Résumé
There are a number of reasons why climate, in certain physical and social environments, could have an impact on the epidemiology of malaria. Events, such as floods or drought, are related to the number of malaria cases and deaths, both seasonally and interannually. At a smaller scale, this study analyses the relation between climate variability and the variability in the number of deaths attributable to malaria in Niakhar, Senegal. The Niakhar area has a population of 30,000 and has been under demographic surveillance system since 1984. The rainfall in this region is highly seasonal, with a rainfall maximum in August and almost no rain between October/November and May/June. In addition to this seasonal cycle, rainfall also varies greatly from year to year (interannual variation). Over the 13 years, there were 661 deaths attributed to malaria with a marked interannual variability (range from 23 to 100, with a median of 43). There was also a strong seasonality in mortality, with nearly all deaths (89.1%) occurring between August and December. The number of deaths peaks in October, two months after the rainfall peak. Standardised monthly values were calculated for each climatic series (rainfall, relative humidity, temperature) as well as standardised five-month and monthly values of the number of deaths attributed to malaria between August and December. Correlation coefficients were calculated between these standardised values. The correlation between the variability in August rainfall and the variability in the number of deaths attributed to malaria between August and December was positive and statistically significant (r = +0.61, p = 0.02). In addition, highly significant cross-correlations were found between monthly rainfall series and monthly mortality series at one- and two-month lag (r = + 0.43, p = 0.0004 for one-month lag; r = + 0.26, p = 0.03 for two-month lag). This correlation is somewhat lower than the correlation of August rainfall alone with August to December mortality, but the result adds confidence to the signal given the increased degrees of freedom in the analysis. Similar, but slightly weaker, results were found when precipitation data were replaced with surface humidity data. Results with temperature were less clear; while temperature could in some circumstances have a direct impact on malaria, in this case here it is possible that the weak negative correlation between malaria deaths and temperature arises mainly because precipitation is physically connected to both the indices, correlating positively with malaria and negatively with temperature. The availability of a continuous demographic and medical survey since 1984 in a region of highly variable rainfall has created a rare opportunity to analyse with some confidence a climate versus malaria relationship. The findings are consistent with our understanding of the proposed link between rainfall and conditions for the reproduction of the malaria vector, leading to a lag time (here of one to two months) between anomalies of rainfall and deaths attributable to malaria. These results may have practical implications in Sub-Saharan regions marked by a great seasonal and interannual variability in rainfall by providing a simple tool to forecast the impact of climate variability on malaria mortality.There are a number of reasons why climate, in certain physical and social environments, could have an impact on the epidemiology of malaria. Events, such as floods or drought, are related to the number of malaria cases and deaths, both seasonally and interannually. At a smaller scale, this study analyses the relation between climate variability and the variability in the number of deaths attributable to malaria in Niakhar, Senegal. The Niakhar area has a population of 30,000 and has been under demographic surveillance system since 1984. The rainfall in this region is highly seasonal, with a rainfall maximum in August and almost no rain between October/November and May/June. In addition to this seasonal cycle, rainfall also varies greatly from year to year (interannual variation). Over the 13 years, there were 661 deaths attributed to malaria with a marked interannual variability (range from 23 to 100, with a median of 43). There was also a strong seasonality in mortality, with nearly all deaths (89.1%) occurring between August and December. The number of deaths peaks in October, two months after the rainfall peak. Standardised monthly values were calculated for each climatic series (rainfall, relative humidity, temperature) as well as standardised five-month and monthly values of the number of deaths attributed to malaria between August and December. Correlation coefficients were calculated between these standardised values. The correlation between the variability in August rainfall and the variability in the number of deaths attributed to malaria between August and December was positive and statistically significant (r = +0.61, p = 0.02). In addition, highly significant cross-correlations were found between monthly rainfall series and monthly mortality series at one- and two-month lag (r = + 0.43, p = 0.0004 for one-month lag; r = + 0.26, p = 0.03 for two-month lag). This correlation is somewhat lower than the correlation of August rainfall alone with August to December mortality, but the result adds confidence to the signal given the increased degrees of freedom in the analysis. Similar, but slightly weaker, results were found when precipitation data were replaced with surface humidity data. Results with temperature were less clear; while temperature could in some circumstances have a direct impact on malaria, in this case here it is possible that the weak negative correlation between malaria deaths and temperature arises mainly because precipitation is physically connected to both the indices, correlating positively with malaria and negatively with temperature. The availability of a continuous demographic and medical survey since 1984 in a region of highly variable rainfall has created a rare opportunity to analyse with some confidence a climate versus malaria relationship. The findings are consistent with our understanding of the proposed link between rainfall and conditions for the reproduction of the malaria vector, leading to a lag time (here of one to two months) between anomalies of rainfall and deaths attributable to malaria. These results may have practical implications in Sub-Saharan regions marked by a great seasonal and interannual variability in rainfall by providing a simple tool to forecast the impact of climate variability on malaria mortality.