Drivers of measles mortality: the historic fatality burden of famine in Bangladesh

Abstract

Measles is a major cause of childhood morbidity and mortality in many parts of the world. Estimates of the case-fatality rate (CFR) of measles have varied widely from place to place, as well as in the same location over time. Amongst populations that have experienced famine or armed conflict, measles CFR can be especially high, although past work has mostly focused on refugee populations. Here, we estimate measles CFR between 1970 and 1991 in a rural region of Bangladesh, which experienced civil war and famine in the 1970s. We use historical measles mortality data and a mechanistic model of measles transmission to estimate the CFR of measles. We first demonstrate the ability of this model to recover the CFR in the absence of incidence data, using simulated mortality data. Our method produces CFR estimates that correspond closely to independent estimates from surveillance data and we can capture both the magnitude and the change in CFR suggested by these previous estimates. We use this method to quantify the sharp increase in CFR that resulted in a large number of deaths during a measles outbreak in the region in 1976. Most of the children who died during this outbreak were born during a famine in 1974, or in the 2 years preceding the famine. Our results suggest that the period of turmoil during and after the 1971 war and the sustained effects of the famine, is likely to have contributed to the high fatality burden of the 1976 measles outbreak in Matlab.

Keywords: case-fatality rate; infectious disease epidemiology; mathematical modelling; measles (rubeola).

Fig. 1.

Deaths due to measles (aggregated to the biweekly level) from January 1970 to September 1991. The three colors represent the three different data collection efforts over this time period (with varying numbers of villages under surveillance).

Fig. 2.

The estimated CFR from the regression of cumulative births against cumulative deaths using three estimation methods (black solid line: spline regression; blue dashed line: loess regression, green dashed line: lowess regression). The case-fatality rate for measles was estimated to be around 0·037 (95% binomial CI 0·025–0·049) in the Matlab area between August 1975 and July 1976 [5], and 0·018 (95% binomial CI 0·013–0·022) in 1980 [27] (indicated by solid red lines). The dashed vertical black lines indicate the onset of the war (1971) and the famine (1974).

Fig. 3.

Estimates of CFR using spline regression (with 5 degrees of freedom) of cumulative births on cumulative simulated deaths. Stochastic simulations of deaths are shown in the top panel of plots. Black line shows the median of 500 simulations; the shadowed region corresponds to the range between the 10th and 90th percentiles of the simulations. Deaths were simulated assuming four scenarios: (a) constant CFR = 0·02; (b) linearly increasing CFR over the time period, (c) linearly decreasing CFR over the time period and (d) non-linear CFR with an increase in the middle of the time period. The bottom panel of plots shows median estimated CFR in black; the shadowed region corresponds to the range between the 10th and 90th percentiles of the estimates. Red line indicates the actual CFR assumed for each simulation scenario.

Fig. 4.

Comparison of reported and simulated measles deaths. Black line shows the median of 1000 simulations using estimates for the transmission rate and CFR for the 1976 outbreak and TSIR estimates of transmission and CFR for the rest of the time series; the shadowed region corresponds to the range between the 10th and 90th percentiles of the simulations. Red line shows the reported deaths due to measles.