Citation

Becker S, Weng S. Int. J. Epidemiol. 1998; 27(5): 814-823.

Affiliation

Department of Population Dynamics, School of Public Health, Johns Hopkins University, Baltimore, MD 21205-2179, USA.

Copyright

(Copyright © 1998, International Epidemiological Association, Publisher Oxford University Press)

DOI

unavailable

PMID

9839738

Abstract

BACKGROUND: Deaths exhibit a seasonal pattern in most parts of the world. Analyses of deaths for the years 1972-1974 from the vital registration system of Matlab, Bangladesh, published in this journal 17 years ago, showed sinusoidal seasonal patterns. As death rates have declined in other nations, the seasonal pattern is attenuated. Death rates have declined substantially in Bangladesh in the past two decades. Thus, the present study examines monthly counts of deaths from Matlab data for a period 15 years later and tests the hypothesis of a decrease or shift in seasonality over time. METHODS: Trigonometric regression models were fit to monthly data by age and cause of death from the Matlab vital registration system for the years 1982-1990. A total of 20,328 death records were available for analyses. RESULTS: In the recent period significant sinusoidal seasonal patterns are found in all but one of the age and cause of death groups. Total deaths peak in the winter as do neonatal deaths but post-neonatal and child deaths are maximum in April and July respectively. Among cause groups, injury deaths (mostly attributed to drowning) show the greatest seasonal swing. The time of peak has only shifted for one age group--neonates--since the 1972-1974 period. The magnitude of the seasonal swing has declined significantly only for the neonatal age group and injury cause of death group. CONCLUSION: Marked seasonal patterns of deaths persist in the Matlab area of Bangladesh even as the level of mortality has declined.This study examined the seasonal variation in mortality by age and cause in Bangladesh during 1982-90. Data were obtained from the International Center for Diarrheal Disease Research, Bangladesh (ICDDR,B) registration system of births, deaths, and migrations. Temperature and rainfall data were obtained from the US National Climate Data Center. The analysis included 20,328 deaths. The patterns were contrasted with findings from 1972-74. Three statistical models were used to adjust monthly counts of death for each age and cause group. Model I used linear regression on time. Model II used simple trigonometric regression. Model III allowed for a peak during late 1983. Total deaths peaked in late November with 11% variance above and below the mean. Neonatal mortality peaked in November with 47% seasonal variation. Model II explained 69% of the monthly variance. Neonatal counts showed greater seasonal variation than rates. Postneonatal mortality peaked in the hot, dry season with 28% seasonal variation. Mortality among children aged 1-4 years peaked in the hot-wet season with 19% seasonal variation. Mortality among children aged 5-14 years peaked in October. Mortality among persons aged 15-44 years showed no seasonal variation. Seasonal variation was greatest among persons aged over 65 years with a 26% variance. 51% of the variation was explained by Model III. Injury and drowning deaths had the greatest seasonal variation and peaked in September. Diarrheal deaths peaked in the hot-wet season with 8% seasonal variation. All other causes and respiratory disease peaked in the cool-dry season with 15-20% seasonal variation.