Methods and tools to evaluate the quality of vital statistics
Small number issues in mortality statistics
Small number issues. In countries or cases where the number of deaths is small, care should be taken to aggregate data across several years, especially when publishing in formats disaggregated by age, sex and COD. Not doing so could have several unintended consequences for both policy-makers and individuals. Policy-makers may see trends in the data that are not real, but rather stem from natural yearly variation that is common in countries with small population sizes. See examples below.
Secondly, there is a confidentiality concern when publishing small numbers of deaths in disaggregation formats. Numbers should not be so small or specific that it would be possible to identify individuals based on other characteristics. For example, if there was only one published female death in the 40–44-year age range, it would be possible for the community to identify who this individual was and thus know what she died from. The decedent and their family have the right to keep this information confidential. In stigmatised deaths – such as suicide or HIV-related deaths – this is especially difficult for the surviving family members.
Countries with small populations often experience year-to-year variations in their vital statistics. Statisticians call this stochastic variation. For example, consider a country with a population size of 10,000, which usually experiences around 100 deaths per year on average, with about 10 of these occurring to children under the age of 1. However, if you look at each year individually, the first year there may have been 9 infant deaths, the next it may have been 12 and the third year perhaps it was 10. The births in this country also vary by year (around 100 per year). This means that there could be large variation in the infant mortality rate, which is defined as the number of deaths of infants under one year old per 1000 live births. In one year, it could be 1.1 infant deaths per 1000 live births, the next year it could be 2.8 and the next year it could be 1.7. Policy-makers may be alarmed when the rate goes up from 11 to 28, but, in actuality, this change is not due to a worsening health situation, but a problem of small numbers of events in a population. By averaging the data across 3 years and comparing a rolling 3-year average over time, a more accurate picture can be attained of trends in vital statistics.