemic in Iran. Finally, 10 main indicators were selected. By applying the AHP method during the collective discussion, a group of twelve experts defined the importance of each indicator of social vulnerability in the COVID-19 pandemic. For determining the importance, the experts completed the pairwise comparison matrix by comparing all indicators in pairs.

The results of the pairwise comparison matrix and relative importance vector are presented in Table 3. According to Figure 1, the most important indicator is population density (ωC1 = 0.229); the second-most important indicator is access to healthcare facilities and relevant services (ωC10 = 0.208); vulnerable group is also the third-most important indicator (ωC9 = 0.206). Further, gender is the least important indicator (ωC2 = 0.014). The sum of the relative importance vectors is equal to 1. The consistency of the pairwise comparison was checked through face-to-face interviews with experts. Next, the consistency ratio was calculated to check the consistency of the judgments. The consistency index was (CI = 0.129), the consistency ratio was (CR = 0.087), which was lower than 0.1. This demonstrates that the pairwise comparison matrix is consistent.