Verbal autopsy: methods in transition

Visceral leishmaniasis-associated mortality in Bangladesh: a retrospective cross-sectional study

OBJECTIVE

To investigate visceral leishmaniasis (VL) deaths and risk factors in two VL endemic areas of Bangladesh.

DESIGN

Retrospective cross-sectional.

SETTING

Two geographically and culturally different VL endemic subdistricts, Godagari in the district of Rajshahi and Trishal in the district of Mymensingh in Bangladesh, August 2009-December 2011.

PARTICIPANTS

51 094 inhabitants from randomly selected Unions in the two subdistricts.

MAIN OUTCOME MEASURES

VL deaths, confirmed independently by qualified physicians using the verbal autopsy procedure ICD10 guideline.

RESULTS

The total number of people screened for VL deaths was 51 094 from 12 032 households from Godagari and Trishal subdistricts . About 16% of the people from Godagari were Tribals. The average age of the study population was 25.6 years (SD 18.4) and 49.7% were females. The VL case fatality rate averaged 6.12% (12/196) including 2/137 in Trishal and 10/59 in Godagari. Most of the VL deaths (9/12, 75%) occurred at home and the rest in tertiary hospitals. None of these deaths had been reported in the national VL surveillance system. The VL case fatality rate in the Tribal ethnic (22.2%) population was about 17 times higher than that in the Bangali ethnic (1.3%) population (p<0.0001). Tribal ethnicity had an 18 times (OR=18.1, 95% CI 3.6 to 90.6) higher risk for VL deaths compared with Bangali ethnicity (p<0.0001).

CONCLUSION

VL deaths were found to be high in the study areas and were under-reported. The Tribal ethnic population was at the highest risk for VL deaths. The national VL Elimination Programme should give special attention to the tribal community in the endemic areas, especially for those in Rajshahi, and should strengthen VL surveillance by including tertiary hospitals in the national surveillance system.

Essential evidence for guiding health system priorities and policies: anticipating epidemiological transition in Africa

Background

Despite indications that infection-related mortality in sub-Saharan Africa may be decreasing and the burden of non-communicable diseases increasing, the overwhelming reality is that health information systems across most of sub-Saharan Africa remain too weak to track epidemiological transition in a meaningful and effective way.

Proposals

We propose a minimum dataset as the basis of a functional health information system in countries where health information is lacking. This would involve continuous monitoring of cause-specific mortality through routine civil registration, regular documentation of exposure to leading risk factors, and monitoring effective coverage of key preventive and curative interventions in the health sector. Consideration must be given as to how these minimum data requirements can be effectively integrated within national health information systems, what methods and tools are needed, and ensuring that ethical and political issues are addressed. A more strategic approach to health information systems in sub-Saharan African countries, along these lines, is essential if epidemiological changes are to be tracked effectively for the benefit of local health planners and policy makers.

Conclusion

African countries have a unique opportunity to capitalize on modern information and communications technology in order to achieve this. Methodological standards need to be established and political momentum fostered so that the African continent's health status can be reliably tracked. This will greatly strengthen the evidence base for health policies and facilitate the effective delivery of services.

Mortality level and predictors in a rural Ethiopian population: community based longitudinal study

Background

Over the last fifty years the world has seen enormous decline in mortality rates. However, in low-income countries, where vital registration systems are absent, mortality statistics are not easily available. The recent economic growth of Ethiopia and the parallel large scale healthcare investments make investigating mortality figures worthwhile.

Methods

Longitudinal health and demographic surveillance data collected from September 11, 2009 to September 10, 2012 were analysed. We computed incidence of mortality, overall and stratified by background variables. Poisson regression was used to test for a linear trend in the standardized mortality rates. Cox-regression analysis was used to identify predictors of mortality. Households located at <2300 meter and ≥2300 meter altitude were defined to be midland and highland, respectively.

Results

An open cohort, with a baseline population of 66,438 individuals, was followed for three years to generate 194,083 person-years of observation. The crude mortality rate was 4.04 (95% CI: 3.77, 4.34) per 1,000 person-years. During the follow-up period, incidence of mortality significantly declined among under five (P<0.001) and 5–14 years old (P<0.001), whereas it increased among 65 years and above (P<0.001). Adjusted for other covariates, mortality was higher in males (hazard ratio (HR) = 1.42, 95% CI: 1.22, 1.66), rural population (HR = 1.74, 95% CI: 1.32, 2.31), highland (HR = 1.20, 95% CI: 1.03, 1.40) and among those widowed (HR = 2.25, 95% CI: 1.81, 2.80) and divorced (HR = 1.80, 95% CI: 1.30, 2.48).

Conclusions

Overall mortality rate was low. The level and patterns of mortality indicate changes in the epidemiology of major causes of death. Certain population groups had significantly higher mortality rates and further research is warranted to identify causes of higher mortality in those groups.

Usefulness of the Population Health Metrics Research Consortium gold standard verbal autopsy data for general verbal autopsy methods

BACKGROUND

Verbal Autopsy (VA) is widely viewed as the only immediate strategy for registering cause of death in much of Africa and Asia, where routine physician certification of deaths is not widely practiced. VA involves a lay interview with family or friends after a death, to record essential details of the circumstances. These data can then be processed automatically to arrive at standardized cause of death information.

METHODS

The Population Health Metrics Research Consortium (PHMRC) undertook a study at six tertiary hospitals in low- and middle-income countries which documented over 12,000 deaths clinically and subsequently undertook VA interviews. This dataset, now in the public domain, was compared with the WHO 2012 VA standard and the InterVA-4 interpretative model.

RESULTS

The PHMRC data covered 70% of the WHO 2012 VA input indicators, and categorized cause of death according to PHMRC definitions. After eliminating some problematic or incomplete records, 11,984 VAs were compared. Some of the PHMRC cause definitions, such as 'preterm delivery', differed substantially from the International Classification of Diseases, version 10 equivalent. There were some appreciable inconsistencies between the hospital and VA data, including 20% of the hospital maternal deaths being described as non-pregnant in the VA data. A high proportion of VA cases (66%) reported respiratory symptoms, but only 18% of assigned hospital causes were respiratory-related. Despite these issues, the concordance correlation coefficient between hospital and InterVA-4 cause of death categories was 0.61.

CONCLUSIONS

The PHMRC dataset is a valuable reference source for VA methods, but has to be interpreted with care. Inherently inconsistent cases should not be included when using these data to build other VA models. Conversely, models built from these data should be independently evaluated. It is important to distinguish between the internal and external validity of VA models. The effects of using tertiary hospital data, rather than the more usual application of VA to all-community deaths, are hard to evaluate. However, it would still be of value for VA method development to have further studies of population-based post-mortem examinations.

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

BACKGROUND

Computer-coded verbal autopsy (CCVA) methods to assign causes of death (CODs) for medically unattended deaths have been proposed as an alternative to physician-certified verbal autopsy (PCVA). We conducted a systematic review of 19 published comparison studies (from 684 evaluated), most of which used hospital-based deaths as the reference standard. We assessed the performance of PCVA and five CCVA methods: Random Forest, Tariff, InterVA, King-Lu, and Simplified Symptom Pattern.

METHODS

The reviewed studies assessed methods' performance through various metrics: sensitivity, specificity, and chance-corrected concordance for coding individual deaths, and cause-specific mortality fraction (CSMF) error and CSMF accuracy at the population level. These results were summarized into means, medians, and ranges.

RESULTS

The 19 studies ranged from 200 to 50,000 deaths per study (total over 116,000 deaths). Sensitivity of PCVA versus hospital-assigned COD varied widely by cause, but showed consistently high specificity. PCVA and CCVA methods had an overall chance-corrected concordance of about 50% or lower, across all ages and CODs. At the population level, the relative CSMF error between PCVA and hospital-based deaths indicated good performance for most CODs. Random Forest had the best CSMF accuracy performance, followed closely by PCVA and the other CCVA methods, but with lower values for InterVA-3.

CONCLUSIONS

There is no single best-performing coding method for verbal autopsies across various studies and metrics. There is little current justification for CCVA to replace PCVA, particularly as physician diagnosis remains the worldwide standard for clinical diagnosis on live patients. Further assessments and large accessible datasets on which to train and test combinations of methods are required, particularly for rural deaths without medical attention.

Performance of four computer-coded verbal autopsy methods for cause of death assignment compared with physician coding on 24,000 deaths in low- and middle-income countries

BACKGROUND

Physician-coded verbal autopsy (PCVA) is the most widely used method to determine causes of death (CODs) in countries where medical certification of death is uncommon. Computer-coded verbal autopsy (CCVA) methods have been proposed as a faster and cheaper alternative to PCVA, though they have not been widely compared to PCVA or to each other.

METHODS

We compared the performance of open-source random forest, open-source tariff method, InterVA-4, and the King-Lu method to PCVA on five datasets comprising over 24,000 verbal autopsies from low- and middle-income countries. Metrics to assess performance were positive predictive value and partial chance-corrected concordance at the individual level, and cause-specific mortality fraction accuracy and cause-specific mortality fraction error at the population level.

RESULTS

The positive predictive value for the most probable COD predicted by the four CCVA methods averaged about 43% to 44% across the datasets. The average positive predictive value improved for the top three most probable CODs, with greater improvements for open-source random forest (69%) and open-source tariff method (68%) than for InterVA-4 (62%). The average partial chance-corrected concordance for the most probable COD predicted by the open-source random forest, open-source tariff method and InterVA-4 were 41%, 40% and 41%, respectively, with better results for the top three most probable CODs. Performance generally improved with larger datasets. At the population level, the King-Lu method had the highest average cause-specific mortality fraction accuracy across all five datasets (91%), followed by InterVA-4 (72% across three datasets), open-source random forest (71%) and open-source tariff method (54%).

CONCLUSIONS

On an individual level, no single method was able to replicate the physician assignment of COD more than about half the time. At the population level, the King-Lu method was the best method to estimate cause-specific mortality fractions, though it does not assign individual CODs. Future testing should focus on combining different computer-coded verbal autopsy tools, paired with PCVA strengths. This includes using open-source tools applied to larger and varied datasets (especially those including a random sample of deaths drawn from the population), so as to establish the performance for age- and sex-specific CODs.

Agreement between physicians and the InterVA-4 model in assigning causes of death: the role of recall period and characteristics specific to the deceased and the respondent

BACKGROUND

In the absence of routine death registration, the InterVA model is a new methodology being used as a physician alternative method to interpret verbal autopsy (VA) data in resource-poor settings. However, various studies indicate that there are significant discrepancies between the two approaches in assigning causes of deaths. This study evaluated the role of recall period and characteristics that were specific to the deceased and the respondent in affecting the level of agreement between the approaches.

METHODS

A population-based cross-sectional study was conducted from March to April, 2012. All adults aged ≥14 years and died between 01 January, 2010, and 15 February, 2012, were included in the study. Data were collected by using a pre-tested and modified WHO designed verbal autopsy questionnaire. The verbal autopsy interviews were reviewed by the InterVA-4 model and the physicians. Cohen's kappa statistic with 95% CI was applied to compare the strength of the agreement between the model and the physician review.

RESULTS

A total of 408 VA interviews were successfully completed and reviewed by the InterVA model and the physicians. Both approaches showed an overall agreement in 294 (72.1%) of the cases [kappa = 0.48, 95% CI: 0.42 - 0.60]. The level of agreement between the approaches was low [kappa ≤0.40] when the deceased was female, 50 and above years old, single, illiterate, rural dweller, belonged to a family of 1-4 people living together, and died at home. This was also true when the recall period was ≤1 year, and the respondent was a relative other than parent/marital partner, lived with the deceased, and had medical information.

CONCLUSION

This study identified important variables affecting the strength of agreement between the InterVA-4 model and the physician in assigning causes of death. The results are believed to significantly contribute to the process of identifying the actual underlying causes of deaths in the population, and may thus serve to promote informed health policy decisions in resource-poor settings.

Household context and child mortality in rural South Africa: the effects of birth spacing, shared mortality, household composition and socio-economic status

BACKGROUND

Household characteristics are important influences on the risk of child death. However, little is known about this influence in HIV-endemic areas. We describe the effects of household characteristics on children's risk of dying in rural South Africa.

METHODS

We use data describing the mortality of children younger than 5 years living in the Agincourt health and socio-demographic surveillance system study population in rural northeast South Africa during the period 1994-2008. Using discrete time event history analysis we estimate children's probability of dying by child characteristics and household composition (other children and adults other than parents) (N=924,818 child-months), and household socio-economic status (N=501,732 child-months).

RESULTS

Children under 24 months of age whose subsequent sibling was born within 11 months experience increased odds of dying (OR 2.5; 95% CI 1.1-5.7). Children also experience increased odds of dying in the period 6 months (OR 2.1; 95% CI 1.2-3.6), 3-5 months (OR 3.0; 95% CI 1.5-5.9), and 2 months (OR 11.8; 95% CI 7.6-18.3) before another household child dies. The odds of dying remain high at the time of another child's death (OR 11.7; 95% CI 6.3-21.7) and for the 2 months following (OR 4.0; 95% CI 1.9-8.6). Having a related but non-parent adult aged 20-59 years in the household reduces the odds (OR 0.6; 95% CI 0.5-0.8). There is an inverse relationship between a child's odds of dying and household socio-economic status.

CONCLUSIONS

This detailed household profile from a poor rural setting where HIV infection is endemic indicates that children are at high risk of dying when another child is very ill or has recently died. Short birth intervals and additional children in the household are further risk factors. Presence of a related adult is protective, as is higher socio-economic status. Such evidence can inform primary health care practice and facilitate targeting of community health worker efforts, especially when covering defined catchment areas.

Revising the WHO verbal autopsy instrument to facilitate routine cause-of-death monitoring

Objective

Verbal autopsy (VA) is a systematic approach for determining causes of death (CoD) in populations without routine medical certification. It has mainly been used in research contexts and involved relatively lengthy interviews. Our objective here is to describe the process used to shorten, simplify, and standardise the VA process to make it feasible for application on a larger scale such as in routine civil registration and vital statistics (CRVS) systems.

Methods

A literature review of existing VA instruments was undertaken. The World Health Organization (WHO) then facilitated an international consultation process to review experiences with existing VA instruments, including those from WHO, the Demographic Evaluation of Populations and their Health in Developing Countries (INDEPTH) Network, InterVA, and the Population Health Metrics Research Consortium (PHMRC). In an expert meeting, consideration was given to formulating a workable VA CoD list [with mapping to the International Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) CoD] and to the viability and utility of existing VA interview questions, with a view to undertaking systematic simplification.

Findings

A revised VA CoD list was compiled enabling mapping of all ICD-10 CoD onto 62 VA cause categories, chosen on the grounds of public health significance as well as potential for ascertainment from VA. A set of 221 indicators for inclusion in the revised VA instrument was developed on the basis of accumulated experience, with appropriate skip patterns for various population sub-groups. The duration of a VA interview was reduced by about 40% with this new approach.

Conclusions

The revised VA instrument resulting from this consultation process is presented here as a means of making it available for widespread use and evaluation. It is envisaged that this will be used in conjunction with automated models for assigning CoD from VA data, rather than involving physicians.

Validating the InterVA model to estimate the burden of mortality from verbal autopsy data: a population-based cross-sectional study

Background

In countries with incomplete or no vital registration systems, verbal autopsy data are often reviewed by physicians in order to assign the probable cause of death. But in addition to being time and energy consuming, the method is liable to produce inconsistent results. The aim of this study is to validate the InterVA model for estimating the burden of mortality from verbal autopsy data by using physician review as a reference standard.

Methods and Findings

A population-based cross-sectional study was conducted from March to April, 2012. All adults aged ≥14 years and died between 01 January, 2010 and 15 February, 2012 were included in the study. The verbal autopsy interviews were reviewed by the InterVA model and physicians to estimate cause-specific mortality fractions. Cohen’s kappa statistic, sensitivity, specificity, positive predictive value, and negative predictive value were applied to compare the agreement between the InterVA model and the physician review. A total of 408 adult deaths were studied. There was a general similarity and just slight differences between the InterVA model and the physicians in assigning cause-specific mortality. Both approaches showed an overall agreement in 298 (73%) cases [kappa = 0.49, 95% CI: 0.37-0.60]. The observed sensitivities and specificities across causes of death categories varied from 13.3% to 81.9% and 77.7% to 99.5%, respectively.

Conclusions

In understanding the burden of disease and setting health intervention priorities in areas that lack reliable vital registration systems, an accurate analysis of verbal autopsies is essential. Therefore, users should be aware of the suboptimal performance of the InterVA model. Similar validation studies need to be undertaken considering the limitation of the physician review as gold standard since physicians may misinterpret some of the verbal autopsy data and finally reach a wrong conclusion of the cause of death.

Why are babies dying in the first month after birth? A 7-year study of neonatal mortality in northern Ghana

Objectives

To determine the neonatal mortality rate in the Kassena-Nankana District (KND) of northern Ghana, and to identify the leading causes and timing of neonatal deaths.

Methods

The KND falls within the Navrongo Health Research Centre’s Health and Demographic Surveillance System (HDSS), which uses trained field workers to gather and update health and demographic information from community members every four months. We utilized HDSS data from 2003–2009 to examine patterns of neonatal mortality.

Results

A total of 17,751 live births between January 2003 and December 2009 were recorded, including 424 neonatal deaths 64.8%(275) of neonatal deaths occurred in the first week of life. The overall neonatal mortality rate was 24 per 1000 live births (95%CI 22 to 26) and early neonatal mortality rate was 16 per 1000 live births (95% CI 14 to 17). Neonatal mortality rates decreased over the period from 26 per 1000 live births in 2003 to 19 per 1000 live births in 2009. In all, 32%(137) of the neonatal deaths were from infections, 21%(88) from birth injury and asphyxia and 18%(76) from prematurity, making these three the leading causes of neonatal deaths in the area. Birth injury and asphyxia (31%) and prematurity (26%) were the leading causes of early neonatal deaths, while infection accounted for 59% of late neonatal deaths. Nearly 46% of all neonatal deaths occurred during the first three postnatal days. In multivariate analysis, multiple births, gestational age <32 weeks and first pregnancies conferred the highest odds of neonatal deaths.

Conclusions

Neonatal mortality rates are declining in rural northern Ghana, with majority of deaths occurring within the first week of life. This has major policy, programmatic and research implications. Further research is needed to better understand the social, cultural, and logistical factors that drive high mortality in the early days following delivery.

Measuring unsafe abortion-related mortality: a systematic review of the existing methods

BACKGROUND

The WHO estimates that 13% of maternal mortality is due to unsafe abortion, but challenges with measurement and data quality persist. To our knowledge, no systematic assessment of the validity of studies reporting estimates of abortion-related mortality exists.

STUDY DESIGN

To be included in this study, articles had to meet the following criteria: (1) published between September 1(st), 2000-December 1(st), 2011; (2) utilized data from a country where abortion is "considered unsafe"; (3) specified and enumerated causes of maternal death including "abortion"; (4) enumerated ≥100 maternal deaths; (5) a quantitative research study; (6) published in a peer-reviewed journal.

RESULTS

7,438 articles were initially identified. Thirty-six studies were ultimately included. Overall, studies rated "Very Good" found the highest estimates of abortion related mortality (median 16%, range 1-27.4%). Studies rated "Very Poor" found the lowest overall proportion of abortion related deaths (median: 2%, range 1.3-9.4%).

CONCLUSIONS

Improvements in the quality of data collection would facilitate better understanding global abortion-related mortality. Until improved data exist, better reporting of study procedures and standardization of the definition of abortion and abortion-related mortality should be encouraged.

Emerging chronic non-communicable diseases in rural communities of Northern Ethiopia: evidence using population-based verbal autopsy method in Kilite Awlaelo surveillance site

INTRODUCTION

In countries where most deaths are outside health institutions and medical certification of death is absent, verbal autopsy (VA) method is used to estimate population level causes of death.

METHODS

VA data were collected by trained lay interviewers for 409 deaths in the surveillance site. Two physicians independently assigned cause of death using the International Classification of Diseases manual.

RESULTS

In general, infectious and parasitic diseases accounted for 35.9% of death, external causes 15.9%, diseases of the circulatory system 13.4% and perinatal causes 12.5% of total deaths. Mortalities attributed to maternal causes and malnutrition were low, 0.2 and 1.5%, respectively. Causes of death varied by age category. About 22.1, 12.6 and 8.4% of all deaths of under 5-year-old children were due to bacterial sepsis of the newborn, acute lower respiratory infections such as neonatal pneumonia and prematurity including respiratory distress, respectively. For 5-15-year-old children, accidental drowning and submersion, accounting for 34.4% of all deaths in this age category, and accidental fall, accounting for 18.8%, were leading causes of death. Among 15-49-year-old adults, HIV/AIDS (16.3%) and tuberculosis (12.8%) were commonest causes of death, whereas tuberculosis and cerebrovascular diseases were major killers of those aged 50 years and above.

CONCLUSION

In the rural district, mortality due to chronic non-communicable diseases was very high. The observed magnitude of death from chronic non-communicable disease is unlikely to be unique to this district. Thus, formulation of chronic disease prevention and control strategies is recommended.

Characteristics, availability and uses of vital registration and other mortality data sources in post-democracy South Africa

The value of good-quality mortality data for public health is widely acknowledged. While effective civil registration systems remains the 'gold standard' source for continuous mortality measurement, less than 25% of deaths are registered in most African countries. Alternative data collection systems can provide mortality data to complement those from civil registration, given an understanding of data source characteristics and data quality. We aim to document mortality data sources in post-democracy South Africa; to report on availability, limitations, strengths, and possible complementary uses of the data; and to make recommendations for improved data for mortality measurement. Civil registration and alternative mortality data collection systems, data availability, and complementary uses were assessed by reviewing blank questionnaires, death notification forms, death data capture sheets, and patient cards; legislation; electronic data archives and databases; and related information in scientific journals, research reports, statistical releases, government reports and books. Recent transformation has enhanced civil registration and official mortality data availability. Additionally, a range of mortality data items are available in three population censuses, three demographic surveillance systems, and a number of national surveys, mortality audits, and disease notification programmes. Child and adult mortality items were found in all national data sources, and maternal mortality items in most. Detailed cause-of-death data are available from civil registration and demographic surveillance. In a continent often reported as lacking the basic data to infer levels, patterns and trends of mortality, there is evidence of substantial improvement in South Africa in the availability of data for mortality assessment. Mortality data sources are many and varied, providing opportunity for comparing results and improved public health planning. However, more can and must be done to improve mortality measurement by improving data quality, triangulating data, and expanding analytic capacity. Cause data, in particular, must be improved.

Shaping healthcare-seeking processes during fatal illness in resource-poor settings. A study in Lao PDR

BACKGROUND

There are profound social meanings attached to bearing children that affect the experience of losing a child, which is akin to the loss of a mother in the household. The objective of this study is to comprehend the broader processes that shape household healthcare-seeking during fatal illness episodes or reproductive health emergencies in resource-poor communities.

METHODS

The study was conducted in six purposively selected poor, rural communities in Lao PDR, located in two districts that represent communities with different access to health facilities and contain diverse ethnic groups. Households having experienced fatal cases were first identified in focus group discussions with community members, which lead to the identification of 26 deaths in eleven households through caregiver and spouse interviews. The interviews used an open-ended anthropological approach and followed a three-delay framework. Interpretive description was used in the data analysis.

RESULTS

The healthcare-seeking behavior reported by caregivers revealed a broad range of providers, reflecting the mix of public, private, informal and traditional health services in Lao PDR. Most caregivers had experienced multiple constraints in healthcare-seeking prior to death. Decisions regarding care-seeking were characterized as social rather than individual actions. They were constrained by medical costs, low expectations of recovery and worries about normative expectations from healthcare workers on how patients and caregivers should behave at health facilities to qualify for treatment. Caregivers raised the difficulties in determining the severity of the state of the child/mother. Delays in reaching care related to lack of physical access and to risks associated with taking a sick family member out of the local community. Delays in receiving care were affected by the perceived low quality of care provided at the health facilities.

CONCLUSIONS

Care-seeking is influenced by family- and community-based relations, which are integrated parts of people's everyday life. The medical and normative responses from health providers affect the behavior of care-seekers. An anthropological approach to capture the experience of caregivers in relation to deciding, seeking and reaching care reveals the complexity and socio-cultural context surrounding maternal and child mortality and has implications for how future mortality data should be developed and interpreted.

Evaluating the performance of interpreting Verbal Autopsy 3.2 model for establishing pulmonary tuberculosis as a cause of death in Ethiopia: a population-based cross-sectional study

Background

In resource- poor settings, verbal autopsy data are often reviewed by physicians in order to assign the probable cause of death. But in addition to being time and energy consuming, the method is liable to produce inconsistent results. The aim of this study is to evaluate the performance of the InterVA 3.2 model for establishing pulmonary tuberculosis as a cause of death in comparison with physician review of verbal autopsy data.

Methods

A population-based cross-sectional study was conducted from March to April, 2012. All adults aged ≥14 years and died between 01 January 2010 and 15 February 2012 were included in the study. Data were collected by using a pre-tested and modified WHO designed verbal autopsy questionnaire. The verbal autopsy interviews were reviewed by the InterVA model and the physicians. Cohen’s kappa statistic, receiver operating characteristic curves, sensitivity, and specificity values were applied to compare the agreement between the InterVA model and the physician review.

Results

A total of 408 adult deaths were studied. The proportion of tuberculosis-specific mortality was established to be 36.0% and 23.0% by the InterVA model and the physicians, respectively. The InterVA model predicted pulmonary tuberculosis as a cause of death with the probability of 0.80 (95% CI: 0.75-0.85). In classifying all deaths as tuberculosis and non-tuberculosis, the sensitivity and specificity values were 0.82 and 0.78, respectively. A moderate agreement was found between the model and physicians in assigning pulmonary tuberculosis as a cause of deaths [kappa= 0.5; 95% CI: (0.4-0.6)].

Conclusions

This study has revealed that the InterVA model showed a more promising result as a community-level tool for generating pulmonary tuberculosis-specific mortality data from verbal autopsy. The conclusion is believed to provide policymakers with a highly needed piece of information for allocating resources for health intervention.

An innovative approach to measuring maternal mortality at the community level in low-resource settings using mid-level providers: a feasibility study in Tigray, Ethiopia

This paper proposes a new, community-based approach to the measurement of maternal mortality, and presents results from a feasibility study in 2010-11 of that approach in rural Tigray, Ethiopia. The study was implemented in three health posts and one health centre with a total catchment area of approximately 22,000 people. Priests, traditional birth attendants and community-based reproductive health agents were responsible for locating and reporting all births and deaths in their areas and assisted mid-level providers in locating key informants for verbal autopsy. Community-based health workers were trained to report all births and deaths to the local health post, where vital registries were kept. Once a month, each health post compiled a list of all deaths of women aged 12-49, which were registered in government logbooks. Nurses and nurse-midwives were trained to administer verbal autopsies on these deaths, and assign primary cause of death using WHO ICD-10 classifications. The study drew on the theory of task-shifting, shifting the task of cause-of-death attribution from physicians to mid-level providers. It aimed to build a sustainable methodology for maximizing existing local health care infrastructure and human capacity, leading to community-based solutions to improve maternal health. While the approach has not yet been implemented outside the initial study area, the results are promising as regards its feasibility.

Strengthening standardised interpretation of verbal autopsy data: the new InterVA-4 tool

Background

Verbal autopsy (VA) is the only available approach for determining the cause of many deaths, where routine certification is not in place. Therefore, it is important to use standards and methods for VA that maximise efficiency, consistency and comparability. The World Health Organization (WHO) has led the development of the 2012 WHO VA instrument as a new standard, intended both as a research tool and for routine registration of deaths.

Objective

A new public-domain probabilistic model for interpreting VA data, InterVA-4, is described, which builds on previous versions and is aligned with the 2012 WHO VA instrument.

Design

The new model has been designed to use the VA input indicators defined in the 2012 WHO VA instrument and to deliver causes of death compatible with the International Classification of Diseases version 10 (ICD-10) categorised into 62 groups as defined in the 2012 WHO VA instrument. In addition, known shortcomings of previous InterVA models have been addressed in this revision, as well as integrating other work on maternal and perinatal deaths.

Results

The InterVA-4 model is presented here to facilitate its widespread use and to enable further field evaluation to take place. Results from a demonstration dataset from Agincourt, South Africa, show continuity of interpretation between InterVA-3 and InterVA-4, as well as differences reflecting specific issues addressed in the design and development of InterVA-4.

Conclusions

InterVA-4 is made freely available as a new standard model for interpreting VA data into causes of death. It can be used for determining cause of death both in research settings and for routine registration. Further validation opportunities will be explored. These developments in cause of death registration are likely to substantially increase the global coverage of cause-specific mortality data.

Relating the construction and maintenance of maternal ill-health in rural Indonesia

Estimates suggest that over 350,000 deaths and more than 20 million severe disabilities result from the complications of pregnancy, childbirth or its management each year. Death and disability occur predominately among disadvantaged women in resource-poor settings and are largely preventable with adequate delivery care. This paper presents the substantive findings and policy implications from a programme of PhD research, of which the overarching objective was to assess quality of, and access to, care in obstetric emergencies. Three critical incident audits were conducted in two rural districts on Java, Indonesia: a confidential enquiry, a verbal autopsy survey, and a community-based review. The studies examined cases of maternal mortality and severe morbidity from the perspectives of local service users and health providers. A range of inter-related determining factors was identified. When unexpected delivery complications occurred, women and families were often uninformed, unprepared, found care unavailable, unaffordable, and many relied on traditional providers. Midwives in villages made important contributions by stabilising women and facilitating referrals but were often scarce in remote areas and lacked sufficient clinical competencies and payment incentives to treat the poor. Emergency transport was often unavailable and private transport was unreliable and incurred costs. In facilities, there was a reluctance to admit poorer women and those accepted were often admitted to ill-equipped, under-staffed wards. As a result, referrals between hospitals were also common. Otherwise, social health insurance, designed to reduce financial barriers, was particularly problematic, constraining quality and access within and outside facilities. Health workers and service users provided rich and explicit assessments of care and outcomes. These were used to develop a conceptual model in which quality and access are conceived of as social processes, observable through experience and reflective of the broader relationships between individuals and health systems. According to this model, differential quality and access can become both socially legitimate (imposed by structural arrangements) and socially legitimised (reciprocally maintained through the actions of individuals). This interpretation suggests that in a context of commodified care provision, adverse obstetric outcomes will occur and recur for disadvantaged women. Health system reform should focus on the unintended effects of market-based service provision to exclude those without the ability to pay for delivery care directly.

Causes of community stillbirths and early neonatal deaths in low-income countries using verbal autopsy: an International, Multicenter Study

OBJECTIVE

Six million stillbirths (SB) and early neonatal deaths (END) occur annually worldwide, mostly in rural settings distant from health facilities. We used verbal autopsy (VA), to understand causes of non-hospital, community-based SB and END from four low-income countries.

STUDY DESIGN

This prospective observational study utilized the train-the-trainer method. VA interviewers conducted standardized interviews; in each country data were reviewed by two local physicians who assigned an underlying causes of deaths (COD).

RESULT

There were 252 perinatal deaths (118 END; 134 SB) studied from pooled data. Almost half (45%) the END occurred on postnatal day 1, 19% on the second day and 16% the third day. Major early neonatal COD were infections (49%), birth asphyxia (26%), prematurity (17%) and congenital malformations (3%). Major causes of SB were infection (37%), prolonged labor (11%), antepartum hemorrhage (10%), preterm delivery (7%), cord complications (6%) and accidents (5%).

CONCLUSION

Many of these SB and END were from easily preventable causes. Over 80% of END occurred during the first 3 days of postnatal life, and >90% were due to infection, birth asphyxia and prematurity. The causes of SB were more varied, and maternal infections were the most common cause. Increased attention should be targeting at interventions that reduce maternal and neonatal infections and prevent END, particularly during the first 3 days of life.

Factors associated with physician agreement and coding choices of cause of death using verbal autopsies for 1130 maternal deaths in India

Background

The Indian Sample Registration System (SRS) with verbal autopsy methods provides estimations of cause specific mortality for maternal deaths, where the majority of deaths occur at home, unregistered. We aim to examine factors that influence physician agreement and coding choices in assigning causes of death from verbal autopsies.

Methodology/Principal Findings

Among adult deaths identified in the SRS, pregnancy-related deaths recorded in 2001–2003 were assigned ICD-10 codes by two independent physicians. Inter-rater reliability was estimated using Landis Koch Kappa classification – poor to fair agreement; >– moderate agreement; >– substantial agreement; >– high agreement. We identified factors associated with physician agreement using multivariate logistic regression. A central consensus panel reviewed cases for errors and reclassified as needed based on 2011 ICD-10 coding guidelines. Of 1130 pregnancy-related deaths, 1040 were assigned ICD-10 codes by two physicians. We found substantial agreement regardless of the woman's residence, whether the death was registered, religion, respondent's or deceased's education, age, hospital admission or gestational age. Physician agreement was not influenced by the above variables, with the exception of greater agreement in cases where the respondent did not live with the deceased, or early gestational age at the time of death. A central consensus panel reviewed all cases and recoded 10% of cases due to insufficient use of information in the verbal autopsy by the coding physicians and rationale for this reclassification are discussed.

Conclusion

In the absence of complete vital registration and universal healthcare services, physician coded verbal autopsies continues to be heavily relied upon to ascertain pregnancy-related death. From this study, two independent physicians had good inter-rater reliability for assigning pregnancy-related causes of death in a nationally-represented sample, and physician coding does not appear to be heavily influenced by case characteristics or demographics.

Socioeconomic factors and all cause and cause-specific mortality among older people in Latin America, India, and China: a population-based cohort study

BACKGROUND

Even in low and middle income countries most deaths occur in older adults. In Europe, the effects of better education and home ownership upon mortality seem to persist into old age, but these effects may not generalise to LMICs. Reliable data on causes and determinants of mortality are lacking.

METHODS AND FINDINGS

The vital status of 12,373 people aged 65 y and over was determined 3-5 y after baseline survey in sites in Latin America, India, and China. We report crude and standardised mortality rates, standardized mortality ratios comparing mortality experience with that in the United States, and estimated associations with socioeconomic factors using Cox's proportional hazards regression. Cause-specific mortality fractions were estimated using the InterVA algorithm. Crude mortality rates varied from 27.3 to 70.0 per 1,000 person-years, a 3-fold variation persisting after standardisation for demographic and economic factors. Compared with the US, mortality was much higher in urban India and rural China, much lower in Peru, Venezuela, and urban Mexico, and similar in other sites. Mortality rates were higher among men, and increased with age. Adjusting for these effects, it was found that education, occupational attainment, assets, and pension receipt were all inversely associated with mortality, and food insecurity positively associated. Mutually adjusted, only education remained protective (pooled hazard ratio 0.93, 95% CI 0.89-0.98). Most deaths occurred at home, but, except in India, most individuals received medical attention during their final illness. Chronic diseases were the main causes of death, together with tuberculosis and liver disease, with stroke the leading cause in nearly all sites.

CONCLUSIONS

Education seems to have an important latent effect on mortality into late life. However, compositional differences in socioeconomic position do not explain differences in mortality between sites. Social protection for older people, and the effectiveness of health systems in preventing and treating chronic disease, may be as important as economic and human development.

Stillbirths and early neonatal mortality in rural Northern Ghana

OBJECTIVE

To calculate perinatal mortality (stillbirth and early neonatal death: END) rates in the Upper East region of Ghana and characterize community-based stillbirths and END in terms of timing, cause of death, and maternal and infant risk factors.

METHODS

Birth outcomes were obtained from the Navrongo Health and Demographic Surveillance System over a 7-year period.

RESULTS

Twenty thousand four hundred and ninty seven pregnant women were registered in the study. The perinatal mortality rate was 39 deaths/1000 deliveries, stillbirth rate 23/1000 deliveries and END rates 16/1000 live births. Most stillbirths were 31 weeks gestation or less. Prematurity, first-time delivery and multiple gestation all significantly increased the odds of perinatal death. Approximately 70% of END occurred during the first 3 postnatal days, and the most common causes of death were birth asphyxia and injury, infections and prematurity.

CONCLUSION

Stillbirths and END remain a significant problem in Navrongo. The main causes of END occur during the first 3 days and may be modifiable with simple targeted perinatal policies.

Birth attendants as perinatal verbal autopsy respondents in low- and middle-income countries: a viable alternative?

OBJECTIVE

To assess the feasibility of using birth attendants instead of bereaved mothers as perinatal verbal autopsy respondents.

METHODS

Verbal autopsy interviews for early neonatal deaths and stillbirths were conducted separately among mothers (reference standard) and birth attendants in 38 communities in four developing countries. Concordance between maternal and attendant responses was calculated for all questions, for categories of questions and for individual questions. The sensitivity and specificity of individual questions with the birth attendant as respondent were assessed.

FINDINGS

For early neonatal deaths, concordance across all questions was 94%. Concordance was at least 95% for more than half the questions on maternal medical history, birth attendance and neonate characteristics. Concordance on any given question was never less than 80%. Sensitivity and specificity varied across individual questions, more than 80% of which had a sensitivity of at least 80% and a specificity of at least 90%. For stillbirths, concordance across all questions was 93%. Concordance was 95% or greater more than half the time for questions on birth attendance, site of delivery and stillborn characteristics. Sensitivity and specificity varied across individual questions. Over 60% of the questions had a sensitivity of at least 80% and over 80% of them had a specificity of at least 90%. Overall, the causes of death established through verbal autopsy were similar, regardless of respondent.

CONCLUSION

Birth attendants can substitute for bereaved mothers as verbal autopsy respondents. The questions in existing harmonized verbal autopsy questionnaires need further refinement, as their sensitivity and specificity differ widely.

Probabilistic methods for verbal autopsy interpretation: InterVA robustness in relation to variations in a priori probabilities

Background

InterVA is a probabilistic method for interpreting verbal autopsy (VA) data. It uses a priori approximations of probabilities relating to diseases and symptoms to calculate the probability of specific causes of death given reported symptoms recorded in a VA interview. The extent to which InterVA's ability to characterise a population's mortality composition might be sensitive to variations in these a priori probabilities was investigated.

Methods

A priori InterVA probabilities were changed by 1, 2 or 3 steps on the logarithmic scale on which the original probabilities were based. These changes were made to a random selection of 25% and 50% of the original probabilities, giving six model variants. A random sample of 1,000 VAs from South Africa, were used as a basis for experimentation and were processed using the original InterVA model and 20 random instances of each of the six InterVA model variants. Rank order of cause of death and cause-specific mortality fractions (CSMFs) from the original InterVA model and the mean, maximum and minimum results from the 20 randomly modified InterVA models for each of the six variants were compared.

Results

CSMFs were functionally similar between the original InterVA model and the models with modified a priori probabilities such that even the CSMFs based on the InterVA model with the greatest degree of variation in the a priori probabilities would not lead to substantially different public health conclusions. The rank order of causes were also similar between all versions of InterVA.

Conclusion

InterVA is a robust model for interpreting VA data and even relatively large variations in a priori probabilities do not affect InterVA-derived results to a great degree. The original physician-derived a priori probabilities are likely to be sufficient for the global application of InterVA in settings without routine death certification.

Validity of verbal autopsy procedures for determining malaria deaths in different epidemiological settings in Uganda

BACKGROUND

Verbal autopsy (VA) procedures can be used to estimate cause of death in settings with inadequate vital registries. However, the sensitivity of VA for determining malaria-specific mortality may be low, and may vary with transmission intensity. We assessed the diagnostic accuracy of VA procedures as compared to hospital medical records for determining cause of death in children under five in three different malaria transmission settings in Uganda, including Tororo (high), Kampala (medium), and Kisoro (low).

METHODS AND FINDINGS

Caretakers of children who died in participating hospitals were interviewed using a standardized World Health Organization questionnaire. Medical records from the child's hospitalization were also reviewed. Causes of death based on the VA questionnaires and the medical records were assigned independently by physician reviewers and then compared. A total of 719 cases were included in the final analysis, 67 in Tororo, 600 in Kampala, and 52 in Kisoro. Malaria was classified as the underlying or contributory cause of death by review of medical records in 33 deaths in Tororo, 60 in Kampala, and 0 in Kisoro. The sensitivity of VA procedures for determining malaria deaths in Tororo was 61% (95% CI 44-78%) and 50% in Kampala (95% CI 37-63%). Specificity for determining malaria deaths in Tororo and Kampala was high (>88%), but positive predictive value varied widely, from 83% in Tororo to 34% in Kampala (difference 49%, 95% CI 31-67, p<0.001). The difference between the cause-specific mortality fraction for malaria as determined by VA procedures and medical records was -11% in Tororo, +5% in Kampala, and +14% in Kisoro.

CONCLUSIONS

Our results suggest that these VA methods have an acceptable level of diagnostic accuracy for determining malaria deaths at the population level in high and medium transmission areas, but not in low transmission areas.

Classifying perinatal mortality using verbal autopsy: is there a role for nonphysicians?

Background

Because of a physician shortage in many low-income countries, the use of nonphysicians to classify perinatal mortality (stillbirth and early neonatal death) using verbal autopsy could be useful.

Objective

To determine the extent to which underlying perinatal causes of deaths assigned by nonphysicians in Guatemala, Pakistan, Zambia, and the Democratic Republic of the Congo using a verbal autopsy method are concordant with underlying perinatal cause of death assigned by physician panels.

Methods

Using a train-the-trainer model, 13 physicians and 40 nonphysicians were trained to determine cause of death using a standardized verbal autopsy training program. Subsequently, panels of two physicians and individual nonphysicians from this trained cohort independently reviewed verbal autopsy data from a sample of 118 early neonatal deaths and 134 stillbirths. With the cause of death assigned by the physician panel as the reference standard, sensitivity, specificity, positive and negative predictive values, and cause-specific mortality fractions were calculated to assess nonphysicians' coding responses. Robustness criteria to assess how well nonphysicians performed were used.

Results

Causes of early neonatal death and stillbirth assigned by nonphysicians were concordant with physician-assigned causes 47% and 57% of the time, respectively. Tetanus filled robustness criteria for early neonatal death, and cord prolapse filled robustness criteria for stillbirth.

Conclusions

There are significant differences in underlying cause of death as determined by physicians and nonphysicians even when they receive similar training in cause of death determination. Currently, it does not appear that nonphysicians can be used reliably to assign underlying cause of perinatal death using verbal autopsy.

Using verbal autopsy to track epidemic dynamics: the case of HIV-related mortality in South Africa

Background

Verbal autopsy (VA) has often been used for point estimates of cause-specific mortality, but seldom to characterize long-term changes in epidemic patterns. Monitoring emerging causes of death involves practitioners' developing perceptions of diseases and demands consistent methods and practices. Here we retrospectively analyze HIV-related mortality in South Africa, using physician and modeled interpretation.

Methods

Between 1992 and 2005, 94% of 6,153 deaths which occurred in the Agincourt subdistrict had VAs completed, and coded by two physicians and the InterVA model. The physician causes of death were consolidated into a single consensus underlying cause per case, with an additional physician arbitrating where different diagnoses persisted. HIV-related mortality rates and proportions of deaths coded as HIV-related by individual physicians, physician consensus, and the InterVA model were compared over time.

Results

Approximately 20% of deaths were HIV-related, ranging from early low levels to tenfold-higher later population rates (2.5 per 1,000 person-years). Rates were higher among children under 5 years and adults 20 to 64 years. Adult mortality shifted to older ages as the epidemic progressed, with a noticeable number of HIV-related deaths in the over-65 year age group latterly. Early InterVA results suggested slightly higher initial HIV-related mortality than physician consensus found. Overall, physician consensus and InterVA results characterized the epidemic very similarly. Individual physicians showed marked interobserver variation, with consensus findings generally reflecting slightly lower proportions of HIV-related deaths. Aggregated findings for first versus second physician did not differ appreciably.

Conclusions

VA effectively detected a very significant epidemic of HIV-related mortality. Using either physicians or InterVA gave closely comparable findings regarding the epidemic. The consistency between two physician coders per case (from a pool of 14) suggests that double coding may be unnecessary, although the consensus rate of HIV-related mortality was approximately 8% lower than by individual physicians. Consistency within and between individual physicians, individual perceptions of epidemic dynamics, and the inherent consistency of models are important considerations here. The ability of the InterVA model to track a more than tenfold increase in HIV-related mortality over time suggests that finely tuned "local" versions of models for VA interpretation are not necessary.

Verbal autopsy-based cause-specific mortality trends in rural KwaZulu-Natal, South Africa, 2000-2009

Background

The advent of the HIV pandemic and the more recent prevention and therapeutic interventions have resulted in extensive and rapid changes in cause-specific mortality rates in sub-Saharan Africa, and there is demand for timely and accurate cause-specific mortality data to steer public health responses and to evaluate the outcome of interventions. The objective of this study is to describe cause-specific mortality trends based on verbal autopsies conducted on all deaths in a rural population in KwaZulu-Natal, South Africa, over a 10-year period (2000-2009).

Methods

The study used population-based mortality data collected by a demographic surveillance system on all resident and nonresident members of 12,000 households. Cause of death was determined by verbal autopsy based on the standard INDEPTH/WHO verbal autopsy questionnaire. Cause of death was assigned by physician review and the Bayesian-based InterVA program.

Results

There were 11,281 deaths over 784,274 person-years of observation of 125,658 individuals between Jan. 1, 2000 and Dec. 31, 2009. The cause-specific mortality fractions (CSMF) for the population as a whole were: HIV-related (including tuberculosis), 50%; other communicable diseases, 6%; noncommunicable lifestyle-related conditions, 15%; other noncommunicable diseases, 2%; maternal, perinatal, nutritional, and congenital causes, 1%; injury, 8%; indeterminate causes, 18%. Over the course of the 10 years of observation, the CSMF of HIV-related causes declined from a high of 56% in 2002 to a low of 39% in 2009 with the largest decline starting in 2004 following the introduction of an antiretroviral treatment program into the population. The all-cause age-standardized mortality rate (SMR) declined over the same period from a high of 174 (95% confidence interval [CI]: 165, 183) deaths per 10,000 person-years observed (PYO) in 2003 to a low of 116 (95% CI: 109, 123) in 2009. The decline in the SMR is predominantly due to a decline in the HIV-related SMR, which declined in the same period from 96 (95% CI: 89, 102) to 45 (95% CI: 40, 49) deaths per 10,000 PYO.

There was substantial agreement (79% kappa = 0.68 (95% CI: 0.67, 0.69)) between physician coding and InterVA coding at the burden of disease group level.

Conclusions

Verbal autopsy based methods enabled the timely measurement of changing trends in cause-specific mortality to provide policymakers with the much-needed information to allocate resources to appropriate health interventions.

Adaptation of a probabilistic method (InterVA) of verbal autopsy to improve the interpretation of cause of stillbirth and neonatal death in Malawi, Nepal, and Zimbabwe

Background

Verbal autopsy (VA) is a widely used method for analyzing cause of death in absence of vital registration systems. We adapted the InterVA method to extrapolate causes of death for stillbirths and neonatal deaths from verbal autopsy questionnaires, using data from Malawi, Zimbabwe, and Nepal.

Methods

We obtained 734 stillbirth and neonatal VAs from recent community studies in rural areas: 169 from Malawi, 385 from Nepal, and 180 from Zimbabwe. Initial refinement of the InterVA model was based on 100 physician-reviewed VAs from Malawi. InterVA indicators and matrix probabilities for cause of death were reviewed for clinical and epidemiological coherence by a pediatrician-researcher and an epidemiologist involved in the development of InterVA. The modified InterVA model was evaluated by comparing population-level cause-specific mortality fractions and individual agreement from two methods of interpretation (physician review and InterVA) for a further 69 VAs from Malawi, 385 from Nepal, and 180 from Zimbabwe.

Results

Case-by-case agreement between InterVA and reviewing physician diagnoses for 69 cases from Malawi, 180 cases from Zimbabwe, and 385 cases from Nepal were 83% (kappa 0.76 (0.75 - 0.80)), 71% (kappa 0.41(0.32-0.51)), and 74% (kappa 0.63 (0.60-0.63)), respectively. The proportion of stillbirths identified as fresh or macerated by the different methods of VA interpretation was similar in all three settings. Comparing across countries, the modified InterVA method found that proportions of preterm births and deaths due to infection were higher in Zimbabwe (44%) than in Malawi (28%) or Nepal (20%).

Conclusion

The modified InterVA method provides plausible results for stillbirths and newborn deaths, broadly comparable to physician review but with the advantage of internal consistency. The method allows standardized cross-country comparisons and eliminates the inconsistencies of physician review in such comparisons.

Trends in causes of death among children under 5 in Bangladesh, 1993-2004: an exercise applying a standardized computer algorithm to assign causes of death using verbal autopsy data

Background

Trends in the causes of child mortality serve as important global health information to guide efforts to improve child survival. With child mortality declining in Bangladesh, the distribution of causes of death also changes. The three verbal autopsy (VA) studies conducted with the Bangladesh Demographic and Health Surveys provide a unique opportunity to study these changes in child causes of death.

Methods

To ensure comparability of these trends, we developed a standardized algorithm to assign causes of death using symptoms collected through the VA studies. The original algorithms applied were systematically reviewed and key differences in cause categorization, hierarchy, case definition, and the amount of data collected were compared to inform the development of the standardized algorithm. Based primarily on the 2004 cause categorization and hierarchy, the standardized algorithm guarantees comparability of the trends by only including symptom data commonly available across all three studies.

Results

Between 1993 and 2004, pneumonia remained the leading cause of death in Bangladesh, contributing to 24% to 33% of deaths among children under 5. The proportion of neonatal mortality increased significantly from 36% (uncertainty range [UR]: 31%-41%) to 56% (49%-62%) during the same period. The cause-specific mortality fractions due to birth asphyxia/birth injury and prematurity/low birth weight (LBW) increased steadily, with both rising from 3% (2%-5%) to 13% (10%-17%) and 10% (7%-15%), respectively. The cause-specific mortality rates decreased significantly due to neonatal tetanus and several postneonatal causes (tetanus: from 7 [4-11] to 2 [0.4-4] per 1,000 live births (LB); pneumonia: from 26 [20-33] to 15 [11-20] per 1,000 LB; diarrhea: from 12 [8-17] to 4 [2-7] per 1,000 LB; measles: from 5 [2-8] to 0.2 [0-0.7] per 1,000 LB; injury: from 11 [7-17] to 3 [1-5] per 1,000 LB; and malnutrition: from 9 [6-13] to 5 [2-7]).

Conclusions

Pneumonia remained the top killer of children under 5 in Bangladesh between 1993 and 2004. The increasing importance of neonatal survival is highlighted by the growing contribution of neonatal deaths and several neonatal causes. Notwithstanding the limitations, standardized computer-based algorithms remain a promising tool to generate comparable causes of child death using VA data.

Performance of InterVA for assigning causes of death to verbal autopsies: multisite validation study using clinical diagnostic gold standards

Background

InterVA is a widely disseminated tool for cause of death attribution using information from verbal autopsies. Several studies have attempted to validate the concordance and accuracy of the tool, but the main limitation of these studies is that they compare cause of death as ascertained through hospital record review or hospital discharge diagnosis with the results of InterVA. This study provides a unique opportunity to assess the performance of InterVA compared to physician-certified verbal autopsies (PCVA) and alternative automated methods for analysis.

Methods

Using clinical diagnostic gold standards to select 12,542 verbal autopsy cases, we assessed the performance of InterVA on both an individual and population level and compared the results to PCVA, conducting analyses separately for adults, children, and neonates. Following the recommendation of Murray et al., we randomly varied the cause composition over 500 test datasets to understand the performance of the tool in different settings. We also contrasted InterVA with an alternative Bayesian method, Simplified Symptom Pattern (SSP), to understand the strengths and weaknesses of the tool.

Results

Across all age groups, InterVA performs worse than PCVA, both on an individual and population level. On an individual level, InterVA achieved a chance-corrected concordance of 24.2% for adults, 24.9% for children, and 6.3% for neonates (excluding free text, considering one cause selection). On a population level, InterVA achieved a cause-specific mortality fraction accuracy of 0.546 for adults, 0.504 for children, and 0.404 for neonates. The comparison to SSP revealed four specific characteristics that lead to superior performance of SSP. Increases in chance-corrected concordance are attained by developing cause-by-cause models (2%), using all items as opposed to only the ones that mapped to InterVA items (7%), assigning probabilities to clusters of symptoms (6%), and using empirical as opposed to expert probabilities (up to 8%).

Conclusions

Given the widespread use of verbal autopsy for understanding the burden of disease and for setting health intervention priorities in areas that lack reliable vital registrations systems, accurate analysis of verbal autopsies is essential. While InterVA is an affordable and available mechanism for assigning causes of death using verbal autopsies, users should be aware of its suboptimal performance relative to other methods.

Effects on the estimated cause-specific mortality fraction of providing physician reviewers with different formats of verbal autopsy data

BACKGROUND

The process of data collection and the methods used to assign the cause of death vary significantly among different verbal autopsy protocols, but there are few data to describe the consequences of the choices made. The aim of this study was to objectively define the impact of the format of data presented to physician reviewers on the cause-specific mortality fractions defined by a verbal autopsy-based mortality-surveillance system.

METHODS

Verbal autopsies were done by primary health care workers for all deaths between October 2006 and September 2007 in a community in rural Andhra Pradesh, India (total population about 180,162). Each questionnaire had a structured section, composed of a series of check boxes, and a free-text section, in which a narrative description of the events leading to death was recorded. For each death, a physician coder was presented first with one section and then the other in random order with a 20- to 40-day interval between. A cause of death was recorded for each data format at the level of ICD 10 chapter headings or else the death was documented as unclassified. After another 20- to 40-day interval, both the structured and free-text sections of the questionnaire were presented together and an index cause of death was assigned.

RESULTS

In all, 1,407 verbal autopsies were available for analysis, representing 94% of all deaths recorded in the population that year. An index cause of death was assigned using the combined data for 1,190 with the other 217 remaining unclassified. The observed cause-specific mortality fractions were the same regardless of whether the structured, free-text or combined data sources were used. At the individual level, the assignments made using the structured format matched the index in 1,012 (72%) of cases with a kappa statistic of 0.66. For the free-text format, the corresponding figures were 989 (70%) and 0.64.

CONCLUSIONS

The format of the verbal autopsy data used to assign a cause of death did not substantively influence the pattern of mortality estimated. Substantially abbreviated and simplified verbal autopsy questionnaires might provide robust information about high-level mortality patterns.

Performance of the Tariff Method: validation of a simple additive algorithm for analysis of verbal autopsies

BACKGROUND

Verbal autopsies provide valuable information for studying mortality patterns in populations that lack reliable vital registration data. Methods for transforming verbal autopsy results into meaningful information for health workers and policymakers, however, are often costly or complicated to use. We present a simple additive algorithm, the Tariff Method (termed Tariff), which can be used for assigning individual cause of death and for determining cause-specific mortality fractions (CSMFs) from verbal autopsy data.

METHODS

Tariff calculates a score, or "tariff," for each cause, for each sign/symptom, across a pool of validated verbal autopsy data. The tariffs are summed for a given response pattern in a verbal autopsy, and this sum (score) provides the basis for predicting the cause of death in a dataset. We implemented this algorithm and evaluated the method's predictive ability, both in terms of chance-corrected concordance at the individual cause assignment level and in terms of CSMF accuracy at the population level. The analysis was conducted separately for adult, child, and neonatal verbal autopsies across 500 pairs of train-test validation verbal autopsy data.

RESULTS

Tariff is capable of outperforming physician-certified verbal autopsy in most cases. In terms of chance-corrected concordance, the method achieves 44.5% in adults, 39% in children, and 23.9% in neonates. CSMF accuracy was 0.745 in adults, 0.709 in children, and 0.679 in neonates.

CONCLUSIONS

Verbal autopsies can be an efficient means of obtaining cause of death data, and Tariff provides an intuitive, reliable method for generating individual cause assignment and CSMFs. The method is transparent and flexible and can be readily implemented by users without training in statistics or computer science.

Family and community concerns about post-mortem needle biopsies in a Muslim society

Background

Post-mortem needle biopsies have been used in resource-poor settings to determine cause of death and there is interest in using them in Bangladesh. However, we did not know how families and communities would perceive this procedure or how they would decide whether or not to consent to a post-mortem needle biopsy. The goal of this study was to better understand family and community concerns and decision-making about post-mortem needle biopsies in this low-income, predominantly Muslim country in order to design an informed consent process.

Methods

We conducted 16 group discussions with family members of persons who died during an outbreak of Nipah virus illness during 2004-2008 and 11 key informant interviews with their community and religious leaders. Qualitative researchers first described the post-mortem needle biopsy procedure and asked participants whether they would have agreed to this procedure during the outbreak. Researchers probed participants about the circumstances under which the procedure would be acceptable, if any, their concerns about the procedure, and how they would decide whether or not to consent to the procedure.

Results

Overall, most participants agreed that post-mortem needle biopsies would be acceptable in some situations, particularly if they benefitted society. This procedure was deemed more acceptable than full autopsy because it would not require major delays in burial or remove organs, and did not require cutting or stitching of the body. It could be performed before the ritual bathing of the body in either the community or hospital setting. However, before consent would be granted for such a procedure, the research team must gain the trust of the family and community which could be difficult. Although consent may only be provided by the guardians of the body, decisions about consent for the procedure would involve extended family and community and religious leaders.

Conclusions

The possible acceptability of this procedure during outbreaks represents an important opportunity to better characterize cause of death in Bangladesh which could lead to improved public health interventions to prevent these deaths. Obstacles for research teams will include engaging all major stakeholders in decision-making and quickly building a trusting relationship with the family and community, which will be difficult given the short window of time prior to the ritual bathing of the body.

The epidemiology of 'bewitchment' as a lay-reported cause of death in rural South Africa

BACKGROUND

Cases of premature death in Africa may be attributed to witchcraft. In such settings, medical registration of causes of death is rare. To fill this gap, verbal autopsy (VA) methods record signs and symptoms of the deceased before death as well as lay opinion regarding the cause of death; this information is then interpreted to derive a medical cause of death. In the Agincourt Health and Demographic Surveillance Site, South Africa, around 6% of deaths are believed to be due to 'bewitchment' by VA respondents.

METHODS

Using 6874 deaths from the Agincourt Health and Socio-Demographic Surveillance System, the epidemiology of deaths reported as bewitchment was explored, and using medical causes of death derived from VA, the association between perceptions of witchcraft and biomedical causes of death was investigated.

RESULTS

The odds of having one's death reported as being due to bewitchment is significantly higher in children and reproductive-aged women (but not in men) than in older adults. Similarly, sudden deaths or those following an acute illness, deaths occurring before 2001 and those where traditional healthcare was sought are more likely to be reported as being due to bewitchment. Compared with all other deaths, deaths due to external causes are significantly less likely to be attributed to bewitchment, while maternal deaths are significantly more likely to be.

CONCLUSIONS

Understanding how societies interpret the essential factors that affect their health and how health seeking is influenced by local notions and perceived aetiologies of illness and death could better inform sustainable interventions and health promotion efforts.

An alternative strategy for perinatal verbal autopsy coding: single versus multiple coders

OBJECTIVE

To determine the comparability between cause of death (COD) by a single physician coder and a two-physician panel, using verbal autopsy.

METHODS

The study was conducted between May 2007 and June 2008. Within a week of a perinatal death in 38 rural remote communities in Guatemala, the Democratic Republic of Congo, Zambia and Pakistan, VA questionnaires were completed. Two independent physicians, unaware of the others decisions, assigned an underlying COD, in accordance with the causes listed in the chapter headings of the International classification diseases and related health problems, 10th revision (ICD-10). Cohen's kappa statistic was used to assess level of agreement between physician coders.

RESULTS

There were 9461 births during the study period; 252 deaths met study enrolment criteria and underwent verbal autopsy. Physicians assigned the same COD for 75% of stillbirths (SB) (K = 0.69; 95% confidence interval: 0.61-0.78) and 82% early neonatal deaths (END) (K = 0.75; 95% confidence interval: 0.65-0.84). The patterns and proportion of SBs and ENDs determined by the physician coders were very similar compared to causes individually assigned by each physician. Similarly, rank order of the top five causes of SB and END was identical for each physician.

CONCLUSION

This study raises important questions about the utility of a system of multiple coders that is currently widely accepted and speculates that a single physician coder may be an effective and economical alternative to VA programmes that use traditional two-physician panels to assign COD.

Capturing the context of maternal deaths from verbal autopsies: a reliability study of the maternal data extraction tool (M-DET)

Background

The availability of quality data to inform policy is essential to reduce maternal deaths. To characterize maternal deaths in settings without complete vital registration systems, we designed and assessed the inter-rater reliability of a tool to systematically extract data and characterize the events that precede a nationally representative sample of maternal deaths in India.

Method/Principal Findings

Of 1017 nationally representative pregnancy-related deaths, which occurred between 2001 and 2003, we randomly selected 105 reports. Two independent coders used the maternal data extraction tool (questions with coding guidelines) to collect information on antenatal care access, final pregnancy outcome; planned place of birth and care provider; community consultation, transport, admission, hospital referral; and verification of cause of death assignment. Kappa estimated inter-rater agreement was calculated and classified as poor (K≤0.4), moderate (K = 0.4-≤0.6), substantial (K = 0.6-≤0.8) and high (K>0.8) using the criteria from Landis & Koch. The data extraction tool had high agreement for gestational age, pregnancy outcome, transport, death en route and admission to hospital; substantial agreement for receipt of antenatal care, planned place of birth, readmission and referral to higher level hospital, and whether or not death occurred in the intrapartum period; moderate to substantial agreement for classification of deaths as direct or indirect obstetric deaths or incidental deaths; moderate agreement for classification of community healthcare consultation and total number of healthcare contacts; and poor agreement for the classification of deaths as sudden deaths and other/unknown cause of death. The ability of the tool to identify the most-responsible-person in labour varied from moderate agreement to high agreement.

Conclusions

This data extraction tool achieved good inter-rater reliability and can be used to collect data on events surrounding maternal deaths and for verification/improvement of underlying cause of death.

The democratic fallacy in matters of clinical opinion: implications for analysing cause-of-death data

Arriving at a consensus between multiple clinical opinions concerning a particular case is a complex issue - and may give rise to manifestations of the democratic fallacy, whereby a majority opinion is misconstrued to represent some kind of "truth" and minority opinions are somehow "wrong". Procedures for handling multiple clinical opinions in epidemiological research are not well established, and care is needed to avoid logical errors. How to handle physicians' opinions on cause of death is one important domain of concern in this respect. Whether multiple opinions are a legal requirement, for example ahead of cremating a body, or used for supposedly greater rigour, for example in verbal autopsy interpretation, it is important to have a clear understanding of what unanimity or disagreement in findings might imply, and of how to aggregate case data accordingly.

In many settings where multiple physicians have interpreted verbal autopsy material, an over-riding goal of arriving at a single cause of death per case has been applied. In many instances this desire to constrain findings to a single cause per case has led to methodologically awkward devices such as "TB/AIDS" as a single cause. This has also usually meant that no sense of disagreements or uncertainties at the case level is taken forward into aggregated data analyses, and in many cases an "indeterminate" cause may be recorded which actually reflects a lack of agreement rather than a lack of data on possible cause(s).

In preparing verbal autopsy material for epidemiological analyses and public health interpretations, the possibility of multiple causes of death per case, and some sense of any disagreement or uncertainty encountered in interpretation at the case level, need to be captured and incorporated into overall findings, if evidence is not to be lost along the way. Similar considerations may apply in other epidemiological domains.

Mortality measurement in transition: proof of principle for standardised multi-country comparisons

Objective

To demonstrate the viability and value of comparing cause-specific mortality across four socioeconomically and culturally diverse settings using a completely standardised approach to VA interpretation.

Methods

Deaths occurring between 1999 and 2004 in Butajira (Ethiopia), Agincourt (South Africa), FilaBavi (Vietnam) and Purworejo (Indonesia) health and socio-demographic surveillance sites were identified. VA interviews were successfully conducted with the caregivers of the deceased to elicit information on signs and symptoms preceding death. The information gathered was interpreted using the InterVA method to derive population cause-specific mortality fractions for each of the four settings.

Results

The mortality profiles derived from 4784 deaths using InterVA illustrate the potential of the method to characterise sub-national profiles well. The derived mortality patterns illustrate four populations with plausible, markedly different disease profiles, apparently at different stages of health transition.

Conclusions

Given the standardised method of VA interpretation, the observed differences in mortality cannot be because of local differences in assigning cause of death. Standardised, fit-for-purpose methods are needed to measure population health and changes in mortality patterns so that appropriate health policy and programmes can be designed, implemented and evaluated over time and place. The InterVA approach overcomes several longstanding limitations of existing methods and represents a valuable tool for health planners and researchers in resource-poor settings.

Moving from data on deaths to public health policy in Agincourt, South Africa: approaches to analysing and understanding verbal autopsy findings

BACKGROUND

Cause of death data are an essential source for public health planning, but their availability and quality are lacking in many parts of the world. Interviewing family and friends after a death has occurred (a procedure known as verbal autopsy) provides a source of data where deaths otherwise go unregistered; but sound methods for interpreting and analysing the ensuing data are essential. Two main approaches are commonly used: either physicians review individual interview material to arrive at probable cause of death, or probabilistic models process the data into likely cause(s). Here we compare and contrast these approaches as applied to a series of 6,153 deaths which occurred in a rural South African population from 1992 to 2005. We do not attempt to validate either approach in absolute terms.

METHODS AND FINDINGS

The InterVA probabilistic model was applied to a series of 6,153 deaths which had previously been reviewed by physicians. Physicians used a total of 250 cause-of-death codes, many of which occurred very rarely, while the model used 33. Cause-specific mortality fractions, overall and for population subgroups, were derived from the model's output, and the physician causes coded into comparable categories. The ten highest-ranking causes accounted for 83% and 88% of all deaths by physician interpretation and probabilistic modelling respectively, and eight of the highest ten causes were common to both approaches. Top-ranking causes of death were classified by population subgroup and period, as done previously for the physician-interpreted material. Uncertainty around the cause(s) of individual deaths was recognised as an important concept that should be reflected in overall analyses. One notably discrepant group involved pulmonary tuberculosis as a cause of death in adults aged over 65, and these cases are discussed in more detail, but the group only accounted for 3.5% of overall deaths.

CONCLUSIONS

There were no differences between physician interpretation and probabilistic modelling that might have led to substantially different public health policy conclusions at the population level. Physician interpretation was more nuanced than the model, for example in identifying cancers at particular sites, but did not capture the uncertainty associated with individual cases. Probabilistic modelling was substantially cheaper and faster, and completely internally consistent. Both approaches characterised the rise of HIV-related mortality in this population during the period observed, and reached similar findings on other major causes of mortality. For many purposes probabilistic modelling appears to be the best available means of moving from data on deaths to public health actions. Please see later in the article for the Editors' Summary.