Modelling determinants of health

Artificial neural networks for infant mortality modelling

This work aims to investigate a simple to use and easy to interpret methodology for assessing the relative importance of input variables in artificial neural networks (ANNs) applied to epidemiological modelling. The independent variables were 43 variables of the social, economic, environmental and health sector of 59 Brazilian municipalities, and the outcomes were infant mortality rates from these municipalities. Two assays were developed for the ANN modelling. On the first, all 43 variables were taken as input; and on the second, input variables were chosen with the help of factor analysis (FA). The relative importance of the input variables was investigated by means of bootstrap replications of the ANN model on the second assay. Further, multiple linear regression models (LRMs) were developed with the same data set and compared to the ANN models. The FA analysis allowed the selection of eight variables for the second assay. The percent of explained variance R(2) on the ANNs was in the range 0.74-0.80, while linear models had R(2)=0.4-0.5. These findings were validated by the bootstrap replications, in which the ANN models remained with higher R(2) and lower mean square error than the LRMs. The analysis of the best (second) ANN model indicated the highest ranking of importance for the variables literacy, agricultural and livestock sector jobs, number of commercial establishments and telephones. The approach presented here successfully integrated a data-oriented model with expert knowledge, indicating the potentiality of ANN modelling in the prediction, planning and assessment of public health actions.

A comparison of empirical models on determinants of infant mortality: a cross-national study on Africa

The goal of achieving health for all by the year 2000 has instigated numerous studies on the determinants of health. In this paper, we re-evaluate two models in which infant mortality--across twenty-eight low- and middle-income African states--is explained by socioeconomic, demographic, medical, environmental and political factors. The results indicate that the GNP per capita, school population as a percentage of the population under 19, population density and the percentage of the population with access to health care together explained 80% of the variations in infant mortality in the sample study. Apart from the GNP per capita and the school population as a percentage of the population under 19 which were negative and statistically significant, variables of importance for health policy, e.g. female literacy rate, water supply, food aid, calorie supply, health care expenditure and the degree of urbanization carried a negative sign but were nonsignificant. We interpreted the above thus: a reduction in infant mortality is feasible only with changes on diverse fronts rather than by marginal improvements in a few determining factors. A comparative test of the replicated model and our proposed model has shown that our model produced a better theoretical and statistical fitting than did the replicated model.

Infant mortality in Sri Lankan households: a causal model

The infant mortality rate in Sri Lanka has fallen precipitiously since World War II, to 44 per 1000 births, a rate that is similar to a number of Western Countries. Yet the aggregated country rate masks wide variations across districts, from a low 21 per 1000 in Jaffna District to a high of 100 per 1000 in Nuwara Eliya District. Such regional variations in infant mortality rates have been shown to be related to a number of social, demographic and intitutional characteristics of each area. To specify such linkages we move, in this paper, from the aggregate level of analysis to the level of individual households in order to look for social, economic and other characteristics associated with infant deaths. Data are derived from a systematic interview of 480 household heads in 12 villages of Sri Lanka, collected in 1980. The most proximate factors, public health or medical variables, that predict infant death in particular households include quality of family nutrition, supervised childbirth, immunization of children and, most significant, whether the family has a sanitary latrine. Yet these medical variables are strongly associated with educational and economic characteristics of the family which, in turn, are predictive of infant mortality. Poor families and poorly educated mothers are less likely to go to hospital for childbirth, have a latrine, etc., and more likely to have had an infant die. Underlying the variations in education and economic status are variations in ethnic group; families with poor sanitation, least education and few economic resources are most likely to be members of the minority communities, Ceylon or Indian Tamils and Muslims. Minority group membership is significantly associated with infant mortality as well. When a set of medical, educational, economic and cultural variables are examined simultaneously within a path model we show that the best causal explanation of infant death in Sri Lankan households says, "Minority group status results in poverty which prevents families from having safe sanitary facilities which causes infant death". Infant mortality in Sri Lanka is thus not simply a medical problem to be dealt with by public health programs, nor is it solely an economic problem that can be solved with creation of jobs, but it is better seen as a problem of the structure of the whole society.