Background
Nepal is a relatively poor country, whose population of approximately 30 million has significant socio-economic disparities and reside in areas of considerable geographic diversity, from remote and mountainous regions to densely-populated cities [
1‐
3]. Like other South Asian countries, Nepal is experiencing an epidemiological transition from high levels of child mortality and infectious disease mortality to the increasing importance of adult and non-communicable disease (NCD) mortality driven by rapid life-style change, unhealthy diet, tobacco use, alcohol consumption and reduced physical activity [
4,
5]. It also recently in 2015 experienced a major earthquake, which resulted in the deaths of 9000 people [
6]. Reliable and timely mortality data are therefore essential to know the level, trends and differentials of key mortality indicators, including adult mortality and life expectancy at birth, which provide evidence to health policymakers to monitor the progress of national and international health goals, including the Sustainable Development Goals (SDGs) [
7‐
10].
Despite this need for reliable mortality indicators, not much is known about mortality levels and differentials in Nepal. A high-quality civil registration and vital statistics (CRVS) system is the optimal source of routine data to provide mortality statistics, with seven out of 17 SDGs best measured using data from a CRVS system [
11,
12]. However, despite many decades of implementation of the CRVS system, national authorities have not previously used death registration data from the CRVS system to make estimates of mortality indicators. A study conducted to assess the quality and completeness of death registration data found that the completeness of the paper-based death registration system (i.e. offline data) is 69% at the national level, being below 50% in recent years in Madhesh and Karnali [
13]. The recently introduced online death registration system still only operates in a minority of districts throughout the country, and so has completeness of just 32% [
13]. Further, a limitation of the offline registration data for the production of mortality indicators is that it does not provide aggregated data by age at death. Further detail of the Nepal CRVS system is provided elsewhere [
13].
In the absence of reliable data sources, the primary estimates of mortality in Nepal are reliant on irregular data from Population Censuses (conducted every 10 years, most recently in 2011) and surveys such as the Demographic and Health Survey (DHS) and the CRVS Survey [
14‐
16]. The most recent data of reported deaths at all ages from one of these sources is from the CRVS Survey 2015/16, which had an estimated completeness of 75% [
13]. Official mortality estimates are made indirectly by the Central Bureau of Statistics (CBS), who estimate the level and patterns of child mortality, life expectancy at birth and maternal mortality ratio by applying demographic methods to different data sources, including the Population Census and DHS [
17,
18]. International groups such as the Global Burden of Disease (GBD) and United Nations World Population Prospects (UNWPP) also estimate mortality indicators of Nepal, at the national level only, based on limited data and large reliance on demographic and statistical modelling, including estimates from other countries [
13,
19,
20]. This has resulted in significant variation in the estimates. For instance, between 1981 and 2011, CBS estimated female life expectancy in Nepal increased from 48.1 to 68.0 years and male life expectancy from 50.9 to 65.4 years [
18]. Similarly, from 1980 to 85 to 2015–20 UNWPP estimated female life expectancy increased from 48.9 to 71.7 years and male life expectancy increased from 48.1 years to 68.8 years, while GBD estimated female life expectancy increased from 59.1 to 73.3 years and male life expectancy 57.7 to 68.7 years respectively between 1990 to 2017 [
4,
21].
In Nepal, there is a lack of mortality indicators at the subnational level, despite increased responsibilities for the new provinces for health program development and policy making. Nepal’s seven provinces and 753 municipalities and rural municipalities were created in 2015 after historic political change following the abolition of the monarchy. The Constitution of Nepal 2015 made the provision that the access to basic health services is a fundamental human right for every citizen, which has been implemented via the promulgation of National Health Policy of 2014, five-year Nepal Health Sector Strategy (2015–2020) and Second Long Term Health Plan (1997–2017) [
22‐
25]. These policy documents envision universal health coverage by means of equitable resource allocation and investment, and so require subnational data to inform local health policy development. There are likely to be considerable differences in mortality indicators between these provinces due to Nepal’s socio-economic disparities. There is a substantial literature on how mortality is associated with different aspects of human development, for example a strong association between socioeconomic status, including family income, and mortality, while income inequality within the population is also correlated with mortality levels [
26‐
31].
Although a complete and well-functioning CRVS system is the gold standard for generating mortality data to measure national and subnational mortality indicators, incomplete death registration can be used to generate mortality indicators using different demographic methods, as used by the UN WPP and GBD in many countries [
4,
21]. Given the policy need for reliable and disaggregated information on mortality levels and trends, as well as the opportunity to better exploit data from the existing, if imperfect, CRVS system in the country, we believe it is imperative to investigate how death registration data and other mortality data sources can be collectively used to estimate levels and subnational differentials in mortality across all ages in Nepal, which has never been conducted before. The first objective of this paper is to estimate levels and subnational differentials in mortality across all ages in Nepal, based on application of a method to derive mortality estimates from incomplete data sources. The second objective is to assess the plausibility of the subnational mortality estimates by assessing the association with monetary poverty headcount rates.
Discussion
This study has estimated three major mortality indicators for Nepal – under-five mortality, adult mortality, and life expectancy – by ecological belts and provinces by analysing the trend of five-year death completeness-adjusted registration and reporting data. The findings show considerable differences in these mortality indicators by province, with the western provinces of Lumbini, Karnali and Sudurpashchim and the ecological belt of Mountain having high mortality and low life expectancy compared with the other areas of the country. We also assessed the plausibility of estimated mortality indicators against the estimated monetary poverty levels in those sub-national regions and found a strong association between them, as expected. This offers some empirical support for the relative pattern of mortality across the country that we have identified.
These indicators are consistent with those made by the GBD and UNWPP. Our estimated adult mortality rates (male 159 per 1000, female 116) are lower than the UN (male 171, female 133), and GBD (male 180, female 130) estimates [
4,
21]. Another source of adult mortality is sibling survival data in the DHS 2016, however the only published estimates from these data are for ages 15–49 years; further, there are methodological challenges of this method, including inherent survival and recall bias and the issue of underreporting [
5,
15,
40]. Our national-level estimate of life expectancy at birth are 69.7 for males and 73.9 for females; male estimates are slightly higher than both UNWPP (68.8) and GBD (68.7) estimates and moderately higher than the UNWPP estimate (71.7) but very close to the GBD estimate (73.3) [
4,
21]. Higher life expectancy and lower mortality estimated by this study is consistent with higher completeness of registration/reporting compared with UNWPP and GBD estimates [
13]. Our estimates are significantly higher than 2011 Population Census-based estimates (male 65.4 and female: 68.0), although the difference would be partly explained by the increase in life expectancy in recent years [
3]. The similarity of our estimates to the UNWPP and GBD provides reassurance in their plausibility given the use of incomplete registration and survey data and supports the extension of the empirical completeness method that uses the age pattern of deaths from a model life table [
36].
Our analysis highlights major differences in adult mortality rates by province, with the highest provincial rate being 74% higher than the lowest rate for males and 78% higher than the lowest rate for males. The poverty rates in Karnali and Sudurpashchim provinces (more than double Province 1) and Mountain belt (almost double Hill and Terai) are all high and correspond with their high mortality/low life expectancy, while Lumbini has a more moderate poverty rate and high mortality/low life expectancy. The close association of our mortality estimates with poverty levels is justified by existing literature about the contribution of higher level-of socio-economic development and lower level of poverty for reducing mortality levels [
29,
41]. Findings consistent with ours have been found by some studies regarding tobacco consumption and other risk factors associated with adult mortality in those subnational areas. For example, a nationally representative study among adults aged 15–49 years found highest male (40%) and female (13%) smoking prevalence in Sudurpashchim and Karnali provinces and the smoking prevalence rate was also found highest in mountain region for both sexes (male 32%, female 9%) [
42]. Likewise, the Nepal Multiple Indicator Cluster Survey 2019 also found that the highest proportion (15%) of females that had ever consumed any tobacco product was in Karnali, while in Sudurpaschim 60% of males had ever consumed tobacco, the second-highest of all provinces [
43].
Moreover, the subnational results are also consistent with the DHS 2016 results showing the percentage of women aged 15–49 years who state that distance to a health facility is a serious problem they face in accessing health care when they are sick. Distance is more an issue in Karnali (72%), Sudurpashchim (70%), and Mountain (66%), where life expectancy is lower, but is less a problem in in Gandaki (38%), Bagmati (43%), Province 1 (50%) and Hill region (51%) where life expectancy is higher [
15]. These results also reflect the difference and difficulties in accessing health services due to different terrain, climate and transport access between provinces and ecological belts [
44]. These findings inform the need to develop interventions to mitigate high levels of smoking in these areas, a major risk factor for chronic respiratory diseases, lung cancer and cardiovascular diseases, to improve access the primary health care, as well as broader efforts to address the structural reasons for higher poverty. Policymakers need to pay close attention to the higher levels of adult mortality, in particular, because they are the most economically active age group, and can be also more vulnerable to both communicable (including HIV/AIDS, Tuberculosis) and non-communicable diseases [
5,
45]. More generally, national, and local governments should focus on health interventions to reduce the mortality burden in these areas.
To our knowledge, this is the first study in Nepal to estimate the subnational mortality differences using routine death registration data. The results presented provide evidence of mortality differentials in Nepal that can be used to assist health policy and that would also be more reliable if there were a complete CRVS system. An improved CRVS system would be able to provide continuous, reliable, and disaggregated mortality statistics to generate evidence for policy interventions and to monitor the progress of SDGs and other health indicators. CRVS data is especially beneficial subnationally, and in Nepal the newly formed provinces are currently dependent on poor quality or non-existent data and are designing local policies without reliable evidence. Improving the CRVS system will have major benefits. Firstly, a large volume of information related to the personal events has been stored at each local level in event registers and notification forms for decades and this could be used to produce reliable mortality statistics at local level. Secondly, the locally produced results could be easy validated and evaluated to ensure their quality on regular basis. Thirdly, local governments can produce and use those statistics as per their specific requirements. This practice will also help to build a system to utilize CRVS system generate vital statistics in local level policies.
Nepal also lacks reliable cause of death data to guide and monitor effective policy interventions and equitable allocation of the resources. Cause of death data are not adequately reported in the national CRVS system and hospital deaths also either do not record a cause of death or use a medical certificate of cause of death that is not aligned with the World Health Organization International Medical Certificate of Cause of Death using the International Classification of Diseases (ICD-10) [
46,
47]. More than 70% of deaths in Nepal occur in the community where only a rudimentary list of causes like ‘natural deaths’ are collected in census and surveys which have very limited statistical utility, and with no routine cause of death data collection [
48,
49]. In this context, modelled global estimates including GBD are only the present alternate source of COD in Nepal [
4].
This study has some limitations that are related to the drawbacks of available mortality data in Nepal. Firstly, to be able to generate age-specific mortality statistics given no age at death in death registration data, we relied on the use of a model life table. The plausibility of resultant national adult mortality and life expectancy estimates with both the GBD and UN however suggests that the choice of model life table is appropriate. Additionally, the empirical completeness method extension employed to integrate the model life table with the empirical completeness method can produce estimates biased by incorrect population data, especially given that a projection from the 2011 Population Census was used. Such an analysis should be repeated once the 2021 Census population figures are available. The under-five mortality used was also an estimate, however it is based on multiple sources of data for each subnational area and scaled to IGME national estimates.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.