[Elaboration and evaluation of a fast detection system of mortality using Fourier analysis. Study of a value with maximal deviation]. / Construcción y evaluación de un sistema de detección rápida de mortalidad mediante análisis de Fourier. Estudio de un valor con desviación máxima.

BACKGROUND: To calculate all causes weekly baseline mortality and an alarm threshold using Fourier analysis. To analyse the largest outlier detected in our series. METHODS: Madrid Undertaker database from October 1988 to September 1993, was used in the analysis. Orthogonal components were detected using Fourier analysis. Expected deaths and confidence limits were fitted using Serfling method. Alarm threshold was placed at a distance of 1.96 standard deviations above baseline. RESULTS: orthogonal frequencies with significant amplitudes corresponding to periods 26, 52, 104 and 156 weeks were detected. The second was the fundamental and its multiples were harmonics. In the time domain, baseline mortality showed a winter peak, declined to a summer plateau and presented its lowest level at the end of August. 21 weeks exceeded the alarm threshold. Of these 17 were related to influenza epidemics. The largest outlier corresponded to a heat wave in July 1991. CONCLUSIONS: A procedure similar to that proposed by R. E. Serfling (1963) to calculate baseline mortality, an alarm threshold and short term extrapolation using Madrid Undertaker database (C. Borrell, 1991) is presented. Madrid Undertaker database provide accurate and timely information about all causes mortality excess in Madrid.