Invited Commentary to the paper ‘Dying to count: mortality surveillance in resource-poor countries’ by Edward Fottrell

Having participated in the formal process of his PhD defence, it was a real pleasure to read the paper by Ed Fottrell discussing challenges, limitations, possibilities and research areas related to mortality data in developing countries 1. In a few pages he has managed to highlight the most important aspects around that topic. The request, quoted from J.W. Lee, former Director General of the World Health Organisation, ‘To make people count, we first need to be able to count people’, is very true, but easier said than done. Fottrell nicely explains why. He lucidly constructs a bridge from describing the data we want, to describing the data we have, and it is most interesting to read about ongoing research activities leading towards making the best use of the data at hand, with their undoubted limitations, to get information, for example on cause-of-death distributions, to a precision and certainty that allow them to be the basis for public health action. His paper, as well his recent PhD thesis 2, shows nicely that being a good statistician who understands sampling surveys and theory of measurement error analysis is necessary but not sufficient in order to make significant contributions to improving mortality surveillance in resource-poor countries. Personal experience in developing countries, the ability to see what can be done and what cannot be done, and to simplify things as much as necessary, but not too much, are just as important.

For public health purposes, information on mortality is necessary. But what degree of precision is needed? What is ‘sufficiently accurate’? How can it be concluded that quality is ‘sufficient’? Fottrell's supervisor wrote in another recent publication: ‘Realistically, there will not be universal vital registration and individually based cause-of-death data on a worldwide basis anytime soon, no matter how useful such information might be in public health terms. Therefore a mixed-methods approach will continue to be used, combining data sources that are most appropriate to their particular settings, and meeting needs at different levels’ 3. The approaches that have been developed and used in the past to obtain the previously mentioned parameters require: 1 refinement – meaning that existing methods may not be fully satisfactory and need updating; and 2 evaluation – the performance of existing methods must be investigated. Another important point, also mentioned in the paper, lies in the fact that insufficient use has been made of the huge amount of data that has been collected. There are many reasons for this. First, in order to analyse health and demographic surveillance system (HDSS) data effectively, efficient working groups must exist, consisting of biostatisticians, database managers and computer scientists, and epidemiologists – the first, to decide on statistical methods and correctly apply them; the second, to efficiently and accurately organise the databases, and the third, to identify questions of interest and to interpret the findings.

In that respect, I may add some personal experiences with data from one particular HDSS, in Nouna, Burkina Faso. This HDSS has existed since 1992 and has fully functioned since 1993. It has grown from covering a population of about 30,000 individuals to a current population under surveillance of about 70,000. More villages have been included, and methods for collecting and organising the data have become more sophisticated over the years. How was that done? Who was financially involved? Have the data been used appropriately? Regarding the latter, Fottrell comments ‘That the value of data lies in their use, not in their collection, does not always seem to be appreciated by surveillance systems’. However, I think it is not so much a lack of appreciation, but more a lack of availability of trained personnel. There is also a direct correlation with lack of funding, as noted below, but in the case of the Nouna HDSS it proved difficult to attract good local statisticians to a working place in the periphery of the country when they could get better paid jobs in the capital city. So, in our case, the funding of a large collaborative research programme by the German Research Foundation since 1999 has allowed us to be more active in collaboratively analysing HDSS data collected over the years, and a number of scientific publications have emerged, as summarised in a monograph 4.

Financing mechanisms for HDSSs is an important point that was not covered in Fottrell's paper. In the case of the Nouna HDSS, financing has always been a critical issue, as described by Diesfeld 5. Data quality and financial support are highly correlated, as can be seen in the data of the Nouna HDSS during the 1990s, when percentages of missing values increased at times when sufficient financial support was lacking. The situation improved after the research grant from the German Research Foundation in 1999 facilitated better surveillance procedures. Thus, the Nouna HDSS, and most others, are constantly working to find sustainable means of guaranteeing uninterrupted, high-quality surveillance of their respective populations, and the INDEPTH network has contributed greatly to achieving high-quality standards in the sites linked to the network. Funding agencies are increasingly acknowledging this fact, but it remains an on-going challenge to obtain long-term funding for the basic maintenance of the surveillance process.

In the early days of most HDSS sites, overall mortality was the main outcome of interest, and only over time has more effort been put into assessing causes of death. The first INDEPTH monograph reported total mortality, mortality by age and life expectancy only. Verbal autopsy developed in the meantime and became the method of choice for cause-of-death statistics in developing countries. Limitations of the verbal autopsy method are obvious, although refinement of the method over the years has led to increases in sensitivity and specificity. Further developments, for example using the InterVA method, have yielded further improvements 678.

The recent advances in cause of death statistics, however, appear to have been focussed more towards communicable diseases. Given recent projections on cause of death distributions in Africa, deaths due to non-communicable diseases will become more frequent, and verbal autopsy-based methods may have serious limitations here. More research is needed – a frequently used but still correct phrase. The author of the paper to which this Commentary pertains, and his collaborators certainly have sufficient fields for future investigations.