In a rural district of Malawi that has embarked on scaling up HIV/AIDS care and ART, there appears to be a very significant linear trend in reduction of all-cause mortality at population level. This is also one of the first reports to use proxy indicators of death such as coffin sales and church funerals both of which show a declining trend over time.
Malawi like other high HIV-prevalence countries in sub-Saharan Africa, has been desperately trying to scale-up HIV/AIDS care and ART as a lifesaving intervention to its population. In this light, the finding of an overall 37% reduction in all-cause mortality at population level is very encouraging. Similar reductions in population level mortality have been reported from Northern Malawi 
, Addis Ababa
and rural South Africa 
. In Botswana, the national program estimated that the annual number of AIDS related deaths halved from a peak of 15,500 in 2003 to 7400 in 2008 (7100 averted deaths) with an 80% treatment coverage
. These reductions are however less dramatic and less rapid than those reported from New York 
and Sau Paulo
(63% in 2 years, 65% in 7 years respectively). The reasons for the slower decline in population deaths in our setting (like other African countries
could include the fact that a) the epidemic peaked earlier compared to the timing of ART roll out, b) the number of AIDS cases are far larger resulting in difficulties for the relatively weak health infrastructure to provide universal and comprehensive HIV/AIDS care c) socio-cultural and economic factors impede access and demand for care and d) the fact that patients present late for ART which compromises their survival even when on the life-saving medication
Although we cannot know with certainty that the observed reduction in deaths is directly linked to HIV/AIDS care and ART roll out, the evidence is however highly suggestive and intuitive 
. In a high HIV prevalence country like Malawi, up to 65% (the great majority) of adult deaths can be attributed to HIV
. A recent study from South Africa 
where the cause of death was known showed that reductions in all-cause mortality were largely attributed to specific reductions in HIV related deaths among people on ART.
Reducing HIV/AIDS related deaths through specific interventions should therefore inevitably have an impact on all-cause mortality. This is also evident from a recent ecological analysis of all cause mortality and HIV treatment 
which demonstrated that when large numbers of people living with HIV receive care and ART in a mature epidemic, the mortality rate or the increase in the rate will decline. Although we do not have specific data to support this, we do not think that there were other factors that could have confounded the results like a significant improvement in social conditions such as employment opportunities or earning capacity of inhabitants. In fact, the prices of petrol, diesel and paraffin have generally shown an upward trend over the study period with a consequent increase in consumer prices 
. These are more likely to negate the overall impact of HIV/AIDS interventions on all-cause mortality. In any case, improvements in socio-economic factors alone, without access to HIV/AIDS treatment are unlikely to have any significant impact in a mature HIV/AIDS epidemic. There were reports of food shortages in 2001 and eventually a famine in 2002
. Although we should have expected to see an increase in deaths during this period, the downward trend in deaths was sustained.
There were also no specific epidemics resulting in exceptional mortality during any particular year of the study period and immunisation coverage in Malawi is generally high and constant 
Although it was only in 2007 that the target of universal access (80% of those estimated to be in urgent need of ART receive treatment) was achieved in Thyolo, mortality reduction was evident prior to this year with the largest annual declines, earlier in the study period. This is probably explained by the fact that mortality is influenced not just by ART alone, but also by other HIV/AIDS related care interventions such as cotrimoxazole prophylaxis 
, availability of drugs for prompt management of opportunistic infections, nutritional support and the presence of community support groups and networks
. This package existed prior to 2004 when ART initiation was started in our setting. Furthermore, MSF started its district level support in 1997 and in particular included, drug support to all peripheral health facilities for three key diseases namely malaria, respiratory disease and diarrhoea as from the year 2000. These interventions might well have contributed to reduction of death rates particularly child mortality in the earlier years. However, this would have had a relatively rapid impact which would have subsequently remained constant as MSF sustained drug support for these diseases during the entire study period (2000–2007). In any case, such interventions in their own right might at best delay, but not prevent mortality in a mature epidemic like in Thyolo where patients need ART.
The strengths of this study are that about half of the total population of a district with half a million inhabitants was included, and deaths were reliably verified through registers and are thus likely to reflect the reality on the ground. We also successfully managed to use three innovative proxy indicators of death in the absence of a formal and functional vital registration system. The limitations of the study are that:- i) it is based on simple observational data with the usual shortcomings; ii) we had missing data from 8% of villages and data on deaths are thus underestimated. However, since these data were missing for the entire study period, they do not influence the trend in mortality; iii) ten villages had no information on deaths for the year 2000. As this is the year of study start, this would have rather negated the impact of ART over the subsequent years and not the contrary; iv) importantly our data were not stratified by age and sex as this information was inconsistently recorded. The concept of age in traditional communities like Thyolo is anyway relative and its accuracy is in doubt. However, lack of age specific mortality data implies that interventions that significantly reduce child mortality will impact on all-cause mortality. Since vaccination coverage was high and specific support for malaria, diarrhoea and respiratory infections were sustained during the entire period this effect should have remained constant. In light of these potential limitations, we have kept the general focus on “trend analysis” of all-cause (crude) mortality so that age-specific and sex-specific mortality as well as any under-reporting bias remains stable during the study period. The reporting is also focused on mean mortality reduction over time; v) finally, we did not have data on the cause of mortality as post-mortems are not done in this setting.
In addition to the reduction in deaths, this study highlights two important additional issues that merit discussion. First, most African countries do not record vital registration data
and where it exists this is limited to health facility based data. However, the latter is not representative as more than half of all deaths that occur do so outside of health facilities and this is typical of other rural African settings
. In the absence of a viable vital registration system or reliable mortality statistics from the health information system, AIDS mortality estimates have been dependent on mathematical models
which could be quite different from the reality on the ground
. In this light, the existing death registration system at the TAs could provide an opportunity to fill this gap.
Second, this system runs parallel to the health management information system (HMIS) and at the moment remains solely for use by TAs. We need to explore whether the TA registration system can be enhanced and linked up with the HMIS to provide timely population level age- and sex-related mortality data that would assist also with child and maternal mortality data. This is particularly pertinent for assessing the impact of national programs and particularly as we approach 2015, reporting against targets in the Millennium Development Goals 4 and 5. Dialogue to explore possible ways to facilitate information transfer (e.g. a system for monthly or even trimestrial data transmission) between the TA systems and HMIS should be explored. This issue would benefit from further evaluation, training and support where necessary as it provides an important potential opportunity to improve information on vital data and reliably assess the impact of health and other interventions at population level.
In a rural district of Malawi that has scaled up HIV/AIDS care and ART, corroborative evidence from registered deaths at the traditional authorities, coffin sales and church funerals show encouraging evidence of a significant downward trend in reduction of population level deaths.
General measures are employed in all Thyolo district health facilities to ensure patient confidentiality, consent for HIV testing, and counselling and support for those who receive a positive HIV test result. The data in this study did not include patient identifiers. This study was formally approved by the District Commissioner of Thyolo who is also head of the Assembly of Traditional authorities. The Malawi National Health Science Research Committee provides general oversight and approval for the collection and use of data for monitoring and evaluation purposes, and approved this study. The study also received ethical approval from Ethics Review Board of the International Union Against TB and Lung Disease, Paris.