Fit
In collaboration with the University of Aberdeen, we are piloting community reporting of maternal deaths though the use of mobile phones. The wide-scale availability of mobile phones in South Africa and the acknowledged need to capture deaths in the community makes this an ideal setting to explore this option. An opportunity to evaluate community-based informant reporting of maternal deaths can be found in the Mpukonyoni Traditional Area in Hlabisa District of KwaZulu Natal. A large part but not all of this area falls within the DSA of the Africa Centre. This would enable the reporting system to be externally validated against the data gathered by the DSA, but would also provide lessons on engagement and set-up with informants unfamiliar with the Africa Centre and demographic surveillance. Although the primary aim of the proposed trial is to assess a system ultimately for reporting maternal deaths, in view of the comparative rarity of these events in short time intervals, it will be necessary to judge reliability of the system initially on the basis of all deaths in the study population.

Methodology
Two phases were used. The first phase involved the development of reporting mechanisms. This included the flow of information from the reporting of a death by family members to the local informants, and the onward passage of these data to produce aggregate statistics at a relevant administrative level. The second phase will involve the validation. The external validation element of the study would make comparisons at individual death and aggregate statistics levels between the informant-based data and the DSA, as well as with the routine HIS data in the case of deaths in health facilities, with burial or cemetery sources in the case of urban areas, and with the National Confidential Enquiry into Maternal Deaths (NCCEMD) for maternal deaths. Internal validation will be based on comparisons between the two informant networks and the size of the capture-recapture adjustment.

Results
Preliminary results show that paper-based reporting was performed by all 29 indunas although with different levels of completeness. However, although coverage of death reporting was high, content errors were common on the forms as there was a lot of missing information. Basically, the missing information syndrome emanated from the source because usually those who are sent to report the death to the induna have only basic information like the name of deceased, the village they come from, the sex etc and they would not have detailed information like was it a maternal death or not, cause of death, place of death etc. The SMS system and cellphone reporting system was not adequately done by the indunas, possibly because of the low literacy levels. One one induna out of the twenty-nine managed to effectively use a cellphone to report deaths in their constituency.

Further analysis of the data will be done in 2010. This will involve matching of deaths reported through independent systems.  Two broadly-independent types of informants are needed to enable internal validation checks, and also apply capture-recapture techniques to minimize underreporting of deaths. The latter requires matching of deaths from the two informants types and this has implications for the identifier information sought.

Timeline
The implementation of the project in the field has been completed and deaths for an 8-month period have been collected. In the following six months, matching of deaths from paper-based, cellphone-based system will be matched and compared. Completeness of reporting for each system will be calculated. From those data, an analytical paper on the performance of the different systems will be produced.