Level of evidence of verbal autopsy – Authors' reply

Most of the 60 million annual global deaths occur in low-income and middle-income countries and most of these occur without medical attention at the time of death, hence the causes of death are unknown.1Jha P Reliable direct measurement of causes of death in low- and middle-income countries.BMC Med. 2014; 12: 19Crossref PubMed Scopus (66) Google Scholar Thus, alternative systems are needed to measure causes of death especially to understand nationally representative patterns of deaths. Verbal autopsies have been used to determine childhood causes of death for decades, and we have extended such measurement for adults in India using careful quality-control efforts and dual physician-coding in the Million Death Study (MDS).2Aleksandrowicz L Malhotra V Dikshit R et al.Performance criteria for verbal autopsy-based systems to estimate national causes of death: development and application to the Indian Million Death Study.BMC Med. 2014; 12: 21Crossref PubMed Scopus (50) Google Scholar A key lesson from the MDS2Aleksandrowicz L Malhotra V Dikshit R et al.Performance criteria for verbal autopsy-based systems to estimate national causes of death: development and application to the Indian Million Death Study.BMC Med. 2014; 12: 21Crossref PubMed Scopus (50) Google Scholar is the need to avoid the fallacy of a “gold standard“of hospital-based causes of death. Hospital-based deaths are a poor choice to validate verbal autopsies since variables such as the age and education of the person, pathogen distribution, and causes of death differ greatly between rural unattended deaths and hospital-based deaths. The findings from the MDS2Aleksandrowicz L Malhotra V Dikshit R et al.Performance criteria for verbal autopsy-based systems to estimate national causes of death: development and application to the Indian Million Death Study.BMC Med. 2014; 12: 21Crossref PubMed Scopus (50) Google Scholar show sharp differences in the proportion of deaths from various causes between hospital and home deaths at ages 5–69 years, even after adjustment for differences in education and age (table). Similarly, deaths caused by malaria rarely occur in hospitals, since malaria is mostly cured when treatment is given. By contrast, malaria remains a common cause of rural, unattended fever deaths among adults in India and in Africa.3Registrar General of India and Centre for Global Health ResearchCauses of death in India, 2010–2013: Sample Registration System. Government of India, New Delhi2016Google Scholar, 4Dhingra N Jha P Sharma VP et al.Adult and child malaria mortality in India: a nationally representative mortality survey.Lancet. 2010; 376: 1768-1774Summary Full Text Full Text PDF PubMed Scopus (203) Google ScholarTableCause-specific mortality fractions for original RGI surveyors versus independent random re-sample teams, and for home versus hospital deaths, with p value tests differences, Million Death Study deaths 2001–03 at ages 5–69 years800 RGI surveyors70 random re-sample teamsp value for differences: re-sample vs RGIHome deaths*This analysis excludes 6549 deaths occurring in other places and 1834 with unknown location. NS=not significant. RGI=Registrar General of India.Hospital deaths*This analysis excludes 6549 deaths occurring in other places and 1834 with unknown location. NS=not significant. RGI=Registrar General of India.p value for differences: hospital vs homeTotal number of deaths63 139181143 97910 779Heart attacks7557 (12·0%)239 (13·2%)NS4964 (11·3%)1590 (14·8%)<0·01Other infectious diseases7005 (11·1%)234 (12·9%)<0·016430 (14·6%)1153 (10·7%)<0·01Tuberculosis5714 (9·0%)139 (7·7%)NS4741 (10·8%)636 (5·9%)<0·01Cancer5511 (8·7%)152 (8·4%)NS4183 (9·5%)984 (9·1%)<0·01Chronic lung disease5494 (8·7%)134 (7·4%)NS4646 (10·6%)515 (4·8%)<0·01Stroke4526 (7·2%)137 (7·6%)NS3290 (7·5%)906 (8·4%)<0·01Suicides2647 (4·2%)60 (3·3%)NS1467 (3·3%)506 (4·7%)NSRenal2511 (4·0%)67 (3·7%)NS848 (1·9%)285 (2·6%)<0·01Liver cirrhosis2463 (3·9%)60 (3·3%)NS1707 (3·9%)545 (5·1%)NSMalaria2094 (3·3%)40 (2·2%)NS1618 (3·7%)326 (3·0%)NSRoad traffic accidents1864 (3·0%)71 (3·9%)NS230 (0·5%)484 (4·5%)<0·01Maternal conditions1053 (1·7%)14 (0·8%)<0·01522 (1·2%)343 (3·2%)<0·01HIV and other sexually transmitted infections439 (0·7%)8 (0·4%)NS359 (0·8%)50 (0·5%)<0·01Ill defined3766 (6·0%)108 (6·0%)NS2576 (5·9%)370 (3·4%)<0·01All other causes10 496 (16·6%)348 (19·2%)<0·016398 (14·5%)2086 (19·4%)<0·01All p values adjusted for age, rural or urban status, and education level.* This analysis excludes 6549 deaths occurring in other places and 1834 with unknown location. NS=not significant. RGI=Registrar General of India. Open table in a new tab All p values adjusted for age, rural or urban status, and education level. Evidence for verbal autopsy should consider whether independent resampling yields similar results. The proportions of major cause of death at ages 5–69 years, and their rank order are very similar between the approximately 3% of deaths that were independently and randomly resampled and those from the main MDS (table). This finding suggests stability and reproducibility of the MDS methods, but makes no specious claim to validity. We have published other performance criteria,2Aleksandrowicz L Malhotra V Dikshit R et al.Performance criteria for verbal autopsy-based systems to estimate national causes of death: development and application to the Indian Million Death Study.BMC Med. 2014; 12: 21Crossref PubMed Scopus (50) Google Scholar including the fact that the MDS is able to make use of more International Classification of Diseases (ICD)-10 codes than alternative classifier systems. As well, the proportion of ill-defined deaths before age 70 years is quite low, but verbal autopsy is less accurate after this age.1Jha P Reliable direct measurement of causes of death in low- and middle-income countries.BMC Med. 2014; 12: 19Crossref PubMed Scopus (66) Google Scholar This finding is consistent with broad public health goals to reduce premature deaths and disability, which in most low-income and middle-income countries arises mostly from mortality before age 70 years.5Norheim OF Jha P Admasu K et al.Avoiding 40% of the premature deaths in each country, 2010–30: review of national mortality trends to help quantify the UN sustainable development goal for health.Lancet. 2015; 385: 239-252Summary Full Text Full Text PDF PubMed Scopus (184) Google Scholar Thus, the evidence is strong for the use of well standardised and reproducible verbal autopsy methods to improve public health planning in countries where medically certified causes of death remain a distant reality. We declare no competing interests. Level of evidence of verbal autopsyVerbal autopsy determines probable causes of death when neither medical records nor formal medical attention are given. Usha Ram and colleagues (December, 2015)1 extrapolated cause-specific mortality risks for men and women aged 15–69 years on the basis of deaths accessed by verbal autopsy during 2001–06 in India.1 But many changes have occurred in the past decade. For example, the maternal mortality ratio for India in 1992 was 437 deaths per 100 000 livebirths; it decreased to 212 deaths per 100 000 livebirths during 2007–09, and it was expected to reach 135 deaths by 2015. Full-Text PDF Open Access

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.