Preference of influenza vaccination among the elderly population in Shaanxi province, China

ABSTRACT

The influenza vaccination uptake rate is low in China. In the current study, we aimed to explore the preferences of influenza vaccination among the Chinese elderly. A discrete choice experiment design was performed to assess their preferences toward five influenza vaccination attributes, including vaccine effectiveness, duration of protection, severe adverse effects, vaccination cost, and vaccination recommendation source. A total of 144 participants aged over 60 years from three cities of Shaanxi province in China were surveyed. A mixed logit model was employed to analyze the data. The elderly population were found to prefer influenza vaccination with a longer duration of protection, followed by lower severe adverse effects, higher vaccine effectiveness, and recommended by healthcare providers. The vaccination cost was the least important attribute. The most considerable marginal willingness to pay for vaccination (CNY220.90) and the highest vaccination choice probability (83.70%) occurred when the duration of protection extended from 3 to 12 months. The present study’s findings would inform decision-makers on implementing appropriate interventions for the increase of influenza vaccination coverage among the elderly in China.

KEYWORDS:

• Discrete choice experiment
• influenza vaccination
• elderly
• preference
• willingness to pay

1. Introduction

Influenza pandemic was listed as one of the ten threats to global health by the World Health Organization in 2019.¹ It was estimated that 291,000–646,000 deaths attribute to seasonal influenza annually.² The influenza-associated mortality was higher among the elderly (2.9–223.5/100,000) due to their weak immune system coupled with chronic diseases compared with other age groups.^2–4

Influenza vaccination is the most effective way to prevent the risk of influenza and its serious complications. The elderly aged over 60 years were recommended to receive influenza vaccination with priority by the Technical Guidelines for Seasonal Influenza Vaccination in China.⁵ Nevertheless, the uptake rate of the influenza vaccine was remarkably lower in China (4%) compared with the United States (59.6%) and Europe (41.8%).^6–8

China is facing the problem of aging as 18.1% of its total population is aged 60 or above (253.88 million) as reported in 2019.⁹ The overall coverage of influenza vaccination among the elderly was 3.8%.¹⁰ The Healthy China Initiative (2019–2030) intends to promote influenza vaccination coverage to a larger extent in the elderly population.¹¹ A few regions like Beijing provided free influenza vaccination for the elderly aged over 60 years, which increased its coverage tremendously (38.7%).¹² Nonetheless, the influenza vaccination policy is different in Shaanxi province, where the elderly need to pay out-of-pocket for the vaccination. Shaanxi province is located in Northwest China, and its economic condition belongs to the middle grade in China.¹³ According to the Shaanxi Statistical Yearbook 2019, there were 6.75 million (17.5%) people aged over 60 years living in Shaanxi province.¹⁴ To optimize interventions for promoting the uptake of influenza vaccination in Shaanxi province, it is of significant interest to determine the vaccination preferences among the elderly that could affect influenza vaccines’ acceptability.

A discrete choice experiment (DCE) is a widely used method to assess preferences in health areas.¹⁵ Former studies used DCE to study the preference regarding influenza vaccination among the elderly in developed countries,¹⁶ which could not adapt to China due to different socioeconomic characteristics. To our knowledge, no study has been previously performed to explore individuals’ preferences regarding influenza vaccination in China. The present study aimed to explore the preferences of elderly living in Shaanxi province of China toward influenza vaccination using the DCE method. The findings will benefit health policy development to increase vaccination uptake and its performance.

2. Materials and methods

2.1. Discrete choice experiment

DCE method is an attribute-based utility measurement technique used to measure individual preference for medical intervention. It is based on the hypothesis that the intervention’s decision can be described by its several key attributes (such as vaccine effectiveness, vaccination cost, and risk of side effects). Individual choice is attributed to the levels (such as 20%, 50%, or 80% of vaccine effectiveness) of these attributes. A series of hypothetical choice sets consisting of different levels of attributes were presented to the enrolled respondents and asked to select their preferred option from each choice set. This will help to determine the relative importance of the attributes and levels.

2.2. Attributes and levels

In the present study, a relevant DCE literature search was initially undertaken to develop appropriate influenza vaccination attributes with their corresponding levels.^16–21 A consultation with three experts expertized in vaccines and two focus groups, including 20 elderly aged over 60 years, was performed to determine the final attributes and levels. Five attributes were selected: (1) vaccine effectiveness; (2) duration of protection; (3) severe adverse events; (4) cost of influenza vaccination; and (5) source of vaccination recommendation. The corresponding levels of each attribute are shown in Table 1.

Table 1. Attributes and levels

Attribute	Level 1	Level 2	Level 3
Vaccine effectiveness (%)	20	50	80
Protection duration (months)	3	6	12
Severe adverse events	1/100,000	10/100,000	100/100,000
Vaccination recommendation	Healthcare providers*	Family/friends	Media**
Vaccination cost (CNY)	0	80	160

*Includes doctors, nurses, and pharmacists; **Includes traditional media (TV and radio) and social media (WeChat and Weibo).

CNY: Chinese yuan.

2.3. DCE and questionnaire design

Based on the combination of the five attributes and their levels, 243 hypothetical influenza vaccination choices (3⁵ = 243) were obtained. Since it was not possible to present an individual with all choice tasks (full factorial design), we performed a fractional factorial design based on orthogonal arrays using IBM SPSS Statistics 21, and 16 hypothetical choices were generated for the present study. The chosen 16 choices were then paired with 8 choice sets. Participants need to choose one from the two profiles in each choice set to acknowledge their preference. Additionally, one choice set as a rationality test was added to the DCE design to check participants’ understanding of this survey.²² A total of 9 choice sets were included in the questionnaire with additional 8 sociodemographic information of the elderly (gender, age, region, occupation, education level, monthly income, health condition with chronic diseases, and living with their children). To examine the questionnaire’s validity, a pilot study was conducted with 30 elderly people, and we found it difficult for the elderly to understand the choice sets and complete the survey. Therefore, the presence of the choice sets in our questionnaire was further designed in a cartoon version with pictograms, as illustrated in Table 2. The investigators explained the graphical representation of choice sets to the elderly before the survey to help them recognize the choice sets’ intention.

Table 2. Illustration of the choice set

2.4. Participants and data collection

In the present study, the maximum level of attributes, the number of choice sets, and the number of alternatives in each choice set were 3, 8, and 2, respectively. Based on Orme’s rule of thumb formula,²³ at least 94 participants (500 × 3 ÷ 8 ÷ 2) were required to accomplish the study. Participants were enrolled from three representative cities of Shaanxi province, including Yan’an (northern region), Xi’an (central region), and Han Zhong (southern region). These cities represent a distinct socioeconomic level in Shaanxi province,^12, and Xi’an has the highest gross domestic product per capita, whereas Han Zhong has the lowest. Four parks located in the east, west, south, and north of each city were randomly selected, where random intercept access was used to seek the elderly who were willing to participate in the survey. The elderly who had poor recognition or refused to give consent were excluded from the study.

Medical students from the Health Science Center of Xi’an Jiaotong University were recruited as investigators for the study. They were trained adequately about the study design and communication skills. The formal survey was conducted in January 2020. All participants were informed about the study’s aim and relevant knowledge of influenza infection and its consequences. Investigators explained each attribute’s meaning to the participants and helped them complete it. Furthermore, all queries of the participants were addressed by data collectors.

2.5. Statistical analysis

Data analyses were performed by STATA 16.0 (StataCorp., LLC, College Station, TX, USA). Random utility theory is an important theoretical basis in designing a DCE and constructing a discrete choice model.^15,24,25 The utility associated with a choice is assumed to be a function of the observed characteristics (attributes and levels) and the unobserved characteristics of that choice. The vaccine effectiveness, duration of protection, severe adverse events, and source of vaccination recommendation were assumed to have a nonlinear association with individual utility. Therefore, they were treated as categorical variables. The cost of vaccination was treated as a continuous variable. A mixed logit model was applied to assess participants’ preferences for each attribute and its level. It fits the utility function as follows:

$\begin{array}{rl}& \mathit{U}={\beta }_0+{\beta }_1\ast \mathrm{Effectiveness}\mathrm{\_}50+{\beta }_2\ast \mathrm{Effectiveness}\mathrm{\_}80\\ & +{\beta }_3\ast \mathrm{Protection}\mathrm{\_}\mathrm{duration}\mathrm{\_}6\\ & +{\beta }_4\ast \mathrm{Protection}\mathrm{\_}\mathrm{duration}\mathrm{\_}12\\ & +{\beta }_5\ast \mathrm{Adverse}\mathrm{\_}\mathrm{effects}\mathrm{\_}1/100,000\\ & +{\beta }_6\ast \mathrm{Adverse}\mathrm{\_}\mathrm{events}\mathrm{\_}10/100,000\\ & +{\beta }_7\ast \mathrm{Vaccination}\mathrm{\_}\mathrm{cost}+{\beta }_8\ast \mathrm{Family}\mathrm{\_}\mathrm{friends}\\ & +{\beta }_9\ast \mathrm{Media}+\mathit{\epsilon }\end{array}$

Where U is an individual utility choice; β₀ is a constant representing the general preference accepting influenza vaccination compared to no influenza vaccination; β₁-β₉ are coefficients reflecting relative preference weights for a given attribute level among various levels of attributes. A significant (p < .05) coefficient indicates that participants differ between one attribute level and the reference level in making stated choices. If the coefficient is positive, it indicates that the attribute positively affects participants’ preferences, and participants are likely to choose the level relative to the reference level. ε is an error term describing the unmeasured variation of participants’ preference.

The equation that was applied to assess participants’ marginal willingness to pay (WTP) for the cost per unit of the attribute is as following:²⁶

$\mathrm{Marginal}\mathrm{WTP}=-\frac{\mathit{\beta }\mathit{attribute}}{\mathit{\beta }\mathit{cost}}$

Where β_attribute is the coefficient of one specific attribute; β_cost is the coefficient of vaccination cost, and its value equals the negative coefficient of cost.

In this study, a hypothetical base-case influenza vaccination strategy is the combination of vaccine effectiveness of 20%, 3-month duration of protection, severe adverse events of 100/100,000, vaccination cost of 160 Chinese yuan (CNY), and vaccination recommendation from healthcare providers (doctors, nurses, and pharmacists). To explore the potential probability of participants’ choice in a given level of vaccination attributes, the following equation was applied:²⁷

$\mathit{p}=\frac{1}{1+{\mathit{e}}^{-\mathit{U}}}$

Where p is the choice probability, U is the utility of influenza vaccination in specific vaccination attributes and levels. The probability variations of participants’ choices were explored when the levels of attributes changed relative to the reference level in the base-case vaccination strategy.

3. Results

3.1. Participants characteristics

In total, 150 elderly people participated in the present study, and 144 (96%) completed the DCE choice with the correct response for the rationality test. Among all respondents, 69 (47.9%) were female, 90 (62.6%) obtained education in secondary school (junior or senior high school level), and 98 (68.1%) reported that they had one or more chronic diseases. The characteristics of the respondents are presented in Table 3.

Table 3. Demographic information

Variable	Frequency (n)	Percentage (%)
Gender
Male	75	52.1
Female	69	47.9
Age (years)
60–69	81	56.3
70–79	44	30.6
>80	19	13.1
Region
Yan’an	50	34.7
Xi’an	50	34.7
Han zhong	44	30.6
Residence
Urban	97	67.4
Rural	47	32.6
Occupation
Retirement	95	66.0
Full-time/part-time job	11	7.6
other	38	26.4
Education level
Primary school or below	27	18.7
Secondary school	90	62.6
College school and above	27	18.7
Monthly income (CNY)
<2,000	57	39.5
2,000–4,000	60	41.7
≥4,000	27	18.8
Health condition with chronic diseases
One and above	98	68.1
None	46	32.0
Live with their children
Yes	67	46.5
No	77	53.5

CNY: Chinese yuan.

3.2. Preferences of vaccination attributes

The results of mixed logit regression analysis are shown in Table 4. There was a significant difference (p < .05) among participants’ preference of influenza vaccination toward 20% versus 80% of vaccine effectiveness, 3- versus 12-month duration of protection, and severe adverse events of 1/100,000 versus 100/100,000, vaccination recommendation from healthcare providers versus media or family/friends, and vaccination cost. The duration of protection was the most important attribute due to the most significant difference (1.216) in preference weights among attribute levels, while the vaccination cost was the least significant attribute for the lowest preference weight (0.006). The positive signs of attribute coefficients demonstrated that participants preferred influenza vaccination with higher effectiveness, longer duration of protection, lower severe adverse events, and higher vaccination cost. The negative coefficient signs of vaccination recommendation indicated that participants preferred recommendation from healthcare providers.

Table 4. Respondents’ preferences for influenza vaccination

Variable	Coefficient	Standard error	P value
Constant	−0.461	0.010	<.001
Vaccine effectiveness (20% as reference level)
50%	0.265	0.189	.16
80%	0.357	0.166	.032
Protection duration (3 months as reference level)
6 months	0.121	0.130	.351
12 months	1.216	0.168	<.001
Severe adverse events (100/100,000 as reference level)
1/100,000	0.826	0.275	.003
10/100,000	0.239	0.201	.235
Source of vaccination recommendation (healthcare providers as reference level)
family or friends	−0.926	0.230	<.001
media	−0.354	0.143	.014
Vaccination cost (CNY)
0–160	0.006	0.001	<.001
Log likelihood	−701.934

CNY: Chinese yuan.

3.3. Trade-off among vaccination attributes

The marginal WTPs for each attribute of influenza vaccination are shown in Table 5. The largest marginal WTP was CNY220.90 when the duration of protection extended from 3 to 12 months. Participants were willing to pay an additional CNY64.81 to increase vaccine effectiveness from 20% to 80%. If the severe adverse events of vaccination were reduced from 100/100,000 to 1/100,000, participants were willing to pay an additional CNY149.97. Compared to the vaccination recommendation from healthcare providers, participants were willing to pay less if it was recommended by media or family/friends.

Table 5. Trade-offs among attributes of influenza vaccination

Attribute	Alternatives	WTP (CNY)	95% CI
Vaccine effectiveness	From 20% to 50%	48.15	−24.58 to 120.88
	From 20% to 80%	64.81	−1.37 to 131
Protection duration	From 3 months to 6 months	22.03	−24.61 to 68.67
	From 3 months to 12 months	220.90	141.19 to 300.60
Severe adverse events	From 100/100,000 to 1/100,000	149.97	34.27 to 265.67
	From 100/100,000 to 10/100,000	43.42	−31.03 to 117.87
Source of vaccination recommendation	From healthcare providers to family or friends	−168.20	−232.10 to −104.30
	From healthcare providers to media	−64.28	−115.55 to −13.01

WTP: willingness to pay; CNY: Chinese Yuan; CI: confidence interval.

3.4. Probability variation of vaccination choice

The probability variations of vaccination choice at different attribute levels are shown in Figure 1. The predicted choice probability was 58.28% in the base-case scenario (20% of vaccine effectiveness, 3-month duration of protection, 100/100,000 severe adverse events, vaccination cost for CNY160, and vaccination recommendation from healthcare providers). With the variation of attribute levels, the highest choice probability (83.70%) occurred when the duration of protection extended from 3 to 12 months, followed by reducing severe adverse events from 100/100,000 to 1/100,000 (77.65%). The change of vaccination recommendation from healthcare providers to family/friends was associated with the lowest choice probability (37.62%).

Figure 1. Probability of vaccination choices at different attribute levels

4. Discussion

To our knowledge, the present study was the first survey on the preferences of influenza vaccination among the elderly aged over 60 years in China. We found that the elderly preferred influenza vaccination with higher vaccine effectiveness, longer protection duration, lower severe adverse events, higher vaccination cost, and vaccination recommendations from healthcare providers. One-way variation of vaccination attribute levels showed that the protective duration of influenza vaccination increasing from 3 to 12 months had considerable impact on the elderly population’s vaccination WTP and probability of vaccination choice.

A study from the Netherlands reported similar findings that the elderly had a higher probability of receiving influenza vaccination if the vaccination was more effective, had lower risks of serious or mild side effects, and had longer protection duration.¹⁶ Furthermore, similar results were found in DCE studies on individuals’ preferences for human papillomavirus (HPV) and rotavirus vaccinations,^20,28 or influenza vaccines for children.²⁹ However, the relative importance of determinants on individuals’ decision for vaccination varied in studies. In our study, the duration of influenza vaccination protection was found to be the most important attribute, while a study from Hong Kong showed that the influenza case-fatality ratio and vaccine efficacy had the most significant effect.³⁰ Vaccine effectiveness was also of utmost importance in a recent study from China with public preference to COVID-19 vaccines, followed by longer protection and fewer adverse events.³¹ Therefore, individuals could have different performance on preference trade-offs of various vaccination programs.

In the present study, the probability of vaccination choice would increase dramatically with the change of protective duration from 3 to 12 months. The elderly considered longer protective effects as the vaccination priority. A meta-analysis on comparison of vaccine effectiveness after influenza vaccination (15–90 days to 91–180 days) found a significant decline (19%-33%) of vaccine effectiveness at 91–180 days.³² Antibodies to the influenza vaccine were expected to wane faster in older populations; therefore, the protective duration among the elderly would be even shorter.³³ The current influenza vaccination is offered every year, and vaccine manufacturers need to make serious efforts to develop influenza vaccines possessing a long-term protective immunity.

Among the three sources of vaccination recommendation, healthcare providers’ recommendation was the most preferred in our study compared with the recommendation from media or family/friends. This finding has also been advocated in previous studies,^34–36 which concluded that healthcare providers’ advice is a significant motivation for the elderly to receive vaccinations. The safety concern of vaccination was sometimes reported by the media, which might produce distrustful emotions for vaccination among the public.^21,37 The recommendation for vaccination from family or friends is not deemed professional and trustworthy. Therefore, healthcare providers’ educational interventions could be a practical approach to improve the uptake rate of influenza vaccination. Notably, the Chinese government is implementing a family doctors’ policy to improve the accessibility of healthcare, especially for the elderly population.³⁸ Family doctors could enhance the influenza knowledge of the elderly and encourage them to get vaccinated before the flu season.

Several DCE studies on vaccination preferences have reported that individuals preferred lower out-of-pocket expenses for vaccination as expected.^29,39–41 Interestingly, elderly people were willing to pay higher for influenza vaccination in our study. It could be explained by the range of vaccination cost included in the DCE design, and the highest cost level (CNY160) was not considered expensive for the elderly since a majority (67.4%) of them were from urban areas. They were not sensitive to the variation of vaccination cost even though it was offered free. Besides, the proportion of educated participants (secondary school, college and above) in our study was high (81.3%), which could be related to the acceptance of higher vaccination cost. Similar findings were reported in a previous DCE study on parental preferences for HPV vaccination, where educated subjects were more inclined to accept vaccination at higher prices in China.⁴² Although, in our study, the vaccination cost had a positive impact on the preferences of the elderly. The attribute coefficient (β = 0.006) was the lowest, implying the least important compared to other attributes.

There were several limitations to the present study. Although the number of participants in our study was small, nevertheless, it satisfied the need for a minimum sample size for a local DCE study. All participants were from Shaanxi province of China; therefore, our study’s findings could not be fully generalized to other regions of China, primarily where immunization programs for influenza vaccination have been implemented. Future studies could expand the surveyed areas and subjects to generate more robust findings for the Chinese elderly’s influenza vaccination preferences. Individuals’ vaccination preferences were associated with their characteristics and vaccination attributes.¹⁶ In light of the limited sample size for subgroup analysis, we could not detect preference variations against the elderly population’s characteristics. The interactive impacts on the preferences on individual characteristics and vaccination attributes were not further explored in the present study. To ensure all participants have a selection within the choice alternatives, we did not include an opt-out alternative in the choice sets. Therefore, the probability of vaccination choice could not represent the actual uptake rate of influenza vaccination in the elderly.

In conclusion, the elderly population in Shaanxi province of China regarded longer vaccination protection duration as the most important attribute for influenza vaccination, followed by lower severe adverse events, higher vaccine effectiveness, and vaccination recommendation from healthcare providers, whilst the vaccination cost was the least important attribute. The present study’s findings would guide policymakers and manufacturers to implement interventions to increase the uptake rate of influenza vaccination in the Chinese elderly.

Disclosure of potential conflicts of interest

The authors declare no conflicts of interest.

Additional information

Funding

This work was supported by the China Postdoctoral Science Foundation under Grant [2018M631179] and Shaanxi Natural Science Foundation under Grant [2020JQ-079].