Verbal autopsy in Vietnam : the degree of agreement on specific cause of death with medical record and families-given cause of death

Abstract

Background: Similar to middle and low-income countries, the accurate of the nationally representative statistics on mortality in Vietnam are deficient because of inadequate routine information systems. There is an effort to solve this problem by introducing the verbal autopsy (VA) method, which is the process of interviewing caregivers on the circumstances leading to death to assess the cause of death (CoD). Studies on CoD are relatively new and scarce in the Vietnamese context given its mortality registration system. In this regard, the reliability study on the VA mortality data is essential for the improvement of data quality on CoDs. This study is designed to evaluate the consistency of the VA derived CoD (VACoD) with the medical record (MCoD) and Families-Given CoD (FGCoD) in three provinces in Vietnam. Methods: This study assesses the consistency of the VACoD in three provinces of Vietnam (Hochiminh City, Binhduong Province and Lamdong Province). The data include all of death cases occurring between 1st January 2008 and 31st December 2009 in selected areas which had verbal autopsy (VA) records together with medical records or death certificates. Furthermore, the study investigates factors that relate to the agreement between these approaches by applying multivariate estimations. Results: Between 2008 and 2009, there were 4,672 VA interview records in these provinces, in which 252 cases had medical records, where 64 cases of these deaths occurred in hospital. There were 3,166 cases that had death certificates. We found that VACoD agreed with MCoD about 56% of 64 deaths. For specific causes of cardiovascular diseases occupied the highest proportion of agreement (90%). With regards to the agreement between VACoD and FGCoD, transport accident and senility represent the highest proportion of agreement (92.6%). In total, nearly a half of 3,166 death cases (47.2%) from VA agreed with those derived from death certificates in terms of causes of death. Moreover, our findings suggest that VA is robust to time after death. We found that the recall period had no effect on the agreement given the duration of the recall period in our study ranged from 9 to 24 months. Conclusion: VA agreed moderately with MCoD and FGCoD in determining the underlying causes of death. We recommend further investigations of the effect of characteristics of respondents on the level of agreement given better data in the future. We hope that our study, albeit small, provide useful results that will stimulate further works for strengthening the national cause-specific mortality statistics in Vietnam and other developing countries alike. Keywords: Cause of death, verbal autopsy, consistency, agreement, medical record, Families-Given cause of death, Vietnam.