ABSTRACT
Preventive health workers rather than general practitioners (GPs) are the principal providers of vaccines in China, which may be a determinant of the unmet demand for vaccination, especially for adults, in recent years. GPs’ preferences had a significant influence on alternative approaches to adult vaccination delivery. To better understand GP’s preferences for adult vaccination services, we employed a discrete choice experiment with seven attributes: income, setting, information system, workshop, workload, performance measurement, and managerial support. Mixed logit models and latent class models were used for statistical analyses. In general, support from primary healthcare managers was the strongest driver of choice, followed by a 10% increase in workload, separate adult vaccination clinic, 5% increase in workload, and independent information system. Monthly income was significantly associated with provision of adult vaccination services. Based on the influence of latent factors, GPs fell into three classes that were correlated with GPs’ years of practice, workload, income satisfaction, and knowledge-attitude-practicescore. Classes 1 and 2 valued the service setting and performance measurement differently, while Class 3 valued the service setting only (preferred separate adult vaccination clinics to provide service). This study generated actionable information to guide innovation in the adult vaccination delivery system in China.
Introduction
Vaccine-preventable diseases among adults, such as influenza and pneumococcal infections, remain major public health problems. It is estimated that there are 291–646 thousand influenza-related deaths and 858–2184 thousand pneumococcal pneumonia-related deaths every year globally.Citation1,Citation2 This is mainly due to the poor vaccination coverage rate among adults. In most developed countries, adult vaccine uptake is below the target level. For example, influenza vaccination coverage among adults is approximately 40% and below 30% in the United States and the European Union, respectively, while the target rate of flu vaccination coverage is 75%.Citation3,Citation4 In developing countries such as China, this coverage rate is even worse, with the overall coverage rate at 1.9%.Citation5
Along with individual choices, the mode of vaccine delivery is a key determinant of vaccine uptake. In developed countries, adult vaccinations are traditionally administered by general practitioners (GPs), physicians, and nurses in primary healthcare (PHC). Recently, some non-traditional settings (e.g., pharmacies, grocery stores, and churches) have been engaged in vaccination service delivery because they are more accessible and convenient than a physician’s office.Citation6,Citation7
In China, the vaccination delivery system differs from that of other countries.Citation8 There are two categories of vaccines: Category A (part of the National Expanded Program on Immunization, EPI) and Category B (not included within the EPI). Category A vaccines are mandatorily provided to children with no charge, and Category B vaccines are usually delivered on request with a user fee. All adult vaccines are Category B vaccines. Although PHC providers are the primary administrators of vaccinations, most of GPs do not participate in vaccination services. There are two types of healthcare workers in PHC settings: GPs and preventive health workers (PHWs). Most of PHC settings are public sectors, and both of PHWs and GPs in it are government-employed worker. GPs and PHWs work in two independent departments (clinical care and public health) within the same PHC setting. These two independent departments are typically located in different buildings or on separate floors in the same building. GPs provide basic medical and public health services, without vaccination services. In China, it is suggested (not mandatory) that all citizens sign with a GP or GP team. Usually, there is no incentive or disincentive for GPs to recommend vaccinations to patients. PHWs oversee public health services, including vaccinations, at the regional level. Previous research has reported that the workload of PHWs is too high to meet the increasing need for vaccination.Citation8
This delivery system has been designed to ensure the implementation of the EPI and has achieved great success in childhood vaccination in the past few decades in China. However, this vaccination service delivery system is insufficient to meet the vaccination demands.Citation9 First, the number of vaccines included in the EPI has increased from four to eleven. Second, some newly available vaccines that were not originally included in the EPI are now recommended, such as human papillomavirus (HPV) and herpes zoster. Third, awareness of both childhood and adult vaccination has continued to improve in recent years in China. Fourth, the pandemic of emerging infectious diseases (e.g., SARS, Ebola, and COVID-19) has increased the demand for vaccines.
Alternative approaches to adult vaccination delivery, such as utilizing GPs, may help to address the gap between demand and supply in China. As health systems grant GPs broader roles and responsibilities, we need to better understand the preferences of GPs for delivering adult vaccinations. A discrete choice experiment (DCE) is widely used in market research to gather information on individuals’ stated preferences by asking them to state their choice over different hypothetical goods or services that are described by selected characteristics (or attributes).Citation10 By presenting a straightforward task, a DCE is likely to more closely resemble a decision made in the real world. A growing number of studies are now using this approach to address health program decisions, including the job preferences of health personnel.Citation11–13 The attributes identified in previous research on health service delivery included four aspects: challenges, need for additional education, compensation, and implementation.Citation14
It is equally appropriate to use a DCE to evaluate preferences for delivering adult vaccination services. This study employed a DCE to reveal the preferences of GPs in delivering adult vaccination services and inter-individual preference heterogeneity in Beijing, China.
Methods
Discrete choice experiments
DCEs were used to measure the preference of GPs for delivering adult vaccination. Relevant attributes and levels were retrieved from the literature review and interviews with an expert group.Citation14 Seven key attributes were identified, each of which had up to three levels. displays these attributes and levels in detail.
Table 1. Attributes and levels.
Combining the seven attributes with each level resulted in 288 possible service alternatives. To reduce the number of choice sets and the burden on respondents, we used an orthogonal experimental design to create the DCE, and we divided the 36 choice sets into three different versions. Each version included 12 choice tasks.Citation15 On each questionnaire, respondents were asked to choose which type of adult vaccination service alternatives they would prefer. Opt-out possibility was also included.
Non-DCE variables
We also collected information on personal factors that may affect the preference for providing vaccination services, including sex, age, years of practice, perceived workload, perceived income satisfaction, and knowledge-attitude-practice (KAP) regarding the six vaccines with the highest uptake rate among adults (influenza, pneumococcal, HPV, herpes zoster, hepatitis B, and COVID-19). For each vaccine type, six questions were used to assess the KAP score. The knowledge part included four questions: “To what extent do you know about immunization schedules, recommendation population, contraindications, and adverse reactions?” Attitude and practice were measured by “To what extent do you agree that vaccine can prevent disease?” and “To what extent do you recommend vaccines to the patient during general practice?,” respectively. All questions on KAP were ranked from 1 to 5 on a Likert scale. The total KAP score was calculated as the mean score of knowledge, adding up each item score of attitude and practice (range: 3–15).
Sampling and participants
The sample was obtained through stratified multistage probability-proportional-to-size (PPS) sampling. Our sample used primary sampling units (PSUs) in PHC settings. First, all PHC settings were sorted (stratified) by functional region (Capital Core, Urban Developed, Urban New, and Ecological Regions). After this sorting (stratification), the number of GPs in each PHC setting and the cumulative number were listed. According to Johnson and Orme,Citation16,Citation17 the minimum sample size was 125. A sample size of 208 GPs was deemed sufficient based on an expected response rate of 60%. The samples were divided into functional regions based on the total number of GPs in each region. The PHC settings were selected for each region using PPS sampling. Participants included GPs working in selected communities in Beijing with at least one year of work experience.
Statistical analyses
Descriptive statistics were used to analyze the characteristics of the responders. Both the mixed logit model (MXL, or random parameter logit model) and latent class model (LCM, or finite mixture analysis) were used to estimate GPs’ preferences and their heterogeneity. Both MXL and LCM were not necessary, considering the assumptions of independence of irrelevant alternatives (IIAs), and provided information on heterogeneity.Citation18 MXL assumes a continuous distribution for each preference and determines how the distribution of heterogeneity is distributed.Citation19 The LCM assumes that there are a finite number of discrete preference subgroups which was done by the software with its algorithms, and is informed by heterogeneity among the latent segments of GPs.Citation19 The mixed logit model is specified as a normally distributed parameter with 500 Halton draws. Based on the Bayesian information criteria (BIC) and Akaike information criteria (AIC), the LCM was specified with three classes, which had significantly better BIC and AIC than the two-class, four-class, and five-class models. For both MXL and LCM, the marginal willingness-to-pay (WTP, the monetary valuation of each attribute) was calculated using the ratios between the coefficients of monthly income and the coefficients of the other attributes.Citation15
Results
Characteristics of participants
A total of 210 GPs from 17 PHC settings were invited to participate. Among them, 165 completed the survey, with a response rate of 78.6% (165/210). As shown in , 69.7% of the respondents were female. The average age of the sample was 40.6 years old. Approximately 80% had more than ten years of practice experience, half (49%) reported a heavy workload, and 77.6% felt satisfied with their income. The average KAP score was 11.41 out of 15.
Table 2. Participant characteristics.
Estimation results
A total of 1980 choice sets were included in the estimation (165 respondents, 12 choice sets each). The opt-out option was chosen for 26.3% of the choice sets. presents the regression results for the MXL and LCM. Using the parameters estimated from MXL and CLM, we can elicit the preference weights.
Table 3. Mixed logit model and latent class logit model of respondent preferences.
The positive sign of the income attribute means that the higher the income, the more likely a GP was to choose that alternative. An alternative with a separate vaccination clinic (0.4), independent information system (0.2), and support by managers (0.8) increased the likelihood of choosing an option. The coefficients for the workshop attributes and performance measurements were not significant. As expected, the higher the workload, the less likely a GP was to choose an alternative. The attribute with the greatest magnitude of association with vaccination service preference was support by managers, and greater workload increase reduced preference. Support by managers increased preference, and a greater workload increased reduced preference.
The results of the CLM are presented in the fourth to sixth columns of . Class 1 was more likely to be driven by higher income and included in the performance measurements, whereas Class 2 showed less preference for these attributes. In Class 3, all attributes were insignificant, except for the setting attributes. We labeled our classes accordingly as the external drive class, internal drive class, and lack of drive class. The external drive class had an average membership probability of 34.7%, internal drive class of 42.3%, and lack of drive class of 23.1%. While we looked at the individual characteristics that determine the membership of classes, we found that the number of practice years, reported light workload, satisfaction with income, and KAP score had the greatest impact. GPs with 10–19 years of practice were more likely than those with less than 10 years of experience to be in the external drive class than the internal drive class. Respondents who reported a light workload and higher KAP scores were more likely to be in the external drive class. GPs who were satisfied with their incomes were more likely to be in the internal drive class.
Willingness to pay
shows the monetary valuations (WTP) for each attribute. In the mixed logit model, respondents were willing to forgo 1.8%, 3.8%, and 6.8% of their current monthly income to provide adult vaccination services with an independent information system, in separate vaccination clinics, and supported by managers. GPs would need to be compensated by 2.7% and 5.3% of their current monthly income if the workload increased by 5% and 10%, respectively, owing to the adult vaccination service. The WTP for the frequency of the workshop and the performance measurement were insignificantly different from zero. In the latent class model, the WTP estimations for each attribute between the three classes were different. The WTP for setting and performance measurements between the external and internal drive classes were opposite. In the lack of drive class, the WTP for all attributes was insignificantly different from zero.
Table 4. Estimated monetary values of aspects of providing vaccination services (% of monthly income), mean (95% CI).
Discussion
In this study, we examined GPs’ preferences and heterogeneity. In general, support from PHC managers was the strongest driver of choice, followed by a separate adult vaccination clinic, and an independent information system, while a 10% and 5% increase in workload were the main barriers to choice. Based on the influence of latent factors, GPs fell into three classes: external drive, internal drive, and lack of drive. The external drive class and internal drive class valued the setting of service and performance measurement differently, while the lack of drive class valued the setting of service only (prefer adult vaccination clinic to provide service).
Our respondents valued support from managers over other attributes, and its monetary value was approximately 7% of their current monthly income. In China, official support from institutions is usually important for programs that make changes in policies and practices concerning PHC.Citation20 In a qualitative case study, Wei et al. (2008) stated that the local manager’s leadership role is critical in generic tuberculosis control programs.Citation21 In a study on pharmacists’ preferences for providing patient-centered services, pharmacists also expressed concerns that they would not receive adequate support from managers or owners.Citation14
The magnitude of the increase in workload was the second most important attribute in the preference for adult vaccination services. Generally, as workload increases, utility declines linearly. A 10% increase in workload is worth 5.3% of the current monthly income. It is reasonable that GPs would worry about the increase in workload because they already have a heavy workload in China.Citation8,Citation22 Nevertheless, in the latent class analysis, we found that respondents with a light workload were more sensitive to changes in the workload. The workload may not seem to increase if GPs provide vaccination services when they interact with patients. Before implementing the policy, policymakers should estimate the workload of delivering routine vaccinations in GP practice.
Respondents’ attitudes toward settings and performance measurements were the opposite. Approximately 42% of respondents preferred providing services in a separate adult vaccination clinic and were included in the performance measurement, while other GPs preferred to provide services in a medical clinic and were not included in the performance measurement. As a result of these opposing perspectives, policymakers should consider trade-offs or offer a menu of programs from which GPs can choose which ones they wish to participate.
Generally, GPs preferred separate information systems. This corresponds to the actual situation with the vaccination information system in China. Vaccination data were recorded with unique identifiers and uploaded in real time to the web-based management platform through separate client application software.Citation23 Therefore, policymakers were not required to consider developing new information systems for vaccination data.
Nearly 80% of respondents preferred services with a higher monthly income. Previous research found that Chinese health providers were less satisfied with income, which was important for recruiting and retaining healthy human resources for PHC.Citation13,Citation24 Although the provider can charge a service fee of approximately 25 RMB for vaccination services, it is too low to provide a sufficient incentive to provide vaccination services.Citation25 Appropriate subsidies or financial incentives may improve the willingness of GPs to provide adult vaccination services.
The LCM captured the heterogeneity in preferences for specific attributes, and thus helped us identify the external, internal, and lack of drive classes. The external drive class was driven by higher income and included performance measurements with an average membership probability of 34.7%. This implies that policymakers can use financial incentives and performance assessments to attract external drive-class GPs in the earlier period of policy. In addition, the heterogeneity in the workplace between the external and internal drive classes highlights that a flexible form of service is a very important consideration in the policy-making process.
This study had several limitations. First, we selected the most relevant attributes based on interviews with vaccination experts, who were unable to cover all aspects of vaccination services. Second, the choice of the opt-out option was slightly high, which may reduce the statistical efficiency. Third, since this study surveyed GPs in Beijing with a relatively small sample size, the generalizability of our findings requires further investigation. Despite these limitations, we believe that this is the first study of GPs’ preference for delivering adult vaccination, which provides new information on GPs’ preferences and WTP for providing adult vaccination services.
Conclusion
This study reveals that PHC managers’ support was the most important attribute in GPs’ preferences for adult vaccination services, followed by an increase in workload, providing services in a separate adult vaccination clinic, and with a separate information system. Monthly income was significantly associated with the provision of adult vaccination services. In addition, we found evidence of preference heterogeneity, which was correlated with years of practice, workload, income satisfaction, and KAP scores of GPs.
Author contributions
ML, DZ, and PH designed the study, performed the statistical analysis, and drafted the manuscript. WL contributed to conducting the study, collecting research data, and preparing the papers. YS, JW, and XW interpreted the data and revised the manuscript. All authors have read and approved the final manuscript.
Ethics approval
This project was approved by the Peking University Health Science Center Institutional Review Board (IRB00001052–20062).
Disclosure statement
No potential conflict of interest was reported by the authors.
Data availability statement
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Additional information
Funding
References
- Iuliano AD, Roguski KM, Chang HH, Muscatello DJ, Palekar R, Tempia S, Cohen C, Gran JM, Schanzer D, Cowling BJ, et al. Estimates of global seasonal influenza-associated respiratory mortality: a modelling study. The Lancet. 2018;391:(10127):1285–6. doi:10.1016/S0140-6736(17)33293-2.
- Wang H, Naghavi M, Allen C, Barber RM, Bhutta ZA, Carter A, Casey DC, Charlson FJ, Chen AZ, Coates MM, et al. Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980–2015: a systematic analysis for the global burden of disease study 2015. The Lancet. 2016;388:(10053):1459–544. doi:10.1016/S0140-6736(16)31012-1.
- USCDC. Flu vaccination coverage, United States, 2016-17 influenza season. [accessed 2022 Feb 20]. https://www.cdc.gov/flu/fluvaxview/coverage-1617estimates.htm.
- Endrich MM, Blank PR, Szucs TD. Influenza vaccination uptake and socioeconomic determinants in 11 European countries. Vaccine. 2009;27(30):4018–24. doi:10.1016/j.vaccine.2009.04.029.
- Yang J, Atkins KE, Feng L, Pang M, Zheng Y, Liu X, Cowling BJ, Yu H. Seasonal influenza vaccination in China: landscape of diverse regional reimbursement policy, and budget impact analysis. Vaccine. 2016;34(47):5724–35. doi:10.1016/j.vaccine.2016.10.013.
- Burson RC, Buttenheim AM, Armstrong A, Feemster KA. Community pharmacies as sites of adult vaccination: a systematic review. Hum Vaccines Immun. 2016;12(12):3146–59. doi:10.1080/21645515.2016.1215393.
- Uscher-Pines L, Harris KM, Burns RM, Mehrotra A. The growth of retail clinics in vaccination delivery in the US. Am J Prev Med. 2012;43(1):63–66. doi:10.1016/j.amepre.2012.02.024.
- Yu M, Yao X, Liu G, Wu J, Lv M, Pang Y, Xie Z, Huang Y. Barriers and facilitators to uptake and promotion of influenza vaccination among health care workers in the community in Beijing, China: a qualitative study. Vaccine. 2022 Mar;40(14):2202–08. S0264-410X(22)00207-9. doi:10.1016/j.vaccine.2022.02.060.
- Lei H, Cao L. Development and application of patterns of vaccination services in China: a literature review. Chin J vaccines Immun. 2021;27:485–90.
- Ryan M. Discrete choice experiments in health care. BMJ. 2004;328(7436):360–61. doi:10.1136/bmj.328.7436.360.
- Mangham LJ, Hanson K, McPake B. How to do (or not to do) … designing a discrete choice experiment for application in a low-income country. Health Policy Plan. 2009;24(2):151–58. doi:10.1093/heapol/czn047.
- Guo N, Zhang G, Zhu D, Wang J, Shi L. The effects of convenience and quality on the demand for vaccination: results from a discrete choice experiment. Vaccine. 2017;35(21):2848–54. doi:10.1016/j.vaccine.2017.04.006.
- Song K, Scott A, Sivey P, Meng Q. Improving Chinese primary care providers’ recruitment and retention: a discrete choice experiment. Health Policy Plan. 2015;30(1):68–77. doi:10.1093/heapol/czt098.
- Grindrod KA, Marra CA, Colley L, Tsuyuki RT, Lynd LD. Pharmacists’ preferences for providing patient-centered services: a discrete choice experiment to guide health policy. Ann Pharmacother. 2010;44(10):1554–64. doi:10.1345/aph.1P228.
- Hensher DA, Rose JM, Greene WH. Applied choice analysis: a primer. Cambridge (NY): Cambridge University Press; 2005.
- Johnson R, Orme B. Getting the most from CBC. Sequim: sawtooth Software Research Paper Series. Sawtooth Software; 2003.
- de Bekker-Grob EW, Donkers B, Jonker MF, Stolk EA. Sample size requirements for discrete-choice experiments in healthcare: a practical guide. Patient–Patient–Centered Outcomes Res. 2015;8(5):373–84. doi:10.1007/s40271-015-0118-z.
- Greene WH, Hensher DA. A latent class model for discrete choice analysis: contrasts with mixed logit. Transp Res Part B Methodol. 2003;37(8):681–98. doi:10.1016/S0191-2615(02)00046-2.
- Boeri M, Saure D, Schacht A, Riedl E, Hauber B. Modeling heterogeneity in patients’ preferences for psoriasis treatments in a multicountry study: a comparison between random-parameters logit and latent class approaches. Pharmaco Econ. 2020;38(6):593–606. doi:10.1007/s40273-020-00894-7.
- Huang W, Long H, Li J, Tao S, Zheng P, Tang S, Abdullah AS. Delivery of public health services by community health workers (CHWs) in primary health care settings in China: a systematic review (1996–2016). Glob Health Res Policy. 2018;3(1):18. doi:10.1186/s41256-018-0072-0.
- Wei X, Walley JD, Liang X, Liu F, Zhang X, Li R. Adapting a generic tuberculosis control operational guideline and scaling it up in China: a qualitative case study. BMC Public Health. 2008;8(1):260. doi:10.1186/1471-2458-8-260.
- Tao T, Zhao Q, Jiang S, Ma L, Wan L, Ma Y, Xu B. Motivating health workers for the provision of directly observed treatment to TB patients in rural China: does cash incentive work? A qualitative study. Int J Health Plann Manage. 2013;28(4):e310–24. doi:10.1002/hpm.2175.
- Wu W, Cao L, Zheng J, Cao L, Cui J, Xiao Q. Immunization information system status in China, 2017. Vaccine. 2019;37(43):6268–70. doi:10.1016/j.vaccine.2019.08.070.
- Shi L, Song K, Rane S, Sun X, Li H, Meng Q. Factors associated with job satisfaction by Chinese primary care providers. Prim Health Care Res Dev. 2014;15(1):46–57. doi:10.1017/S1463423612000692.
- Beijing Price Monitoring Center. Notice of the Beijing municipal development and reform commission, the Beijing municipal health and family planning commission and the Beijing municipal human resources and social security bureau on setting the price of the second type of vaccination service for Beijing municipality (Beijing municipal development and reform [2018] no.362). Published 2018. [accessed 2022 May 18]. http://www.beijingprice.cn/c/2018-03-06/502119.shtml.