Data sources and methods of analysis
2.6 Variables used
The variables that are available from the TDHS data have been identified as important to the present study. These variables are defined into two broad categories: dependent and explanatory variables. The dependent variable, health status of children, is expressed in terms of nutritional status and prevalence of diarrhoea among children, whereas the explanatory variables are composed of a set of variables including demographic and socio-economic variables, household sanitation, maternal anthropometry, use of health services, and access to health facilities. These variables are defined below.
Nutritional status was based only on the physical growth (weight and height measurements) of children from the cross-sectional survey. This type of data allows us to identify the nature and extent of protein-energy malnutrition in the community (Mora, 1989: 133). Three nutritional indices were used: weight-for- height, weight-for-age, and height-for-age. Both the overall distributions and cut off points of the indices were expressed as the standard deviation units or z-score values relative to the standard deviation median of the WHO-NCHS reference population. Choices of the indices, cut-off points, and reference population are discussed in Section 2.4.
The use of the z-scores has statistical meaning as they take into account the co-efficient of measurement variation which varies with the age of children (United Nations, 1990: 48). In the present study, after the preliminary analysis of the nutritional status data, which are continuous variables, the variables were treated as dichotomous, with two values 'well-nourished' and 'under-nourished'. In order to assess malnutrition, cut-off points need to be used to estimate the prevalence of anthropometric abnormality. The conventional cut-off point, which is applied in the
present study, is -2 standard deviation units (z-scores) from the median reference population. Only height-for-age and weight-for-age are further used in the multiple logistic analysis. Weight-for-height is left out of the multiple analysis due to the prevalence of low weight-for-height or wasting among the study population being very low (3.9 per cent). Children whose z-scores fall below -2 standard deviation units are classified as under-nourished (coded 1) and those above -2 SD as well nourished (coded 0).
Prevalence of diarrhoea. Diarrhoeal disease is one of the most common illnesses in young children in developing countries. Prevalence of diarrhoea often reflects poor environmental conditions, including social, physical and biological, in which the child lives (Black, 1984; Black et al., 1984; Tomkins and Watson, 1989; Briscoe, 1991). In this study, diarrhoea is defined as the proportion of children who experienced diarrhoea in the two weeks prior to the survey. The prevalence of a disease is thought to be an appropriate measure of the morbidity of that disease when using a cross-sectional survey in which there is no other information available, such as number of episodes and duration of disease. Black (1984) concludes from a study of childhood diarrhoea in Bangladesh that specific types of diarrhoea have a higher incidence and longer duration among children from low-income households. He points out differentials in the prevalence of diarrhoeal disease associated with socio-economic status of the household, which may reflect differences in quality of child care practises such as preparation of weaning food, feeding patterns, and access to clean water. On the other hand, the differentials may reflect the poorer nutritional status of children from low-income households, a factor known to be closely associated with more prolonged diarrhoea and other infectious diseases. On the basis of the availability of the data, reported diarrhoea morbidity is identified in terms of prevalence or the proportion of children aged 0-59 months who experienced a diarrhoea episode in the two weeks prior to the survey. This variable was coded 1 for children who experienced diarrhoea during the two weeks prior to the survey and otherwise as 0.
Demographic variables. Six variables describing demographic characteristics of children and women were examined: age of child, sex of child, age of woman, birth order, family size, and breastfeeding.
All children under five years of age were classified into five age groups: 0-5, 6-11, 12-23, 24-47, and 48-59 months. However, for the analysis of malnutrition, only three age groups were identified: 6-17, 18-23, and 24-36 months.
Age of woman was classified into three broad age groups on the basis that each group actually contained equal numbers of women: young women (15-24 years), middle age women (25-29 years), and older women (30-49 years).
Birth order is considered to be one of the factors influencing child health and utilisation of health services, which captures a woman's past experiences in child bearing and child rearing. Four categories were classified: first, second, third, and fourth or higher birth order.
Family size was also classified into four categories: one, two, three, four or more children.
Breastfeeding is recognised as an important factor directly influencing child health and nutritional status. The TDHS survey collected a wide range of information on breastfeeding, but the length of full breastfeeding and supplementary feeding was not available. The analysis was restricted to whether the children still breastfed or not.
Socio-economic variables. Four variables were identified: household possessions, co-residence with parents, place of residence, women's education, women's occupation, husbands' education, and husband's occupation.
Information on household possessions available from the TDHS, including ownership of radio, television set, refrigerator, and motorcycle were used in this study. Where there is no income information, the possession of these household items can be used as proxy for economic status of the household. Apart from representing economic status of the households, possessions of household items also
represents access to information (ownership of radio and television set) and hygienic practices (ownership of refrigerator). Thus, no attempt has been made to construct an economic index from these variables. The variables are analysed individually in both bivariate and multivariate analyses.
This study intends to examine to what extent co-residence with parents in the family influences the health of young children. In accordance with this concern three variables are identified: currently living with parents, couple's mother still alive and couple's father still alive. In Thai society, the extended family is traditional and remains in existence in most rural villages. However, recent rapid economic and social changes have brought about changing social and family structure. In urban areas in particular the emergence of the nuclear family is notable. In circumstances where family ties are strong, the influence of old people on daily life is obvious.
Education, especially of mothers, is viewed as important in health behaviour and health practises which have a great influence on the health and survival of young children (Caldwell, 1979; Schultz, 1984). Level of education is used in many studies as a measure of a woman's resources for nurturing her children. In Thailand, school attainment is relatively high. However, the majority only finish compulsory education. Taking into account only level of education as a measure of woman's resources in this circumstance may not be worthwhile. Fortunately, the TDHS survey also provides information on literacy for both women and their husbands. Therefore, in this study, education of both women and their husbands was examined. Level of education and literacy were combined to form an education indicator and classified as 'illiteracy', 'primary education and read with difficulty', 'primary education and literate', and 'secondary or higher education'.
The occupation of women and their husbands is another important variable which greatly influences the health and well-being of their offspring. It is more likely that children whose parents are engaged in better paid jobs and socially recognised jobs such as professionals, can afford to have better nutrition and health care than children of blue-collar parents. Women's work, however, may have an impact on
child health as well as utilisation of health services through lack of time for child care and feeding, particularly breast-feeding. Occupation was classified into three categories: non-agricultural work, agricultural work, and not currently working.
Environmental variables. Environmental factors have long been regarded as important factors directly and indirectly influencing child health. The analysis was restricted to the following variables: source of drinking water and ownership of latrine. The variable 'source of drinking water' is classified into three categories: pipe or tap water, rain water, and well or others. Although the TDHS collected information on type of toilet facility, there is not much variation in toilet type. Therefore, the variable 'ownership of latrine' identifies whether the household has a toilet regardless of type.
Accessibility to health service. Data on access to health and social service facilities are taken from the TDHS community questionnaire. Due to lack of information on traditional health care services, the present study focuses on the availability of modern public health services. Three variables are examined in this regard: type of the nearest public health service, distance (in kilometres) to the nearest hospital and/or health centre, and time (in minutes) taken to the nearest hospital and/or health centre. Although the TDHS community data also provide information on private health facilities such as private clinics and drugstores, these facilities are concentrated in urban areas; neither is reported as being located in the rural communities. Therefore, private health facilities were not analysed in the present study.
'Type of the nearest public health facility' is a combined variable taken from detailed information about the closest hospital and health centre to the cluster or locality. This information was only collected where the clusters were classified as rural areas, and where information was collected from key informants such as village headman, village health volunteers and leaders of women's groups. Towns where the data were not available were assumed to have a whole range of health service facilities, both public and private.
Utilisation of health service was taken as a measure of preventive health behaviour. Four measures were assessed: whether women reported receiving a tetanus toxoid injection during pregnancies leading to living children under five years of age, whether women received prenatal care during pregnancies leading to living children aged under five, type of attendant at births during the five years prior to the survey, and ownership of child immunisation or health record cards.
The TDHS collected data on pregnant women who had been given tetanus injections during the previous five years. Although the WHO states that life-long protection of tetanus injections is achieved after five doses of vaccine (WHO, 1987), the TDHS indicator based on vaccinating all pregnant women does not take into account whether they had been ever partially or fully immunised. Asking for information about injections in the last five years could have introduced a recall bias. Women may have had difficulty remembering if they had had an injection earlier, even on a birth specific basis. Moreover, the women may have confused tetanus with other kinds of injection given in that period.
The variable describing prenatal care is confined to all pregnancies which took place between 1982 and 1987, as is choice of birth attendant. These variables are classified into two categories: pregnant women who visited trained western health providers (physician or nurse/midwife), or pregnant women who visited traditionally trained practitioners, including those who did not receive modern health care services.
In analysing use of health services, the ownership of a health record or immunisation card for children is also important. In Thailand, a card is to be issued by health providers where children receive preventive health services such as immunisation and nutrition surveillance. By and large it reflects the use of preventive health care and contact with modern trained health providers. However, since the card must be kept at home and is supposed to be presented at the next service, many children, especially older children, are likely to lose their cards. Most health interventions are concentrated in the first few months of life. For example,
immunisation is scheduled at ages two, four, six, and nine months and weight measurement takes place every four to six months. These immunisation cards are kept at home and it is a parents, especially a mother's, responsibility to show them to a health worker wherever services are provided. The health workers are supposed to issue a new card to the child if the card is lost. However, in practice this may not happen and some parents may be charged for a new card. Thus, ownership of a health record card may represent under-reporting of use of preventive health services. The present study investigates the relationship between health outcomes and utilisation of preventive health services; therefore, the variable 'ownership of a health record card' is expressed in terms of a child reported as ever having a card (regardless of whether a card could be presented to the interviewer or not) or not having a card at all.
The immunisation of children against the six immunisable diseases, tuberculosis, diphtheria, pertussis, tetanus, poliomyelitis and measles, should be completed in accordance with a fixed schedule and doses. For example, a child should be given vaccine against tuberculosis (BCG) at age 0-1 month and only one dose of vaccination is required. In order to construct an index of immunisation, details such as information on age of children, type of immunisation and date of immunisation are needed. Although these data can be obtained from the TDHS, the percentage of children able to present documents on such details such as health card was relatively small (24 per cent out of 682 children aged 0-59 months had health cards). The immunisation status of children was examined in terms of whether children were given at least one immunisation regardless of type, or whether they were never given any immunisation at all. However, due to the small number of cases, immunisation status will not be used to examine the health outcomes.
Maternal anthropometric measurements, both weight and height, were collected in the TDHS survey. The data were used to assess the nutritional status of women, expressed in terms of both their present and past nutritional status. Height in centimetres represents the past nutritional status of women. In the bivariate analysis, height was treated as a categorical variable. Thus, the cut-off points were
assigned to stratify height into four categories based on the distribution of the population in each category: less than 150.0, 150.0-154.9, 155.0-159.9, and 160.0 centimetres or taller. Height of women is treated as a continuous variable when running the multiple logistic analyses. Weight was used to assess the composition of a body. Like height, weight was treated as a categorical variable in the bivariate analysis and a continuous variable in the multiple analysis. Weight was classified into four categories: less than 45.0, 45.0-49.9, 50.0-54.9, and 55.0 kilograms or heavier. The current nutritional status for women was expressed in terms of the body mass index or obesity. The index is obtained from women's weight in kilograms divided by the corresponding height in square metres (kg/m^). Based on the WHO recommended cut-off points as reported in Jelliffe (1966), the standard for body mass index of adult women can be classified of woman adult into four groups of nutritional status: undernourished (0-18.6 kg/m^), normal nutrition (18.7-23.8 kg/m^), overweight (23.9-28.5 kg/m^), and obese (greater than 28.6 kg/m^). However, obesity is rather rare among Thai adults. Therefore, three categories of body mass index are used in the study. That is, under-nourished, normal, and over nourished (overweight and obese).
In summary, this chapter has discussed the source of data used in the study and methods of analysis. Two sources of data were chosen to use in this thesis: the TDHS, and field research from a case study village in the Northeast. The definition and measurement of all the variables were also included. It is anticipated that several sets of variables including socio-economic, demographic and environmental conditions, maternal anthropometry, availability of health services, and utilisation of health services were among many to influence the health status of children. These variables were drawn from the TDHS data, while the information on cultural beliefs and practices was mainly obtained from the field research. Two approaches were used to enable the researcher to understand the complexity of health and health behaviour in the Northeast Thailand. The Northeast and the case study village will be described in Chapter Three.