Study area
Ouagadougou, the capital city of Burkina Faso, is experiencing an urban population boom, with expected increases in poverty and health disparities typical of many other major towns in sub-Saharan Africa [12, 13]. Between the years 1980 and 2005, the population of Ouagadougou grew more than four-fold, from 280,000 to 1,200,000 inhabitants. More than 25.0% of the population currently live in irregular peri-urban settlements of the town characterized by poorly constructed houses, poor sanitary conditions and lack of all services.
Like the general health services, oral health care in Burkina Faso has been organized through a public/private mixture of providers based on an out-of-pocket payment system. The national health insurance system does not cover any costs of curative dental care. The dentist to population ratio for the whole country is one per 220,000. Oral health facilities in the capital city are somewhat better equipped and definitely more numerous than in other parts of the country. In 2005, 24 oral health care services were established in Ouagadougou: 14 facilities within the private health sector, 4 within the public health sector and 6 within the nongovernmental not-for-profit health sector [14]. In parallel, a large number of pharmacies, street-market traders and traditional healers play a significant role in providing oral health care.
Study design and sample
To tackle the complex health problems resulting from environmental and social changes in an urban setting, the whole research approach has to be multi-disciplinary in accordance with epidemiological and geographical constraints.
Because of heterogeneity of ecological situations within the capital city and the absence of recent socio-demographic census data, a two-stage stratified sampling technique was used. The purpose was to identify various ecological settings representative of the distinct levels of urbanization in Ouagadougou. The stratification process was based on two criteria: building density, by comparing "average building density" with "high building density"; and settlement subdivision, by comparing "regular settlements" with "irregular settlements". Building density, the first criterion of stratification, was evaluated using a SPOT 5 (Satellite Pour l'Observation de la Terre) panchromatic satellite image of Ouagadougou dated November 26, 2002. In brief, ERDAS Imagine© software was used to locate buildings, and Arc GIS 8.2© made it possible to calculate the density of the built-up area. The process was supplemented and validated by field investigation. The localization of settlement subdivisions, the second criterion of stratification, was established using administrative information, the SPOT 5 image and field observations. The regular settlements are characterized by a network of hierarchical streets and the presence of basic services and infrastructures as electricity and water. In contrast, the irregular settlements are defined by an important but disorganized network of tracks and the absence of services and infrastructures. At the conclusion of this exercise, urban stratification comprised four types of urban situations: "regular settlements with average building density", "regular settlements with high building density", "irregular settlements with average building density" and "irregular settlements with high building density" [15].
In each of the four principal areas identified, two sub-areas were chosen by convenience sampling as representative of their geographical location in the city (central or peripheral) and age of housing construction (old or recent). At the second level of sampling, a random selection of households was chosen. In regular sub-areas, yards were randomly selected through cadastral maps. In irregular sub-areas, random sampling could not be applied because there was no cadastral map, so households were selected by the "door-to-door" approach used in the standard Expanded Program of Immunisation (EPI) sampling method [16]. In brief, from each randomly selected starting point, the closest household was chosen and checked for compliance with the inclusion criteria. From this first household, the interviewers moved to the next household according to well-defined rules [16] until four households had been randomly chosen. The same process was repeated from other starting points until the required number of participants was obtained.
As the whole research program focused on the impacts of urban life on disparities in health, an "urban criterion", the duration of residence in Ouagadougou of the head of the household, was taken into consideration. In each randomly selected yard, households were eligible if the head of the household had been born in Ouagadougou or had lived in Ouagadougou for more than five years. In addition, the householder had to be over 34 years old to comply with the WHO standard monitoring group for health conditions among adults. In each qualified household, eligible participants were women of any age and men of 35 years or older. All participants were required to have kinship or marital relationships with the head of the household.
Data collection
Data were collected by face-to-face field interviews, based on structured questionnaires, conducted either in households or in temporary medical examination centres. Interviewers fluent in the local language were given an intensive three-day field-training program. The survey instrument was evaluated for face validity and pilot tested before use.
Conceptual framework
In 1973, Andersen and Newman [17] proposed a framework for evaluating the utilization of health care. This model assumes that a person's use of health services is a function of predisposing, enabling and need factors. Predisposing characteristics include gender, marital status, educational level, occupation, length of time in the community and health beliefs. Health beliefs – such as attitudes, values and knowledge of the dental care delivery system – are often influenced by cultural values. Enabling resources refer to attributes specific to the individual or the community (e.g. income, social network, access to regular source of care). Need variables reflect illness levels that require the use of services. Needs can be perceived by the individual and are influenced by cultural beliefs and values (e.g. perceived health status, disease severity, limitation of activity) [18]. In 1997, an expanded version of this model was developed as a conceptual framework for the WHO International Collaborative Study of Oral Health Outcomes (ICS II), which was undertaken in selected industrialized countries [4]. This theoretical model was derived by integrating existing oral health behaviour and oral health status models with the general health model of Andersen and Newman. In the present study, we adapted the individual-level determinants proposed in both models to an urban setting of Burkina Faso.
We defined a set of independent individual-level variables that may influence utilization of oral health care: (1) predisposing socio-demographic and health beliefs factors, which can be either modifiable (e.g. education, marital status or health attitudes such as perceived general health status and seriousness of oral disease, importance of oral health, benefits of brushing, or perceived barriers to obtaining oral health care) or non-modifiable (e.g. age or sex); (2) enabling characteristics, which refer to specific attributes of the individual or the community in which the individual lives (e.g. level of income, residence, family size, integration in urban life, social support) and may affect ability to access the health care system; and (3) need factors, which reflect the perceived need for oral health care (patient's perceptions of illness, impairment of quality of life) or the self assessment of health status.
Furthermore, because of the complexity of assessing the socio-economic level of households, a composite index for material living conditions of households was computed using the SPSS 13.0. Module of Multiple Correspondence Analysis. This index was based on ownership of certain household assets such as refrigerator, television, motorcycle and car. Finally, the dependent variables were: whether individuals had experienced an oral health problem during the past twelve months; whether they practised self-medication; and whether they made a dental visit. No distinction was made between preventive and curative visits.
Statistical analysis
The data were processed and analyzed by SPSS 13.0. First, characteristics of the study sample were described, then frequency distributions were used to highlight the socio-demographic status of participants who had or not experienced an oral problem within the past twelve months. Secondly, chi-square tests were used to detect statistically significant proportions of people who used oral health facilities only, people who practised self-medication only, and people who did not seek treatment at all, in relation to various independent variables. Thirdly, logistic regression was performed in order to estimate the relative risk (Odds Ratio) of the independent variables explaining the use of oral health services within one year. The full regression model was specified on the basis of the theoretical model (the Anderson and Newman model). Our general approach was to test the predisposing characteristics first, then the enabling variables, and then the need variables. Estimates are presented with 95% confidence intervals. Results were weighted using the sampling proportions in order to reflect the population in the strata studied. There were no major differences in estimates when the sample data rather than weighted data were used. In addition, a multi-colinearity diagnostic statistic was derived from the Variance Inflation Factor for each variable found to be significantly associated with our dependent variable.