The relative prevalence of disease symptoms for ill persons: evidence from Benin

In developing countries, the usual modelling of the correlates of health problems is not a good fit for the health phenomena encountered and the available data. Indeed, three common situations occur: (a) it is often the observed symptoms that are used to determine medical interventions instead of specific disease diagnostics or general health indicators; (b) the ill persons described by the data are often affected by multiple health problems; and (c) the correlates of the full spectrum of all symptoms need to be considered together. In this paper, these issues are dealt with by proposing a statistical approach based on competing scores of symptoms that explain their relative prevalence among the observed ill persons. Using multinomial logit models, the relative prevalence of four symptoms was estimated for four age classes of ill persons in Benin. Socio-demographic characteristics, household equipment and consumption behaviour are shown to influence the relative prevalence of symptoms and therefore could be used to decide what treatment to use. Moreover, living standards and economic activities are important and the pattern of symptoms among poor or agricultural ill persons differs from that of the rich or the non-peasants. The proposed method can be used to assist the definition of target groups and to guide the allocation of scarce resources in poor countries.