The intentions to get vaccinated against influenza and actual vaccine uptake among diabetic patients in Ningbo, China: identifying motivators and barriers

ABSTRACT

Background: Health authorities recommend influenza vaccination to diabetic patients. Nevertheless, the vaccination coverage of adults was low in China. This study aimed to estimate influenza vaccination intentions and actual uptake among diabetic patients in China and identify the motivators and barriers associated with vaccination.

Methods: During Nov 2016–Jan 2017, 1960 diabetic patients were invited to participate in the investigation about influenza vaccination based on health belief model (HBM). To link vaccination intention to behavior, a follow-up survey was conducted to collect vaccination records of the 2016–2017 season. Predictors of the motivation to obtain influenza vaccination were assessed using logistic regressions.

Results: 1914 diabetic patients completed the survey. 46.13% participants reported intentions to be vaccinated against influenza and 7.84% actually received vaccination. In the multivariate models, while all the domains of HBM constructs were associated with intentions, age, increasing numbers of comorbidity, urban residents, perceived susceptibility, perceived benefits, free vaccination, and vaccination history displayed positive associations with vaccine uptake while increasing income and perceptions of barriers were negative predictors of vaccine uptake. Besides, interactions between perceived susceptibility and healthcare workers’ (HCW’s) recommendation, perceived severity and benefits, perceived severity and health-seeking behaviors, perceived benefits, and influence of family/friends had a significant positive effect modification on the vaccine uptake.

Conclusion: To improve diabetic patients’ influenza vaccination and close the intention-behavior gap, multipronged strategies are required not only to increase vaccination intention by promoting HCW’s recommendation to improve perceptions about influenza vaccination but also facilitate a follow through on initial intentions by implementing free influenza vaccination program funded by the government.

KEYWORDS:

• Seasonal influenza vaccine
• diabetes
• vaccination intention
• vaccine uptake

Introduction

During the last few decades, diabetes has become a significant and increasing burden in China, which now has the world’s largest diabetes epidemic due to rapidly increasing prevalence from 2.5% in 1994 to 10.9% in 2013.^1–3 Diabetic patients are at an increased risk of hospitalization and death attributable to influenza infections, typically associated with an increased severity and recurrence of the infection. The economic burden of influenza-associated medical care among diabetic patients is substantial. The mean costs per episode among patients with chronic disease were much higher than that among patients without chronic disease in China, especially the inpatients costs.⁴

Annual seasonal influenza vaccination is recommended to diabetic patients as the most effective strategy to prevent influenza infection and reduce related severe complications by the World Health Organization (WHO) and the Chinese Center for Disease Prevention and Control (China CDC).^5,6 A previous study has found that diabetic patients have optimal B cell responses to influenza vaccine, which suggested a valid protection against influenza.⁷ Furthermore, current evidence suggests that influenza vaccination could reduce winter mortality and hospitalization within diabetic patients, and even offer benefits in reducing community-acquired lower respiratory tract infection, heart failure, acute myocardial infarction, and stroke.^8–11

In China, the influenza vaccine is not included in the national expanded program on immunization (EPI) and the reimbursement policies vary in different regions.¹² Influenza vaccine uptake remains low and the coverage rate ranges from 0.3 to 35.5% among Chinese adults with chronic diseases,^12–16 which is far below the level of high-income countries.^17–19 And unlike the high-income countries, the influenza vaccination coverage of population with chronic diseases is even a bit lower than the general population in China^20,21 while the vaccination coverage also varies with chronic disease species and regions.²² Besides, more remarkable, there is a significant gap between intention and actual behavior to get vaccinated. A recent investigation showed that the willingness for influenza vaccination was 45% in general population²³ while the actual vaccination coverage was 9.4%, which was reported by a meta-analysis in China.²⁰ These indicate that there are barriers existed from intention to behavior besides the cognitive factors, which affected the vaccination willingness, such as not receiving recommendation from doctor and not having cost-free vaccination.²⁰ Previous studies have examined factors associated with intention of influenza vaccination among older adults or healthcare workers in China,^16,20 However, less is known about influenza vaccination intention, actual uptake and the related factors among patients with chronic diseases.²⁰

Ningbo is an economically developed coastal city located in the middle of the east coastline of China and has two influenza activity peaks in November to February and July to September.²⁴ In Ningbo city, the influenza vaccine could be reimbursed by basic social medical insurance (BSMI). Besides, free influenza vaccine has been recommended to patients with chronic diseases who have purchased for community health services by general practitioners in Dongqianhu national tourism vacation zone since 2016. However, the influenza vaccination rate is still far from satisfaction among patients with chronic diseases^13,25 compared to high-income countries,^17,19,26 whereas the influenza vaccine hesitancy, defined as delay in acceptance or refusal of vaccination despite availability of vaccination services,²⁷ is widespread among patients with chronic diseases, which is a significant threat to the efforts made to reduce the burden of pandemic and seasonal influenza.^27,28 Consequently, it is important to understand the predictors and barriers during individual influenza vaccination decision-making process, in order to establish effective interventions to improve the vaccine uptake. In spite of the large number of studies conducted to understand the current influenza vaccination coverage and assess predictors of the influenza vaccination in China, a few were about diabetic patients.²⁰ In addition, previous studies focused on the cognitive factors associated with influenza vaccination willingness or uptake, such as perceived severity of disease and perceived effectiveness and safety of vaccination, and neglect the intention-behavior gap and the effects of barriers existed from willingness to behavior.²⁹

The health belief model (HBM), which has been widely applied to the studies of influenza vaccination intentions or behaviors in many countries,^30–34 consists of six main components: perceived susceptibility of the disease, perceived severity of the disease, benefits of the recommended behavior, barriers related to the health behavior, self-efficacy that refers to the belief that one can complete the health behavior successfully, and cues to action including the factors that promote adoption of the health behavior.³⁵ Using the theoretical framework of the HBM, we conducted a cross-sectional investigation among participants diagnosed with diabetes in Ningbo aimed to assess the influenza vaccination intention and a combination of cognitive and behavioral variables as a baseline. Then, we additionally conducted a follow-up survey using Ningbo adult vaccination database to estimate the subsequent actual vaccination coverage among diabetes patients and examine the motivators and barriers associated with the influenza vaccination using both cognitive and behavioral insights. Specifically, we evaluated the effects of the externals, such as cost of vaccination, healthcare-seeking behavior as well as family and doctors’ recommendation, and the potential effect modification of interactions between internal and external factors.

Methods

Study design and sample

The baseline investigation was conducted in Ningbo from November 2016 to January 2017 and the influenza vaccination data follow-up was collected from July to August 2017. The participants were included if they met the follow inclusion criteria: (1) diagnosed with diabetes by one-level hospitals or above; (2) ≥18 years of age; (3) attending the diabetes health management program in Ningbo and visit community health center regularly. The diabetes health management program is a government-provided program under the Chinese basic public health service, which provides free healthcare service to diabetic patients at least four times in a year.³⁶

A sample size of 1400 participants was calculated using the formula:$n=\frac{z_{\alpha }^2\times pq}{d^2}$, based on an estimated influenza vaccination coverage among diabetic patients of 6.6%, which was determined from a recent study in eastern China,¹⁶ a p-value of 0.05, and a study power of 80%.

In Ningbo, 152 CHCs in ten districts and three special economic zones provide primary healthcare and disease-prevention services to diabetic patients living nearby, including physical examinations, vaccination, and health information. In our study, a multistage stratified sampling method was used to select a set of representative samples. For sampling, two CHCs were selected from each district except Haishu district (Haishu district was excluded because a similar survey was previously conducted there.) through quota sampling and Qianhu hospital in Dongqianhu National tourism vacation zone was also selected as free influenza vaccination project area. Therefore, totally 18 CHCs and Qianhu hospital were included in our study. The quota of each CHC or hospital was determined and we allocated 100 questionnaires to each CHC and 160 questionnaires to Qianhu hospital. Then, the study population was selected through a convenient sampling method. Diabetic patients having a regular visit for health service in the selected CHCs or hospitals would be eligible to participate if he/she met the inclusion criteria. During the survey period from November 2016 to January 2017, the eligible participants would be informed about the study by their general practitioners. If they agreed to be interviewed, they would be recruited for this study. Finally, among 1960 diabetic patients who were invited to participate in our investigation, 1947 of them agreed and 13 refused; 1914 participants completed the questionnaire with a response rate of 97.65%, and 33 participants were excluded because of incomplete answers.

Theoretical framework

We used the HBM as the theoretical framework of this study. In the recent years, the HBM has been extensively used to examine the vaccination beliefs or behaviors as well as identify motivators and barriers.^30–33 The HBM used in this study included six constructs to predict vaccination intention and behavior: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, self-efficacy, and cues to action.³⁵ Perceived susceptibility was patients’ self-assessments regarding themselves as susceptible to the disease. Perceived severity addressed patients’ beliefs about potentially serious consequences of the disease. Perceived benefits were patients’ perceptions that the vaccination would reduce the susceptibility or severity of the disease or lead to other positive outcomes for them. Perceived barriers were related to patients’ negative beliefs and concerns about vaccination. Self-efficacy measured the patients’ confidence that they could successfully complete vaccination despite considered barriers. Cues to action were specific cues which could impact the final intention or behavior of vaccination and these cues to action can be internal or external. A recent study³⁵ indicated it is plausible that interaction effects between the HBM items in addition to a direct effects of the HBM constructs on vaccination. Therefore, in our study, we measured both direct effects of the HBM constructs and the effect modification of interactions between the HBM items.

Questionnaire

The questionnaire was comprised of a total 34 items. It contained 25 questions based on the six key domains of the HBM, which mainly contained the following fields of inquiry: (1) seven demographic variables were collected, including age, gender, marital status, monthly income, highest level of education attainment, occupation and comorbidity diagnosed by one-level hospitals or above, containing chronic respiratory disease, chronic heart disease, cerebrovascular disease, hypertension and cancer; (2) 17 questions based upon the HBM model (Table 3): 1) three statements addressed the participants’ perceived susceptibility of influenza; 2) two statements addressed the participants’ perceived severity of influenza infection; 3) two statements addressed the participants’ perceived benefits of influenza vaccination; 4) two statements addressed the barriers to influenza vaccination, including whether influenza is preventable and the worries about the safety of influenza vaccine; 5) three statements addressed the self-efficacy of the participants to execute the vaccination behavior, including the payment for vaccine, access of vaccine, and healthcare-seeking behavior if they develop any symptoms of influenza-like illness; 6) five statements assessed the participants’ cues to action, including recommendation from doctor, family/friends and media/internet, influenza vaccination history, and purchase of health services of community family physician; (3) one question was used to directly ask participants their intention to receive the influenza vaccine during 2016–2017 influenza season(1 = Have been vaccinated; 2 = Have the intention of vaccination; 3 = Have the intention of vaccination if the vaccine is free; 4 = Have the intention of vaccination if the cost of vaccine is covered by my/family’s insurance; 5 = No clear decision; 6 = No intention) and later, their influenza vaccination record would be verified in the Ningbo adult vaccination database in the follow-up survey. The questionnaire was pretested among a convenience sample of 160 diabetic patients in Qianhu hospital to assess readability, comprehension of instructions, and clarity. According to the feedback from the pretest, minor modifications were made to improve readability and clarity of the questionnaire. The Cronbach’s α coefficient of the questionnaire was 0.73, which indicated good internal consistency reliability.

Data collection

Written notifications were sent to all the selected CHCs. The general practitioners at each selected CHC were recruited and trained for 1 d by the professional investigators from Ningbo Municipal Center for Disease Prevention and Control (NBCDC) and served as interviewers, since the general practitioners had their own fixed diabetic patients and were visited by the patients at least once a quarter so that they had established a relationship of trust with their patients. Face-to-face interviews were performed during the diabetic patients’ visits. The questionnaire items were explained to the respondents by the interviewers. Each interview lasted approximately 30 min. In our study, if a respondent claimed the intention to get vaccinated against influenza, he/she would be asked to sign the informed consent of vaccination and make an appointment to get vaccinated at the CHC where the interview was conducted; if a respondent claimed the intention when the vaccination was free or covered by insurance, he/she would be given an informed consent of vaccination and asked to think it through carefully after going home. Multiple checks of the questionnaires were conducted and we excluded the data if there were incomplete answers to any measures of this study. After all the questionnaires were completed and influenza vaccination was offered, investigators of the NBCDC conducted the follow-up survey from July to August 2017 using the Ningbo adult vaccination database to collected the participants’ influenza vaccination records of 2016–2017 influenza season (August 2016 to April 2017).

Statistical analysis

All the data of the HBM-based variables were coded into binary variables (1 = Yes/0 = Not sure or no) and grouped according to domains. The intention of vaccination was classified into four-point ordinal variable (3 = High intention (have been vaccinated or have the intention of vaccination); 2 = Moderate intention (have the intention if the cost of vaccine is free or covered by insurance); 1 = No clear decision; 0 = No intention). Pearson’s correlation was used for domains with two items and the Cronbach’s α coefficient was calculated for domains with three or more items. The scores of the HBM items in perceived susceptibility, perceived severity, and perceived benefits were summed to create new three-point Likert scaled (low/moderate/high) variables for domains with α coefficient or Pearson’s correlation >0.5. All the items in domains of barriers, self-efficacy, and cues to action were kept as separate variables. Frequency and percentages were calculated for all the variables. Cross tabulation was used to assess the association between the intention of vaccination and the actual vaccine uptake during 2016–2017 influenza season. Binary and multivariate logistic regressions were used to estimate the effects of demographic variables and the HBM-based variables on participants’ intentions of vaccination and actual vaccine uptake. Crude and adjusted odds ratio (OR) and 95% confidence interval (95%CI) were calculated. In the multivariate logistic regression analysis, a fully adjusted multivariate model was used including all the demographic variables and HBM constructs, and we introduce the effect modification of interactions between the HBM items into the multivariate logistic regression model. We calculated variance inflation factors to estimate the multicollinearity of input factors (VIF = 1.70) that indicated these variables did not have the serious multiple collinearity. All the statistical analyses were performed using Stata 12.0 (Stata Corp, Texas, USA). A two-tailed p-value of <0.05 was considered statistically significant.

Ethics statement

The study was approved by the Institutional Review Board of Ningbo Municipal Center for Disease Control and Prevention (No. 201811). Informed consent was obtained from all the participants before the interview.

Results

Demographic characteristics of participants

A total of 1914 diabetic patients completed the questionnaire. Of these participants, average age was 65.13 ± 9.75 years old, 740 (38.66%) were male, 804 (42.01%) were from rural areas, and 1652 (86.31%) were married. Most participants had primary school or below educational background (67.14%), followed by secondary school graduates (30.77%), and tertiary school graduates (2.09%). The occupation included 119 (6.22%) executives and professionals, 477 (24.92%) clerks and workers, 983 (51.36%) farmers, and 335 (17.50%) economically inactive. 75.71% had a monthly income under ￥3000 and 24.29% had a monthly income more than ￥3000. 603 (31.50%) participants reported suffering from diabetes only, while 1095 (57.21%) and 216 (11.29%) participants reported one kind or at least two kinds of comorbidities besides diabetes. Details are shown in Table 1.

Table 1. Characteristics of survey participants

Demographics	Total (N = 1914)
Age (mean, SD)	65.13 ± 9.75
Gender
Male	740 (38.66)
Female	1174 (61.34)
Monthly income (CNY)
<3000	1449 (75.71)
≥3000	465 (24.29)
Highest level of education attainment
Primary or below	1285 (67.14)
Secondary	589 (30.77)
Tertiary or above	40 (2.09)
Occupation
Executives and professionals	119 (6.22)
Clerks and workers	477 (24.92)
Farmers	983 (51.36)
Economically inactive	335 (17.50)
Marital status
Single/divorced/widowed	262 (13.69)
Married	1652 (86.31)
Region
Rural	804 (42.01)
Urban	1110 (57.99)
Numbers of Comorbidities^a
0	603(31.50)
1	1095(57.21)
≥2	216(11.29)

Data are reported as number of participants(%).

^aComorbidities included chronic respiratory disease, chronic heart disease, cerebrovascular disease, hypertension and cancer.

Intention to receive influenza vaccine and actual vaccination coverage rate

As shown in Figure 1 and Table 2, 883 (46.13%) participants reported the intention to receive the influenza vaccination in the 2016–2017 season, including 274 (14.32%) had a high intention and 609 (31.82%) had a moderate intention to receive vaccination if the vaccination could be funded by medical insurance, employer, or government; 341 (17.82%) participants indicated that they had not made a clear decision to be vaccinated and 690 (36.05%) showed no intention. Among the groups with different demographic characteristics, the percentages of diabetic patients who had the intention to be vaccinated were comparable which ranged from 40.30 to 56.02% (Figure 1). Nonetheless, the actual influenza vaccination coverage rate of diabetic patients was much lower compared to the intention (Table 2). 150 (7.84%) participants got vaccinated against influenza in the 2016–2017 season. The vaccination coverage rate differed by intention (χ² = 464.88, p < .001). In high intention group, the coverage rate was 39.42%, which was much higher than 6.73, 0.29, and 0% in moderate intention, no clear decision, and no intention groups, respectively (Table 2). The coverage rate also increased with age. A total of 9.34% of those aged 65 and over and 6.06% of those aged under 65 were vaccinated against influenza in the 2016–2017 season. The coverage rate of the diabetic patients who had one comorbidity (11.28%) was 4.53 times as likely as those who had diabetes only (2.49%), and those who had at least two comorbidities (11.11%) were 4.46 times as likely as those who had diabetes only (Figure 1).

Figure 1. (a) Intention to receive influenza vaccine among diabetic patients, by select demographics, the 2016/2017 influenza season. (b) Influenza vaccination coverage of diabetic patients, by select demographics, the 2016/2017 influenza season

Table 2. Crosstab of diabetic patients’ intention to get vaccinated and their actual vaccination behavior

	No intention	No clear decision	Moderate intention	High intention	Total
Not vaccinated	690 (100.00)	340 (99.71)	568 (93.27)	166 (60.58)	1764 (92.16)
Vaccinated	0 (0.00)	1 (0.29)	41 (6.73)	108 (39.42)	150 (7.84)
Total	690 (100.00)	341 (100.00)	609 (100.00)	274 (100.00)	1914 (100.00)

Data are reported as number of participants (%).

Health belief model domains about influenza and vaccination

Survey items were grouped according to the HBM framework and the responses presented by intention and vaccination status in Table 3. Among all the participants, the responses to “Perceived Susceptibility” items ranged from 25.03 to 30.51%: frequency of influenza-like illness (N = 513, 26.80%), susceptibility to influenza (N = 479, 25.03%), and susceptibility of diabetics (N = 584, 30.51%). 747 (39.03%) participants responded to two “Perceived Severity” items “severe infection” and “glucose control.” 475 (24.82%) and 378 (19.75%) participants agreed to “Perceived benefits” items “prevent influenza” and “prevent acute complications,” respectively. As for “Barries” items, 880 (45.98%) perceived “unpreventable of influenza” and 510 (26.65%) were concerned about “safety of vaccine.” 778 (40.65%) participants reported that their influenza vaccination could be funded by medical insurance,, employer or the government; most participants (N = 1579, 79.36%) thought it was convenient to get the influenza vaccine and 837 (43.73%) participants indicated that they would go to hospital if they had symptoms of influenza-like illness. 655 (34.22%), 301 (15.73%), and 479 (33.38%) participants received the recommendations from doctors, family/friends, or media respectively. 268 (14%) reported influenza vaccination history and 1230 (64.26%) participants purchased health services of community family physician. Those reporting higher perceptions of susceptibility, severity, benefits, self-efficacy, and cues to action were more likely to have the intentions to be vaccinated and actually take the influenza vaccine, while those reporting higher perceptions of barriers were less likely to willing to be vaccinated or take action to get vaccination.

Table 3. The Summary Health Belief Model responses of survey participants

				Intention to receive influenza vaccination				Influenza vaccination in 2016/2017
HBM Constructs	Item	Level	Total	High (N = 274)	Moderate (N = 609)	No clear (N = 341)	No (N = 690)	Vaccinated (N = 150)	Unvaccinated (N = 1764)
Perceived Susceptibility
Frequency of influenza-like illness	I often had influenza-like illness last year (≥3 times).	Yes	513	99 (19.30)	199 (38.79)	81 (15.79)	134 (26.12)	48 (9.36)	465 (90.64)
		No/not sure	1401	175 (12.49)	410 (29.26)	260 (18.56)	556 (39.69)	102 (7.28)	1299 (92.72)
Susceptibility to influenza	Compared with peers, I am more likely to get influenza.	Yes	479	113 (23.59)	199 (41.54)	63 (13.15)	104 (21.71)	59 (12.32)	420 (87.68)
		No/not sure	1435	161 (11.22)	410 (28.57)	278 (19.37)	586 (40.84)	91 (6.34)	1344 (93.66)
Susceptibility of diabetics	Diabetic patients were more likely to get infection.	Yes	584	131 (22.43)	236 (40.41)	90 (15.41)	127 (21.75)	77 (13.18)	507 (86.82)
		No/not sure	1330	143 (10.75)	373 (28.05)	251 (18.87)	563 (42.33)	73 (5.49)	1257 (94.51)
Perceived Severity
Severe infection	Diabetic patients would get severely sick and even lead to death with influenza infection.	Yes	747	141 (18.88)	266 (35.61)	134 (17.94)	206 (27.58)	77 (10.31)	670 (89.69)
		No/not sure	1167	133 (11.40)	343 (29.39)	207 (17.74)	484 (41.47)	73 (6.26)	1094 (93.74)
Glucose control	Influenza would lead to blood glucose oscillations among diabetic patients.	Yes	747	147 (19.68)	271 (36.28)	127 (17.00)	202 (27.04)	80 (10.71)	667 (89.29)
		No/not sure	1167	127 (10.88)	338 (28.96)	214 (18.34)	488 (41.82)	70 (6.00)	1097 (94.00)
Perceived Benefits
Prevent influenza	Influenza vaccination can prevent influenza infection.	Yes	475	142 (29.89)	157 (33.05)	63 (13.26)	113 (23.79)	104 (21.89)	371 (78.11)
		No/not sure	1439	132 (9.17)	452 (31.41)	278 (19.32)	577 (40.10)	46 (3.20)	1393 (96.80)
Prevent acute complications	Influenza vaccination can prevent acute complications of diabetes.	Yes	378	71 (18.78)	153 (40.48)	64 (16.93)	90 (23.81)	53 (14.02)	325 (85.98)
		No/not sure	1536	203 (13.22)	456 (29.69)	277 (18.03)	600 (39.06)	97 (6.32)	1439 (93.68)
Barriers
Unpreventable of influenza	I do not think influenza can be prevented.	Yes	880	68 (7.73)	250 (28.41)	148 (16.82)	414 (47.05)	10 (1.14)	870 (98.86)
		No/not sure	1034	206 (19.92)	359 (34.72)	193 (18.67)	276 (26.69)	140 (13.54)	894 (86.46)
Safety of vaccine	I was concerned about the safety of the influenza vaccine.	Yes	510	63 (12.35)	140 (27.45)	78 (15.29)	229 (44.90)	32 (6.27)	478 (93.73)
		No/not sure	1404	211 (15.03)	469 (33.40)	263 (18.73)	461 (32.83)	118 (9.05)	1286 (90.95)
Self-efficacy
Payment for vaccine	The influenza vaccine is covered by my/family’s insurance or free for me.	Yes	778	188 (24.16)	428 (55.01)	161 (20.69)	1 (0.13)	115 (14.78)	663 (85.22)
		No/not sure	1136	86 (7.57)	181 (15.93)	180 (15.85)	689 (60.65)	35 (3.08)	1101 (96.92)
Accessibility to vaccine	It is convenient to get the influenza vaccine.	Yes	1519	258 (16.98)	522 (34.36)	287 (18.89)	452 (29.76)	137 (9.02)	1382 (90.98)
		No/not sure	395	16 (4.05)	87 (22.03)	54 (13.67)	238 (60.25)	13 (3.29)	382 (96.71)
Health seeking behaviors	I would go to hospital If I develop symptoms of influenza-like illness, like fever, cough.	Yes	1077	156 (14.48)	385 (35.75)	201 (18.66)	335 (31.10)	100 (9.29)	977 (90.71)
		No/not sure	837	118 (14.10)	224 (26.76)	140 (16.73)	355 (42.41)	50 (5.97)	787 (94.03)
Cues to action
Doctor Recommendation	The healthcare provider recommends that I get vaccinated.	Yes	655	142 (21.68)	263 (40.15)	124 (18.93)	126 (19.24)	111 (16.95)	544 (83.05)
		No/not sure	1259	132 (10.48)	346 (27.48)	217 (17.24)	564 (44.80)	39 (3.10)	1220 (96.90)
Family/friends influence	My friends and/or family recommend that I get vaccinated.	Yes	301	72 (23.92)	121 (40.20)	30 (9.97)	78 (25.91)	41 (13.62)	260 (86.38)
		No/not sure	1613	202 (12.52)	488 (30.25)	311 (19.28)	612 (37.94)	109 (6.76)	1504 (93.24)
Media influence	I get influenza recommendation from the media and/or internet.	Yes	479	94 (19.62)	191 (39.87)	79 (16.49)	115 (24.01)	76 (15.87)	403 (84.13)
		No/not sure	1435	180 (12.54)	418 (29.13)	262 (18.26)	575 (40.07)	74 (5.16)	1361 (94.84)
Vaccination history	I have a history of influenza vaccination.	Yes	268	121 (45.15)	105 (39.18)	19 (7.09)	23 (8.58)	128 (47.76)	140 (52.24)
		No/not sure	1646	153 (9.30)	504 (30.62)	322 (19.56)	667 (40.52)	22 (1.34)	1624 (98.66)
Health services of family physician	I have purchased health services of community family physician.	Yes	1230	187 (15.20)	382 (31.06)	216 (17.56)	445 (36.18)	123 (10.00)	1107 (90.00)
		No/not sure	684	87 (12.72)	227 (33.19)	125 (18.27)	245 (35.82)	27 (3.95)	657 (96.05)

Data are reported as number of participants (%).

Motivators and barriers associated with influenza vaccination intention and uptake based on health belief model

1. Motivators and barriers associated with influenza vaccination intention

In unadjusted models, age, region, comorbidity, and HBM constructs which included perceived susceptibility, severity, benefits, self-efficacy and cues to action (except “health services of family physician”) were associated with the influenza vaccination intention (Table 4).

Table 4. Demographic characteristics and HBM constructs associated with intention and receipt of 2016–2017 seasonal influenza vaccine among diabetic patients in Ningbo city

		Intention to accept influenza vaccination^a				Receipt of influenza vaccination in 2016/2017^b
Items	Level	COR (95%CI)	p-Value	AOR (95%CI)	p-Value	COR (95%CI)	p-Value	AOR (95%CI)	p-Value
Age		1.01 (1.00–1.02)	0.003	1.00 (0.99–1.01)	0.489	1.03 (1.01–1.05)	0.001	1.04 (1.00–1.08)	0.026
Gender	Male	1.00	-	1.00	-	1.00	-	1.00	-
	Female	0.89 (0.75–1.05)	0.165	0.97 (0.80–1.20)	0.807	0.79 (0.56–1.10)	0.163	1.00 (0.56–1.78)	0.995
Monthly income (CNY)	<3000	1.00	-	1.00	-	1.00	-	1.00	-
	≥3000	0.94 (0.78–1.14)	0.540	0.76 (0.58–1.00)	0.047	1.15 (0.79–1.67)	0.481	0.39 (0.18–0.86)	0.020
Highest level of education attainment	Primary or below	1.00	-	1.00	-	1.00	-	1.00	-
	Secondary	1.11 (0.93–1.32)	0.253	0.84 (0.66–1.07)	0.160	1.09 (0.76–1.56)	0.651	0.56 (0.27–1.16)	0.118
	Tertiary or above	1.45 (0.83–2.53)	0.191	0.71 (0.35–1.45)	0.349	1.75 (0.67–4.57)	0.253	0.37 (0.06–2.29)	0.285
Occupation	Executives and professionals	1.00	-	1.00	-	1.00	-	1.00	-
	Clerks and workers	0.84 (0.58–1.22)	0.359	1.07 (0.69–1.65)	0.766	0.66 (0.37–1.17)	0.156	0.76 (0.27–2.18)	0.617
	Farmers	0.72 (0.51–1.02)	0.063	1.25 (0.79–1.98)	0.340	0.37 (0.21–0.65)	0.001	0.55 (0.17–1.82)	0.329
	Economically inactive	0.95 (0.65–1.39)	0.785	1.23 (0.76–1.99)	0.401	0.38 (0.19–0.73)	0.004	0.57 (0.16–2.00)	0.379
Marital status	Single/divorced/widowed	1.00	-	1.00	-	1.00	-	1.00	-
	Married	0.87 (0.69–1.10)	0.240	0.82 (0.62–1.08)	0.162	1.26 (0.75–2.13)	0.383	1.71 (0.72–4.04)	0.222
Region	Rural	1.00	-	1.00	-	1.00	-	1.00	-
	Urban	1.42 (1.20–1.68)	<0.001	1.02 (0.83–1.25)	0.839	16.84 (7.84–36.16)	<0.001	10.23 (4.01–26.06)	<0.001
Number of Comorbidities^c	0	1.00	-	1.00	-	1.00	-	1.00	-
	1	1.31 (1.09–1.57)	0.003	1.01 (0.82–1.25)	0.891	4.42 (2.55–7.65)	<0.001	4.66 (1.95–11.15)	0.001
	≥2	1.95 (1.47–2.59)	<0.001	1.23 (0.89–1.71)	0.217	4.90 (2.52–9.53)	<0.001	4.49 (1.52–13.22)	0.006
Health Belief Model Constructs
Perceived Susceptibility	Low	1.00	-	1.00	-	1.00	-	1.00	-
	Moderate	1.95 (1.59–2.38)	<0.001	2.09 (1.53–2.85)	<0.001	2.27 (1.48–3.47)	<0.001	3.56 (1.03–12.29)	0.044
	High	3.40 (2.77–4.18)	<0.001	3.57 (2.56–4.99)	<0.001	2.95 (1.96–4.44)	<0.001	4.87 (1.44–16.48)	0.011
Perceived Severity	Low	1.00	-	1.00	-	1.00	-	1.00	-
	Moderate	1.71 (1.39–2.11)	<0.001	2.19 (1.18–4.07)	0.013	1.36 (0.86–2.14)	0.188	0.06 (0.00–1.51)	0.087
	High	2.21 (1.82–2.69)	<0.001	1.50 (0.84–2.68)	0.174	2.12 (1.45–3.09)	<0.001	0.61 (0.04–9.83)	0.725
Perceived Benefits	Low	1.00	-	1.00	-	1.00	-	1.00	-
	Moderate	2.20 (1.82–2.66)	<0.001	1.70 (1.03–2.79)	0.037	5.70 (3.75–8.65)	<0.001	0.55 (0.18–1.67)	0.293
	High	3.52 (2.61–4.76)	<0.001	1.01 (0.09–10.74)	0.996	11.13 (6.86–18.07)	<0.001	14.39 (1.86–111.15)	0.011
Barriers
Unpreventable of influenza	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	0.42 (0.36–0.50)	<0.001	0.80 (0.64–1.00)	0.055	0.07 (0.04–0.14)	<0.001	0.21 (0.08–0.53)	0.001
Safety of vaccine	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	0.67 (0.55–0.80)	<0.001	0.21 (0.11–0.41)	<0.001	0.73 (0.49–1.09)	0.127	0.47 (0.25–0.89)	0.021
Self-efficacy
Payment for vaccine	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	13.25 (10.84–16.20)	<0.001	5.71 (3.95–8.26)	<0.001	5.46 (3.69–8.06)	<0.001	3.61 (1.94–6.71)	<0.001
Accessibility to vaccine	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	3.45 (2.78–4.29)	<0.001	2.66 (1.64–4.32)	<0.001	2.91 (1.63–5.20)	<0.001	0.25 (0.05–1.20)	0.082
Health seeking behaviors	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	1.44 (1.22–1.70)	<0.001	1.21 (1.00–1.46)	0.049	1.61 (1.13–2.29)	0.008	0.76 (0.29–1.96)	0.566
Cues to action
Doctor Recommendation	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	2.78 (2.33–3.31)	<0.001	1.72 (1.36–2.16)	<0.001	6.38 (4.37–9.32)	<0.001	0.38 (0.14–1.03)	0.057
Family/friends influence	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	2.15 (1.71–2.70)	<0.001	1.63 (1.24–2.14)	<0.001	2.18 (1.48–3.19)	<0.001	0.36 (0.09–1.46)	0.152
Media influence	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	1.97 (1.64–2.38)	<0.001	0.78 (0.61–0.99)	0.044	3.47 (2.47–4.87)	<0.001	1.24 (0.66–2.31)	0.498
Vaccination history	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	8.00 (6.17–10.36)	<0.001	3.88 (2.85–5.29)	<0.001	67.49 (41.59–109.53)	<0.001	67.08 (32.74–137.47)	<0.001
Health services of family physician	No/not sure	1.00	-	1.00	-	1.00	-	1.00	-
	Yes	1.04 (0.88–1.23)	0.669	0.24 (0.11–0.54)	0.001	2.70 (1.76–4.15)	<0.001	1.75 (0.89–3.47)	0.106

COR: Crude odds ratio; AOR: Adjusted odds ratio; CI: Confidence interval.

^aCOR and AOR were calculated by binary and multivariate ordered logistic regression models, respectively. In the multivariate ordered logistic regression analysis, the adjusted factors included demographic characteristics, other HBM constructs, and interactions between the HBM items.

^bCOR and AOR were calculated by binary and multivariate logistic regression models, respectively. In the multivariate logistic regression analysis, the adjusted factors included demographic characteristics, other HBM constructs, and interactions between the HBM items.

^cComorbidities included chronic respiratory disease, chronic heart disease, cerebrovascular disease, hypertension, and cancer.

In the adjusted model that controlled for demographic characteristics, other HBM constructs and interactions between the HBM items, other HBM constructs and interactions between the HBM items, perceived susceptibility (Moderate: OR 2.09, 95%CI 1.52–2.85; High: OR 3.57, 95%CI 2.56–4.99), perceived severity (Moderate: OR 2.19, 95%CI 1.18–4.07), perceived benefits (Moderate: OR 1.70, 95%CI 1.03–2.79), free access of vaccine (OR 5.71, 95%CI 3.95–8.26), convenient accessibility (OR 2.66, 95%CI 1.64–4.32), health-seeking behavior (OR 1.21, 95%CI 1.00–1.46), doctor recommendation (OR 1.72, 95%CI 1.36–2.16), family/friends’ influence (OR 1.63, 95%CI 1.24–2.14), and vaccination history (OR 3.88, 95%CI 2.85–5.29) were significantly associated with influenza vaccination intention as motivators. Monthly income (OR 0.76, 95%CI 0.58–1.00), worries about safety of vaccine (OR 0.21, 95%CI 0.11–0.41), media influence (OR 0.78, 95%CI 0.61–0.99), and purchasing health services of family physician (OR 0.24, 95%CI 0.11–0.54) were significantly associated with decreased intention of influenza vaccination. In addition, as summarized in Table 5, we also found interactions between perceived susceptibility and payment for vaccine (OR 1.88, 95%CI 1.14–3.11), perceived severity and perceived benefits (Moderate*High: OR 13.77, 95%CI 1.26–149.95; High*Moderate: OR 0.46, 95%CI 0.27–0.80), perceived severity and accessibility to vaccine (OR 0.39, 95%CI 0.19–0.83), perceived benefits and safety of vaccine (OR 3.43, 95%CI 1.59–7.42), and perceived benefits and purchasing health services of family physician (OR 4.26, 95%CI 1.73–10.53).

• (2)Motivators and barriers associated with actual influenza vaccine uptake

Table 5. Interactions between the HBM items associated with intention and receipt of 2016–2017 seasonal influenza vaccine among diabetic patients in Ningbo city

		Intention to accept influenza vaccination^a		Receipt of influenza vaccination in 2016/2017^b
Items	Level	AOR (95%CI)	p-Value	AOR (95%CI)	p-Value
Perceived Susceptibility*Payment for vaccine	Moderate*Yes	1.88 (1.14–3.11)	0.014	-	-
Perceived Susceptibility*Doctor Recommendation	Moderate*Yes	-	-	4.32 (1.06–17.60)	0.041
Perceived Severity*Perceived Benefits	Moderate*Moderate	-	-	8.08 (1.67–39.14)	0.009
	Moderate*High	13.77 (1.26–149.94)	0.031	-	-
	High* Moderate	0.46 (0.27–0.80)	0.006	13.82 (1.48–129.01)	0.021
Perceived Severity*Accessibility to vaccine	Moderate*Yes	0.39 (0.19–0.83)	0.015	-	-
Perceived Severity* Health seeking behaviors	Moderate*Yes	-	-	10.45 (2.22–49.06)	0.003
Perceived Benefits*Safety of vaccine	Moderate*Yes	3.43 (1.59–7.42)	0.002	-	-
Perceived Benefits*Family/friends influence	Moderate*Yes	-	-	8.89 (1.68–46.98)	0.010
Perceived Benefits*Health services of family physician	Moderate*Yes	4.26 (1.73–10.53)	0.002	-	-

AOR: Adjusted odds ratio; CI: Confidence interval.

^aAOR were calculated by the multivariate ordered logistic regression models, the adjusted factors included demographic characteristics, other HBM constructs, and interactions between the HBM items.

^bAOR were calculated by the multivariate logistic regression model, the adjusted factors included demographic characteristics, other HBM constructs, and interactions between the HBM items.

With regard to actual influenza vaccine uptake, occupation, region, comorbidity, and HBM constructs which included perceived susceptibility, severity, benefits, self-efficacy and cues to action (except worries about safety of vaccine) were associated with the influenza vaccine uptake in the unadjusted models (Table 4).

In the adjusted model that controlled for demographic characteristics, other HBM constructs and interactions between the HBM items, age (OR 1.04, 95%CI 1.00–1.08), urban residents (OR 10.23, 95%CI 4.01–26.06), comorbidity (one comorbidity: OR 4.66, 95%CI 1.95–11.15; at least two comorbidities: OR 4.49, 95%CI 1.52–13.22), perceived susceptibility (Moderate: OR 3.56, 95%CI 1.03–12.29; High: OR 4.86, 95%CI 1.44–16.48), perceived benefits (High: OR 14.39, 95%CI 1.86–111.15), free access of vaccine (OR 3.61, 95%CI 1.94–6.71), and vaccination history (OR 67.08, 95%CI 32.74–137.47) were significantly associated with increased influenza vaccination uptake. Monthly income (OR 0.39, 95%CI 0.18–0.86), perceptions of barriers including influenza, was unpreventable (OR 0.21, 95%CI 0.08–0.53) and worries about safety of vaccine (OR 0.47, 95%CI 0.25–0.89) were significantly associated with influenza vaccine uptake as barriers. In addition, there were interactions between perceived susceptibility and doctor recommendation (OR 4.32, 95%CI 1.06–17.60), perceived severity and perceived benefits (Moderate*Moderate: OR 8.08, 95%CI 1.67–39.14; High*Moderate: OR 13.82, 95%CI 1.48–129.01), perceived severity and healthcare-seeking behaviors (OR 10.45, 95%CI 2.22–49.06), perceived benefits and family/friends influence (OR 8.89, 95%CI 1.68–46.98), which had an effect modification on influenza vaccination.

Discussion

Seasonal influenza vaccination is generally recommended to people with chronic medical conditions by the WHO and the China CDC,^5,6 including diabetic patients. Large-scale seasonal influenza vaccination programs can potentially be cost-effective in reducing influenza-associated burden³⁷ in eastern China, where the influenza-associated excess respiratory mortality was higher due to high population density and high influx of people from other provinces for migrant work.³⁸ Vaccination coverage is the pivotal determinant for the impact of such programs. In spite of the large number of studies conducted to describe the influenza vaccination coverage rate in China, a few were about diabetic patients.²⁰ Our study demonstrated the intention and actual uptake of influenza vaccination among diabetic patient in eastern China and assessed the motivators and barriers associated with vaccination based on the HBM framework.

In line with previous studies conducted in other countries,^39,40 our study found sizable gap between self-reported influenza vaccination intentions and later vaccine uptake. In our study, despite nearly half of diabetic patients had the intention to get influenza vaccination, the actual vaccination coverage rate of diabetic patients in Ningbo remained very low (7.84%) compared to high-income countries where the coverage rates were estimated to be over 50%,^19,41 and even surpassed 75% (the target of coverage rate among diabetic patients by the WHO)⁴² in the recent years. Moreover, our estimated influenza coverage of diabetic patients was slightly lower than those of the general population but much higher than the chronic disease population in China reported by Wang Q et al.²⁰ that was opposite of most high-income countries where the influenza vaccination rates are higher in high-risk adults than non-risk adults.⁴³ In Ningbo city, the coverage rate varied by age, region, and comorbidity status though these measures did not have any significant effects on vaccination intentions in the fully adjusted multivariate model. The coverage rate increased with age that might be partially attributed to the general recommendation and extensive promotion of influenza vaccination among adults aged over 60.¹² Our study revealed that increased number of comorbidities might increase the likelihood of influenza vaccination as the coverage rate of diabetic patients with at least one comorbidity was four times higher than those had diabetes only which was consisted with other studies in other countries.^26,44 Compared to rural population, urban patients were more likely to get vaccinated against influenza. The following reasons may contribute to this. First, the travel distance to nearest vaccination clinic was shorter in the urban area where the vaccination clinics were more intensive⁴⁵ and distance to the nearest clinic was a significant factor increasing outpatient healthcare-seeking behavior of residents.⁴⁶ Second, the reimbursement policy and promotion program of influenza vaccine were distinguishing in different regions. In the urban region of Ningbo, there were government-funded influenza vaccination programs for the patients with chronic disease who purchased for community health services in Dongqianhu town and community-based intervention requiring community healthcare workers to recommend influenza vaccination during routine chronic disease follow-up visits for diabetic patients in Yinzhou and Jiangbei districts.¹³ In addition, unexpectedly, our study indicated that low income was not a barrier for influenza vaccination. On the contrary, the low-income group of diabetic patient was more likely to get vaccinated.

Our study indicated that the vaccination intention was strongly related to the actual vaccine uptake notwithstanding the significant gap between the two. The increasing level of vaccination intention was associated with the rising rate of vaccine uptake. With regards to the factors influencing the vaccination intention and uptake, our results support findings from previous studies and add some new insights. Consistent with other studies using the HBM constructs,^30,31 our study suggested that the perceived susceptibility of influenza and benefits of vaccination displayed significant positive associations with both vaccination intention and vaccine uptake, and had much greater effects on the latter. Perceiving the severity of influenza among diabetic patients was a motivator increasing the vaccination intention directly and it played an active role in increasing vaccination coverage by interaction with perceiving benefits of vaccination and having active healthcare-seeking behaviors. We observed the interactions of perceiving the severity of influenza and benefits of vaccination on both vaccination intention and actual uptake. And interestingly, our results indicated that the efforts of this interaction was positive on vaccination intention if the diabetic patient highly valued the benefits of vaccination and had a moderate level of disease severity perception, but the efforts became negative if the severity of influenza for diabetic patients was deeply perceived and the benefits of vaccination was not valued well.

As for the self-efficacy domain of HBM constructs, the payment of influenza vaccination was one of the most essential factors affecting both intention and later uptake of vaccine. In Ningbo, the cost of influenza vaccination was about ￥40–65 in the 2016–2017 season, which was affordable for general population. Hence, in our study, the income level was not an obstacle of influenza vaccination. However, providing free influenza vaccination was still a great motivation for sharply increasing the vaccination intention and uptake. 31.82% of the diabetic patients interviewed reported intention to be vaccinated if the vaccination could be funded by medical insurance, employer, or the government. Actually, 14.78% of the respondents who reported free access got vaccinated that was 4.80 times as likely as those who should pay for the vaccination by themselves. In Beijing, where provided free influenza vaccines to people aged ≥60 years, the free vaccination policy greatly increased the vaccine uptake rate of elder population.⁴⁷ Therefore, an organized free influenza vaccination program is needed in Ningbo to drive the coverage rate of in vulnerable populations higher. In addition, the healthcare-seeking behavior was another motivator for increasing both vaccination intention and uptake. We observed great efforts of interaction between perceiving the severity of influenza and healthcare-seeking behaviors on actual vaccine uptake. The diabetic patients who perceived the severity of influenza for themselves and had active healthcare-seeking behaviors were more likely to actually take action to get vaccinated.

In line with the cues to actions domain of HBM constructs reported in previous studies,^30,48 recommendations from healthcare workers and family/friends were significant motivators for influenza vaccination. And more remarkable, rather than working directly on the actual vaccine uptake, we found that the recommendations from healthcare workers and family/friends improved diabetic patients’ perceptions of influenza susceptibility and benefits of vaccination to raise the vaccination coverage rate of target population. Besides, although we did not observe the effect of media on actual vaccine uptake, our results indicated that media possibly had negative effects on diabetic patients’ vaccination intention, which should not be overlooked. In the recent years, as a number of news regarding the serious adverse events with vaccination were reported by media,⁴⁹ the public concerns and distrust about vaccine was increasing and the hesitancy and unwillingness of influenza vaccination consequently grew larger. In fact, influenza vaccination was generally safe according to the surveillance data for adverse events following immunization.⁵⁰ Efforts should be made to gain positive media exposure for influenza vaccination. In addition, the most significant predictor for influenza vaccine uptake in this study was having influenza vaccination history, which was also significant on intentions. 47.76% of the responders received influenza vaccine before got vaccinated in the 2016–2017 season, while only 1.34% of those who were without influenza vaccine history were vaccinated. This result suggested that targeting the population without influenza vaccination history as the main promotion target in order to expand the coverage of influenza vaccine.

Concerns about the safety of influenza vaccine were a common barrier reported in other studies.^20,28 Besides, our study found the perception that influenza was unpreventable was also a significant barrier to take action to get vaccinated.

The strengths and limitations

To our knowledge, this was the first study that explored the relationship between the intention of influenza vaccination and actual uptake among diabetic patients in China. The actual vaccine uptake was estimated more accurately using the vaccination records in Ningbo adult vaccination database compared to self-report vaccination from the survey in many other studies of China.^20,21 Moreover, unlike other studies using the HBM,^30,31,33,48 we added the interaction effects between the HBM items and the HBM constructs so that it could provide some insight for policy-makers and make us better understand how the motivators and barriers functioned as a causal chain. There are some limitations in this study. First, the sample may not adequately represent the real population of diabetic patients due to selection bias. We conducted the investigation among the diabetic patient attending the diabetes health management program in Ningbo and visiting community health center regularly. It was possible that diabetic patients who were elder and concerned about the health were more likely to meet the inclusion criteria. Therefore, we included the age and healthcare-seeking behaviors as covariates in the analysis model. Second, the investigation was conducted from November 2016 to January 2017 while the influenza vaccine was provided to the diabetic patients since September 2016, which may have resulted in biased estimates of initial intention before vaccination. However, our estimation of percentage of participants who had intention to be vaccinated was similar to another study conducted in China.⁵¹ Third, we used Yes/No or not sure instead of a Likert scale in the HBM constructs, which might not fully reflect the various levels of agreement or disagreement. Fourth, we used face-to-face interview in our study which may affect the reliability of the study and make us overestimate the intention of vaccination due to the interviewer effect. Therefore, in order to reduce the interviewer effect, we hired the general practitioners who were familiar with the participants owing to regular outpatient visits as the interviewers. The interviewers explained the questionnaire purpose and items to participants. If a participant claimed the intention to get vaccinated against influenza, appointment of vaccination would be made. And if a participant claimed the intention when the vaccination was free or covered by insurance, the interviewer would ask him/her to consider about influenza vaccination carefully after going home.

Conclusions

The intentions of influenza vaccination were not strong among the diabetic patients. Much less, the actual uptake was much lower due to sizable intention–behavior gap. To improve diabetic patients’ uptake of influenza vaccine, some efforts should be made: first, free influenza vaccination program funded by the government would be needed urgently for diabetic patients. Second, the messages conveyed to diabetic patients should emphasize the cause of influenza, how it is developed and the possible complications for diabetic patients, and the effectiveness and safety of vaccination against influenza. Third, vaccination campaigns should focus on the healthcare workers who could spread the knowledges and perceptions about influenza vaccine during routine diabetes follow-up visits, given the importance of the recommendations of these professionals on vaccine uptake. We should also gain help from news media to guide public voice speaking for influenza vaccination.

Authors’ contribution

Lixia Ye and Jun Cui designed the study with guidance from Ting Fang, Hui Li, Hongjun Dong, and Guozhang Xu. Lixia Ye, Jun Cui, Hui Li, Ting Fang, Rui Ma, Guanghui Zhu, Yexiang Sun, and Pingping Li coordinated the study and assisted the data collection. Lixia Ye, Jun Cui, and Ting Fang carried out data analysis and were responsible for writing draft of the manuscript. Yexiang Sun and Guozhang Xu helped revise drafts of the manuscript. All authors have read and approved the final manuscript.

Disclosure of potential conflicts of interest

The authors have no conflicts of interest to declare.

Acknowledgments

We would like to especially thank Jicheng Cao for his help and constructive suggestions during the drafting of this article. We thank all the study participants. We also thank those who worked on the study. In addition, we acknowledge Yumin Tao and Ying Dong for helping us solve the problems during the study.

Additional information

Funding

This work was supported by Ningbo Health Branding Subject Fund [ppxk2018–10]; Zhejiang Provincial Key Medical Discipline “Field Epidemiology” program [07–013]; The Science and Technology Planning Project of Ningbo [2017C50063]. The views expressed in this report are those of the authors and not necessarily the official position of the Centers for Disease Control and Prevention.