Comparison of physician-certified verbal autopsy with computer-coded verbal autopsy for cause of death assignment in hospitalized patients in low- and middle-income countries: systematic review

Pregnancy-related mortality up to 1 year postpartum in sub-Saharan Africa: an analysis of verbal autopsy data from six countries

OBJECTIVE

To compare the causes of death for women who died during pregnancy and within the first 42 days postpartum with those of women who died between >42 days and within 1 year postpartum.

DESIGN

Open population cohort (Health and Demographic Surveillance Systems).

SETTING

Ten Health and Demographic Surveillance Systems (HDSS) in The Gambia, Kenya, Malawi, Tanzania, Ethiopia and South Africa.

POPULATION

2114 deaths which occurred within 1 year of the end of pregnancy where a verbal autopsy interview was conducted from 2000 to 2019.

METHODS

InterVA5 and InSilicoVA verbal autopsy algorithms were used to attribute the most likely underlying cause of death, which were grouped according to adapted International Classification of Diseases-Maternal Mortality categories. Multinomial regression was used to compare differences in causes of deaths within 42 days versus 43-365 days postpartum adjusting for HDSS and time period (2000-2009 and 2010-2019).

MAIN OUTCOME MEASURES

Cause of death and the verbal autopsy Circumstances of Mortality Categories (COMCATs).

RESULTS

Of 2114 deaths, 1212 deaths occurred within 42 days postpartum and 902 between 43 and 365 days postpartum. Compared with deaths within 42 days, deaths from HIV and TB, other infectious diseases, and non-communicable diseases constituted a significantly larger proportion of late pregnancy-related deaths beyond 42 days postpartum, and health system failures were important in the circumstances of those deaths. The contribution of HIV and TB to deaths beyond 42 days postpartum was greatest in Southern Africa. The causes of pregnancy-related mortality within and beyond 42 days postpartum did not change significantly between 2000-2009 and 2010-2019.

CONCLUSIONS

Cause of death data from the extended postpartum period are critical to inform prevention. The dominance of HIV and TB, other infectious and non-communicable diseases to (late) pregnancy-related mortality highlights the need for better integration of non-obstetric care with ante-, intra- and postpartum care in high-burden settings.

Community perception of causes of death using verbal autopsy for diabetes mellitus in Saudi Arabia

Diabetes mellitus is a serious global health issue which significantly impacts public health and socioeconomic development. Exploring how the community perceives the causes of death and their associated risk factors is crucial for public health. This study combines verbal autopsy (VA) with the Type 2 Diabetes Mellitus (T2DM) register to explore community perceptions of causes of death and associated influential factors in Makkah province, Saudi Arabia. 302 VA interviews were conducted with relatives or caregivers of deceased who died between 2018 and 2021 based on T2DM medical register from Alnoor Specialist Hospital in Makkah City, Saudi Arabia. Cause-specific mortality fractions (CSMFs) obtained from the VA using the InterVA-5 model were utilized to assess community perception. We used a multivariable logistic regression model to determine factors influencing community perceptions of causes of death. Lin's CCC with 95% CI was used to analyze the concordance for the CSMFs from verbal autopsy causes of death (VACoD) as a presumed reference standard and family-reported causes of death (FRCoD). The outcomes of this study demonstrate a generally broad spectrum of community perceived mortalities, with some critical misconceptions based on the type of death and other vital events like marital status, with an overall CCC of 0.60 (95% CI: 0.20-1.00; p = 003). The study findings demonstrate that community perception is weak if the deceased was male compared to female (aOR: 0.52; 95% CI: 0.26-1.03) and if the deceased was > = 80 years compared to 34-59 years (aOR: 0.48; 95% CI: 0.16-1.38), but it significantly improves among married compared to single (aOR: 2.13; 95% CI: 1.02-4.42). Exploring community perception of causes of death is crucial as it provides valuable insights into the community's understanding, beliefs, and concerns regarding mortality. Higher or lower community perception is attributed to how people may perceive risk factors associated with the causes of death, which can guide public health planning and interventional programs. The study findings further emphasize the need to employ robust and standardized VA methods within the routine medical services for a systemized assessment of families' reported causes of death.

Cardiovascular disease mortality based on verbal autopsy in low- and middle-income countries: a systematic review

Objective

To conduct a systematic review of verbal autopsy studies in low- and middle-income countries to estimate the fraction of deaths due to cardiovascular disease.

Method

We searched MEDLINE®, Embase® and Scopus databases for verbal autopsy studies in low- and middle-income countries that reported deaths from cardiovascular disease. Two reviewers screened the studies, extracted data and assessed study quality. We calculated cause-specific mortality fractions for cardiovascular disease for each study, both overall and according to age, sex, geographical location and type of cardiovascular disease.

Findings

We identified 42 studies for inclusion in the review. Overall, the cardiovascular disease cause-specific mortality fractions for people aged 15 years and above was 22.9%. This fraction was generally higher for males (24.7%) than females (20.9%), but the pattern varied across World Health Organization regions. The highest cardiovascular disease mortality fraction was reported in the Western Pacific Region (26.3%), followed by the South-East Asia Region (24.1%) and the African Region (12.7%). The cardiovascular disease mortality fraction was higher in urban than rural populations in all regions, except the South-East Asia Region. The mortality fraction for ischaemic heart disease (12.3%) was higher than that for stroke (8.7%). Overall, 69.4% of cardiovascular disease deaths were reported in people aged 65 years and above.

Conclusion

The burden of cardiovascular disease deaths outside health-care settings in low- and middle-income countries is substantial. Increasing coverage of verbal autopsies in these countries could help fill gaps in cardiovascular disease mortality data and improve monitoring of national, regional and global health goals.

Evaluation of methods for assigning causes of death from verbal autopsies in India

Background

Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (COD) in countries where medical certification of death is low. Computer-coded verbal autopsy (CCVA), an alternative method to PCVA for assigning the COD is considered to be efficient and cost-effective. However, the performance of CCVA as compared to PCVA is yet to be established in the Indian context.

Methods

We evaluated the performance of PCVA and three CCVA methods i.e., InterVA 5, InSilico, and Tariff 2.0 on verbal autopsies done using the WHO 2016 VA tool on 2,120 reference standard cases developed from five tertiary care hospitals of Delhi. PCVA methodology involved dual independent review with adjudication, where required. Metrics to assess performance were Cause Specific Mortality Fraction (CSMF), sensitivity, positive predictive value (PPV), CSMF Accuracy, and Kappa statistic.

Results

In terms of the measures of the overall performance of COD assignment methods, for CSMF Accuracy, the PCVA method achieved the highest score of 0.79, followed by 0.67 for Tariff_2.0, 0.66 for Inter-VA and 0.62 for InSilicoVA. The PCVA method also achieved the highest agreement (57%) and Kappa scores (0.54). The PCVA method showed the highest sensitivity for 15 out of 20 causes of death.

Conclusion

Our study found that the PCVA method had the best performance out of all the four COD assignment methods that were tested in our study sample. In order to improve the performance of CCVA methods, multicentric studies with larger sample sizes need to be conducted using the WHO VA tool.

Levels, trends, causes, place and time of, care-seeking for, and barriers in preventing indirect maternal deaths in Bangladesh: An analysis of national-level household surveys

Background

Eclampsia, haemorrhage, and other direct causes are the primary burden of maternal mortality in Bangladesh, often reducing attention given to indirect maternal deaths (IMDs). However, Sustainable Development Goals may not be achieved without actions to prevent IMDs. We examined the levels, trends, specific causes, timing, place, and care-seeking, and explored the barriers to IMD prevention.

Methods

We used three nationally representative surveys conducted in 2001, 2010, and 2016 to examine levels and trends in IMDs. The analysis of specific causes, timing, and place of IMDs, and care-seeking before the deaths was based on 37 IMDs captured in the 2016 survey. Finally, we used thematic content analysis of the open history from the 2016 survey verbal autopsy (VA) questionnaire to explore barriers to IMD prevention.

Results

After increasing from 51 deaths per 100 000 live births in 2001 to 71 in 2010, the indirect maternal mortality ratio (IMMR) dropped to 38 deaths per 100 000 live births in 2016. In 2016, the indirect causes shared one-fifth of the maternal deaths in Bangladesh. Stroke, cancer, heart disease, and asthma accounted for 80% of the IMDs. IMDs were concentrated in the first trimester of pregnancy (27%) and day 8-42 after delivery (32%). Public health facilities were the main places for care-seeking (48%) and death (49%). Thirty-four (92%) women who died from IMDs sought care from a health facility at least once during their terminal illness. However, most women experienced at least one of the "three delays" of health care. Other barriers were financial insolvency, care-seeking from unqualified providers, lack of health counselling, and the tendency of health facilities to avoid responsibilities.

Conclusions

IMMR remained unchanged at a high level during the last two decades. The high concentration of IMDs in pregnancy and the large share due to chronic health conditions indicate the need for preconception health check-ups. Awareness of maternal complications, proper care-seeking, and healthy reproductive practices may benefit. Improving regular and emergency maternal service readiness is essential.

Assessment of the underlying causes of adult deaths using a short version of verbal autopsy in Xaiyabouli Province, Lao People's Democratic Republic

BACKGROUND

In developing countries, it is difficult to collect the data of the underlying cause of death (UCOD), especially when a death does not occur in a health facility. This study aimed to develop a short version of verbal autopsy (VA) and identify the UCOD of adults in Lao People's Democratic Republic (Lao PDR).

METHODS

A short version of VA for deaths outside health facilities was developed. This study included all deaths of people aged 15 years old or older in Xaiyabouli Province in 2020. Socio-demographic factors, place of death, and UCOD of the deceased were collected from health facilities or from family members using a questionnaire including the short VA form. UCOD was compared between home deaths and hospital deaths, between the age group of 15-59 years old and the age group ≥ 60 years old, and between males and females.

RESULTS

Of all the 1,235 deaths included in this study, 1,012 deaths (81.9%) occured at home and 223 deaths (18.1%) at hospitals. The most common UCOD was senility (13.3%), followed by heart/renal failure (10.5%), pneumonia (9.6%) and traffic accident (7.1%). Compared to hospital deaths, home deaths had more people who were females, 75 years old or older, and Lao-Tai. Home deaths had more deaths than hospital deaths due to accident/injury (16.0% vs. 8.1%), tumor (4.7% vs. 1.8%), and senility (16.2% vs. 0%); fewer deaths due to heart/renal disease (15.1% vs. 32.3%), respiratory disease (12.2% vs. 18.8%), liver/gastro-intestine disease (5.3% vs. 9.0%), and infection (3.1% vs. 14.3%). The age group of 15-59 years had more deaths in the categories of accident/injury (28.1% vs. 4.4%), liver/gastro-intestine disease (8.1% vs. 4.4%), infection (7.2% vs. 3.5%), and tumor (6.0% vs. 2.8%). Males had more deaths due to tumor (5.2% vs. 3.0%) and fewer natural deaths (11.2% vs. 15.9%) than females.

CONCLUSIONS

The major UCOD category was heart/renal disease in the adult generation in Xaiyabouli Province. Cost-effective interventions based on the multisectoral noncommunicable disease prevention plan should be appropriately implemented. Mortality surveillance using the short VA tool should be conducted for all home deaths in Lao PDR.

Population cause of death estimation using verbal autopsy methods in large-scale field trials of maternal and child health: lessons learned from a 20-year research collaboration in Central Ghana

Low and middle-income countries continue to use Verbal autopsies (VAs) as a World Health Organisation-recommended method to ascertain causes of death in settings where coverage of vital registration systems is not yet comprehensive. Whilst the adoption of VA has resulted in major improvements in estimating cause-specific mortality in many settings, well documented limitations have been identified relating to the standardisation of the processes involved. The WHO has invested significant resources into addressing concerns in some of these areas; there however remains enduring challenges particularly in operationalising VA surveys for deaths amongst women and children, challenges which have measurable impacts on the quality of data collected and on the accuracy of determining the final cause of death. In this paper we describe some of our key experiences and recommendations in conducting VAs from over two decades of evaluating seminal trials of maternal and child health interventions in rural Ghana. We focus on challenges along the entire VA pathway that can impact on the success rates of ascertaining the final cause of death, and lessons we have learned to optimise the procedures. We highlight our experiences of the value of the open history narratives in VAs and the training and skills required to optimise the quality of the information collected. We describe key issues in methods for ascertaining cause of death and argue that both automated and physician-based methods can be valid depending on the setting. We further summarise how increasingly popular information technology methods may be used to facilitate the processes described. Verbal autopsy is a vital means of increasing the coverage of accurate mortality statistics in low- and middle-income settings, however operationalisation remains problematic. The lessons we share here in conducting VAs within a long-term surveillance system in Ghana will be applicable to researchers and policymakers in many similar settings.

Comparing quantitative and qualitative verbal and social autopsy tools: does a qualitative supplement improve understanding of the social determinants of under-five deaths in the slums of Kampala, Uganda?

Background

Understanding biological causes of death and sociocultural factors influencing outcomes is critical to reducing mortality in low-resource settings. Verbal and Social Autopsy instruments (VASAs) query family members about events leading to an individual’s death, resulting in quantitative, categorical data. This study sought to determine the value of a supplemental in-depth qualitative interview (VASA-QUAL).

Methods

This cross-sectional study was conducted in two slum neighborhoods in Kampala, Uganda, among families who lost a child under five within the preceding six months. A trained, local researcher conducted the quantitative VASA and then administered the VASA-QUAL to family members. Quantitative data were analyzed using Stata V16.0; qualitative data were transcribed into English and analyzed using NVivo V12.0. The biomedical cause of death was determined using a panel of physicians to code verbal autopsy items. Quantitative VASA variables were compared with qualitative variables from the VASA-QUAL using a rubric of indicators derived from the Pathways to Survival framework. Kappa statistics and percent agreement were calculated to compare quantitative and qualitative data. Three coders independently rated whether qualitative data provided additional information that improved understanding of the cause of death.

Results

48 VASAs were conducted (child age range: 1 month to 52 months). Agreement on key indicators ranged from 81.2% (place of death) to 93.8% (recognition of illness), with Kappa coefficients ranging from -0.038 to 0.368. The qualitative component added or clarified information about pediatric illness and care-seeking across all indicators, including recognition of illness (94.0%), care-seeking decisions (79.0%), whether home care was provided (73.0%), and choice of outside care (85.0%). Qualitative interviews frequently included symptoms missing or denied in the quantitative VASA and clarified the chronological order of symptoms. Many qualitative interviews described complicated mechanisms of decision-making not captured in the quantitative survey. Both agreement across data types and whether meaningful information was added by the qualitative data varied by cause of death, although our sample size limited our ability to conduct statistical analysis in this regard.

Conclusions

Supplementing quantitative VASA tools with an in-depth VASA-QUAL interview provided important additional information, but not consistently across indicators or causes of death. Despite challenges associated with feasibility, supplemental qualitative interviews may be an important tool for understanding the complexity of events leading up to childhood deaths.

Mixed-methods analysis of select issues reported in the 2016 World Health Organization verbal autopsy questionnaire

BACKGROUND

Use of a standardized verbal autopsy (VA) questionnaire, such as the World Health Organization (WHO) instrument, can improve the consistency and reliability of the data it collects. Systematically revising a questionnaire, however, requires evidence about the performance of its questions. The purpose of this investigation was to use a mixed methods approach to evaluate the performance of questions related to 14 previously reported issues in the 2016 version of the WHO questionnaire, where there were concerns of potential confusion, redundancy, or inability of the respondent to answer the question. The results from this mixed methods analysis are discussed across common themes that may have contributed to the underperformance of questions and have been compiled to inform decisions around the revision of the current VA instrument.

METHODS

Quantitative analysis of 19,150 VAs for neonates, children, and adults from five project teams implementing VAs predominately in Sub-Saharan Africa included frequency distributions and cross-tabulations to evaluate response patterns among related questions. The association of respondent characteristics and response patterns was evaluated using prevalence ratios. Qualitative analysis included results from cognitive interviewing, an approach that provides a detailed understanding of the meanings and processes that respondents use to answer interview questions. Cognitive interviews were conducted among 149 participants in Morocco and Zambia. Findings from the qualitative and quantitative analyses were triangulated to identify common themes.

RESULTS

Four broad themes contributing to the underperformance or redundancy within the instrument were identified: question sequence, overlap within the question series, questions outside the frame of reference of the respondent, and questions needing clarification. The series of questions associated with one of the 14 identified issues (the series of questions on injuries) related to question sequence; seven (tobacco use, sores, breast swelling, abdominal problem, vomiting, vaccination, and baby size) demonstrated similar response patterns among questions within each series capturing overlapping information. Respondent characteristics, including relationship to the deceased and whether or not the respondent lived with the deceased, were associated with differing frequencies of non-substantive responses in three question series (female health related issues, tobacco use, and baby size). An inconsistent understanding of related constructs was observed between questions related to sores/ulcers, birth weight/baby size, and diagnosis of dementia/presence of mental confusion. An incorrect association of the intended construct with that which was interpreted by the respondent was observed in the medical diagnosis question series.

CONCLUSIONS

In this mixed methods analysis, we identified series of questions which could be shortened through elimination of redundancy, series of questions requiring clarification due to unclear constructs, and the impact of respondent characteristics on the quality of responses. These changes can lead to a better understanding of the question constructs by the respondents, increase the acceptance of the tool, and improve the overall accuracy of the VA instrument.

Performance evaluation of machine learning and Computer Coded Verbal Autopsy (CCVA) algorithms for cause of death determination: A comparative analysis of data from rural South Africa

Computer Coded Verbal Autopsy (CCVA) algorithms are commonly used to determine the cause of death (CoD) from questionnaire responses extracted from verbal autopsies (VAs). However, they can only operate on structured data and cannot effectively harness information from unstructured VA narratives. Machine Learning (ML) algorithms have also been applied successfully in determining the CoD from VA narratives, allowing the use of auxiliary information that CCVA algorithms cannot directly utilize. However, most ML-based studies only use responses from the structured questionnaire, and the results lack generalisability and comparability across studies. We present a comparative performance evaluation of ML methods and CCVA algorithms on South African VA narratives data, using data from Agincourt Health and Demographic Surveillance Site (HDSS) with physicians' classifications as the gold standard. The data were collected from 1993 to 2015 and have 16,338 cases. The random forest and extreme gradient boosting classifiers outperformed the other classifiers on the combined dataset, attaining accuracy of 96% respectively, with significant statistical differences in algorithmic performance (p < 0.0001). All our models attained Area Under Receiver Operating Characteristics (AUROC) of greater than 0.884. The InterVA CCVA attained 83% Cause Specific Mortality Fraction accuracy and an Overall Chance-Corrected Concordance of 0.36. We demonstrate that ML models could accurately determine the cause of death from VA narratives. Additionally, through mortality trends and pattern analysis, we discovered that in the first decade of the civil registration system in South Africa, the average life expectancy was approximately 50 years. However, in the second decade, life expectancy significantly dropped, and the population was dying at a much younger average age of 40 years, mostly from the leading HIV related causes. Interestingly, in the third decade, we see a gradual improvement in life expectancy, possibly attributed to effective health intervention programmes. Through a structure and semantic analysis of narratives where experts disagree, we also demonstrate the most frequent terms of traditional healer consultations and visits. The comparative approach also makes this study a baseline that can be used for future research enforcing generalization and comparability. Future study will entail exploring deep learning models for CoD classification.

Verbal Autopsy as a Tool for Defining Causes of Death in Specific Healthcare Contexts: Study of Applicability through a Traditional Literature Review

Autopsy examination, the gold standard for defining causes of death, is often difficult to apply in certain health care settings, especially in developing countries. The COVID-19 pandemic and its associated difficulties in terms of implementing autopsy examinations have made the need for alternative means of determining causes of death even more evident. One of the most interesting alternatives to the conventional autopsy is the verbal autopsy, a tool that originated in Africa and Asia in the 1950s and consists of a structured interview with the deceased's family members concerning the symptoms manifested by the person and the circumstances of death. In the early 1990s, the first doubts emerged about the validity of verbal autopsies, especially about the real reliability of the cause of death identified through this tool. The objective of the review was to identify studies that had assayed the validity of verbal autopsies through a rigorous comparison of the results that emerged from it with the results of conventional autopsies. When starting from an initial pool of 256 articles, only 2 articles were selected for final review. These are the only two original research articles in which a verbal autopsy validation process was performed by employing the full diagnostic autopsy as the gold standard. The two papers reached opposite conclusions, one suggesting adequate validity of verbal autopsy in defining the cause of death and the other casting serious doubts on the real applicability of this tool. Verbal autopsy undoubtedly has extraordinary potential, especially in the area of health and demographic surveillance, even considering the implementation that could result from the use of artificial intelligence and deep learning. However, at present, there appears to be a lack of solid data to support the robust reliability of this tool in defining causes of death.

Assessing the Diagnostic Accuracy of Physicians for Home Death Certification in Shanghai: Application of SmartVA

Approximately 30% of deaths in Shanghai either occur at home or are not medically attended. The recorded cause of death (COD) in these cases may not be reliable. We applied the Smart Verbal Autopsy (VA) tool to assign the COD for a representative sample of home deaths certified by 16 community health centers (CHCs) from three districts in Shanghai, from December 2017 to June 2018. The results were compared with diagnoses from routine practice to ascertain the added value of using SmartVA. Overall, cause-specific mortality fraction (CSMF) accuracy improved from 0.93 (93%) to 0.96 after the application of SmartVA. A comparison with a “gold standard (GS)” diagnoses obtained from a parallel medical record review investigation found that 86.3% of the initial diagnoses made by the CHCs were assigned the correct COD, increasing to 90.5% after the application of SmartVA. We conclude that routine application of SmartVA is not indicated for general use in CHCs, although the tool did improve diagnostic accuracy for residual causes, such as other or ill-defined cancers and non-communicable diseases.

Integrating community-based verbal autopsy into civil registration and vital statistics: lessons learnt from five countries

This paper describes the lessons from scaling up a verbal autopsy (VA) intervention to improve data about causes of death according to a nine-domain framework: governance, design, operations, human resources, financing, infrastructure, logistics, information technologies and data quality assurance. We use experiences from China, Myanmar, Papua New Guinea, Philippines and Solomon Islands to explore how VA has been successfully implemented in different contexts, to guide other countries in their VA implementation. The governance structure for VA implementation comprised a multidisciplinary team of technical experts, implementers and staff at different levels within ministries. A staged approach to VA implementation involved scoping and mapping of death registration processes, followed by pretest and pilot phases which allowed for redesign before a phased scale-up. Existing health workforce in countries were trained to conduct the VA interviews as part of their routine role. Costs included training and compensation for the VA interviewers, information technology (IT) infrastructure costs, advocacy and dissemination, which were borne by the funding agency in early stages of implementation. The complexity of the necessary infrastructure, logistics and IT support required for VA increased with scale-up. Quality assurance was built into the different phases of the implementation. VA as a source of cause of death data for community deaths will be needed for some time. With the right technical and political support, countries can scale up this intervention to ensure ongoing collection of quality and timely information on community deaths for use in health planning and better monitoring of national and global health goals.

Verbal autopsy models in determining causes of death

OBJECTIVES

To systematically review current practices, strengths and limitations of existing VA approaches to increase understanding of health system stakeholders and researchers.

METHODS

The review was conducted and reported based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines, in which articles were systematically obtained from the PubMed and SCOPUS online databases. The search was limited to English language journal articles published between 2010 and 2020. The review identified 5602 articles and after thorough scrutiny, 25 articles related to VA approaches were included.

RESULTS

(1) InterVA and Tariff are widely used VA models; (2) Bayes rule is the most common and successful algorithm; (3) the lack of standardised datasets and metrics to evaluate models creates bias in determining VA model performance; (4) performance of the models trained using in-hospital data cannot be replicated in community death; (5) the performance of models among physicians and computer-coded algorithms differs with variation in settings.

CONCLUSION

The physician-certified verbal autopsy (PCVA) approaches are more effective in determining community CoD while computerised coding of verbal autopsy (CCVA) models perform well when the underlying CoD are reliably established using hospital data where data are trained in a similar environment to the target population. Our study recommends the use of hybrid models that combine strengths from various models and using an open standards dataset that includes death from different settings.

The ethical implications of verbal autopsy: responding to emotional and moral distress

Background

Verbal autopsy is a pragmatic approach for generating cause-of-death data in contexts without well-functioning civil registration and vital statistics systems. It has primarily been conducted in health and demographic surveillance systems (HDSS) in Africa and Asia. Although significant resources have been invested to develop the technical aspects of verbal autopsy, ethical issues have received little attention. We explored the benefits and burdens of verbal autopsy in HDSS settings and identified potential strategies to respond to the ethical issues identified.

Methods

This research was based on a case study approach centred on two contrasting HDSS in Kenya and followed the Mapping-Framing-Shaping Framework for empirical bioethics research. Data were collected through individual interviews, focus group discussions, document reviews and non-participant observations. 115 participants were involved, including 86 community members (HDSS residents and community representatives), and 29 research staff (HDSS managers, researchers, census field workers and verbal autopsy interviewers).

Results

The use of verbal autopsy data for research and public health was described as the most common potential benefit of verbal autopsy in HDSS. Community members mentioned the potential uses of verbal autopsy data in addressing immediate public health problems for the local population while research staff emphasized the benefits of verbal autopsy to research and the wider public. The most prominent burden associated with the verbal autopsy was emotional distress for verbal autopsy interviewers and respondents. Moral events linked to the interview, such as being unsure of the right thing to do (moral uncertainty) or knowing the right thing to do and being constrained from acting (moral constraint), emerged as key causes of emotional distress for verbal autopsy interviewers.

Conclusions

The collection of cause-of-death data through verbal autopsy in HDSS settings presents important ethical and emotional challenges for verbal autopsy interviewers and respondents. These challenges include emotional distress for respondents and moral distress for interviewers. This empirical ethics study provides detailed accounts of the distress caused by verbal autopsy and highlights ethical tensions between potential population benefits and risks to individuals. It includes recommendations for policy and practice to address emotional and moral distress in verbal autopsy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12910-021-00683-7.

Validating reported cause of death using integrated electronic health records from a nation-wide database

BACKGROUND

To compare the underlying cause of death reported by the Israeli Central Bureau of Statistics (CBS) with diagnoses in the electronic health records (EHR) of a fully integrated payer/provider healthcare system.

METHODS

Underlying cause of death was obtained from the CBS for deaths occurring during 2009-2012 of all Clalit Health Service members in Israel. The final cohort consisted of members who had complete medical records. The frequency of a supportive diagnosis in the EHR was reported for 10 leading causes of death (malignancies, heart disease, cerebrovascular disease, diabetes, kidney disease, septicemia, accidents, chronic lower respiratory disease, dementia and pneumonia and influenza).

RESULTS

Of the 45 680 members included in the study, the majority of deaths had at least one diagnosis in the EHR that could support the cause of death. The lowest frequency of supportive diagnosis was for septicemia (52.2%) and the highest was for malignancies (94.3%). Sensitivity analysis did not suggest an alternative explanation for the missing documentation.

CONCLUSIONS

The underlying cause of death coded by the CBS is often supported by diagnoses in Clalit's EHR. Exceptions are septicemia or accidents that cannot be anticipated from a patient's EHR, and dementia which may be under-reported.

Limitations to current methods to estimate cause of death: a validation study of a verbal autopsy model

Background: Accurate information on causes of death (CoD) is essential to estimate burden of disease, track global progress, prioritize cost-effective interventions, and inform policies to reduce mortality. In low-income settings, where a significant proportion of deaths take place at home or in poorly-resourced peripheral health facilities, data on CoD often relies on verbal autopsies (VAs). Validations of VAs have been performed against clinical diagnosis, but never before against an acceptable gold standard: the complete diagnostic autopsy (CDA). Methods: We have validated a computer-coded verbal autopsy method -the InterVA- using individual and population metrics to determine CoD against the CDA, in 316 deceased patients of different age groups who died in a tertiary-level hospital in Maputo, Mozambique between 2013 and 2015.   Results: We found a low agreement of the model across all age groups at the individual (kappa statistic ranging from -0.030 to 0.232, lowest in stillbirths and highest in adults) and population levels (chance-corrected cause-specific mortality fraction accuracy ranging from -1.00 to 0.62, lowest in stillbirths, highest in children). The sensitivity in identifying infectious diseases was low (0% for tuberculosis, diarrhea, and disseminated infections, 32% for HIV-related infections, 33% for malaria and 36% for pneumonia). Of maternal deaths, 26 were assigned to eclampsia but only four patients actually died of eclampsia. Conclusions: These findings do not lead to building confidence in current estimates of CoD. They also call to the need to implement autopsy methods where they may be feasible, and to improve the quality and performance of current VA techniques.

Limitations to current methods to estimate cause of death: a validation study of a verbal autopsy model

Background: Accurate information on causes of death (CoD) is essential to estimate burden of disease, track global progress, prioritize cost-effective interventions, and inform policies to reduce mortality. In low-income settings, where a significant proportion of deaths take place at home or in poorly-resourced peripheral health facilities, data on CoD often relies on verbal autopsies (VAs). Validations of VAs have been performed against clinical diagnosis, but never before against an acceptable gold standard: the complete diagnostic autopsy (CDA). Methods: We have validated a computer-coded verbal autopsy method -the InterVA- using individual and population metrics to determine CoD against the CDA, in 316 deceased patients of different age groups who died in a tertiary-level hospital in Maputo, Mozambique between 2013 and 2015.   Results: We found a low agreement of the model across all age groups at the individual (kappa statistic ranging from -0.030 to 0.232, lowest in stillbirths and highest in adults) and population levels (chance-corrected cause-specific mortality fraction accuracy ranging from -1.00 to 0.62, lowest in stillbirths, highest in children). The sensitivity in identifying infectious diseases was low (0% for tuberculosis, diarrhea, and disseminated infections, 32% for HIV-related infections, 33% for malaria and 36% for pneumonia). Of maternal deaths, 26 were assigned to eclampsia but only four patients actually died of eclampsia. Conclusions: These findings do not lead to building confidence in current estimates of CoD. They also call to the need to implement autopsy methods where they may be feasible, and to improve the quality and performance of current VA techniques.

Limitations to current methods to estimate cause of death: a validation study of a verbal autopsy model

Background: Accurate information on causes of death (CoD) is essential to estimate burden of disease, track global progress, prioritize cost-effective interventions, and inform policies to reduce mortality. In low-income settings, where a significant proportion of deaths take place at home or in poorly-resourced peripheral health facilities, data on CoD often relies on verbal autopsies (VAs). Validations of VAs have been performed against clinical diagnosis, but never before against an acceptable gold standard: the complete diagnostic autopsy (CDA).

Methods: We have validated a computer-coded verbal autopsy method –the InterVA- using individual and population metrics to determine CoD against the CDA, in 316 deceased patients of different age groups who died in a tertiary-level hospital in Maputo, Mozambique between 2013 and 2015.  

Results: We found a low agreement of the model across all age groups at the individual (kappa statistic ranging from -0.030 to 0.232, lowest in stillbirths and highest in adults) and population levels (chance-corrected cause-specific mortality fraction accuracy ranging from -1.00 to 0.62, lowest in stillbirths, highest in children). The sensitivity in identifying infectious diseases was low (0% for tuberculosis, diarrhea, and disseminated infections, 32% for HIV-related infections, 33% for malaria and 36% for pneumonia). Of maternal deaths, 26 were assigned to eclampsia but only four patients actually died of eclampsia.

Conclusions: These findings do not lead to building confidence in current estimates of CoD. They also call to the need to implement autopsy methods where they may be feasible, and to improve the quality and performance of current VA techniques.

Community surveillance and response to maternal and child deaths in low- and middle-income countries: A scoping review

BACKGROUND

Civil registration and vital statistics (CRVS) systems do not produce comprehensive data on maternal and child deaths in most low- and middle-income countries (LMICs), with most births and deaths which occur outside the formal health system going unreported. Community-based death reporting, investigation and review processes are being used in these settings to augment official registration of maternal and child deaths and to identify death-specific factors and associated barriers to maternal and childcare. This study aims to review how community-based maternal and child death reporting, investigation and review processes are carried out in LMICs.

METHODS

We conducted a scoping review of the literature published in English from January 2013 to November 2020, searching PubMed, EMBASE, PsycINFO, Joanna Briggs, The Cochrane Library, EBM reviews, Scopus, and Web of Science databases. We used descriptive analysis to outline the scope, design, and distribution of literature included in the study and to present the content extracted from each article. The scoping review is reported following the PRISMA reporting guideline for systematic reviews.

RESULTS

Of 3162 screened articles, 43 articles that described community-based maternal and child death review processes across ten countries in Africa and Asia were included. A variety of approaches were used to report and investigate deaths in the community, including identification of deaths by community health workers (CHWs) and other community informants, reproductive age mortality surveys, verbal autopsy, and social autopsy. Community notification of deaths by CHWs complements registration of maternal and child deaths missed by routinely collected sources of information, including the CRVS systems which mostly capture deaths occurring in health facilities. However, the accuracy and completeness of data reported by CHWs are sub-optimal.

CONCLUSIONS

Community-based death reporting complements formal registration of maternal and child deaths in LMICs. While research shows that community-based maternal and child death reporting was feasible, the accuracy and completeness of data reported by CHWs are sub-optimal but amenable to targeted support and supervision. Studies to further improve the process of engaging communities in the review, as well as collection and investigation of deaths in LMICs, could empower communities to respond more effectively and have a greater impact on reducing maternal and child mortality.

Does recall time matter in verbal autopsies? Evidence from urban informal settlements in Nairobi, Kenya

Background: To assign a cause of death to non-medically certified deaths, verbal autopsies (VAs) are widely used to determine the cause of death. The time difference between the death and the VA interview, also referred to as recall time, varies depending on social and operational factors surrounding the death. We investigated the effect of recall time on the assignment of causes of death by VA. Methods: This is a secondary analysis of 2002-2015 survey data of the Nairobi Urban Health Demographic Surveillance System (NUHDSS). The independent variable recall time was derived from the date of death and the date when the VA was conducted. Univariate and multivariate logistic regression methods were used to calculate odds ratios of assigning a cause of death in defined categories of recall time. Results: There were 6218 deaths followed up between 2002 and 2016, out of which 5495 (88.3%) had VAs done. Recall time varied from 1-3001 days (median  92 days, IQR 44-169 days). Majority of the VAs (45.7%) were conducted between 1-3 months after death. The effect of recall time varied for different diseases. Compared to VAs conducted between 1-3 months, there was a 24% higher likelihood of identifying HIV/AIDS as the cause of death for VAs conducted 4-6 months after death (AOR 1.24; 95% CI 1.01-1.54; p-value = 0.043) and a 40% increased chance of identifying other infectious diseases as the cause of death for VAs conducted <1 month after death (AOR 1.4; 95% CI 1.02-1.92, p-value = 0.024). Conclusions: Recall time affected the assignment of VA cause of death for HIV/AIDS, other infectious diseases,maternal/neonatal and indeterminate causes. Our analysis indicates that in the urban informal setting, VAs should be conducted from one month up to 6 months after the death to improve the probability of accurately assigning the cause of death.

Ascertaining cause of mortality among middle-aged and older persons using computer-coded and expert review verbal autopsies in the China Health and Retirement Longitudinal Study

BACKGROUND

Verbal autopsy is designed to ascertain causes of death that are not registered or certified. Verbal autopsy has been validated in multiple settings but has not been as widely evaluated for older populations as for younger age groups.

OBJECTIVE

This study aims to provide empirical evidence of the value of verbal autopsy interviews in the context of population-based surveys of older adults by comparing the cause-of-death assignments derived from two methods of interpreting verbal autopsy data.

METHODS

Data used in this study come from the China Health and Retirement Longitudinal Study, a nationally representative longitudinal survey of older Chinese. We compared 407 causes of death determined using InterVA, which is a computer-coded method, and causes of death as assigned by experts; then evaluated factors that affect the results of the two approaches.

RESULTS

Among the 407 deaths, neoplasms, cardiac disease, and stroke are the leading causes of death according to both approaches. The consistency of the two approaches is about 45% at the individual level. The primary reason for the mismatch is that no cause of death could be assigned for more than 25% of the sample based on expert review. A higher likelihood of mismatch is associated with advanced age and a long period between death and verbal autopsy interview.

CONCLUSION

Both approaches identify the same leading causes of death at the aggregate level, but consistency is relatively low at the individual level. InterVA works well when causes of death are characterized by distinctive signs and symptoms. Grouping the various causes of death with shared etiology or common risk factors may help improve the quality of the ascertainment of causes of death. Open-ended narratives are helpful because they provide information about the circumstances surrounding the death that are not available in the structured verbal autopsy interviews.

COVID-19 Home Deaths without Medical Assistance in Northeastern Brazil

Since its beginning in Wuhan, China, in December 2019, the disease caused by COVID-19 has reached more than 27 million confirmed cases and more than 880 thousand deaths worldwide by early September 2020. Although it is known that some of these deaths may have been influenced by the overload of health systems, the world medical literature lacks data on deaths due to COVID-19 in patients who have not received medical assistance. We conducted a retrospective transversal study to report the clinical and epidemiological profile of the first 200 consecutive cases of home deaths without medical assistance caused by COVID-19 diagnosed by verbal autopsy and real-time PCR in samples of postmortem nasopharyngeal swabs, in the state of Ceara, in Northeastern Brazil. The data show a slightly increased prevalence of cases in males (57%) and an average age of 76.8 years. Previous comorbidities were reported in 85.5% of cases, the most common being cardiovascular disease (45%), neurological disease (30%), and diabetes (29%). The main symptoms reported were dyspnea (79%), fever (75.5%), cough (69%), and fatigue (42.5%). The average time between the onset of illness and death was 7.3 days, being statistically shorter in patients who had previous comorbidities (P = 0.0215). This is the first study to evidence the clinical and epidemiological characteristics of COVID-19 home deaths without medical assistance, which may represent a considerable portion of the pandemic burden, especially in the context of health system overload.

Non-communicable diseases are the leading cause of mortality in rural Birbhum, West Bengal, India: a sex-stratified analysis of verbal autopsies from a prospective cohort, 2012-2017

OBJECTIVES

There is a dearth of data on causes of death in rural India, which impedes identification of public health priorities to guide health interventions. This study aims to offer insights from verbal autopsies, to understand the pattern and distribution of causes of death in a rural area of Birbhum District, West Bengal, India.

DESIGN

Causes of death data were retrieved from a prospective vital event surveillance system.

SETTING

The Birbhum Population Project, a Health and Demographic Surveillance System, West Bengal, India.

PARTICIPANTS

Between January 2012 and December 2017, all deaths were recorded.

MAIN OUTCOME MEASURES

Trained Surveyors tracked all deaths prospectively and used a previously validated verbal autopsy (VA) tool to record causes of death. Experienced physicians reviewed completed VA forms, and assigned cause of death using the 10th version of International Classification of Diseases. In addition to cause-specific mortality fraction, cause-specific crude death rate (CDR) among males and females were estimated.

RESULTS

A total of 2320 deaths (1348 males and 972 females) were recorded. An estimated CDR was 708/100 000. Over half of all deaths (1176 deaths, 50.7%) were attributed to non-communicable diseases (NCDs), with nearly 30% of all deaths attributed to circulatory system disorders; whereas 24.2% and 3.9% deaths were due to cerebrovascular diseases and ischaemic heart disease, respectively. Equal percent (13%) of males died from external causes and from infectious and parasitic diseases, and 11% died from respiratory system-related diseases. Among females, 12% died from infectious and parasitic diseases. Among children aged 0-4 years, 50% of all male deaths and 45% of all female deaths were attributed to conditions in the perinatal period.

CONCLUSIONS

NCDs are the leading cause of death among adults in a select population of rural Birbhum, India. Health programmes for rural India should prioritise plans to mitigate deaths due to NCDs.

Does recall time matter in verbal autopsies? Evidence from urban informal settlements in Nairobi, Kenya

Background: To assign a cause of death to non-medically certified deaths, verbal autopsies (VAs) are widely used to determine the cause of death. The time difference between the death and the VA interview, also referred to as recall time, varies depending on social and operational factors surrounding the death. We investigated the effect of recall time on the assignment of causes of death by VA. Methods: This is a secondary analysis of 2002-2015 survey data of the Nairobi Urban Health Demographic Surveillance System (NUHDSS). The independent variable recall time was derived from the date of death and the date when the VA was conducted. Univariate and multivariate logistic regression methods were used to calculate odds ratios of assigning a cause of death in defined categories of recall time. Results: There were 6218 deaths followed up between 2002 and 2016, out of which 5495 (88.3%) had VAs done. Recall time varied from 1-3001 days (mean 115.5 days, sd216.8). Majority of the VAs (45.7%) were conducted between 1-3 months after death. The effect of recall time varied for different diseases. Compared to VAs conducted between 1-3 months, there was a 24% higher likelihood of identifying HIV/AIDS as the cause of death for VAs conducted 4-6 months after death (AOR 1.24; 95% CI 1.01-1.54; p-value = 0.043) and a 40% increased chance of identifying other infectious diseases as the cause of death for VAs conducted <1 month after death (AOR 1.4; 95% CI 1.02-1.92, p-value = 0.024). Conclusions: Recall time affected the assignment of VA cause of death for HIV/AIDS, other infectious diseases and maternal/neonatal causes. Our analysis indicates that in the urban informal setting, VAs should be conducted from one month up to 6 months after the death to improve the probability of accurately assigning the cause of death.

Unraveling Specific Causes of Neonatal Mortality Using Minimally Invasive Tissue Sampling: An Observational Study

BACKGROUND

Postmortem minimally invasive tissue sampling (MITS) is a potential alternative to the gold standard complete diagnostic autopsy for identifying specific causes of childhood deaths. We investigated the utility of MITS, interpreted with available clinical data, for attributing underlying and immediate causes of neonatal deaths.

METHODS

This prospective, observational pilot study enrolled neonatal deaths at Chris Hani Baragwanath Academic Hospital in Soweto, South Africa. The MITS included needle core-biopsy sampling for histopathology of brain, lung, and liver tissue. Microbiological culture and/or molecular tests were performed on lung, liver, blood, cerebrospinal fluid, and stool samples. The "underlying" and "immediate" causes of death (CoD) were determined for each case by an international panel of 12-15 medical specialists.

RESULTS

We enrolled 153 neonatal deaths, 106 aged 3-28 days. Leading underlying CoD included "complications of prematurity" (52.9%), "complications of intrapartum events" (15.0%), "congenital malformations" (13.1%), and "infection related" (9.8%). Overall, infections were the immediate or underlying CoD in 57.5% (n = 88) of all neonatal deaths, including the immediate CoD in 70.4% (58/81) of neonates with "complications of prematurity" as the underlying cause. Overall, 74.4% of 90 infection-related deaths were hospital acquired, mainly due to multidrug-resistant Acinetobacter baumannii (52.2%), Klebsiella pneumoniae (22.4%), and Staphylococcus aureus (20.9%). Streptococcus agalactiae was the most common pathogen (5/15 [33.3%]) among deaths with "infections" as the underlying cause.

CONCLUSIONS

MITS has potential to address the knowledge gap on specific causes of neonatal mortality. In our setting, this included the hitherto underrecognized dominant role of hospital-acquired multidrug-resistant bacterial infections as the leading immediate cause of neonatal deaths.

Malaria infection, disease and mortality among children and adults on the coast of Kenya

BACKGROUND

Malaria transmission has recently fallen in many parts of Africa, but systematic descriptions of infection and disease across all age groups are rare. Here, an epidemiological investigation of parasite prevalence, the incidence of fevers associated with infection, severe hospitalized disease and mortality among children older than 6 months and adults on the Kenyan coast is presented.

METHODS

A prospective fever surveillance was undertaken at 6 out-patients (OPD) health-facilities between March 2018 and February 2019. Four community-based, cross sectional surveys of fever history and infection prevalence were completed among randomly selected homestead members from the same communities. Paediatric and adult malaria at Kilifi county hospital was obtained for the 12 months period. Rapid Diagnostic Tests (CareStart™ RDT) to detect HRP2-specific to Plasmodium falciparum was used in the community and the OPD, and microscopy in the hospital. Crude and age-specific incidence rates were computed using Poisson regression.

RESULTS

Parasite prevalence gradually increased from childhood, reaching 12% by 9 years of age then declining through adolescence into adulthood. The incidence rate of RDT positivity in the OPD followed a similar trend to that of infection prevalence in the community. The incidence of hospitalized malaria from the same community was concentrated among children aged 6 months to 4 years (i.e. 64% and 70% of all hospitalized and severe malaria during the 12 months of surveillance, respectively). Only 3.7% (12/316) of deaths were directly attributable to malaria. Malaria mortality was highest among children aged 6 months-4 years at 0.57 per 1000 person-years (95% CI 0.2, 1.2). Severe malaria and death from malaria was negligible above 15 years of age.

CONCLUSION

Under conditions of low transmission intensity, immunity to disease and the fatal consequences of infection appear to continue to be acquired in childhood and faster than anti-parasitic immunity. There was no evidence of an emerging significant burden of severe malaria or malaria mortality among adults. This is contrary to current modelled approaches to disease burden estimation in Africa and has important implications for the targeting of infection prevention strategies based on chemoprevention or vector control.

Regularized Bayesian transfer learning for population-level etiological distributions

Computer-coded verbal autopsy (CCVA) algorithms predict cause of death from high-dimensional family questionnaire data (verbal autopsy) of a deceased individual, which are then aggregated to generate national and regional estimates of cause-specific mortality fractions. These estimates may be inaccurate if CCVA is trained on non-local training data different from the local population of interest. This problem is a special case of transfer learning, i.e., improving classification within a target domain (e.g., a particular population) with the classifier trained in a source-domain. Most transfer learning approaches concern individual-level (e.g., a person’s) classification. Social and health scientists such as epidemiologists are often more interested with understanding etiological distributions at the population-level. The sample sizes of their data sets are typically orders of magnitude smaller than those used for common transfer learning applications like image classification, document identification, etc. We present a parsimonious hierarchical Bayesian transfer learning framework to directly estimate population-level class probabilities in a target domain, using any baseline classifier trained on source-domain, and a small labeled target-domain dataset. To address small sample sizes, we introduce a novel shrinkage prior for the transfer error rates guaranteeing that, in absence of any labeled target-domain data or when the baseline classifier is perfectly accurate, our transfer learning agrees with direct aggregation of predictions from the baseline classifier, thereby subsuming the default practice as a special case. We then extend our approach to use an ensemble of baseline classifiers producing an unified estimate. Theoretical and empirical results demonstrate how the ensemble model favors the most accurate baseline classifier. We present data analyses demonstrating the utility of our approach.

Validation of verbal autopsy methods for assessment of child mortality in sub-Saharan Africa and the policy implication: a rapid review

Reliable data on the cause of child death is the cornerstone for evidence-informed health policy making towards improving child health outcomes. Unfortunately, accurate data on cause of death is essentially lacking in most countries of sub-Saharan Africa due to the widespread absence of functional Civil Registration and Vital Statistics (CRVS) systems. To address this problem, verbal autopsy (VA) has gained prominence as a strategy for obtaining Cause of Death (COD) information in populations where CRVS are absent. This study reviewed publications that investigated the validation of VA methods for assessment of COD. A MEDLINE PubMed search was undertaken in June 2018 for studies published in English that investigated the validation of VA methods in sub-Saharan Africa from 1990-2018. Of the 17 studies identified, 9 fulfilled the study inclusion criteria from which additional five relevant studies were found by reviewing their references. The result showed that Physician-Certified Verbal Autopsy (PCVA) was the most widely used VA method. Validation studies comparing PCVA to hospital records, expert algorithm and InterVA demonstrated mixed and highly varied outcomes. The accuracy and reliability of the VA methods depended on level of healthcare the respondents have access to and the knowledge of the physicians on the local disease aetiology and epidemiology. As the countries in sub-Saharan Africa continue to battle with dysfunctional CRVS system, VA will remain the only viable option for the supply of child mortality data necessary for policy making.

Automated versus physician assignment of cause of death for verbal autopsies: randomized trial of 9374 deaths in 117 villages in India

BACKGROUND

Verbal autopsies with physician assignment of cause of death (COD) are commonly used in settings where medical certification of deaths is uncommon. It remains unanswered if automated algorithms can replace physician assignment.

METHODS

We randomized verbal autopsy interviews for deaths in 117 villages in rural India to either physician or automated COD assignment. Twenty-four trained lay (non-medical) surveyors applied the allocated method using a laptop-based electronic system. Two of 25 physicians were allocated randomly to independently code the deaths in the physician assignment arm. Six algorithms (Naïve Bayes Classifier (NBC), King-Lu, InSilicoVA, InSilicoVA-NT, InterVA-4, and SmartVA) coded each death in the automated arm. The primary outcome was concordance with the COD distribution in the standard physician-assigned arm. Four thousand six hundred fifty-one (4651) deaths were allocated to physician (standard), and 4723 to automated arms.

RESULTS

The two arms were nearly identical in demographics and key symptom patterns. The average concordances of automated algorithms with the standard were 62%, 56%, and 59% for adult, child, and neonatal deaths, respectively. Automated algorithms showed inconsistent results, even for causes that are relatively easy to identify such as road traffic injuries. Automated algorithms underestimated the number of cancer and suicide deaths in adults and overestimated other injuries in adults and children. Across all ages, average weighted concordance with the standard was 62% (range 79-45%) with the best to worst ranking automated algorithms being InterVA-4, InSilicoVA-NT, InSilicoVA, SmartVA, NBC, and King-Lu. Individual-level sensitivity for causes of adult deaths in the automated arm was low between the algorithms but high between two independent physicians in the physician arm.

CONCLUSIONS

While desirable, automated algorithms require further development and rigorous evaluation. Lay reporting of deaths paired with physician COD assignment of verbal autopsies, despite some limitations, remains a practicable method to document the patterns of mortality reliably for unattended deaths.

TRIAL REGISTRATION

ClinicalTrials.gov , NCT02810366. Submitted on 11 April 2016.

Engaging communities in collecting and using results from verbal autopsies for child deaths: an example from urban slums in Freetown, Sierra Leone

BACKGROUND

Verbal autopsies (VAs) can provide important epidemiological information about the causes of child deaths. Though studies have been conducted to assess the validity of various types of VAs, the programmatic experience of engaging local communities in collecting and using VA has received little attention in the published literature. Concern Worldwide, an international non-governmental organization (NGO), in collaboration with the Ministry of Health and Sanitation (MOHS), has implemented a VA protocol in five urban slums of Freetown, Sierra Leone. This paper provides VA results and describes lessons learned from the VA process.

METHODS

Under-five child deaths were registered by Community Health Workers (CHWs) in five urban slums between 2014 and 2017, and a specially trained local clinician used a VA protocol to interview caretakers. Symptoms were analysed using InterVA-4 computerized algorithm, a probabilistic expert-driven model to determine the most likely cause of death. Themes in care-seeking were extracted from multiple-choice and open-ended questions. VAs were implemented in collaboration with the community and the results were shared with community stakeholders in participatory review meetings.

RESULTS

Main challenges included limitations in death notification and capacity to conduct VA for all notified deaths. A total of 215 VA were available for analysis. Among 79 neonatal deaths aged 0-27 days, the most common cause of death was neonatal pneumonia (55%); among 136 children deaths aged 1-4 years, the most common causes were malaria (56%) and pneumonia (41%). Key themes in care-seeking identified included use of traditional medicine (14% of deaths), absence of care-seeking (23% of deaths), and difficultly reaching the health facility (8% of deaths that occurred at home) during fatal illness.

CONCLUSIONS

Conducting VAs as a collaborative process with communities is challenging but can provide valuable data that can be used for local-level decision-making. The findings have practical implications for engaging the community and CHWs in reducing the number of these preventable deaths through expanded efforts at prevention, early and appropriate treatment, and reduction of barriers to care-seeking. A functional end-to-end VA system can enhance meaningful routine vital events monitoring by community, national, and international stakeholders.

Automated verbal autopsy classification: using one-against-all ensemble method and Naïve Bayes classifier

Verbal autopsy (VA) deals with post-mortem surveys about deaths, mostly in low and middle income countries, where the majority of deaths occur at home rather than a hospital, for retrospective assignment of causes of death (COD) and subsequently evidence-based health system strengthening. Automated algorithms for VA COD assignment have been developed and their performance has been assessed against physician and clinical diagnoses. Since the performance of automated classification methods remains low, we aimed to enhance the Naïve Bayes Classifier (NBC) algorithm to produce better ranked COD classifications on 26,766 deaths from four globally diverse VA datasets compared to some of the leading VA classification methods, namely Tariff, InterVA-4, InSilicoVA and NBC. We used a different strategy, by training multiple NBC algorithms using the one-against-all approach (OAA-NBC). To compare performance, we computed the cumulative cause-specific mortality fraction (CSMF) accuracies for population-level agreement from rank one to five COD classifications. To assess individual-level COD assignments, cumulative partially-chance corrected concordance (PCCC) and sensitivity was measured for up to five ranked classifications. Overall results show that OAA-NBC consistently assigns CODs that are the most alike physician and clinical COD assignments compared to some of the leading algorithms based on the cumulative CSMF accuracy, PCCC and sensitivity scores. The results demonstrate that our approach improves the performance of classification (sensitivity) by between 6% and 8% compared with other VA algorithms. Population-level agreements for OAA-NBC and NBC were found to be similar or higher than the other algorithms used in the experiments. Although OAA-NBC still requires improvement for individual-level COD assignment, the one-against-all approach improved its ability to assign CODs that more closely resemble physician or clinical COD classifications compared to some of the other leading VA classifiers.

Automated verbal autopsy classification: using one-against-all ensemble method and Naïve Bayes classifier

Verbal autopsy (VA) deals with post-mortem surveys about deaths, mostly in low and middle income countries, where the majority of deaths occur at home rather than a hospital, for retrospective assignment of causes of death (COD) and subsequently evidence-based health system strengthening. Automated algorithms for VA COD assignment have been developed and their performance has been assessed against physician and clinical diagnoses. Since the performance of automated classification methods remains low, we aimed to enhance the Naïve Bayes Classifier (NBC) algorithm to produce better ranked COD classifications on 26,766 deaths from four globally diverse VA datasets compared to some of the leading VA classification methods, namely Tariff, InterVA-4, InSilicoVA and NBC. We used a different strategy, by training multiple NBC algorithms using the one-against-all approach (OAA-NBC). To compare performance, we computed the cumulative cause-specific mortality fraction (CSMF) accuracies for population-level agreement from rank one to five COD classifications. To assess individual-level COD assignments, cumulative partially-chance corrected concordance (PCCC) and sensitivity was measured for up to five ranked classifications. Overall results show that OAA-NBC consistently assigns CODs that are the most alike physician and clinical COD assignments compared to some of the leading algorithms based on the cumulative CSMF accuracy, PCCC and sensitivity scores. The results demonstrate that our approach improves the performance of classification (sensitivity) by between 6% and 8% compared with other VA algorithms. Population-level agreements for OAA-NBC and NBC were found to be similar or higher than the other algorithms used in the experiments. Although OAA-NBC still requires improvement for individual-level COD assignment, the one-against-all approach improved its ability to assign CODs that more closely resemble physician or clinical COD classifications compared to some of the other leading VA classifiers.

Socio-anthropological methods to study the feasibility and acceptability of the minimally invasive autopsy from the perspective of local communities: lessons learnt from a large multi-centre study

The minimally invasive autopsy (MIA), an innovative approach for obtaining post-mortem samples of key organs, is increasingly being recognized as a robust methodology for cause of death (CoD) investigation, albeit so far limited to pilot studies and research projects. A better understanding of the real causes of death in middle- and low-income countries, where underlying causes of death are seldom determined, would allow improved health planning, more targeted prioritization of available resources and the implementation of coherent public health policies. This paper discusses lessons learnt from the implementation of a Feasibility and Acceptability (F&A) study evaluating the MIA approach in five countries: Gabon, Kenya, Mali, Mozambique and Pakistan. This article reports the methodological choices made to document sociocultural and religious norms around death, to examine community and relatives' attitudes and perceptions towards MIA, and to identify factors motivating the MIA's acceptance and refusal. We used ethnography, grounded theory and framework method approaches. In-depth and semi-structured interviews and focus group discussions with key informants, including next of kin of deceased individuals and healthcare providers, were conducted. Participant observation and direct observation of procedures and ceremonies around death were organized in all study sites. In Mozambique, MIA procedures were observed and case studies conducted. The implementation of this F&A protocol has provided critical lessons that could facilitate the future implementation of post-mortem procedures for CoD investigation. These include the need for early community engagement, staff training and preparedness, flexibility to adapt the protocol, gathering qualitative data from diverse sources, and triangulation of the data. We have applied a rigorous, effective and culturally sensitive methodological approach to assess the F&A of MIA in resource-constrained settings. We strongly recommend that such an approach is applied in settings where MIAs or similar post-mortem sensitive procedures are to be introduced.

Integrating community-based verbal autopsy into civil registration and vital statistics (CRVS): system-level considerations

Background: Reliable and representative cause of death (COD) statistics are essential to inform public health policy, respond to emerging health needs, and document progress towards Sustainable Development Goals. However, less than one-third of deaths worldwide are assigned a cause. Civil registration and vital statistics (CRVS) systems in low- and lower-middle-income countries are failing to provide timely, complete and accurate vital statistics, and it will still be some time before they can provide physician-certified COD for every death.

Proposals: Verbal autopsy (VA) is a method to ascertain the probable COD and, although imperfect, it is the best alternative in the absence of medical certification. There is extensive experience with VA in research settings but only a few examples of its use on a large scale. Data collection using electronic questionnaires on mobile devices and computer algorithms to analyse responses and estimate probable COD have increased the potential for VA to be routinely applied in CRVS systems. However, a number of CRVS and health system integration issues should be considered in planning, piloting and implementing a system-wide intervention such as VA. These include addressing the multiplicity of stakeholders and sub-systems involved, integration with existing CRVS work processes and information flows, linking VA results to civil registration records, information technology requirements and data quality assurance.

Conclusions: Integrating VA within CRVS systems is not simply a technical undertaking. It will have profound system-wide effects that should be carefully considered when planning for an effective implementation. This paper identifies and discusses the major system-level issues and emerging practices, provides a planning checklist of system-level considerations and proposes an overview for how VA can be integrated into routine CRVS systems.

Performance of verbal autopsy methods in estimating HIV-associated mortality among adults in South Africa

INTRODUCTION

Verbal autopsy (VA) can be integrated into civil registration and vital statistics systems, but its accuracy in determining HIV-associated causes of death (CoD) is uncertain. We assessed the sensitivity and specificity of VA questions in determining HIV status and antiretroviral therapy (ART) initiation and compared HIV-associated mortality fractions assigned by different VA interpretation methods.

METHODS

Using the WHO 2012 instrument with added ART questions, VA was conducted for deaths among adults with known HIV status (356 HIV positive and 103 HIV negative) in South Africa. CoD were assigned using physician-certified VA (PCVA) and computer-coded VA (CCVA) methods and compared with documented HIV status.

RESULTS

The sensitivity of VA questions in detecting HIV status and ART initiation was 84.3% (95% CI 80 to 88) and 91.0% (95% CI 86 to 95); 283/356 (79.5%) HIV-positive individuals were assigned HIV-associated CoD by PCVA, 166 (46.6%) by InterVA-4.03, 201 (56.5%) by InterVA-5, and 80 (22.5%) and 289 (81.2%) by SmartVA-Analyze V.1.1.1 and V.1.2.1. Agreement between PCVA and older CCVA methods was poor (chance-corrected concordance [CCC] <0; cause-specific mortality fraction [CSMF] accuracy ≤56%) but better between PCVA and updated methods (CCC 0.21-0.75; CSMF accuracy 65%-98%). All methods were specific (specificity 87% to 96%) in assigning HIV-associated CoD.

CONCLUSION

All CCVA interpretation methods underestimated the HIV-associated mortality fraction compared with PCVA; InterVA-5 and SmartVA-Analyze V.1.2.1 performed better than earlier versions. Changes to VA methods and classification systems are needed to track progress towards targets for reducing HIV-associated mortality.

Pneumonia mortality and healthcare utilization in young children in rural Bangladesh: a prospective verbal autopsy study

Background

The present study aimed to examine the risk factors for death due to pneumonia in young children and healthcare behaviors of the guardians for children in rural Bangladesh. A prospective autopsy study was conducted among guardians of children aged 4 weeks to 59 months in Mirzapur, Bangladesh, from 2008 to 2012.

Results

Pneumonia was the primary cause of death, accounting for 26.4% (n = 81) of all 307 deaths. Of the pneumonia deaths, 58% (n = 47) deaths occurred in younger infants (aged 4 weeks to < 6 months) and 24.7% (n = 20) in older infants (aged 6–11 months). The median duration of illness before pneumonia death was 8 days (interquartile range [IQR] 3–20 days). Prior to death, 91.4% (n = 74) children with pneumonia sought treatment, and of those who sought treatment, 52.7% (n = 39) sought treatment ≥ 2 days after the onset of disease. Younger infants of 4 weeks to < 6 months old were at 5.5-time (95% confidence interval [CI] 2.5, 12.0) and older infants aged 6–11 months were at 3-time (1.2, 7.5) greater risk of dying from pneumonia than older children aged 12–59 months. Children with a prolonged duration of illness (2–10 days) prior to death were at more risk for death by pneumonia than those who died from other causes (5.8 [2.1, 16.1]). Children who died from pneumonia sought treatment 3.4-time more than children who died from other causes. Delayed treatment seeking (≥ 2 days) behavior was 4.9-time more common in children who died from pneumonia than those who died from other causes. Children who died from pneumonia more often had access to care from multiple sources (5.7-time) than children who died from other causes.

Conclusions

Delay in seeking appropriate care and access to multiple sources for treatment are the underlying risk factors for pneumonia death in young children in Bangladesh. These results indicate the perplexity in guardians’ decisions to secure appropriate treatment for children with pneumonia. Therefore, it further underscores the importance of focusing on mass media coverage that can outline the benefits of seeking care early in the progression of pneumonia and the potential negative consequences of seeking care late.

Electronic supplementary material

The online version of this article (10.1186/s41182-018-0099-4) contains supplementary material, which is available to authorized users.

Validation of the verbal autopsy questionnaire for adult deaths in Iran

Background: Verbal Autopsy Questionnaire (VA) is an important tool to estimate the cause of death among those populations lacking an appropriate death registration system. In this study, the validity and reliability of verbal autopsy were assessed. Methods: The Persian version of the questionnaire was prepared using the translation and back- translation method. In the first and second phases of the study, 213 and 198 families of deceased persons accepted an invitation to complete the questionnaire. A physician determined the cause of death. These causes were compared with the registered cause of death on the death certificate. Sensitivity, specificity, positive predictive value (PPV), as well as the kappa statistic (between the first verbal autopsy questioning and death registry) were calculated to determine the validity of the questionnaire. Kappa statistic was also used to determine the reliability between the first and second questioning. Results: The sensitivity of the questionnaire varied from 75% among deaths due to diabetes to 100% Due to breast cancer. The specificity of the questionnaire for all causes of death was higher than 97%. PPV varied from 62.5% to 100% for all causes of death. The kappa statistic between causes of death derived from death certificate, and the first VA questioning for all causes of death was above 0.7 (p<0.001), and it was above 0.78 for the first and second verbal autopsy questioning (p<0.001). Conclusion: Although the Verbal Autopsy Questionnaire does not fully identify all causes of death, it can be a useful tool for diagnosing causes of death for those deceased persons who have no death certificate.

Evaluation of data sources and approaches for estimation of influenza-associated mortality in India

BACKGROUND

No estimates of influenza-associated mortality exist for India.

OBJECTIVE

To evaluate national mortality and viral surveillance data from India for assessing their appropriateness in estimating influenza-associated mortality using varied analytic approaches.

METHODS

We reviewed influenza virus surveillance data from a national influenza surveillance network. We also reviewed national mortality data from Civil Registration System (CRS), Medical Certification of Cause of Death (MCCD) and the Sample Registration System (SRS). We compared and scored the different sources of mortality data using specific criteria, including the process of cause of death assignment, sample size, proportion of ill-defined deaths, representativeness and availability of time series data. Each of these 5 parameters was scored on a scale from 1 to 5. To evaluate how to generate an influenza-associated mortality estimate for India, we also reviewed 4 methodologic approaches to assess the appropriateness of their assumptions and requirements for these data sets.

RESULTS

The influenza virus surveillance data included year-round sample testing for influenza virus and was found to be suitable for influenza mortality estimation modelling. Based on scoring for the 5 mortality data criteria, the SRS data had the highest score with 20 of 25 possible score, whereas MCCD and CRS scored 16 and 12, respectively. The SRS which used verbal autopsy survey methods was determined to be nationally representative and thus adequate for estimating influenza-associated mortality. Evaluation of the modelling methods demonstrated that Poisson regression, risk difference and mortality multiplier methods could be applied to the Indian setting.

CONCLUSION

Despite significant challenges, it is possible to estimate influenza-associated mortality in India.

Determining the Cause of Death: Mortality Surveillance Using Verbal Autopsy in Indonesia

In the absence of a vital registration and health information systems, Indonesia does not have complete, accurate, and continuous data to summarize the mortality statistics of the population, nor determine the exact cause of death. Verbal autopsies performed in a community-based mortality surveillance have been used to provide information on the cause of deaths in such context. However, physician review of verbal autopsy can be expensive, time-consuming, and give inconsistent results, raising concern about its reliability. We used the Purworejo Health and Demographic Surveillance System's (HDSS) mortality data collected between 2000 and 2002 and assigned causes of death for all age groups using Interpreting Verbal Autopsy-4, analytic software that applies a probabilistic model. A total of 1,999 deaths were identified among 55,581 individuals surveyed in 14,409 households; 830 deaths were able to be recorded using the standardized World Health Organization (WHO) verbal autopsy questionnaire. We calculated the proportion of different causes of death and its incidence rate (IR) ratios with 95% confidence interval (CI) to compare the IR per person-years-observation (PYO). The IR of stroke was 126.7 per 100,000 PYO (95% CI: 109.7, 143.7); acute respiratory infection including pneumonia was 70.8 per 100,000 PYO (95% CI: 58.1, 83.5); and the IR of other and unspecified cardiac diseases was 57.7 per 100,000 PYO (95% CI: 46.2, 69.2). Stroke was indicated as the leading cause of death among elderly people aged 50 years and above. Meanwhile, pneumonia as a communicable disease was found to be the most common cause of death among both 0-14-year-old children and elderly people.

The Global Network Neonatal Cause of Death algorithm for low-resource settings

AIM

This study estimated the causes of neonatal death using an algorithm for low-resource areas, where 98% of the world's neonatal deaths occur.

METHODS

We enrolled women in India, Pakistan, Guatemala, the Democratic Republic of Congo, Kenya and Zambia from 2014 to 2016 and tracked their delivery and newborn outcomes for up to 28 days. Antenatal care and delivery symptoms were collected using a structured questionnaire, clinical observation and/or a physical examination. The Global Network Cause of Death algorithm was used to assign the cause of neonatal death, analysed by country and day of death.

RESULTS

One-third (33.1%) of the 3068 neonatal deaths were due to suspected infection, 30.8% to prematurity, 21.2% to asphyxia, 9.5% to congenital anomalies and 5.4% did not have a cause of death assigned. Prematurity and asphyxia-related deaths were more common on the first day of life (46.7% and 52.9%, respectively), while most deaths due to infection occurred after the first day of life (86.9%). The distribution of causes was similar to global data reported by other major studies.

CONCLUSION

The Global Network algorithm provided a reliable cause of neonatal death in low-resource settings and can be used to inform public health strategies to reduce mortality.

Burden, timing and causes of maternal and neonatal deaths and stillbirths in sub-Saharan Africa and South Asia: protocol for a prospective cohort study

Objectives

The AMANHI mortality study aims to use harmonized methods, across eleven sites in eight countries in South Asia and sub–Saharan Africa, to estimate the burden, timing and causes of maternal, fetal and neonatal deaths. It will generate data to help advance the science of cause of death (COD) assignment in developing country settings.

Methods

This population–based, cohort study is being conducted in the eleven sites where approximately 2 million women of reproductive age are under surveillance to identify and follow–up pregnancies through to six weeks postpartum. All sites are implementing uniform protocols. Verbal autopsies (VAs) are conducted for deaths of pregnant women, newborns or stillbirths to confirm deaths, ascertain timing and collect data on the circumstances around the death to help assign causes. Physicians from the sites are selected and trained to use International Classification of Diseases (ICD) principles to assign CODs from a limited list of programmatically–relevant causes. Where the cause cannot be determined from the VA, physicians assign that option. Every physician who is trained to assign causes of deaths from any of the study countries is tested and accredited before they start COD assignment in AMANHI.

Importance of the AMANHI mortality study

It is one of the first to generate improved estimates of burden, timing and causes of maternal, fetal and neonatal deaths from empirical data systematically collected in a large prospective cohort of women of reproductive age. AMANHI makes a substantial contribution to global knowledge to inform policies, interventions and investment decisions to reduce these deaths.

Measuring mortality due to HIV-associated tuberculosis among adults in South Africa: Comparing verbal autopsy, minimally-invasive autopsy, and research data

BACKGROUND

The World Health Organization (WHO) aims to reduce tuberculosis (TB) deaths by 95% by 2035; tracking progress requires accurate measurement of TB mortality. International Classification of Diseases (ICD) codes do not differentiate between HIV-associated TB and HIV more generally. Verbal autopsy (VA) is used to estimate cause of death (CoD) patterns but has mostly been validated against a suboptimal gold standard for HIV and TB. This study, conducted among HIV-positive adults, aimed to estimate the accuracy of VA in ascertaining TB and HIV CoD when compared to a reference standard derived from a variety of clinical sources including, in some, minimally-invasive autopsy (MIA).

METHODS AND FINDINGS

Decedents were enrolled into a trial of empirical TB treatment or a cohort exploring diagnostic algorithms for TB in South Africa. The WHO 2012 instrument was used; VA CoD were assigned using physician-certified VA (PCVA), InterVA-4, and SmartVA-Analyze. Reference CoD were assigned using MIA, research, and health facility data, as available. 259 VAs were completed: 147 (57%) decedents were female; median age was 39 (interquartile range [IQR] 33-47) years and CD4 count 51 (IQR 22-102) cells/μL. Compared to reference CoD that included MIA (n = 34), VA underestimated mortality due to HIV/AIDS (94% reference, 74% PCVA, 47% InterVA-4, and 41% SmartVA-Analyze; chance-corrected concordance [CCC] 0.71, 0.42, and 0.31, respectively) and HIV-associated TB (41% reference, 32% PCVA; CCC 0.23). For individual decedents, all VA methods agreed poorly with reference CoD that did not include MIA (n = 259; overall CCC 0.14, 0.06, and 0.15 for PCVA, InterVA-4, and SmartVA-Analyze); agreement was better at population level (cause-specific mortality fraction accuracy 0.78, 0.61, and 0.57, for the three methods, respectively).

CONCLUSIONS

Current VA methods underestimate mortality due to HIV-associated TB. ICD and VA methods need modifications that allow for more specific evaluation of HIV-related deaths and direct estimation of mortality due to HIV-associated TB.

Verbal autopsy-assigned causes of death among adults being investigated for TB in South Africa

Background

Adults being investigated for TB in South Africa experience high mortality, yet causes of death (CoD) are not well defined. We determined CoD in this population using verbal autopsy (VA), and compared HIV- and TB-associated CoD using physician-certified verbal autopsy (PCVA) and InterVA-4 software.

Methods

All contactable consenting caregivers of participants who died during a trial comparing Xpert MTB/RIF to smear microscopy were interviewed using the WHO VA tool. CoD were assigned using PCVA and InterVA-4. Kappa statistic (K) and concordance correlation coefficient (CCC) were calculated for comparison.

Results

Among 231 deaths, relatives of 137 deceased were interviewed. Of the 137 deceased 76 (55.4%) were males, median age 41 years (IQR 33–50). PCVA assigned 70 (51.1%) TB immediate CoD (44 [62.8%] pulmonary TB; 26 [37.1%] extra-pulmonary TB); 21 (15.3%) HIV/AIDS-related; and 46 (33.5%) other CoD. InterVA-4 assigned 48 (35.0%) TB deaths; 49 (35.7%) HIV/AIDS-related deaths; and 40 (29.1%) other CoD. Agreement between PCVA and InterVA-4 CoD was slight at individual level (K=0.20; 95% CI 0.10–0.30) and poor at population level (CCC 0.67; 95% CI 0.38–0.99).

Conclusions

TB and HIV are leading CoD among adults being investigated for TB. PCVA and InterVA agreement at individual level was slight and poor at population level. VA methodology needs further development where TB and HIV are common.

Main causes of death in Dande, Angola: results from Verbal Autopsies of deaths occurring during 2009-2012

BACKGROUND

The Dande Health and Demographic Surveillance System (HDSS) located in Bengo Province, Angola, covers nearly 65,500 residents living in approximately 19,800 households. This study aims to describe the main causes of deaths (CoD) occurred within the HDSS, from 2009 to 2012, and to explore associations between demographic or socioeconomic factors and broad mortality groups (Group I-Communicable diseases, maternal, perinatal and nutritional conditions; Group II-Non-communicable diseases; Group III-Injuries; IND-Indeterminate).

METHODS

Verbal Autopsies (VA) were performed after death identification during routine HDSS visits. Associations between broad groups of CoD and sex, age, education, socioeconomic position, place of residence and place of death, were explored using chi-square tests and fitting logistic regression models.

RESULTS

From a total of 1488 deaths registered, 1009 verbal autopsies were performed and 798 of these were assigned a CoD based on the 10(th) revision of the International Classification of Diseases (ICD-10). Mortality was led by CD (61.0 %), followed by IND (18.3 %), NCD (11.6 %) and INJ (9.1 %). Intestinal infectious diseases, malnutrition and acute respiratory infections were the main contributors to under-five mortality (44.2 %). Malaria was the most common CoD among children under 15 years old (38.6 %). Tuberculosis, traffic accidents and malaria led the CoD among adults aged 15-49 (13.5 %, 10.5 % and 8.0 % respectively). Among adults aged 50 or more, diseases of the circulatory system (23.2 %) were the major CoD, followed by tuberculosis (8.2 %) and malaria (7.7 %). CD were more frequent CoD among less educated people (adjusted odds ratio, 95 % confidence interval for none vs. 5 or more years of school: 1.68, 1.04-2.72).

CONCLUSION

Infectious diseases were the leading CoD in this region. Verbal autopsies proved useful to identify the main CoD, being an important tool in settings where vital statistics are scarce and death registration systems have limitations.

Validating hierarchical verbal autopsy expert algorithms in a large data set with known causes of death

BACKGROUND

Physician assessment historically has been the most common method of analyzing verbal autopsy (VA) data. Recently, the World Health Organization endorsed two automated methods, Tariff 2.0 and InterVA-4, which promise greater objectivity and lower cost. A disadvantage of the Tariff method is that it requires a training data set from a prior validation study, while InterVA relies on clinically specified conditional probabilities. We undertook to validate the hierarchical expert algorithm analysis of VA data, an automated, intuitive, deterministic method that does not require a training data set.

METHODS

Using Population Health Metrics Research Consortium study hospital source data, we compared the primary causes of 1629 neonatal and 1456 1-59 month-old child deaths from VA expert algorithms arranged in a hierarchy to their reference standard causes. The expert algorithms were held constant, while five prior and one new "compromise" neonatal hierarchy, and three former child hierarchies were tested. For each comparison, the reference standard data were resampled 1000 times within the range of cause-specific mortality fractions (CSMF) for one of three approximated community scenarios in the 2013 WHO global causes of death, plus one random mortality cause proportions scenario. We utilized CSMF accuracy to assess overall population-level validity, and the absolute difference between VA and reference standard CSMFs to examine particular causes. Chance-corrected concordance (CCC) and Cohen's kappa were used to evaluate individual-level cause assignment.

RESULTS

Overall CSMF accuracy for the best-performing expert algorithm hierarchy was 0.80 (range 0.57-0.96) for neonatal deaths and 0.76 (0.50-0.97) for child deaths. Performance for particular causes of death varied, with fairly flat estimated CSMF over a range of reference values for several causes. Performance at the individual diagnosis level was also less favorable than that for overall CSMF (neonatal: best CCC = 0.23, range 0.16-0.33; best kappa = 0.29, 0.23-0.35; child: best CCC = 0.40, 0.19-0.45; best kappa = 0.29, 0.07-0.35).

CONCLUSIONS

Expert algorithms in a hierarchy offer an accessible, automated method for assigning VA causes of death. Overall population-level accuracy is similar to that of more complex machine learning methods, but without need for a training data set from a prior validation study.

Measuring maternal, foetal and neonatal mortality: Challenges and solutions

Levels and causes of mortality in mothers and babies are intrinsically linked, occurring at the same time and often to the same mother-baby dyad, although mortality rates are substantially higher in babies. Measuring levels, trends and causes of maternal, neonatal and foetal mortality are important for understanding priority areas for interventions and tracking the success of interventions at the global, national, regional and local level. However, there are many measurement challenges. This paper provides an overview of the definitions and indicators for measuring mortality in pregnant and post-partum women (maternal and pregnancy-related mortality) and their babies (foetal and neonatal mortality). We then discuss current issues in the measurement of the levels and causes of maternal, foetal and neonatal mortality, and present options for improving measurement of these outcomes. Finally, we illustrate some important uses of mortality data, including for the development of models to estimate mortality rates at the global and national level and for audits.

The quality and diagnostic value of open narratives in verbal autopsy: a mixed-methods analysis of partnered interviews from Malawi

BACKGROUND

Verbal autopsy (VA), the process of interviewing a deceased's family or caregiver about signs and symptoms leading up to death, employs tools that ask a series of closed questions and can include an open narrative where respondents give an unprompted account of events preceding death. The extent to which an individual interviewer, who generally does not interpret the data, affects the quality of this data, and therefore the assigned cause of death, is poorly documented. We aimed to examine inter-interviewer reliability of open narrative and closed question data gathered during VA interviews.

METHODS

During the introduction of VA data collection, as part of a larger study in Mchinji district, Malawi, we conducted partner interviews whereby two interviewers independently recorded open narrative and closed questions during the same interview. Closed questions were collected using a smartphone application (mobile-InterVA) and open narratives using pen and paper. We used mixed methods of analysis to evaluate the differences between recorded responses to open narratives and closed questions, causes of death assigned, and additional information gathered by open narrative.

RESULTS

Eighteen partner interviews were conducted, with complete data for 11 pairs. Comparing closed questions between interviewers, the median number of differences was 1 (IQR: 0.5-3.5) of an average 65 answered; mean inter-interviewer concordance was 92% (IQR: 92-99%). Discrepancies in open narratives were summarized in five categories: demographics, history and care-seeking, diagnoses and symptoms, treatment and cultural. Most discrepancies were seen in the reporting of diagnoses and symptoms (e.g., malaria diagnosis); only one pair demonstrated no clear differences. The average number of clinical symptoms reported was 9 in open narratives and 20 in the closed questions. Open narratives contained additional information on health seeking and social issues surrounding deaths, which closed questions did not gather.

CONCLUSIONS

The information gleaned during open narratives was subject to inter-interviewer variability and contained a limited number of symptom indicators, suggesting that their use for assigning cause of death is questionable. However, they contained rich information on care-seeking, healthcare provision and social factors in the lead-up to death, which may be a valuable source of information for promoting accountable health services.