Abstract
There has been no systematic evaluation of Viet Nam’s A6 mortality reporting system. An evaluation was undertaken in 3 provinces in Viet Nam. Deaths identified in the A6 system were compared with deaths identified by an independent consensus panel to determine the percentage completeness of the A6 system. Verbal autopsies (VAs) were conducted for all identified deaths from the consensus panels, and the sensitivity and positive predictive value of the A6 system was assessed. The sensitivity of the A6 system varied depending on the cause of death, with the sensitivity of the system being excellent for injury (sensitivity = 75.4%), cancer (sensitivity = 66.9%), and circulatory diseases (sensitivity = 63.1%). The A6 mortality reporting system performs well in relation to its completeness and classification of 3 leading causes of death—namely, circulatory disease, cancer, and injury. With further enhancements and ongoing support from government and donor agencies, the A6 system will be a valuable resource.
Keywords
Background
Mortality reporting systems have provided valuable data for monitoring the public’s health globally, 1 and they contribute to monitoring the progress of key health policies such as the United Nations Millennium Development Goals for 2015 of reducing child and maternal mortality and mortality from HIV/AIDS, tuberculosis, and malaria. 2 It is important to note that mortality reporting systems are the cornerstone of a country’s health information system; yet in many low- and middle-income countries, mortality reporting systems either do not exist, or if they exist, have significant limitations, including misclassification of the cause of death, underreporting, lack of timeliness, and/or incomplete capture of mortality. 1
In many counties, civil registration and vital statistics systems suffice as the key measure for mortality because data on deaths recorded for legal purposes tends to capture the majority of deaths. 3 In countries such as Viet Nam, where civil registration and vital statistics systems operated by the Ministry of Justice have been established for more than 50 years, it has become apparent that these systems have limitations. For example, recent studies of the Vietnamese civil registration system identified limitations, including a lack of completeness, particularly for early neonatal deaths and deaths of temporary residents and/or migrants. 4 Also, the death certificate provided by the authority does not require the signature of the medical doctor, and therefore, the cause of death is poorly recorded. 4 In parallel with the civil registration and vital statistics, the General Office for Population and Family Planning reports child and maternal mortality and groups many causes of death, including those resulting from crime; hence, the accuracy and completeness of the causes of death provided by these 2 mortality registration systems is unreliable for national mortality reporting. 4
Aware of the limitations of relying on a civil registration and vital statistics system as a national health information system, the Ministry of Health of the Socialist Republic of Viet Nam established in 1992 the A6 mortality reporting system. 5 The A6 mortality reporting system relies on commune-level health officials providing basic demographic data and information on the cause of death, which is recorded in an official book referred to as the A6. The data from the A6 are collated by the district-level health service, and the information is then forwarded to the provincial and central-level governments. The commune-level officials play a significant role in maintaining the current mortality reporting system and, in turn, are able to actively use the information gained to plan commune-level health services.
To date, data from the A6 mortality system has been used to enumerate the status of Viet Nam’s civil registration system, 4 to assess mortality patterns,6-9 and to assess the validity of the A6 system to capture injury-related deaths. 10 However, no systematic evaluation of the A6 system has been undertaken in relation to all-cause mortality or to enumerate (subject to its robustness) age-standardized mortality rates (ASRs) for all causes of death. The proposed study, therefore, assesses the completeness, sensitivity, and positive predictive value of the system in relation to all-cause mortality and estimates the age-specific mortality rates for Viet Nam.
Methods
The evaluation of the A6 system was undertaken in 3 provinces in Viet Nam—namely, Bac Ninh (north area), Lam Dong (central area), and Ben Tre (south area) provinces (see Figure 1). Viet Nam has an estimated population of 86 million in 2008, 11 and these provinces were randomly selected to represent the geographic distribution of Viet Nam with Bac Ninh in the northern red delta river region, Lam Dong in the central highlands region, and Ben Tre in the southern Mekong delta region; this has been described elsewhere. 10 Bac Ninh had a total population of approximately 1 million, with 7 districts and 125 communes; Lam Dong had a population of 1.2 million, with 8 districts and 145 communes; and Ben Tre had a population of 1.4 million with 7 districts and 160 communes in 2008.12,13 In each province, 3 districts were randomly selected to represent rural, suburban, and city areas, and all communes in these districts participated in the evaluation (n = 140). All methods and instrumentation used for this evaluation were approved by both the ethics committee of the Hanoi Medical University and the Human Research Ethics Committee of the University of Sydney.
Study locations: Bac Ninh in the north, Lam Dong in the center, and Ben Tre in the south.
Completeness of the system (as described by Silva) 14 was assessed by comparing all deaths reported in the A6 book from the communes (n = 140) for the observation period January 1, 2008, to December 31, 2009 (for 1 district in the Lam Dong Province, only data for the calendar period January 1 to December 31, 2008, were available) with those reported from an independent consensus panel established for this study in each commune. To do this, a list of all deaths was compiled from the A6 records comprising the age and sex of the deceased, the date of death, and the cause of death.
Because all deaths are compiled and reviewed (n = 5613) at the commune level each year by commune representatives of the Ministry of Justice and Ministry of Population and Family Planning, we also used these listings as independent sources based on which to assess the reported deaths with the deaths compiled from the A6 records. For the purpose of this study, both the commune justice officer and the commune population family planning officer were sent a form requesting the following data from their independent systems: the name, age, sex, date of death, and cause of death of the deceased. Consensus panels comprising the commune health officer, the commune justice officer, and the commune population family planning officer were established (a total of 420 commune-level staff), and each member participated in a series of workshops in which each death and all children born during the study period (January 1, 2008, to December 31, 2009) were reviewed, and a consensus was achieved on the list of all deaths in their communes for the observation period. A comparison of the deaths captured in the A6 system with those identified from the consensus panel was undertaken by the authors to estimate the percentage completeness of the A6 system.
To assess the sensitivity, specificity, and positive predictive value of the A6 system, an external gold standard classification of the cause of death was needed. For the purpose of this study, the standardized and validated verbal autopsy (VA) developed by WHO was utilized. 15 To administer the VA, a team of commune health staff (n = 140) were trained in the administration of the VA, and then, each field worker visited the household of the deceased and administered 1 of 3 VA questionnaires; separate questionnaires were used for deaths in each of 3 age groups (0- to 28-day-old infant, child 29 days to <5 years old, and 5 years onward). The instruments included a series of questions relating to various signs and symptoms of different diseases and injury along with an open narrative section that was completed with the aid of the field worker. Cause of death assignment for the VAs was undertaken, independently, by 2 experienced physicians, each of whom had more than 15 years of clinical experience and who had previously worked in allocating causes of death from surveillance systems. An underlying cause of death (the disease or injury that initiated the events directly leading to death) for all cases was assigned wherever possible (92%, n = 5168 cases). The immediate (the disease, condition, or complication that occurred closest to the time of death) and contributory (a condition that originated in the underlying cause and terminated in the immediate cause) causes of death were also assigned. The cause of death was assigned from the 10th version of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) 16 and comprised approximately 54 groups of causes of death that could be confidently assigned by the experienced physicians based on the information typically provided in a VA. Assigning the cause of death was enhanced by utilizing a previously developed series of algorithms.15,17,18 A further enhancement to the classification of cause of death was the review of the medical/health records of the deceased; the majority of deaths had either a hospital or commune health station record or a private medical doctor record (65.6%, n = 3684). The mortality review, which was also undertaken by the physicians, provided assistance when the underlying cause of death differed from the immediate cause of death. The mortality review also assisted in cases where there was disagreement between the allocated causes of death by each physician. The mortality review of medical records was examined for the last health event that might be related to the cause of death, such as the underlying cause of death or immediate cause of death or the contributing cause of death. If the mortality review disagreed, that is, if there was disagreement between the 2 reviewers, we undertook a telephone interview again with the family member who provided care to the patient, and following the interview, a final classification of the cause of death was made. 10 For the κ estimation, we followed the ICD-10 coding, which comprised 18 groups of cause of death.
Statistical Analysis
To determine the agreement between the 2 methods of determination of the cause of death by ICD-10 chapters, the κ statistic was used. Causes of death reported by A6 and determined by the VA were categorized as 3 groups—namely, underlying cause of death (code 1), immediate causes of death (code 2), and the other (code 3), including the contributing cause of death or cause by ICD-10 chapters. The κ statistic measure of agreement is scaled to be 0 when the agreement is what would be expected to be observed by chance and 1 when there is perfect agreement: 0.00 to 0.20, poor; 0.21 to 0.40, fair; 0.41 to 0.60, moderate; 0.61 to 0.80, substantial; 0.81 to 1.00, almost perfect. 19 The standard approach for estimating the sensitivity, specificity, and positive predictive value of the A6 system14,20 was utilized. The VA method (the best available) was taken as the standard, and the study test was the A6 system. All analyses were undertaken using STATA software version 10.0. 19
Comparison of the age and sex distribution of the selected provinces was compared with the age and sex distribution for the entire Vietnamese population, and deaths were converted to crude mortality rates and ASRs (the direct standardization method using the World Population was undertaken) 14 by age and sex. The denominator for the mortality rates comprised the population of the 140 communes, including permanent residents and nonpermanent residents of the commune who had lived in the commune for more than 6 months at the time of data collection. The data were obtained from the annual population statistics from the 140 communes for 2008-2009 and verified against the estimates from the 2009 census. 11 Persons who had lived in the commune for less than 6 months were excluded from the denominator (n = 3).
Results
As highlighted in Figure 2, the population demographics from the 3 selected provinces reflect the age and sex distribution of the Vietnamese population as reported in the National Census 2009. 11
The study and national census populations pyramid.
A total of 5613 deaths (2014 cases in Bac Ninh, 1636 cases in Lam Dong, and 1963 cases in Ben Tre Provinces) were identified (a total of 5273 deaths were identified from the A6 system and 340 cases identified by the independent consensus panel), and 323 causes of death were classified according to the ICD-10 codes. Average completeness was 93.7%: Bac Ninh, 95.2%; Lam Dong, 94.6%; and Ben Tre, 92.2%.
Table 1 presents the agreement between the A6 mortality reporting system and the VA method for classifying the cause of death. It is evident from Table 1 that agreement was excellent for the following ICD-10 codes—namely, abnormality (Q00-Q99) with 91.67% agreement, injury (S00-Y89) with 85.46%, cancer (C00-D48) with 78.76%, and circulatory diseases (I00-I99) with 71.12% agreement. These 4 ICD-10 groups were responsible for two-thirds (66.6%, 3732 of 5613 cases) of all mortality. It is important to note that the agreement was poor for blood disease (25.0%, 4 cases, ICD-10: D50-D89), skin diseases (40.0%, 10 cases, ICD-10: L00-L99), and endocrine (46.3%, 153 cases, ICD-10: E00-E90). However, these 3 groups were responsible for only 3% of all causes of death (167 cases of 5613).
κ Agreement Between Verbal Autopsy (VA) and the A6 System by 18 Group Causes.
Abbreviation: ICD-10, International Statistical Classification of Diseases and Related Health Problems.
As expected, the sensitivity of the A6 system varied depending on the cause of death, with the sensitivity of the system being excellent for injury (S00-Y89) as a cause of death (sensitivity= 75.4%), followed by cancer (C00-D48; sensitivity = 66.9%), and circulatory diseases (I00-I99; sensitivity = 63.1%). The specificity of the system was also excellent for the 3 causes of death—namely, 98.4%, 94.5%, and 84.2%, respectively (see Table 2). The positive predictive value for the same 3 causes ranged from good to excellent—namely, 88.4%, 77.9%, and 62.7%, respectively.
Sensitivity, Specificity, and Predictive Values for 3 Leading Causes of Deaths. a
Proportion of deaths resulting from injury (13.6%), cancer (22.2%), and circulatory diseases (29.9%), accounting for 65.7% of all deaths.
The leading causes of deaths were circulatory diseases (I00-I99), with the estimated ASR of 114.3 per 100 000, followed by cancer (C00-D48, ASR 96.1 per 100 000), and injury (S00-Y89, ASR 52.3 per 100 000). Male to female ratios for the 3 groups were 1.2, 2.0, and 2.7, respectively. The total number of deaths caused by infectious diseases (bacterial diseases [A00-A99], viral and parasitic [B00-B99]) were 232 (4.1% of all cases, ASR 16.0 per 100 000). In all, 579 cases had a classification of a symptom (R00-R99), accounting for 10.3% of all deaths (see Table 3). The maternal mortality rate (as derived from this study) was 9.8 per 100 000 live births, and the newborn mortality rate was 275.5 per 100 000 live births. The crude mortality rate for tuberculosis (A15-A19) was 9.1 per 100 000, for HIV/AIDS (B20-B24) 4.6 per 100 000 population, and diabetes (E10-E14) 9.4 per 100 000 person-years (data not shown in tables).
Mortality Rates by Sex and 18 ICD-10 Cause-of-Death Groups.
Abbreviations: ICD-10, International Statistical Classification of Diseases and Related Health Problems; ASR, age-standardized mortality rate.
Mortality rates were estimated with a denominator of 20 307 cases of alive newborn: the number of newborn deaths was 56 (mortality rate = 275.8 per 100 000 live births). The maternal mortality rate was 9.8 per 100 000.
Discussion
The A6 mortality reporting system captures almost 94% of deaths (a total of 5273 deaths were identified from the A6 system, with a further 340 cases identified by the independent consensus panel). The system also captures and accurately classifies the 3 leading causes of deaths in Viet Nam (66% of all deaths)—namely, circulatory diseases, cancer, and injury.
This is an important finding, given that the A6 reporting system is a low-cost system, highlighting the fact that the system could operate as a valid system for enumerating all-cause mortality and could comprise the foundation on which future sentinel surveillance and associated research could be conducted. Particularly important is the fact that the system captures the majority of deaths and certainly considerably more than the criteria used in the recent Global Burden of Disease Study, which required a percentage completeness criteria of 70% for inclusion of global and regional mortality systems. 21
The overall sensitivity, specificity, and positive predictive value of the system indicate that the A6 system is performing well for the leading causes of death, which comprise two-thirds of all deaths in Viet Nam. However, there are a number of causes of death for which the A6 system is not sensitive—namely, blood diseases (ICD-10: D50-D89; κ agreement of 25%) and skin diseases (ICD-10: L00-L99; κ agreement of 40%). Among the 4 cases caused by blood diseases, 3 were reported as being incorrect malignant neoplasms of the lymphoid, hematopoietic, and related tissue (C96) of the liver and intrahepatic bile ducts (C22) and septicemia (A41). Similarly, observations for skin diseases (L00-L99) were misreported as not skin diseases in the A6 system.
Variation in the sensitivity of causes of deaths is not unexpected; the sensitivity of mortality data systems reported by the US Veterans Affairs have been reported to vary between 12.0% and 92.1%. 22 What the variation in the A6 system highlights, however, is that efforts are needed to enhance the classification of the cause of death for a considerable number of causes of death and, particularly, deaths related to bacterial diseases (ICD-10: A00-A99), urology (ICD-10: N00-N99), blood diseases (ICD-10: D50-D89), skin diseases (ICD-10: L00-L99), and symptoms (ICD-10: R00-R99); see Table 1.
The study provided a valuable enumeration of the leading causes of death. For both the male and female population, the leading cause of death was circulatory disease, with the mortality rate consistent with previously reported estimates.9,23 The estimated number of deaths caused by cancer (the second leading cause of death identified in the A6 system) was consistent with the estimates reported for Viet Nam for all cancers reported by the International Agency for Research on Cancer (IARC) 24 ; the third most common cause of death was injury, and the estimate in this study concurs with a previous study reported in the Ba Vi District of Ha Tay Province (FilaBavi) 9 as well as with an annual national injury survey published by the Ministry of Health for the period 2005 to 2009.13,25-27
The current study was the largest study undertaken in Viet Nam that evaluated the mortality reporting system and enumerated the causes of death by ICD-10 chapters. 16 Until now, the most recent study was the National Health Suvey 28 (n = 2481), which was undertaken 12 years ago and a VA study (of only 189 deaths) in FilaBavi.9,23
It is important to consider the limitations of this study when interpreting the findings—namely, the study relied on the recall of events by household members (34% of the 5613 cases had not been admitted to hospital) 10 and by accessing hospital records to develop the VA databases; each has well-documented limitations in relation to capturing the leading causes of death.9,18 As highlighted, the study was undertaken in only 3 provinces (there are a total of 63 provinces in Viet Nam), yet despite the few provinces captured, we selected provinces that represented the geographic distribution of Viet Nam, with Bac Ninh in the northern red delta river region, Lam Dong in the central highlands region, and Ben Tre in the southern Mekong delta region. It is important to note that the demographic features of the study population reflect the population demographics for Viet Nam as reported in the national census for Viet Nam (2009). 11
This study highlights that for the leading causes of disease, the A6 mortality reporting system is a robust health information system. What is important is that the study highlights aspects in which the system is not sensitive and that will require enhancements, particularly the provision of training for health workers who complete the classification of all-cause mortality at the commune level. Too often, surveillance systems in low-income countries are established in association with funded projects without regard for currently operating systems that could benefit from the funding to provide the enhancements needed to ensure it can deliver, in a sustained way, as a key health information system for the country. 10 With financial support to ensure the necessary enhancements can be implemented, the A6 mortality reporting system has the rudiments of an excellent health information system for capturing all-cause mortality in Viet Nam.
Footnotes
Acknowledgements
We deeply appreciate the Management/Advisory Board for the study namely, doctors Nguyen Duc Hinh, Nguyen Thi Bich Lien, Tran Nhu Nguyen, Nguyen Phuong Hoa, Nguyen Thi Lua, and Lai Thi Minh Hang (HMU); Nguyen Dinh Tam (Bac Ninh); Pham Thi Bach Yen (Lam Dong); and Nguyen Van Thang (Ben Tre). We acknowledge the numerous officers of the 3 provincial health departments, officers of the 9 district health centers, and the 420 commune health workers and justice and population family planning officers who participated in the study.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was supported by an AusAID Research and Development Grant. MS is supported by a National Health and Medical Research Council (Australia) Fellowship. The final manuscript was completed at the Harvard School of Public Health with sponsorship from UICC under the academic supervision of Prof Walter C. Willett to Dr Le Tran Ngoan.