Discrete choice experiments (DCEs) are a quantitative method aiming at eliciting stated preferences. This method draws on Lancaster’ consumer theory which assumes health-care interventions and policies are the combinations of attributes, and individuals’ choice on these goods is based on various levels of attributes[28]. The DCE model has been widely used to predict the probability of take-up of various contract service plans and to determine preferences for goods services in lieu of observations on real-world market interactions.
DCE Questionnaire Design
To select representative attributes that could clearly depict and capture resident preferences for family doctor teams under the FDCS, we developed a DCE questionnaire through qualitative methods, which included a literature review as well as interviews with key informants . We first reviewed international and domestic literatures on primary health care providers and patients' choices of doctors to identify which attributes were highly relevant to our study. A pilot study recruited 3 rural residents who have signed FDCS, 2 village doctors and 2 township hospital managers from Zhangqiu county located at central of Shandong province, to encourage them to share views on 1) how has family doctor service mechanism been implemented; 2) influencing factors for residents considered to sign family doctors with contract services; 3) hinder factors which deter awareness and acceptance of this service scheme; 4) policy recommendations to increase family doctor registration rate. Using semi-structured interview, we collected data about what factors influence rural residents most when they sign family doctors. A DCE workshop with 2 DCE experts was also conducted in March 23-25, 2018. The DCE experts gave valuable suggestions on attributes description, determine the levels for each attributes, and experiment designs. Combined with literature review results and the common suggestions raised by FGD participants, five determinants which impact rural residents’ decision making most were selected.
The five attributes of FDCS contracts described below were determined to be most relevant to uptake in our setting. A full description of the attribute selection and questionnaire implementation process is available in the Appendix.
(1) Contract Price: This attribute refers to the annual signing expenses for an individual resident. After we reviewed public policies and guidelines on FDCS enacted by central and local governments, three levels were specified for this attribute: 0CNY, 100CNY and 200 CNY per year [29, 30].
(2) Availability of medicines: Medicine availability refers to the ability to obtain affordable medicines necessary to maintain one's health[31]. We selected this attribute to indicate the accessibility of health services provided by the contracted family doctor We divide this attribute into two levels in our questionnaire: shortage and sufficient.
(3) Insurance reimbursement rate: While health insurance was recently universalized in China, insurance reimbursement rates vary by plan and scheme. Previous studies have shown that a close relationship between medical insurance and patient choice of medical treatment[32]. Referring to reimbursement guidelines from the Shandong health commission, we divide this attribute into three levels in our questionnaire: standard reimbursement, 5% more than standard, and 10% more than the standard reimbursement rate.
(4) Competence of the family doctor. The competence and skill of the physician is considered of great importance to patients[25, 33, 34]. This attribute refers to the resident’s attention to physician credentials and perceived competence when selecting a family doctor. We divide this attribute into three levels in our questionnaire: low, medium and high.
(5) Attitude of the family doctor. Many studies have shown a correlation between doctors' attitudes and patients' medical behaviors[35-37]. Thus, we sought to investigate the relative importance of perceived attitude in the decision to sign a family doctor team. Three levels were divided in the research: poor, normal and good.
Consistency test was performed to ensure each respondent were making realistic trade-offs and checking validity of this research. In this study, one repeated choice set question was added in each version of the questionnaire to check preference consistency of each respondent. We excluded the information for respondent who failed the consistency test.
Data Collection
This study was conducted in Shandong province, the second largest province in China. Within Shandong, 3 cities—Binzhou, Zibo, and Liaocheng—located in the northeast, central, and west regions of the province, respectively, were selected as study sites. Multi-stage random sampling was used to choose a sample of respondents representative of rural residents in each selected city. To do so, 2 counties in each city were first chosen at random. Within each county, 5 townships (the administrative level below the county) and 24 households in each townships were chosen randomly. In this study, the questionnaire was administered to 720 residents aged 18 and above, which is higher than the 600 observations recommended as sufficient for preference heterogeneity analysis[38] . For the 720 questionnaires , 20 of them were incomplete, then we dropped these ineligible surveys. There were 91 surveys failed to pass the consistency test in the questionnaire and were excluded. Finally, 609questionnaires were included in the statistical analysis..
Data was collected in this study through a DCE questionnaire administered by teams of trained enumerators at study households. Since most of the respondents were low educated, face-to-face interview method was applied to ensure each respondent clearly understand the whole survey. At the beginning of each interview, enumerators described the purpose of the study and sought participant consent. Following consent, a brief introduction of the FDCS, recent public health policies launched by the government, and attributes in each choice set were explained. Then, a one-page introduction of the task with warm-up choice question was followed to check if the respondent could fully understand the questionnaire and make tradeoff in each pair-wise choice set. Each participant was asked to imagine different hypothetical scenarios in which different family doctor contract service plans are registered to enhance their health status. They were then asked to make discrete choices between 10 pair-wise combinations of scenarios. On average, it took around 50 minutes to complete the whole questionnaire and the survey returned to interviewer immediately. A sample questionnaire choice is shown in Table 1.
Table 1 An example of a DCE question.
Attributes
|
Contract plan 1
|
Contract plan 2
|
Cost of the contract
|
200CNY/year
|
100CNY/year
|
Availability of medicine
|
Easy
|
Difficult
|
Reimbursement rate
|
Standard
|
10% more
|
Competence of family doctor
|
Medium
|
Low
|
Attitude of family doctor
|
Good
|
Normal
|
Which contract plan would you choose?
|
□
|
□
|
Pleases consider you are going to enroll in the contract service of family doctor for yourself. Of the following two contract plans, which contract plan would you choose?
Statistical Analysis
Data were first double-entered and coded using Epidata version 3.1, and the final data was then transferred to STATA 14.2 for all statistical analyses.
Random utility theory provided the theoretical foundation for the analysis of DCEs data[39]. Mixed logit models were used to estimate the utility of registering with one contract plan[39].We assumed that respondents were relatively homogenous on demographic measures, hence their preference would be associated with choice variables. The utility function as specified as follows: (See Equation 1 in Supplementary Files)
All attributes were dummy coded except for the costs of the contract, which was specified as a continuous variable to facilitate the calculation of willingness to pay (WTP).WTP was calculated to measure the trade-offs among various contract attributes. WTP was estimated as the ratio of the coefficient to the negative of the coefficient on the contract cost attribute. The coefficients indicated the relative importance of the worst values for the categorical variables.