Study design and sampling
China and Vietnam were chosen as study countries because they have similar social economic background and are facing similar challenges in financing their health system. Interestingly, they have however chosen different health insurance systems which may have different implications on equity in health care.
Two provinces from each country (Shandong and Ningxia from China, Hai Duong and Bac Giang from Vietnam) were selected for this study. The selection of provinces was based on three criteria: 1) one province represented relatively developed areas within the country (Shangdong and Hai Duong), and the other one represented less developed areas (Ningxia and Bac Giang); 2) all study sites had established a rural health insurance system; 3) local governments were capable and willing to cooperate with the study. In each province, 2 districts from Vietnam and 3 counties from China were selected for the study using similar selection criteria as those of provinces.
The study used a survey and qualitative methods to collect and analyze data. Data from the household survey was used to analyze and compare the utilization of health services between health insurance members and non-members in different income groups. This was then triangulated and explained through focus group discussions (FGD) with health insurance members and in-depth interviews with health insurance managers, local government leaders and administrators.
Sample size for household survey was calculated based on estimation and comparison of utilization rate of inpatient services between different economic groups. In China, a sample size of 22,008 individuals (11,004 per province) was estimated based on inpatient utilization rate in low income group (3.3%) and high income group (4.2%) from 2003 National Health Service Survey. In Vietnam, a smaller sample size of 7518 individuals was estimated based on an expected inpatient utilization rate of 5.2%, also with its smaller population size in consideration.
Multistage sampling processes were used in household survey. In China, we selected 3 townships from each county, 3 villages from each township, and then a systematic random sample of 100 households in each village were selected based on the household registration. Similar processes were used in Vietnam: 4 communes from each district, 3 villages from each commune, and a systematic random sample of 50 households from each village were selected from a village household registration list. In total, 6,147 households (22,636 individuals) in China and 2,397 households (8,983 individuals) in Vietnam were interviewed.
Purposive samples were selected for FGDs and in-depth interviews. Gender, health insurance membership, and location were taken into consideration when selecting the respondents to capture a wide range of experiences and views. In total, in China 26 FGDs were held with male and female NCMS members and non-members, 89 in-depth interviews with patients with catastrophic medical expenditure, health providers, health insurance policy makers and managers. In Vietnam, 26 FGDs were convened with members of different health insurance schemes and non-members. Sixteen in-depth interviews were conducted with health managers and health insurance managers.
Data collection
Data collection was conducted from May to July 2006 by researchers from both countries. Standard structured questionnaires were developed in English for discussion and to ensure the maximum possible comparability between the two countries, and then translated into local languages for data collection. Questions related to this paper included: demographic information on individuals and households, rural health insurance membership, reported health service utilization (of outpatient service by those reporting illness in the last 4 weeks, and inpatient service in the last year). After receiving training, school teachers in Vietnam and postgraduate students in China acted as interviewers to visit the selected households and conduct the interviews. Completed questionnaires were carefully checked by quality supervisors immediately after the interview for quality assurance.
For qualitative study, semi-structured topic guides for FGDs and in-depth interviews were developed in a similar process as the questionnaire. Senior qualitative researchers from partner institutions acted as facilitators and interviewers. The interviews with health managers were conducted to explore rural health insurance policies, design and implementation processes. Selected rural residents were asked questions about factors affecting their health seeking behaviour, their perceptions and experiences of health insurance, and the reasons of choosing their health insurance membership.
Data analysis
The quantitative data analysis focuses on a comparison of outpatient and inpatient service utilization between health insurance members and non-members in different income groups. In this study, outpatient service utilization rate is defined as the number of people who used outpatient services in the last 4 weeks as a percentage of total number of individual interviewees. Inpatient service utilization rate is defined as the percentage of total respondents who were hospitalized in the last 12 months. In calculating these two indicators, the numerators were number of patients, rather than number of service episodes, that is, even if a patient had more than one outpatient visits or hospitalization episodes, s/he was only counted once in the analysis. Reported household annual income per capita was used as a proxy for socio-economic status. Income groups were defined based on the reported household annual income per capita. From the lowest to the highest annual income per capita in each country, we divided the whole country sample into three equally sized groups: low income group, middle income group and high income group. Analysis was conducted using SPSS 14.0 and Stata 8.0.
All interviews were tape recorded with the permission of participants and were then transcribed and word processed and entered into MaxQDA. The 'framework' approach was used to analyze the qualitative data [
15]. Researchers read through the transcriptions and listed the recurring viewpoints and the common themes from the data. This formed the basis of the thematic framework. Each segment of the text was categorised and coded using this framework. Segments relevant to each theme were then charted to identify majority and minority views, enabling interpretation and explanation. Findings relevant to views and experiences of health insurance and reasons for service utilization and non-utilization were compared among different sampled groups.