Bulletin of the World Health Organization

Mortality in the Islamic Republic of Iran, 1964–2004

Ardeshir Khosravia, Richard Taylora, Mohsen Naghavib, Alan D Lopeza

Introduction

Reliable information on a population’s levels and patterns of causes of death is essential for health planning and priority-setting for interventions to control diseases and injuries, and for population-based evaluation of health programmes.1 The ultimate data source is an accurate and complete national vital registration system. Unfortunately, WHO reports that complete mortality data are available routinely for only a minority of countries worldwide. Completeness is highest (> 90%) in developed countries (e.g. in Europe and North America) and lowest in developing countries, particularly in Africa.2,3

Mortality data sources

As in many other developing countries, the Islamic Republic of Iran has incomplete mortality information.4–6 Currently, the National Organization for Civil Registration (NOCR) and the Ministry of Health and Medical Education (MOH&ME) operate death registration systems (Fig. 1).

Fig. 1. Overview of death registration systems of various organizations in the Islamic Republic of Iran
Fig. 1. Overview of death registration systems of various organizations in the Islamic Republic of Iran

The NOCR was established in 1918 and is responsible for the registration of births, marriages, divorces and deaths.7 In 1984 the Iranian Parliament revised and passed civil registration legislation requiring a report to be sent to the NOCR within ten days of any Iran resident’s death. In addition, deaths must be confirmed by a physician or two witnesses.7

Of the four vital events registered by the NOCR, mortality data are the least complete and of the poorest quality.6 This may be due in part to unregistered births when the infants die soon after birth. Recent assessments suggest some progress in capturing and registering deaths but problems remain, including delayed registration and inaccurate recording of causes of death. As a result, previous studies generally have not used this data source for analysing Iranian mortality.6,8

Until 2002, the MOH&ME Deputy of Research and Technology (DRT) operated a system to record the causes of deaths.9–11 Between 1966 and 1995, mortality data based on cemetery records were collected in a sample of 24 cities. In 1995, the system was redesigned to cover the entire country.9,12 This did not allow delayed registration of deaths and thus was more timely than the NOCR data. However, the causes of death were based on cemetery records and therefore were susceptible to bias (e.g. information bias due to misclassification).8 No study has assessed the validity of this data. The system was discontinued after the launch of the MOH&ME Deputy of Health (DH) programme.

The DH established a comprehensive programme for death registration in order to improve district health networks’ registration and certification of all deaths by age, sex, cause and place of residence. This was piloted in Busher Province in 1997. To improve the level of registration, different sources of mortality data in a district (e.g. NOCR, cemetery and hospital) are cross-checked for omissions and duplication. The annual report of mortality data in ten provinces for 2000 compared these results with those from the NOCR and the DRT and found more reliable completeness and classification of the causes of death.13 For instance, the DH recorded 6 681 more deaths than the NOCR and 15 390 more than the DRT.13 The DH system is being implemented progressively and currently covers 29 provinces. It is expected that all 30 provinces will be covered by the end of 2007.

The Iranian Vital Horoscope (VH) is another mortality and population source that operates in rural areas. Rural health workers (behvarz) conduct periodic censuses and household visits to record information in family files. A notice pinned to the wall of each Health House (the basic primary health-care unit, covering between one and four villages) displays this up-to-date information on the number of births, deaths and population and family planning activities within the rural community. The VH is sent to the District Health Centre at year’s end; these data are entered into a customized computer programme and forwarded to Tehran for aggregation at national level.14 Since 1988, the Deputy of Health has analysed and published the results of this innovative system.15

Other mortality data sources include censuses and surveys conducted by the Statistical Centre of Iran (SCI), and MOH&ME surveys. Mortality estimates obtained from censuses or surveys using direct methods (retrospective questions about deaths in households during the previous 12 months) suffer from well-known problems of undercounting and misreporting of age and sex, particularly for adult mortality. For this reason, indirect demographic techniques have often been used to estimate the level of mortality and to create life tables. The level of child mortality and information on children ever born and children surviving (CEBCS) have been estimated by collecting and analysing data according to the Brass methodology.4,6 The results of this analysis and data sources are shown in Fig. 2.

Methods

A systematic review was conducted to obtain all published and unpublished documents in Farsi and/or English concerning child mortality in Iran between 1973 and 2004. Unpublished data and government reports were obtained from various organizations that collect mortality data, such as the NOCR, SCI and MOH&ME.4,5,16–18 The SCI carried out the most comprehensive assessment of infant mortality trends from 1956 to 2021, based on data from various studies conducted between 1973 and 2001 using Brass’s indirect CEBCS method (Table 1, available at: http://www.who.int/bulletin).4 Adult mortality was estimated as the risk of death between ages 15 and 60 based on the following available data sources:6,16–27 Table 1 Studies on Iranian mortality, 1973–2001 Year Researcher/survey Organization Indexes Study method 1973–1976 Baseline population growth survey Statistical Centre of Iran a, b c, d, e 1973 Iran’s vital indexes survey Tehran University a, b f, d, e 1977 Fertility survey Statistical Centre of Iran a, b c, d, e 1978 Farhani, EM, (abridged life table of Iran, 1956-1966) – b g 1982 Iran’s life tables based on 1976 census Statistical Centre of Iran b g 1984 Mortality and fertility survey MOH&ME a, b f, e 1986 Census Statistical Centre of Iran a, b f, d, e 1987 Socioeconomic characteristics of Iranian households Statistical Centre of Iran a, b c, d, e 1990 Measuring population change Statistical Centre of Iran a c, d 1991 Birth survey NOCR a f, d 1991–1994 Intercensal population survey Statistical Centre of Iran a, b c, d, e 1994 Mortality and fertility survey MOH&ME a h 1992–1995 Socioeconomic characteristic of Iranian households Statistical Centre of Iran a, b c, d, e 1996 Mortality and fertility survey MOH&ME a, b f, d, e 1998–1999 Measuring population growth survey Statistical Centre of Iran a, b c, d, e 2000 Zanjani and Nourollahi – b g 2000 Iranian DHS MOH&ME and Statistical Centre of Iran a,b f, h, d, e 2001–2003 Socioeconomic characteristics of Iranian households Statistical Centre of Iran a, b c, d a Infant mortality.b Adult mortality.c Multiple-round surveys (retrospective method, questioned member of household on number of deaths at different stages of the study).d Children ever born/children surviving.e Retrospective study for adult mortality (deaths in household during last 12 months).f Single-round survey (retrospective method, questioned on number of deaths in household in past 12 months).g Life tables based on census data.h Direct information. DRT: death registration system for 2001 (entire country except Tehran Province); DH: VH data for 1992–2004; death registration in 18 provinces in 2000–2001, 23 in 2003, and 29 (of 30) in 2004; and NOCR-registered deaths in 2002 and 2003 for the whole country. To remove natural disaster effects on background mortality rates, 28 745 deaths from the Bam earthquake in 2003 were excluded from the analysis.The most recent population census in the Islamic Republic of Iran was carried out in 1996 and no up-to-date population estimates by age or sex were available.28 As a result, we used the United Nations (UN) estimated population size by age and sex to calculate national age-specific death rates from 2001 onwards. We applied the observed ratio in the Iranian Demographic and Health Survey (DHS) 2000 to the under-5 population estimated by the UN in order to separate this population into those aged 0–1 and 1–4 years.18

Analytical methods

Estimation of child mortality from previous studies

With relatively few studies on child mortality over the past three decades, trends in infant mortality were assessed first as estimates of these were more readily available. The results of all studies carried out between 1973 and 2001, and of MOH&ME death registrations between 1992 and 2004, were collected and located in time between 1958 and 2004 (Fig. 2). Given the methodological limitations of some estimation techniques and small sample sizes, some estimates were judged to be implausible and needed to be censored. Standard statistical methods were used to censor implausible data points and infant mortality for 2004 was predicted from the fitted trend line (Fig. 2).29 Child mortality by sex was then estimated from overall infant mortality based on observed age-sex patterns from the Iranian DHS.

Assessment of 2004 death registration data

Brass’s growth balance method was used to adjust for undercounting in mortality data for those aged over 5. This assumes that a population is stable and closed to migration, and that the partial birth rate has a linear relationship with the partial death rate (deaths above age x in population aged x and over) for ages over 5. Standard procedures were applied to estimate the slope of the fitted line describing the relationship between the birth and death rates.30,31 We applied this technique to all data sources for the period 2001–2004, and used the average completeness estimate (65.0% for females and 89.8% for males) to adjust the DH registration data to estimate adult mortality in 2004. Estimated completeness for each data source is summarized in Table 2.

Comparison of adjusted and registered child mortality from death registration systems suggests that the completeness of data for both sexes was about 59% for the DRT (2001), 50% for NOCR (2002–2003) and 61% for DH (2001–2004) (Table 2). Completeness of death registration data for females in all systems and all age groups was lower. Completeness of death registration for age groups above 5 years was estimated at 74% for DRT, 74% for NOCR (2002–2003) and 79% for DH (average completeness 2001–2004). Since each system had similar estimated completeness, we used the average assessment of underreporting to correct reported mortality rates.

Estimation of adult mortality

We began the assessment of the risk of adult mortality between ages 15 and 60 (45 Q 15), the key summary indicator, by assessing the trends. We did this by examining results of all studies between 1973 and 2000 and of deaths registered between 1992 and 2004 (Table 1, available at: http://www.who.int/bulletin, Fig. 3, Fig. 4). Estimates that were clear outliers were censored based on plausibility or known methodological problems with estimation techniques or data sources. A logistic curve was fitted to the remaining points and trends were estimated. Current (2004) levels were estimated based on the predicted value of the fitted line for 2004 (Fig 3, Fig. 4).

Constructing 2004 life tables

Life tables based on adjusted data from death registration systems were constructed for both sexes according to data from 2004 DH death registration in 29 provinces. Child mortality was estimated based on the methods described earlier, and all death rates at ages 5 and over were adjusted based on the correction factor from Brass’s growth balance method.

An alternative approach predicts the level of adult mortality corresponding to observed (estimated) child mortality. Our prediction model is based on a modification of Brass’s relational model life table. Brass proposed that the logit of the survival values at each age x (lx) in a given life table has a linear relationship with the logit of the survival values in a standard life table.32 As departures from the standard lead to a systematic bias in predictive ability, particularly among children and older adults, a modified model based on two additional correction factors for child and adult mortality, has been proposed.33 This enables us to predict the level of adult mortality based on the estimate of child mortality, and compare this with the observed values in the life tables based on adjusted data (as before). Using the estimated level of child mortality for 2004, the modified logit life table system was also used to estimate a life table for the Islamic Republic of Iran for 2004.33,34

Results

Child mortality estimates from previous studies

Infant mortality has decreased sharply over the past three decades: from around 154 per 1000 live births in 1964 to about 26 per 1000 (95% confidence interval, CI: 19-36) in 2004. This implies child mortality levels of 30 and 34 per 1000 live births for females and males respectively.

Adult mortality

After censoring the outlying data, various mortality data sources suggest that adult mortality has been decreasing over the past four decades, particularly for females. The predicted adult mortality risk for females in 1960 was around 0.382 (95% CI: 0.277-0.52), dropping to 0.108 (95% CI: 0.08-0.146) by 2004. The male index was 0.370 (95% CI: 0.274-0.499) in 1960, declining to 0.160 (95% CI: 0.122-0.211) in 2004. In the adjusted life tables for 2004, based on DH data, adult mortality risk was 0.124 for females and 0.175 for males (Table 3). Based on the modified logit life table system (using estimated child mortality for 2004), the predicted value of adult mortality was 0.128 (mean) for females, and 0.224 (mean) for males. The predicted levels of adult female mortality in the trend analysis and from the registered mortality data, corrected for undercounting, agree quite closely with predictions from applying model life tables based on observed values for child mortality. For males, the estimated level (0.160–0.175) is significantly lower than what is expected from model life tables (0.224) given the observed level of male child mortality (Table 3).

Estimation of life expectancy in 2004

We chose the adjusted (higher) levels of adult mortality as the best estimate, given our preference for using local evidence over models. This suggests that life expectancy at birth in 2004 was 71.2 years for females and 68.7 years for males (Table 3, Table 4(a) and Table 4(b), available at: http://www.who.int/bulletin). By comparison, the values based on the modified logit life table system suggested a larger sex differential, with implied life expectancies of 72.5 years for females and 66.8 for males (Table 3).

Discussion

Lack of documentation about methods and data sources is of major concern when studying mortality. This study presents the first comprehensive analysis of mortality levels and trends in the Islamic Republic of Iran, using different methods and sources of published and unpublished information and registration data. Previous attempts by organizations such as the UN and WHO have used substantially poorer and less complete data sets and have not been explicit about methods.

Review of infant mortality trends based on various sources (surveys, censuses, registration systems) indicates that over the past three decades this indicator has declined dramatically. A key factor in this decline may have been the expansion of the Iranian health network and access to primary health care, particularly in rural areas.35

Studies to estimate infant and child mortality using indirect demographic methods, such as the CEBCS technique, generally yield more plausible estimates than other methods. Although these methods are valuable in situations where death registration is incomplete, the results depend on the quality of the reported information. In addition, this method assumes stable fertility and child mortality during recent decades. The fertility rate in the Islamic Republic of Iran has increased dramatically over the past three decades,31,36 so caution is necessary when interpreting these estimates of child mortality.

The few studies carried out to estimate adult mortality used the retrospective method and have serious problems with underrecorded deaths. Therefore, this study used demographic methods to estimate mortality levels and construct life tables. While Iranian death registration systems undercount deaths (particularly of females), our study suggests that these data sources may be useable in combination with other sources, such as survey or census data, to estimate adult mortality levels.

Our estimates of child mortality are somewhat lower than those of the UN and WHO (36–42 per 1000 live births). Our estimates of adult mortality risk in 2004 (0.124 and 0.175 for females and males respectively) are very nearly identical to WHO’s figure (0.125) for females but somewhat lower than that for males (0.201).37 Recent data showing a decline in child mortality may have been unavailable to WHO; this could explain the difference in these estimates. More comprehensive and recent data from the MOH&ME undoubtedly have helped in reducing uncertainly regarding adult mortality levels.

Despite the NOCR’s long history of death registration, underreporting of death is common, particularly for females.6 Delayed registration by individuals is this system’s chief limitation: often there is a year or more between the date of death and its registration. The main aim of the DH registration system is to collate cause of death data from different sources. This system shows improving completeness and coverage, and the results of this study suggest sufficient completeness of data for use in mortality analysis, with appropriate corrections.

Although we have attempted a comprehensive evaluation of trends and levels of mortality, our study has several limitations. First, up-to-date population data by age and sex were not available for this study. The UN estimate of the Iranian population (2000–2005) was used although its method is not clear.28 Second, since there were few sources of data on child mortality levels, we estimated these indirectly using infant mortality data and patterns observed in the Iranian DHS of 2000. Third, as the DH death registration system is not yet operational in Tehran province (population 12 million) we have assumed that its pattern of mortality is the average of all other provinces. Fourth, it is uncertain whether the Brass growth balance method is appropriate for estimating underregistration of deaths, given its assumption that populations should be stable and closed to migration. While there is little international migration to and from the Islamic Republic of Iran, the assumption of a stable population is unlikely to hold.

In conclusion, the Islamic Republic of Iran might be considered to be at the end of the second stage of health and demographic transition. Decreasing child and infant mortality indicate declines in infectious and nutritional diseases. It is probable that the leading causes of death will shift to chronic diseases in adults. Female mortality is now lower than that of males at all ages, confirming the pattern observed in countries undergoing epidemiological transition. However, increases in male mortality from traffic accidents are a particular public health concern.

Undoubtedly, high-quality and complete national vital registration is the best data source to measure mortality. The Islamic Republic of Iran has made considerable progress towards this, but a focus on better registration of adult mortality, particularly of females, is an urgent priority for public health policy. Direct measurement of undercounting in the vital registration system would increase confidence in the data. A specific investigation to measure the completeness of data from death registration systems by conducting surveys (e.g. capture-recapture method) and subsequent improvements in infrastructure, capacity-building and resources for death registration are priorities. ■

References

Affiliations

  • School of Population Health, University of Queensland, Herston 4006, Brisbane QLD, Australia.
  • Ministry of Health and Medical Education, Tehran, Islamic Republic of Iran.
Share
Email Twitter Facebook Google Delicious LinkedIn
More...