If someone dies without any medical attention, there is often a problem of determining what the cause of death was. Verbal autopsy (VA) methods are designed to collect cause-of-death information from populations where many deaths occur outside of health facilities and where death certification is weak or absent. It is estimated that around 40 million deaths each year are not medically certified. Verbal autopsy is a process which has been in use for many years, particularly in poorer societies where many deaths don't have a cause assigned. A VA consists of an interview with a relative or carer of a recently deceased individual in order to gather information on the signs and symptoms the decedent presented with prior to death. WHO has produced standard questionnaires for undertaking verbal autopsies.

The details gathered during a verbal autopsy are used to determine and assign a likely cause-of-death. At a population level this information can be invaluable to help guide prioritisation and direct health policy and services although this CAN NOT be USED to determine the cause of death at individual level.

Once the basic VA interview has been done, next steps can either involve physicians looking at the interview data case-by-case, and assigning causes of death, or the whole dataset can be fed into a model like InterVA-5 to automatically assign causes of death. Physician coding is slow and expensive, and there are worries about consistency between physicians and over time. Models like InterVA-5 are fast, cheap and completely consistent. But note that neither physicians nor models will be 100% correct for all cases.

Verbal Autopsy is a valuable research method to foster evidence-informed responses for populations affected by humanitarian crises. When coordinated and communicated effectively, data generated through these methods can help to identify levels, causes and circumstances of deaths among vulnerable groups, and can enable planning and allocating resources effectively, potentially improving health system resilience to future crises.

VA should be viewed as complementary to MCCOD, not as a replacement for it. Therefore, interventions to support and extend MCCOD should continue to be supported, while also promoting VA implementation for those cases in which MCCOD is not possible.

Verbal autopsy is:

• The only option available for determining major causes of death in settings without physician certification.
• A method for estimating the fraction of deaths in the population due to major causes of health concern (CSMFs).
• An approach based on a health rather than a clinical logic to ascertain COD, by way of a structured interview with caregivers of the deceased.
• A method for ascertaining a statistically probable COD at the individual level.
• A means of generating plausible, population-level data on causes of death in the community that can be aggregated into broad ICD codes.
• A stimulus for improving the identification, notification and reporting of deaths in the community.

Verbal autopsy is not:

• Designed to generate a COD that is legally equivalent to MCCOD at the individual level.
• A replacement for physician-certified COD, which can generate information on causes of death to a high degree of detail.
• Able to generate the detailed COD categories contained in the ICD.

How can we ensure that physician-certified deaths are accurate and recorded in accordance with the international standard and why is that important?

• Medical certification of cause of death training must be offered at key learning points in a physician’s career.
• Improvements in death certification will be underpinned by the medical community’s understanding of the importance of this task and the obligation physicians have towards their patients for recording their deaths correctly.

What are mortality decision tables and how they are useful in mortality coding?

• Coders in many countries are not medically trained. In mortality coding, understanding causal relationships between the conditions reported in death certificates is very important for the correct application of coding rules.
• Coders will need to undergo training in the new version of ICD, as the classification is very different to ICD-10.

Are registration and notification the same thing?

No, but they are often confused.

Registration is the act of formally registering an event at a civil registration office. At this point, details of the event are entered into the official civil register by the Registrar.

Notification is the capture and onward transmission of minimum essential information on the fact of birth or death by a designated agent or official of the CRVS or other government (eg health department) system using a CRVS authorised notification form (paper or electronic) with that transmission of information being sufficient to support immediate policy action to prevent premature deaths and the eventual registration and certification of the vital event.

The 10 CRVS milestones framework is a useful diagram to understand the CRVS system as a whole.

What are the strategies I can use to improve notification of deaths?

• By utilizing the health sector which is an increasingly active player in CRVS through its role in the notification of births and deaths, events that increasingly occur within health institutions or under the care of health personnel.

What is registration completeness?

• The completeness of registration is defined as the percentage of actual births or deaths in a population that are registered. Put another way, it is the number of registered births or deaths divided by the actual number of births or deaths in a population.

How do I measure the completeness of death registration for national and subnational populations?

• Completeness of death registration can be measured at the national and subnational level using a new empirical method that utilises data that are commonly available at the national and subnational level

How can I send the outputs from automated VA to an existing database platform such as DHIS2 so the results can be readily analysed?

• The introduction of routine verbal autopsy into CRVS systems has created a need for automated processing of verbal autopsy (VA) data into health system databases. Verbal autopsy data analysed using InterVA and InSilicoVA can be integrated into DHIS2 using the OpenVA pipeline, while SmartVA can be integrated into DHIS2 using smartva-dhis2. Both integration platforms are currently configured for DHIS2 but use an intermediate storage database that can be configured for other destination database platforms.
• Alternatively, desktop software for the VA algorithms are available for manual use.
• Read more about the OpenVA Pipeline or smartva-dhis2.

How are deaths in the community integrated to produce vital statistics reports?

• When community deaths are captured using verbal autopsy (VA), it is critical to consider the importance of drawing them from a nationally representative sample. This will provide the most accurate representation to allow VA data to be meaningfully integrated into the CRVS system. Countries may wish to use civil registration administrative units to identify the sample, to facilitate data administration.
• Measures should be put into place at various stages of the collection and production cycle of vital statistics. At the first stage where a death notification is generated and a VA is administered, data quality checks should focus on accuracy and completeness of the individual record. As the data become aggregated and move down the line to the agency responsible for reporting vital statistics, the data quality focus will be on accuracy, consistency and plausibility of the data.