Incorporating verbal autopsy into the civil registration and vital statistics system
Stages and steps for implementing verbal autopsy in country
VA should be implemented in a staged manner, with evaluation at each of the stages.
Ensuring that the CRVS and health system are ready for the implementation of VA is a crucial stage in the process. Although system level changes may occur gradually, a broad understanding and agreement of what will be required to introduce VA and to sustainably scale it up needs to be agreed upon before VA activities take place. Initial stages in VA implementation (pretest or pilot) may commence before all systems are ready to accept VA implementation, and the speed of scale-up may to some extent be dictated by the speed of the necessary system changes.
Form national steering committees and review institutional set up and processes for notification, registration and certification of death
a) Set up a CRVS committee and/or national mortality committee and a VA technical committee. These committees will be responsible for ensuring that broad systems issues are put in place in order that VA can be implemented. They will also be responsible for the design of the VA roll-out, monitoring and evaluating the intervention and taking decisions on subsequent phases of implementation.
b) Apply enterprise architecture methods to develop business process maps to establish how VA will fit into the current design of the CRVS. Often this involves introducing strategies to strengthen existing notification and registration processes.
c) Conduct a legal review to identify and remove barriers to introducing VA to establish cause of death in a community or for deaths without a medical certificate of cause of death. This is where the status of the VA information, who has access to the information and what the information can be used for is confirmed.
Establishment of data transfer system for VA information and data sharing between relevant agencies, underpinned by current legislation:
- Data from VA needs to be available for use by agencies responsible for compiling statistics or designing health programs. This means that VA data may need to be available to more than one agency. The agencies most often involved in CRVS are civil registration and health and the national statistics office.
- The IT capacity at different levels of the system needs to be considered, because automated VA implies the use of a tablet (at community level), the transfer of the interview data over a mobile or wireless network and the compiling and analysing of data at a centralised point
Identification of a cadre of VA interviewers and their supervisors and mechanism for training and supporting these workers on an ongoing basis:
- The choice of VA interviewer is often a practical decision based on the cadre of workers available in the community. This may include nurses, midwives and community health workers. The ability of these workers to take on the task of conducting VA interviews in addition to their existing workload needs to be explored.
- The mechanism for supervising staff who are conducting VA interviews is crucial for maintaining the quality of the information being collected. Supervisors of VA interviewers need to have the skills and time to support them in this difficult task.
Stage 2: Pretest
Pretesting the questionnaire and methods in country is an essential stage in the process of VA implementation. The primary purpose of the pretest is to ensure the questionnaire is well translated and the VA data collection methods are acceptable to the VA interviewer and the community.
Translating and testing the questionnaire. This is an iterative process with several elements:
- A forward and back translation of the VA questionnaire is necessary to ensure that the meanings of the questions have not been lost during the translation process.
- Medical professionals of the country need to review the questionnaire to ensure the translated medical terms are sufficiently specific to allows identification of symptoms and diseases.
- Front-line workers should review the questionnaire to ensure the words used are understandable to the community that will be interviewed.
- Ideally, cognitive testing is undertaken to ensure a common understanding and agreement of the different stakeholders on the terms used in the questionnaire.
- Testing the questionnaire during a real verbal autopsy interview in the community will uncover problems in the terminology or with the interview questions.
Pretest training of VA interviewers and data collection
A forward and back translation of the VA questionnaire is necessary to ensure that the meanings of the questions have not been lost during the translation process.
Medical professionals of the country need to review the questionnaire to ensure the translated medical terms are sufficiently specific to allows identification of symptoms and diseases.
Front-line workers should review the questionnaire to ensure the words used are understandable to the community that will be interviewed.
Ideally, cognitive testing is undertaken to ensure a common understanding and agreement of the different stakeholders on the terms used in the questionnaire.
Testing the questionnaire during a real verbal autopsy interview in the community will uncover problems in the terminology or with the interview questions.
VA interviewers should complete a five-day VA interview training course. This course covers all aspects of conducting a VA interview, including how to approach a family for a VA interview, how to conduct a sensitive VA interview, the questions that form the VA questionnaire for adults, children and neonates, how to conduct an interview on a tablet and then save and send the information.
The VA training should provide the opportunity for the participants to conduct a VA interview in the field and for a debriefing on the experience as well as preparation for the field testing of the questionnaire.
Data collection for the pretest will be on deaths identified through the health system or other appropriate authority. A minimum of 100 deaths (ideally 60 adults, 20 children and 20 neonates) are recommended to adequately field test the questionnaire.
Evaluation of the pretest focuses on the quality of the translation and the acceptability of the methods to the community and to the VA interviewers. It provides information to inform the pilot phase of the implementation, including necessary changes to the questionnaire or training and the need for community education. Questions that can be answered from the pretest include:
- Problems with the comprehension of questions in the questionnaire either due to translation or cultural differences
- Average time to conduct a VA Interview
- Acceptability of the VA Interview and methods (using a tablet) to the community
- Problems with the functioning of tablets and adequacy of IT support for VA
- Adequacy of the training provided for the confidence of the interviewer to perform this task
- Assessment of the cause of death distribution coming from the VA information for any obvious errors.
Stage 3: Pilot
The pilot stage involves a rollout of the VA activities to a select area and activities are carried out as they would be done in a routine system. The primary purpose of the pilot is to test the operational aspects of the VA implementation and assess the plausibility of the COD data arising from VA.
Plan pilot activities
Decide on where to pilot the VA activities. Pilot areas may be based on practical considerations. There may also be some attempt to select diverse sites that may present different challenges to the implementation of VA to provide ‘proof of concept’ that VA methods will work in different areas of the country.
Design VA standard operating procedures for the pilot. The pilot will be trialling the collection of VA as part of the routine work of the health worker identified as the VA interviewer. It will be necessary to clearly define the role of the VA interviewer and of the supervisor for the pilot. This may also include new or strengthened procedures for notification or registration of a death.
IT and system protocols should be reviewed. With a larger-scale data collection over a more dispersed area, different methods for data transfer and analysis may be employed. It will be important to have a feedback mechanism to inform VA interviewers of the results of their work.
Step 8 & 9
Step 8: Pilot training and data collection
- Pilot training will involve all aspects of the VA interview as noted in the pretest training. It may also involve discussion of the importance of notification and registration of death and the role of the VA interviewer in collecting these data.
- Data collection for the pilot will be over an extended period, ideally collecting data from more than 1000 VAs, preferably more. It should be conducted on deaths that have occurred since the start of the VA pilot. The timeframe of the pilot needs to be long enough to allow VA interviewers and their supervisors to understand the operational implications of VA. The number of VAs collected needs to be sufficient for a preliminary assessment of the causes of death arising from VA.
- Training of IT personnel in regional and remote locations may be necessary to ensure IT support for VA interviewers and their supervisors.
Step 9: Evaluation of the VA pilot
Evaluation of the pilot gives an opportunity to look at the feasibility of VA implementation more broadly. The primary focus is on the operational aspects of the implementation and on the COD data that are being generated from the VA interviews. Questions that can be answered from the pilot include:
- Can VA activities be integrated into the routine work of the staff identified to conduct the VA interviews?
- Is the IT support adequate at all levels of the system, and what are the IT challenges associated with routine collection of VA data?
- Are the supervisory and monitoring systems in place and sustainable to ensure the quality of the VA interview on an ongoing basis?
- Does the COD analysis of the VA interview data reflect the known mortality pattern and epidemiological environment in the country?
- What are the challenges facing a large-scale rollout of VA in the country?
Stage 4: Scale-up using phased approach
Following the evaluation of the pilot, the national committee(s) need to decide whether to proceed with a rollout of VA. It is likely that this will happen in a phased manner to allow for gradual systems change.
Decide on whether and how to rollout VA. Systems-level considerations for the rollout of VA include:
The scope of activities. Countries may decide to choose a representative sample of administrative regions for the rollout of VA. Since VA is primarily a method to provide information on causes of death in a population, such a sample might provide sufficient information for prioritising health programs in the country.
The ongoing training opportunities for VA. Staff changes necessitate ongoing training opportunities, and existing staff will need refresher training. The country will need to decide on the mechanism for this and whether to retain a national training team or to delegate to regional authorities. Integration of VA training into existing curricula of the health staff responsible for VA is another way to maintain high standards of VA as implementation is rolled out.
Sustaining supervision, monitoring and evaluation. The information on COD from VA is primarily reliant on the quality of VA interviews. The continued and periodic monitoring of VA interviews and of the COD results from these interviews is essential and involves staff at different levels of the system. Putting in place robust but sustainable plans for monitoring and evaluation will be important for maintaining quality.
The IT infrastructure and mechanism for transmitting and analysing data needs to be reviewed for large scale rollout. Mechanisms for ensuring VA data are transmitted to the server for analysis and for the results of the VA data to be fed back to the relevant partners need to be determined. Ensuring that VA information is integrated into existing databases for the CRVS and/or health information system may require innovative solutions.
Since VA data are only one source of COD information, the country needs to decide how to synthesise these data with other sources of mortality data – for example, information from medical certification of cause of death (MCCOD). The purpose of VA is to provide a country with a more complete understanding of COD information, particularly where many deaths occur in the community without MCCOD. So, ultimately, this information needs to be incorporated into population statistics and used for health programming purposes.
Lessons learned from the Beijing verbal autopsy (VA) pilot
Dr Peng Yin from the Chinese Center for Disease Control and Prevention discusses the barriers to verbal autopsy in Beijing, VA data quality and next steps for implementation. Filmed in February 2018.