Skip to main content

Existing recommendations

(go to Outline)

What level of mortality is high? At what level should you do something? And when mortality reaches this level, what should you do?

These are very important questions. It is worthless to measure mortality rates if you are not ready and able to do something if the rate is elevated. This "something" might be as concrete as to implement specific preventive or curative health programmes or it may be as general as advocacy directed to donors to increase the resources devoted to a specific emergency-affected population.

Although it is difficult to give hard and fast rules, past recommendations have stated that an emergency exists, and therefore urgent action is needed, if the mortality rate is elevated to more than double the baseline pre-emergency level. But, of course, this raises the question of what is the baseline level? Earlier recommendations assumed that in stable populations in most less-developed countries, the baseline crude mortality rate is 0.5 deaths per 10,000 per day. Therefore, the cut-off defining emergency levels of mortality is 1.0 per 10,000 per day. Later recommendations state that the assumed baseline mortality rates depends on the specific population.

Regardless, decisions about when to respond and what response to implement cannot be made using only arbitrary quantitative cut-off values. Such planning must take into account trends in mortality rates over time, other measurements of health and nutritional status, available resources, other circumstances, and many other factors.

When evaluating an estimate of mortality rate from a survey, you can initially determine if the point estimate exceeds some threshold value. However, determining the importance of the difference between the point estimate and the threshold is often very difficult. Because survey estimates of mortality rate are, by definition, based on a sample of the population, they are subject to sampling error. As a result, a confidence interval is needed to convey the level of uncertainty introduced by this sampling error. (See section on Sampling error for more information.)

If the point estimate of the mortality rate is greater than the threshold defining an emergency, but the confidence intervals overlap this threshold, there is at least a 2.5% chance that the point estimate only exceeds the threshold because of sampling error. However, the importance of this must be judged using logic and common sense; there is no statistical formula for judgement. If you are using the threshold of 1.0 death/10,000/day, and the point estimate and confidence interval are 1.1 (0.3, 1.8), there is a very high probability that this point estimate is greater than 1.0 only because of sampling error. If, on the other hand, the point estimate and confidence interval are 1.7 (0.9,2.4), the likelihood is much lower that the point estimate exceeds the threshold only because of sampling error. In determining the action to be taken on the basis of such estimates, you must consider the consequences of two very different mistakes: 1) taking no action if the mortality is truly elevated versus 2) the consequence of taking action if the mortality is not truly elevated.

The survey done in Badghis Province, Afghanistan in March and April 2002 estimated that the age-specific mortality rate for children under 5 years of age was 2.5 deaths/10,000/day with a 95% confidence interval of 1.8, 2.5. We will use 2.0 deaths/10,000/day as the threshold to define an emergency for age-specific mortality rate for children under 5 years of age.

1

How would you interpret this age-specific mortality rate for children less than 5 years of age?

a)
b)
c)
d)
Please select an answerIncorrect. The point estimate is 2.5, definitely greater than the threshold of 2.0.Incorrect. Just because the lower end of the confidence interval falls below the threshold does not invalidate the elevated point estimate. In fact, the lower end of the confidence interval is not that far below the threshold, so it is still likely that the mortality rate in the population is higher than 2.0Incorrect. This thinking is too simplistic. The point estimate is not statistically significantly above the threshold, so there is still a relatively high likelihood that the estimate of mortality so high only because of sampling error. You should gather additional data to determine if the situation is as grave as indicated by this estimate, and, if so, what should be done about it.Correct. This finding may indicate a problem which deserves further investigation.
Check your answer

The survey in Badghis Province, Afghanistan also estimated that the crude mortality was 0.72 deaths/10,000/day with a 95% confidence interval of 0.49, 0.96. We will use the emergency threshold for the crude mortality rate of 1.0 death/10,000/day.

2

True or false?

The point estimate is not higher than the threshold, and the confidence interval does not overlap the threshold. Therefore, this is not an emergency and NO intervention at all is needed.

a)
b)

Correct. We cannot really be sure that NO intervention is needed based only on the crude mortality rate. Other indicators of health status need to be considered. In fact, other information collected by the same survey showed:

So obviously the health status of this population is not very good even though the crude mortality does not exceed the threshold defining an emergency.

Incorrect. We cannot really be sure that NO intervention is needed based only on the crude mortality rate. Other indicators of health status need to be considered. In fact, other information collected by the same survey showed:

So obviously the health status of this population is not very good even though the crude mortality does not exceed the threshold defining an emergency.

Your answer has been saved.
Check your answer