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.

Geospatial, racial, and educational variation in firearm mortality in the USA, Mexico, Brazil, and Colombia, 1990-2015: a comparative analysis of vital statistics data.

BACKGROUND: Firearm mortality is a leading, and largely avoidable, cause of death in the USA, Mexico, Brazil, and Colombia. We aimed to assess the changes over time and demographic determinants of firearm deaths in these four countries between 1990 and 2015. METHODS: In this comparative analysis of firearm mortality, we examined national vital statistics data from 1990-2015 from four publicly available data repositories in the USA, Mexico, Brazil, and Colombia. We extracted medically-certified deaths and underlying population denominators to calculate the age-specific and sex-specific firearm deaths and the risk of firearm mortality at the national and subnational level, by education for all four countries, and by race or ethnicity for the USA and Brazil. Analyses were stratified by intent (homicide, suicide, unintentional, or undetermined). We quantified avoidable mortality for each country using the lowest number of subnational age-specific and period-specific death rates. FINDINGS: Between 1990 and 2015, 106·3 million medically-certified deaths were recorded, including 2 472 000 firearm deaths, of which 851 000 occurred in the USA, 272 000 in Mexico, 855 000 in Brazil, and 494 000 in Colombia. Homicides accounted for most of the firearm deaths in Mexico (225 000 [82·7%]), Colombia (463 000 [93·8%]), and Brazil (766 000 [89·5%]). Suicide accounted for more than half of all firearm deaths in the USA (479 000 [56·3%]). In each country, firearm mortality was highest among men aged 15-34 years, accounting for up to half of the total risk of death in that age group. During the study period, firearm mortality risks increased in Mexico and Brazil but decreased in the USA and Colombia, with marked national and subnational geographical variation. Young men with low educational attainment were at increased risk of firearm homicide in all four countries, and in the USA and Brazil, black and brown men, respectively, were at the highest risk. The risk of firearm homicide was 14 times higher in black men in the USA aged 25-34 years with low educational attainment than comparably-educated white men (1·52% [99% CI 1·50-1·54] vs 0·11% [0·10-0·12]), and up to four times higher than in comparably-educated men in Brazil, Colombia, and Mexico. In the USA, the risk of firearm homicide was more than 30 times higher in black men with post-secondary education than comparably educated white men. If countries could achieve the same firearm mortality rates nationally as in their lowest-burden states, 1 777 800 firearm deaths at all ages and in both sexes could be avoided, including 1 028 000 deaths in men aged 15-34 years. INTERPRETATION: Firearm mortality in the USA, Mexico, Brazil, and Colombia is highest among young adult men, and is strongly associated with race and ethnicity, and low education levels. Reductions in firearm deaths would improve life expectancy, particularly for black men in the USA, and would reduce racial and educational disparities in mortality. FUNDING: Canadian Institutes of Health Research and the University of Toronto Connaught Global Challenge.

Divergent trends in ischaemic heart disease and stroke mortality in India from 2000 to 2015: a nationally representative mortality study.

INTRODUCTION: India accounts for about a fifth of cardiovascular deaths globally, but nationally representative data on mortality trends are not yet available. In this nationwide mortality study, we aimed to assess the trends in ischaemic heart disease and stroke mortality over 15 years using the Million Death Study. METHODS: We determined national and subnational cardiovascular mortality rates and trends by sex and birth cohort using cause of death ascertained by verbal autopsy from 2001 to 2013 among 2·4 million households. We derived mortality rates for ischaemic heart disease and stroke by applying mortality proportions to UN mortality estimates for India and projected the rates from 2000 to 2015. FINDINGS: Cardiovascular disease caused more than 2·1 million deaths in India in 2015 at all ages, or more than a quarter of all deaths. At ages 30-69 years, of 1·3 million cardiovascular deaths, 0·9 million (68·4%) were caused by ischaemic heart disease and 0·4 million (28·0%) by stroke. At these ages, the probability of dying from ischaemic heart disease increased during 2000-15, from 10·4% to 13·1% in men and 4·8% to 6·6% in women. Ischaemic heart disease mortality rates in rural areas increased rapidly and surpassed those in urban areas. By contrast, the probability of dying from stroke decreased from 5·7% to 5·0% in men and 5·0% to 3·9% in women. A third of premature stroke deaths occurred in the northeastern states, inhabited by a sixth of India's population, where rates increased significantly and were three times higher than the national average. The increased mortality rates of ischaemic heart disease nationally and stroke in the northeastern states were higher in the cohorts of adults born in the 1970s onwards, than in earlier decades. A large and growing proportion of the ischaemic heart disease nationally and stroke deaths in high-burden states reported earlier diagnosis of cardiovascular disease, but low medication use. INTERPRETATION: The unexpectedly diverse patterns of cardiovascular mortality require investigation to identify the role of established and new cardiovascular risk factors. Secondary prevention with effective and inexpensive long-term treatment and adult smoking cessation could prevent substantial numbers of premature deaths. Without progress against the control of cardiovascular disease in India, global goals to reduce non-communicable diseases by 2030 will be difficult to achieve. FUNDING: Fogarty International Center of the US National Institutes of Health, Dalla Lana School of Public Health, University of Toronto, Indian Council of Medical Research, and the Disease Control Priorities.

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.

Exploring the social and neighbourhood predictors of diabetes: a comparison between Toronto and Chicago.

OBJECTIVES: This report examined the impact and extent that spatial access to primary care physicians (PCPs) and social neighbourhood-/community-level factors have on diabetes prevalence for Toronto and Chicago. METHODS: The two-step floating catchment area method was used to compute spatial access scores. Bivariate correlation and multivariate linear regression identified the factors that were associated with, and/or predicted, diabetes prevalence. RESULTS: Potential spatial access to PCPs had no strong associations with diabetes prevalence. Low socio-economic status factors and certain ethnic groups were strongly associated with diabetes prevalence for both cities. For Toronto, South American place of birth, households below poverty and high school-level education predicted diabetes prevalence. African ethnicity and households below poverty predicted diabetes prevalence for Chicago. CONCLUSION: Although this report found no strong association between diabetes prevalence and access to PCPs, contextual factors significant in past individual-level diabetes studies were associated with diabetes prevalence at the neighbourhood/community level for Toronto and Chicago.