2.1 Study Design

During the third wave of COVID-19 in Iran and the onset of vaccination against COVID-19 with globally approved vaccines, we conducted a cross-sectional survey using various data collection methods, including in-person and online questionnaires, from 1 to 20 February 2021.

2.2 Study Instrument

The survey instrument was an updated self-administered questionnaire (Online Resource 1) that was based on a previous study for eliciting public preference for a hypothetical COVID-19 vaccine [13]. The questionnaire had three sections: socioeconomic and demographic information of respondents and their households; the health status of respondents; COVID-19 risk perception and personal experience with the virus and stated preference for COVID-19 vaccines. However, since the previous study was conducted before developing and approving COVID-19 vaccines, we improved and updated the vaccine preference section by considering recently approved vaccines worldwide and a scenario for a domestically produced vaccine in future months.

To do this, a sequential questioning approach was taken, i.e., a three-layered question was asked to understand respondents' preference for contributing to vaccine financing as follows.

Initially, respondents were asked if they would pay for vaccines if they were not available for free, with different response options ranging from yes at any price to an undecided or definite refusal (Q3: Layer 1).

Respondents showing a positive WTP for vaccination were provided details about the approved vaccines, including vaccine names, originating country (Russia, China, US, and forthcoming Iranian vaccines), needed dose, total cost in US dollars, and the international and local agencies that provided the vaccine approval. They were then asked about their preferences regarding financial contributions, choosing among three alternatives that spanned from full or partial government financial contribution to a complete user-fee payment option (Q3: Layer 2)

For respondents selecting government contribution options, a follow-up question examined their preference based on a payment scale concerning the proportion of the vaccine cost that the government should bear, providing choices of 25%, 50%, and 75% (Q3: Layer 3)

Furthermore, to have a better picture of Iranians’ WTP for the domestic vaccine, for respondents who had stated a preference for partial contribution in vaccine financing (25%, 50%, or 75%), the weighted mean of their willingness to contribute for two doses was calculated in three price scenarios based on the lowest, average, and highest vaccine prices that were presented to the respondents for the Chinese, Russian, and FDA-approved vaccines in the survey questionaries, accordingly. It should be mentioned that this approach was employed because during the study period, the Iranian vaccine was still in production and did not have an established price.

Considering the highlighted changes, before the launch of the survey the face validity of this questionnaire was assessed by experts through a focus group consisting of two pharmacoeconomists, a community-medicine specialist, an epidemiologist, and three health/pharmaceutical policymakers. The questionnaire was revised in line with these comments. Lastly, the clarity, comprehensibility, and appropriateness of the final questionnaire were confirmed and a pilot study was then administered to 50 respondents to check the reliability of the questionnaire (Cronbach’s alpha = 0.86).

2.3 Sample

The target population was adults aged 18 years and above. For the face-to-face survey that was conducted in Shiraz, the capital of the Fars province (south of Iran), a random sampling method was employed considering location, age and gender proportion. Hence, Shiraz was divided into five geographical clusters, including North, South, East, West, and the Center. To define the proportion sample size of each geographical zone, the pedestrian traffic at the three main streets of each zone during rush hour was measured.

There was no sampling framework for the online survey, however at the beginning of the survey, a focal point (typically a health economist academic colleague from a collaborating research center at the university in each province) was selected as the start point to disseminate the questionnaire link. Hence, the questionnaire was distributed randomly through a snowball method. Participants could anonymously respond or forward the link. It should be mentioned an invitation letter and a written consent form that included information about research purposes and ethical issues were provided. The final sample consisted of 2071 respondents, after excluding incomplete and invalid responses. As shown in Table 1, the face-to-face survey constituted 60% of the overall sample, while the remaining 40% of the population sample was obtained through an online survey. Details regarding the distribution of the online survey sample are presented in Table A1 (Online Resource 2).

Table 1 Sample size and distribution2.4 Statistical Analysis

We modeled determinants of public preference for financing Iranian, Chinese, Russian, and FDA-approved vaccines using ordered probit models; an ordered-probit analysis requires specifying a latent variable. In our model, the latent variable is a continuous variable representing people's willingness to contribute to financing available vaccines. The observed variable is five categories of people's willingness to finance vaccines from 0 to 100% of the price in 25% increments. Five categories resulted from participant responses to a three-part question: first, WTP for vaccination; second, preference for financing options, including full or partial government contribution, or user fee payment; and third, for those choosing government contribution, preference for the proportion of vaccine cost, with options of 25%, 50%, or 75%. Individuals without an interest in financing the vaccine were included in the first category (0); individuals who were interested in vaccination, even if they had to pay the full cost, were in the last category (100%); and for the remaining individuals, preference for government contribution was asked on a 3-point scale, i.e. 25%, 50%, and 75% of the vaccine price, which indicated to us those who were interested in paying 75%, 50%, or 25% of the price, respectively. Public preference for financing different vaccines is determined by Eq. 1.

$$^=_+_Hj+_Hs+_D+__+__+__+__+_I+_L+\gamma DEI+\delta _+u$$

(1)

where \(u\) is normally distributed with mean zero and variance normalized to unity. Equation 1 can be written in compact form as \(^=\beta X+u\). The regressors listed in Eq. 1 are as follows.

Hj is a dummy variable that is equal to 1 if the respondent is working in the health sector, or a value equal to 0 otherwise.

Hs gives the self-reported health status of participants in ascending order.

D is a 5-level Likert item indicating the perceived risk of COVID-19 in ascending order.

\(_\) is a 5-level Likert item indicating participants' perceived exposure to COVID-19 in ascending order. Similarly, \(_\) measures participants' perceived exposure to family members.

\(_\) is a dummy variable that takes a value of 1 if the respondent has non-communicable diseases, and a value of 0 otherwise. Similarly, \(_\) measures whether family members have non-communicable diseases.

I is a 4-level Likert item indicating the participant's history of infection with COVID-19 in ascending order from no such record to be hospitalized.

L is a dummy variable that takes a value of 1 if the respondent has lost a family member due to COVID-19, and a value of 0 otherwise.

The vector \(_\) contains a number of subject-specific control variables as follows.

Income is the participants' income with five categories, starting at none and moving to < 30 million Rials, > 30 and < 70 million Rials, > 70 and < 100 million Rials, and > 100 million Rials.

Age gives the self-reported age of participants.

Gender is a dummy variable and takes a value of 0 for males and 1 for females.

Education gives self-reported education levels in ascending order.

Family size is the size of the respondents' families.

Head is a dummy variable that takes a value of 1 for family heads and a value of 0 otherwise.

City is a dummy variable that takes a value of 1 if the participant is living in a city and a value of 0 otherwise.

Insurance is a dummy variable that takes a value of 1 if the participant has basic insurance and a value of 0 otherwise.

Insurance supplement is a dummy variable that takes a value of 1 if the participant has supplementary insurance and a value of 0 otherwise.

\(D_\) is the interaction term between \(D\) and \(_.\)

Introducing constants \(_\),\(_\), \(_\), and \(_\) (to be determined in the ordered probit regression analysis), the ordered dependent variable is related to the latent variable in Eq. 1, as shown in Eqs. 2–6 below.

Using our assumptions on the error term and the expression in Eq. 1, denoting probability by P and the cumulative distribution function of the normal distribution by \(\phi \), we have (Eqs. 7–11):

$$P\left(f=0\right)=P\left(^<_\right)=P\left(u<_-\beta X\right)=\phi (_-\beta X)$$

(7)

$$P\left(f=1\right)=P\left(_<f}^<_\right)=\phi \left(_-\beta X\right)-\phi (_-\beta X)$$

(8)

$$P\left(f=2\right)=P\left(_<f}^<_\right)=\phi \left(_-\beta X\right)-\phi (_-\beta X)$$

(9)

$$P\left(f=3\right)=P\left(_<f}^<_\right)=\phi \left(_-\beta X\right)-\phi (_-\beta X)$$

(10)

$$P\left(f=4\right)=P\left(_<f}^\right)=1-\phi (_-\beta X)$$

(11)

Thus, we analyzed how different factors influence people's likelihood of choosing to cover 0%, 25%, 50%, 75%, or 100% of the vaccine price. The ordered probit regression coefficient is the marginal index effect on the average value of the latent variable conditional on the data, E[f^* |.]. The marginal probability effects describe the change in probability of being in a particular category, P(f = k),k = 0,1,2,3,4, which arises from a unit change in an explanatory variable.