Search results

The search identified 634 articles. 362 publications were identified through PubMed, while 132, 117 and 23 publications were identified through Scopus, Web of Science and POPLINE respectively. An additional eight papers were identified through hand searches. After title and abstract review, 26 publications were considered eligible for the review (figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Literature search and review process.

Study characteristics

From the selected papers, 14 were research papers.22 25 37–49 Seven were quantitative, from which the rating according to the EPHPP tool was strong for two articles, moderate for four, and weak for one.38–40 43 45–47 According to the CASP tool, the one qualitative study identified had a low risk of bias.49 Three papers were considerations of contextual factors for systems integrating VA,22 25 44 and two investigated feasibility issues (one of community informants and one, the cost of VA in a mortality surveillance system).42 48 One was a pilot study that identified weaknesses in an existing maternal death review system and introduced a community-based process to improve the recording of maternal deaths.37 From the remaining papers, four were HDSS profiles,50–53 four were correspondences/commentaries,54–57 two were information sheets,41 58 one was a summary of study results59 and one was a research proposal.60 The studies were conducted in 15 LMICs: 10 in Africa,37 40 41 43 47 51–53 59 60 8 in Asia,25 39 42 44 46 49 50 55 2 in the Middle East45 48 and 1 in South America38 (figure 4). In five studies, the country was not specified.22 54 56–58 The finding that the majority of studies took place in LMICs was not surprising as VA is predominantly used in countries where CRVS systems are incomplete or absent. Ten studies used a WHO VA tool,38 39 43 45–48 56 58 60 two an adapted version developed by INDEPTH,52 53 two used the PHMRC tool49 59 and one applied a study-specific tool (Vital Registration and VA tool).40 Eleven studies mentioned the use of a VA tool but gave no specifics on which.22 25 37 41 42 44 50 51 54 55 57 For data collection with VA, five stated the use of electronic devices,40 49 52 56 60 two used paper-based questionnaires,42 50 while the remaining studies did not state how data were collected. Eight mentioned the additional use of questionnaires investigating social determinants.37 40 43 44 47 48 53 59 Specifics of how governments were involved in the process varied considerably. In eight studies, government involvement was in more than one element of the process: five studies described involvement in planning and research programme implementation,37–39 49 52 and in three studies, in implementation and scale up.45 48 59 In a further nine studies, government involvement was described in a single element of the process: five in implementation,25 41 43 46 47 two in scale up,44 60 one in funding support,50 and one through a working relationship which was stated but not described,53 respectively (online supplementary table 1). Twelve studies were conducted within established routine health information systems,25 38 40 41 46 49–53 59 60 and eight introduced or established new methods of data collection on cause of death.37 39 42–45 47 48

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Countries in which verbal autopsy (VA) studies were conducted and reported in scientific literature 2010-2017.

Implementation features

VA was used to ascertain cause of death predominantly in studies where no other form of certification was available22 54 58 60 or for deaths outside facilities,25 40 45–49 investigating mainly maternal, child and neonatal deaths.37 41 43–45 47 48 53 59 Other studies used VA in demographic surveillance sites in conjunction with epidemiological and demographic research,50–53 and for recording and investigating deaths in HDSS for the purposes of informing health policy and planning.38 40 42 46 50–52 55 60

Through methodological advances, and often with reference to the use of CCVA, VA was stated to have become a valued and feasible approach for use outside research settings, for wider application in health systems and policy as well as in national CRVS systems.22 25 38 40 42 46 48 49 54–56 58 60 This was supported by the study from Joshi et al,42 which investigated the cost of a VA-based mortality surveillance system in rural India. The study concluded that the approach was cost-effective (US$0.10 per capita per year), and that with data collection on electronic devices the cost could be further reduced.42 The authors demonstrated the use of VA in a mortality surveillance system as feasible, and as capable of providing valuable, timely evidence on cause of death for policymakers and health professionals.42

Other studies did not directly investigate costs but stated that cost savings could be achieved by conducting VA on a representative sample of all deaths or in a selection of registration administrative areas.22 As noted by Sankoh and Byass,56 the time and resource implications of establishing and maintaining CRVS systems inclusive of VA are inevitable, however the benefits that functional CRVS systems confer on people and societies are considerable. These include enabling access to public goods, facilitating development across all sectors and delivering valuable information for health policy and planning.56 In this sense, the cost of not collecting vital information is acknowledged as far higher. Government funding and commitment, as well as support and governance through an independent body, were also recommended  to ensure longevity, continued evolution and provision of academic freedom to achieve this.43 44 50

Several studies described VA as an applicable tool that adds value across intersectoral domains and beyond public health.38 40–42 44 50 59 60 These included ministries of health, local governments, specialised agencies (eg, national and international organisations), national health insurance, local health authorities, health facilities, community workers as well as departments of justice and interior.22 43 54 Several authors recommended an interdisciplinary and intersectoral approach that combines empirical and analytical efforts with different organisations, stakeholders and the community to enable functional registration in poorly resourced health systems.22 37 44 47 53 59 In the paper by Bensaïd et al,59 an intersectoral collaboration is described to demonstrate the power of implementing surveillance and response interventions in settings with high levels of mortality. The collaborative effort was initiated by the government of Niger and implemented by the National Statistics Institute and a VASA working group including the MOH, the Niger country office of UNICEF and other partners in partnership with the John Hopkins School of Public Health, which provided technical assistance on behalf of WHO/UNICEF and supported by CHERG.59

Furthermore, communities were described as a critical foundation on which to build CRVS systems and ensure sustainability.37 44 52 54 So-called hybrid approaches, using a combination of reporting through community-based systems, for example, health facilities, burial authorities, key informants as well as official registration, were detailed.22 38 58 Several authors pointed out, however, that communities should not be ‘used’ as passive data collectors but instead adopt roles as active partners in the surveillance and response processes.37 44 48 54 Authors also noted that this approach has the potential to generate self-sustaining sources of information on mortality, improve trust in health systems, and counteract the potential for ’community fatigue' from repeated data collection cycles.37 52

In this sense, active community involvement can enhance understandings of, and ultimately transform social exclusion from, access to health systems by shifting control over the production, use and exchange of knowledge on health towards those most directly affected.54 One example of such an approach was a pilot study by Bayley et al,37 which investigated a community-linked maternal death review (CLMDR) to measure and prevent maternal mortality.37 The study demonstrated that CLMDR improved identification and review of deaths, providing opportunities for education and stimulating action in facilities. Due to the success of this approach in a sample site in Malawi, the MOH initiated a nationwide roll-out of the programme.37

Studies employing VA to examine facility-based deaths were also identified. VA was described to become particularly important where postmortem pathological examinations are not available, not mandatory, unreliable, or death certificates are issued for non-specific causes, which are of little value for public health decision making.22 38 55 58 De Savigny et al22 and Bayley et al37 note the importance of investigations of this nature when recording maternal deaths that were not attended by trained providers.

VA in health systems and policy

Of the papers reviewed, 18 explicitly acknowledged a close intersection between the use of VA methods, and the functions of the health system to improve population health. This was achieved via: (1) active engagement of the government in the process,39 43 46 47 49 55 59 60 for example, high-level engagement of the Niger government and stakeholders for dissemination of VASA study results to support policy development,59 (2) demands from, or working relationship with, governments/health authorities,38 45 48 50 52 53 for example, the request from the Ministry of Public Health in Afghanistan to expand an intervention programme to reduce maternal deaths,45 and (3) purposeful dissemination of information to governmental, national and international agencies,37 40 41 51 58 for example, VA within the Millennium Global Village-Network provided feedback to local providers.40

In combination with questionnaires about social determinants, including as part of SA processes, VA was also used to identify social factors contributing to deaths and to determine gaps in health systems.44 52–54 SA has been reported as effective in supporting a more comprehensive analysis of the extent to which interventions achieve intended effects or help to explain why actions to improve people’s health are not taken by community members or health authorities.44 47 53

The studies demonstrated that when other routinely collected data, for example, surveys on demographic characteristics, are supplemented with VA data on a regular basis, they can provide information that can help to enhance health planning, inform policies and contribute to health system strengthening. However, to do so, data need to be shared and interpreted at all levels, from communities to systems.38 40–42 44 46 47 52 54 59 60

Strengths and limitations of VA for health systems and policy

According to the WHO, the gold standard for cause of death reporting is certification by a medical practitioner trained using the ICD system.61 However, in LMICs there is often insufficient supply of appropriate medical equipment, trained forensic experts and pathologists. Cultural norms and traditions can also interfere with timing and performance of autopsies.38 46 As de Savigny et al22 state, the absence of locally relevant gold standard data presents challenges in conducting locally applicable validation studies to determine the external validity of VA results. As stated above, VA has become increasingly standardised in recent years and more frequently applied globally, allowing for comparison with other standardised methods, for example, hospital records.56 58 The literature also acknowledged, however, the value of adaptations to context, and that standardisation may inadvertently compromise accuracy, completeness and validity, limiting relevance to local policy and planning.22 49 54 58

Use of mobile devices and electronic databases, ‘mHealth’, was a relatively new development affecting how VA is used. Electronic platforms were frequently described as able to enhance data sharing and link data of a range of types and sources: demographic, epidemiological, mortality, morbidity, clinical, laboratory, household, environmental, health system for a more holistic view.57 National electronic records systems, ‘SmartCare’ in Zambia,62 for example, have been reported to create new opportunities to enhance the timeliness of data collection and analysis, among others with VASA, enabling real-time data capture.22 40 47 52 53 60 These advances were reported to assist in the detection of rapid changes in health systems and disease burdens, as well as track key indicators related to local, regional and global goals and associated targets,48 59 aiding in efficient decision making and delivery of care.25 40

Encouraging the use of VA in countries lacking adequate or representative data via the expansion of HDSS, and with VA complementing hospital data were also seen as promising approaches.22 38 41 De Savigny et al22 recommend that data gathered with VA in the community should be analysed with physician-certified hospital deaths. To realise potential, it was further recognised that expansion of infrastructure and collaboration between research and service organisation and delivery are necessary.54 In South Africa, for example, the Department of Science and Technology has recently launched an initiative to consolidate three existing HDSS sites and expand the platform with three further urban and rural nodes, to cover over 1% of the national population.63