The infrastructure of a demographic surveillance site in a district in northern Vietnam has been used for this study. Details of the district and the surveillance site are presented below. The sampling procedures and other methodological issues are also discussed.
The setting and the interviews
This study was conducted in the Bavi district of Ha Tay province in Vietnam. Bavi district is situated in the northwest part of Vietnam, about 60 km west of Hanoi. It has a population of 235,000. There are three major ethnic groups in the district: Kinh (91%), Muong (8%) and Dao (1%). There are also some families of the Tay, Hoa and Khmer tribal groups. The district is divided into 32 communes including one small town. Farming is the predominant occupation. The average annual income in terms of rice production was about 290 kilograms/person/year in 1999, which corresponds to 50 USD per capita per year [13, 14].
There is a demographic surveillance site in Bavi: The Epidemiological Field Laboratory for the Health Systems Research Project (FilaBavi) in Vietnam. FilaBavi is a joint project between Hanoi Medical University, Karolinska Institute, University of Umeå, and the Nordic School of Public Health in Sweden. The aims of the project are to implement a longitudinal epidemiological surveillance system to generate basic health and health care data, to supply information for health planning, serve as a background and a sampling frame for specific studies (especially intervention studies), and to constitute a setting for epidemiological training of research students. In 1999 a baseline household survey was conducted followed by quarterly surveillance of vital events and complete re-surveys every two years [14].
The infrastructure of FilaBavi was utilised for the study presented in this paper. The total FilaBavi sample consists of 11,089 households. The households were selected using a multistage sampling procedure. At the first stage, 67 population clusters were selected using probability proportional to size. These clusters had 11,089 households and 51,024 individuals [14]. Assuming α level of 5% and 50% probability that a household will have an episode of illness in a year, the required sample size becomes 576. To ensure adequate sample size, one out of every 18 households was randomly selected from the original sample for the purpose of this study. The procedure generated a sample of 629 households.
The study units of the FilaBavi project are households. The heads of households were interviewed at monthly intervals during July 2001 to June 2002. If the head of the household could not be contacted, another adult was interviewed. These household representatives provided information on the household's health situation, health care utilisation, health expenditures and total expenditures. For information on illnesses the respondents were specifically asked if the household member in question had seen a medically trained person (doctor, nurse, health worker or such) and if so, which diagnose had been made. For all questions related to female-specific diagnoses, the interviewers were instructed to interview the patient directly.
Households kept daily notes of their health situation and health care payments including illness events of every person in the household. During the first week of each month, the interviewer conducted an interview based on the daily notes from the previous month. The interviews were carried out by 42 qualified interviewers employed by the larger FilaBavi project. All interviewers had completed high school education and were inhabitants of the Bavi district. The interviewers used a structured questionnaire and were given special training on data collection strategies for collecting information on income, expenditure and illnesses. Ten per cent of the questionnaires were randomly selected for re-interviews before the data entry.
Illnesses
This study defines an illness episode as a report of at least one of the following conditions: having stayed in bed; having been restricted from normal activities (e.g. work, school); having been able to do normal activities but with reduced capacity for at least one day and/or having to pay out-of-pocket for health services. An illness episode is concluded when normal activities (with normal capacity) are resumed. All episodes of illness that occurred during the month prior to the interview were recorded. An individual could thus have several illness episodes in a month [15].
In the interview questionnaire, respondents were specifically asked whether members of the household had the following symptoms or illnesses: cough; fever; difficulty breathing; headache; abdominal pain; intestinal disorder; pain in bones or joints; injury/accident; hypertension; heart illness. Respondents were also asked to specify whether members of the household had any other symptoms or illnesses.
From the respondents' answers, illnesses were classified into communicable, non-communicable, other and mixed illnesses.
Communicable illnesses include respiratory infection (throat, flu, and cough), fever, diarrhoea and the following symptoms/illnesses when combined with fever: abdominal pain; pimples; illnesses of the teeth, liver, kidney, eyes or skin; difficulty breathing; headache; stomach ache; gynaecological problems.
Non-communicable illnesses include diabetes, goitre, cancer, neurological problems, rheumatologic problems, hypertension, heart illness, allergic problems, and the following symptoms/illnesses when not combined with fever: illnesses of the liver, kidney, eyes, or skin; difficulty breathing; headache; stomach ache; gynaecological problems.
Other illnesses include symptoms/illnesses not included in the above list, such as toothache without fever, vertigo etc.
Mixed illnesses include symptoms/illnesses belonging to two or more of the above categories.
Expenditures and the classification of income groups
The expenditures recorded in the study are the total financial outlays that the households had each month for food, health care and other means. Health care payments include medical expenses (for consultations, tests, x-rays, drugs and beds) as well as non-medical expenses (for travel, food and other related means).
The households were classified according to their economic situation in two ways. First, total household expenditure quintiles were used. Expenditures, rather than income, are commonly used as a measure of socio-economic status in developing countries for several reasons. First, household expenditures tend to vary less than income, and second, households may be less willing to state their true income or may underestimate their total income [14].
In order to test the robustness of our expenditure classifications we also used a socioeconomic classification determined by local leaders. Households were ultimately classified into rich, medium and poor based on standards set by the Ministry of Labour, Invalids and Social Affairs.
Catastrophic health care spending
To identify the households that had catastrophic health care spending we used the methodology developed by Xu et al [16–18]. According to this approach, catastrophic spending occurs when health care expenditure for a household exceeds 40% of the households' capacity to pay.
A household's capacity to pay (CTPi) was calculated in the following way:
CTP
i
= TEXP
i
- SE
(45–55)i
If FEXP
i
≻ SE
(45–55)i
CTP
i
= TEXP
i
- FEXP
i
If FEXP
i
≺ SE
(45–55)i
TEXP denotes total expenditure and FEXP denotes food expenditure. SE stands for subsistence expenditure and is the average food expenditure for households whose food expenditure share of total expenditure is in the 45th to 55th percentile.
Microsoft ACCESS was used for data entry and data analyses were performed using SPSS software.