IJE Advance Access originally published online on December 20, 2006
International Journal of Epidemiology 2007 36(2):374-384; doi:10.1093/ije/dyl257
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Epidemiologic transition interrupted: a reassessment of mortality trends in Thailand, 1980–2000
1Department of Population and International Health, Harvard School of Public Health, USA.
2Institute for Population and Social Research, Mahidol University, Thailand.
3Setting Priorities using Information on Cost-Effectiveness (SPICE) Project, Ministry of Public Health, Thailand.
4School of Population Health, University of Queensland, Australia.
* Corresponding author. School of Population Health, University of Queensland, Herston Road, Herston, Queensland, Australia 4006. E-mail: a.lopez{at}sph.uq.edu.au
 |
Abstract |
|---|
 Top Abstract IntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
Background In the late 1980s and early 1990s a generalized HIV epidemic affected Thailand which was relatively well controlled by an intensive national campaign by the mid 1990s. The extent to which the epidemic has slowed or possibly reversed the epidemiological transition in Thailand is relatively unknown.
Methods Under-five mortality rates (U5MR) were determined from various sources and weighted least squares regression conducted to determine U5MR over the years 1980–2000. Direct and indirect estimates of the completeness of death registration were used to estimate mortality levels in those aged more than 5 years for the 1980–90 and 1990–2000 periods. Life tables were constructed using the various estimates to determine changes in life-expectancy between the two time periods.
Results U5MR in Thailand is estimated to have been 58/1000 live births in 1980, declining to 30 in 1990 and to 23 in 2000. The vital registration system clearly underestimates U5MR. Successive surveys of Population Change (SPC) imply coverage of death registration improving from 75–77% in 1985–86 to 95% in 1995–96, partly due to a reliance on self-reported registration in the latter survey. In contrast, the General Growth Balance–Synthetic Extinction Generations (GGB–SEG) method suggests coverage worsening from 78–85% in 1980–90 to 64–72% in 1990–2000. Life tables based on SPC adjustments show continued declines in female, and to a lesser extent, male adult mortality with corresponding increases in life-expectancy at birth of around 6 years for both sexes from 1980–90 to 1990–2000. In contrast, the indirect adjustments suggest a substantial increase in male adult mortality with female adult mortality unchanged; life expectancy decreased by 4 years for males and was only marginally higher in females.
Conclusion Given the conflicting evidence a definitive assessment of mortality change in Thailand between 1980 and 2000 is difficult to make. Indirect adjustments, based on demographic methods point to a major reversal in mortality decline among males, and a slowing in females. If adult mortality registration has declined, and given the continued under-registration of infant and child deaths, remedial measures are urgently required if the mortality system is to better inform and monitor health development in Thailand.
Keywords Mortality, Thailand, cause of death, vital registration
Accepted 21 October 2006
|
Introduction |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
Epidemiological transition broadly describes the process of transition from a high mortality, infectious disease-dominated mortality regime to a low mortality, degenerative disease dominated regime.1 Populations of Europe and of European origin passed through this transition in the first half of the 20th century. The majority of Latin American and East and Southeast Asian populations are in an advanced stage of transition, whereas much of Africa and South Asia are in earlier stages.
Such transitions have tended to be thought of as irreversible. This perception is partly the result of descriptions of fertility changes in the 20th century, which have so far generally been irreversible, though sometimes incorporating pauses or temporary plateaus such as the 1950s baby boom in developed countries. Concerns about the adverse mortality effects of economic restructuring programmes in the 1980s, and the simultaneous emergence of HIV/AIDS, have led to the realization that the epidemiological transition might be halted. This has been well-documented in Eastern Europe over the past two decades, primarily as a result of increases in chronic disease and injury mortality,2,3 but also in Africa due to HIV/AIDS.3–5 In extreme cases this may lead to the balance of disease shifting back from degenerative to infectious. Most national populations severely affected by the HIV epidemic are in sub-Saharan Africa but basic data in these countries (with the possible exception of South Africa) are inadequate to monitor the situation. One country where it can be well-studied is Thailand, where basic demographic data have generally been judged to be of reasonable quality,6–8 and an HIV epidemic (though not of the magnitude of those affecting many countries of sub-Saharan Africa) affected the country in the 1980s and 1990s. Cause of death data from the Thai civil registration system, however, remain of questionable quality with a large proportion of deaths coded to ill-defined causes (28–36% over the period 1996–2000 with no discernable trend over time; author's calculations), and even those deaths with a defined cause may not always be correctly coded.9
The objectives of this article are to review mortality statistics available for Thailand from 1980 to 2000 to evaluate, and as necessary adjust the data for shortcomings, and to use life tables for the periods 1980–90 and 1990–2000 in combination with patterns of deaths by age, sex and calendar year to describe the mortality reversal in Thailand in the 1990s. Understanding age- and sex-specific mortality rates can help in understanding the epidemiological transition as much as studying incomplete and uncertain cause of death data.
|
Data sources |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
Sources of data needed for monitoring mortality conditions at the national level are accurate records of deaths by age, sex and cause (normally from a civil registration system) and periodic censuses to provide estimates of exposure time in the form of person-years lived in each age–sex-time period. In countries where these sources are inadequate, alternative data collection strategies such as periodic sample surveys are often implemented to supplement conventional data.
Civil registration
Thailand has a long history of recording vital events, but surveys indicate that not all deaths are recorded, especially for young children. Some of the problems seem to be with the system itself. For example, there was a sharp decline for both male (13%) and female (11%) deaths from 1996 to 1997, followed by a sharp increase (males: 18%, females: 16%) from 1998 to 1999. Such sudden changes in mortality are implausible and most likely result from a breakdown in the processing of reported deaths; indeed, the process for death reporting was changed in 1996.
Population censuses
Recent Thai population censuses were carried out in 1980, 1990 and 2000. The 1990 and 2000 censuses collected information both on the number of children ever born and on surviving children of women of reproductive age (15–49 years), and these data provide a basis for the indirect estimation of under-five mortality levels and trends.10
Periodic sample surveys
The National Statistical Office conducts periodic intercensal surveys [Surveys of Population Change, (SPC)], most recently during 1985–86 and 1995–96 (the 2005–06 survey was still in the field at the time of writing). The 1985–86 survey followed the ‘dual record’ methodology, whereby vital events were independently identified, both by 3-monthly household surveys conducted over a period of 15 months and by the regular civil registration system in each sample area. Events from the two sources were then matched, to identify those recorded by both and those recorded by only one system. Assuming that omission from one source is independent of the probability of omission from the other, the completeness of registration can be estimated—75–77% for deaths in the 1985–86 SPC.11 The 1995–96 SPC used only the 3-monthly surveys, and estimated coverage of civil registration (95%) through a survey question as to whether each death recorded by the survey had been reported to the registrars. Both SPCs collected information on the number of children ever born and on surviving children of women of reproductive age (15–49 years).
Other sources
The 1987 Thailand Demographic and Health Survey (DHS) collected full birth histories, from which both levels and age patterns of child mortality can be estimated for the 15 years prior to the survey.10
|
Methods |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
Child mortality measurement
A variety of data exist for arriving at estimates of infant and child mortality. The civil registration system provides an annual series of estimates from 1975 to 2004 of the infant mortality rate (IMR), the ratio of deaths under age 1 in a year to the births in that year, and of the under-5 mortality rate (U5MR), i.e. the probability of dying at the exact age of 5 years. Data on birth histories from the 1987 DHS provide retrospective, direct estimates of IMR, 4q1 (the probability of dying between the ages of 1 and 5 years) and U5MR for 5-year periods prior to the survey using standard life table methods. Data from the 1985–86 and 1995–96 SPC provide prospective life table estimates of IMR, 4q1 and U5MR for the survey year. The 1990 and 2000 census data and SPC data on children ever born and on surviving children of women of reproductive age provide indirect estimates of probabilities of dying by exact ages of childhood for specific reference dates,10 assuming that the ‘West’ pattern of mortality from the Coale–Demeny model life tables applies.12 (Hill13 found that the direct estimates of IMR and 4q1 from the birth histories from the 1980 Survey of Fertility in Thailand and the 1987 DHS both followed the ‘West’ pattern closely.) To determine the level of child mortality in Thailand over time we use weighted least squares regression, the weights used for the various sources of child mortality are shown in Table 1 and represent judgements about the relative value of each source for estimating mortality.
|
Adult mortality measurement
Information about adult mortality is more limited than for child mortality. The two types of data available are registered deaths by age and sex from the civil registration system and the estimates of age–sex-specific mortality rates from the two SPCs. A summary indicator of registered (uncorrected) adult mortality, namely, the probability of dying between the ages of 15 and 59 years (45q15) is shown in Figure 1, for the period 1980–2004. This clearly shows the major problem that affected the registration system in the years 1997 and 1998 as well as the impact of the HIV epidemic on male mortality, although the actual levels remain uncertain.
|
The dramatic impact of the HIV epidemic on male mortality is further emphasized in Figure 2, which shows ratios of mortality risk in each year from 1990 to 2002 to the average risk for 1985–89 for selected age groups. Whereas the risk ratios for males [Figure 2a] aged 5–14 years are largely <1.0, the ratios for males aged 15–24 years, 25–34 years, (especially) and 35–44 years rise steeply from 1992 until the late 1990s, to peak at about two to two and a half times their values in the late 1980s. For females [Figure 2b], the effect is somewhat delayed, to about 1995, and is evident for the age groups 15–24 and 25–34 years with some indication of rising risks late in the period among 35–44-year-olds.
|
While the registered mortality from Figures 1 and 2 demonstrates the impact of the HIV epidemic on adult mortality, it may well underestimate or overestimate the extent of the epidemic due to differential levels of coverage of mortality over time. Unbiased estimates of age-specific mortality rates require that events (deaths) and exposure time (population at risk) be consistently recorded. A number of demographic methods have been proposed to evaluate (and under certain assumptions, to adjust for) the consistency of such coverage.15–18 These methods compare the age pattern of deaths with the age pattern of the population, a relationship determined by demographic regularities. The general growth balance (GGB)17 and synthetic extinct generations (SEG)15 methods are both appropriate for application to closed, non-stable populations. Hill19 uses simulations to show that the GGB methods are more sensitive to possible errors in age reporting than the SEG methods, whereas the SEG methods are much more sensitive to any changes in census coverage. He suggests a two-stage process, by which the GGB method is applied first to estimate any change in census coverage, and the SEG method is then applied to data after adjusting the census numbers for possible coverage change.19 We use the GGB and the SEG, as well as the two-stage GGB–SEG combined procedure to assess the completeness of the mortality registration system. All three methods require census data for the start and end of the period being studied. Since we do not have such information after the 2000 census, we cannot assess coverage of registered deaths beyond 2000.
|
Results |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
Child Mortality
Table 2 and Figure 3 summarize the available estimates of child mortality. The second and third columns of the table show the IMR and U5MR derived from civil registration by year. The registered IMR declined steadily through the 1980s, falling from about 13/1000 live births around 1980 to around 8 by 1990, an annual rate of decline of almost 5%, and continued to decline to around 6/1000 by 2000, an annual rate of decline of <3% (reaching a very low level of 4/1000 in 1997). Although the trends in the IMR look plausible over the entire period, the levels are suspiciously low, lower than the corresponding rates in the United States for example. The trend in the recorded U5MR is rather different, and the levels somewhat higher: a rapid decline from 1980 to 1990, of over 7% per annum, but no clear trend in the 1990s, stabilizing at around 10/1000. The large differences between the IMR and the U5MR are most plausibly explained by a large under-registration of infant deaths relative to those occurring between the ages of 1 and 5.
|
|
The indirect U5MR estimates from the 1990 and 2000 censuses are shown in columns 3 and 4 of Table 2. The points reinforce the trend from civil registration of gradual decline in the 1980s followed by a period of little change in the 1990s. Figure 2 shows that the indirect estimates of U5MR from the 1990 census for the late 1970s and early 1980s are well below other estimates, except those from vital registration, for overlapping time periods. The indirect estimates from the 2000 census are some 15/1000 higher than those from the 1990 census for the second half of the 1980s, and substantially higher than the registration estimates.
The DHS direct estimates for the mid-1980s are more or less in line with the 2000 census indirect estimates. The birth history-based U5MR estimates from the 1987 DHS indicate substantially higher levels than vital registration (53/1000 for the period 1978–82 and 45/1000 for the period 1983–87) although they also suggest a downward trend in the 1970s and early 1980s.
The SPC of 1985–86 and 1995–96 provide both direct point estimates of U5MR and indirect estimates covering the decade or so prior to each survey. The direct and indirect estimates from the SPCs are largely independent of one another. The 1985–86 SPC gives a direct U5MR estimate of 47 for the period 1985–86, with indirect estimates that track the DHS indirect estimates. The 1995–96 SPC gives a direct U5MR estimate of 33 for the period 1995–96, with indirect estimates that agree with the 2000 census estimates.
How can sense be made of all these different estimates? It is clear that the level, if not the trend, of U5MR from vital registration, and the indirect estimates from the 1990 census, is too low. Direct estimates from DHS and indirect and direct estimates from the 1985–86 SPC are consistent with each other and with the indirect estimates from the 2000 census. For the 1990s, the 1995–96 SPC gives a direct estimate of 33/1000, while the indirect estimates from the 1995–96 SPC and the 2000 census are between 15 and 20/1000 in the 1990s, well below the SPC 1995–96 direct estimate. Using a weighted regression model (Table 1) with a spline at 1990 to account for a different rate of change in the 1990s, and excluding estimates from the vital registration and 1990 census, U5MR is estimated to have been 58/1000 in 1980, declining to 30 in 1990 and to 23 in 2000 (Figure 3).
Adult mortality
Plots of entry minus growth rates against death rates for population segments aged 
5 and 80 (the GGB method) are shown in Figure 4 for the 1980–90 intercensal period and Figure 5 for the 1990–2000 intercensal period. Summary results from the application of the method are shown in Table 3.
|
|
|
The GGB points fit straight lines quite closely, indicating that the assumptions underlying the methodology are not seriously violated (Figures 4 and 5). The straight line fit is particularly good for the 1990–2000 intercensal period; the 1980–90 relationships have clear distortions at the points for lower death rates (open-ended age segments starting at younger ages), possibly reflecting some net international migration at these ages.
Table 4 summarizes the estimated coverage of vital registration from the indirect GGB, SEG and GGB–SEG combined applications as well as the direct mid-decade estimates from the two SPC surveys. There is substantial difference between the 1985–86 SPC estimates of coverage (75–77%) and the indirect GGB estimates (85–91%). The SEG estimates of completeness for the 1980s as a whole (106–115%) are clearly implausible. This is largely due to differences in what is being measured. The SPC ‘dual record’ methodology aims to estimate absolute coverage, whereas the GGB and SEG methodology estimate coverage of deaths relative to census coverage. The GGB intercepts in Table 3 indicate that the 1980 census was less complete than the 1990 census (accounting for the high SEG estimate of coverage for the 1980s), which in turn was slightly less complete than the 2000 census. The GGB and SEG methods are therefore measuring completeness of death recording relative to incomplete censuses, and thus give higher estimates than the 1985–86 absolute estimates. The combined GGB–SEG method produces estimates more in line with the direct 1985–86 SPC.
|
The SPCs suggest coverage of death registration improved to 95% by 1995–96. In contrast, the indirect methods show a substantial worsening of death registration coverage with estimates ranging from 64% to 72% for the 1990s. There are a number of reasons for this. The 1995–96 SPC is likely to have over-estimated coverage as it relied on self-reported death registration rather than using a dual estimation methodology to estimate coverage as used in the earlier 1985–86 survey.20 The 1995–96 SPC would also not have captured the major system problems evident in 1997–98 (Figures 1 and 2). Importantly, even if the time period for the 1995–96 SPC had included the 1997–98 system problems it is likely that these would not have been picked up due to the reliance on self-report, as these system problems are more likely to have been due to data collation issues rather than individuals not registering deaths.
Life tables and epidemiological transition in Thailand, 1980–2000
Based on both the registered, uncorrected data with subsequent corrections using the coverage estimates from both the direct SPCs and indirect techniques using the GGB and combined GGB–SEG methods (Table 4), we have constructed life tables for the two intercensal periods 1980–90 and 1990–2000. U5MR estimates were derived from the registered (uncorrected) data, direct mid-decade SPC estimates and predicted U5MR (Table 2 and Figure 3). Table 5 summarizes the key mortality indicators.
|
Registered, uncorrected adult mortality (45q15) shows a 4% increase for males from the 1980s to the 1990s, and a 15% decrease for females. It is important to note that this, and all the other estimates in Table 5, are decade averages: the true impact of the HIV epidemic on Thai males was substantially greater given that there was a large decline in 45q15 during the 1980s, followed by a rise and fall in the 1990s (Figures 1 and 2). Despite the increase in male adult mortality risk, male expectation of life at birth based on registered, uncorrected data increased by 1.4 years from the 1980s to 1990s; female expectation of life at birth increased by 2.6 years [Table 5; Panel (a)]. The increases in life expectancy are partly the result of declining mortality in childhood, and partly an artefact of the system problems that affected death registration at all ages during the 1997–98 period (Figures 1 and 2).
Adjusting the 1980s and 1990s decennial registered deaths by the direct SPC estimates of death registration coverage would imply a 15% decrease in adult mortality for males from the 1980s to the 1990s, and a 30% decrease for females; male expectation of life at birth was 6 years higher for males and 7 years higher for females in the 1990s than in the 1980s [Table 5; Panel (b)]. Clearly this adjustment approach is problematic, as the 1995–96 SPC survey estimates of vital registration coverage would not have captured the system problems in 1997–98 (Figures 1 and 2) and would underestimate mortality for the 1990s. These estimates, however, are similar to the official SPC estimates of life expectancy at birth for the mid-decade years for males of 63.8 years for 1985–86 and 69.9 years for 1995–96, although quite different to the SPC survey estimates for females of 68.9 years for 1985–86 and 74.9 years for 1995–96.21
The adjustments using the indirect methods suggest that male 45q15 increased by 33–35% from the 1980s to the 1990s, while the female value remains essentially unchanged; male expectation of life at birth was around 4 years lower in the 1990s than in the 1980s; female expectation of life at birth increased by only 0.1–0.2 years [Table 5; Panels (c) and (d)]. This is despite significant declines in U5MR over the same time period.
|
Discussion and implications for the Thai health system |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
These conflicting estimates hinder a definitive assessment of mortality change in Thailand between 1980 and 2000. While the unadjusted data indicate that there was some reversal of the epidemiological transition in Thai males and a slowing in Thai females, the SPC surveys suggest little or no impact with continuing and substantial improvements in both male and female life-expectancy. The male life expectancy increases suggested by the SPC surveys (6.1 years from the 1985–86 to 1995–96 survey), however, are unlikely given that a generalized HIV epidemic affected the male population over this period. The change in the method of determining completeness of death registration in the 1995–96 survey has clearly obscured a true assessment of coverage over time. If, on the other hand, the indirect adjustments are true, the HIV epidemic has had a devastating effect with a major reversal in males, and a slowing in females, of the epidemiological transition in Thailand.
Our analysis suggests that child mortality decline appears to have slowed but not stopped in the 1990s, while the mortality of young adults, first males and then females increased by 200–300%, implying a dramatic interruption of epidemiological transition in the 1990s. Without any data adjustment, the HIV epidemic increased mortality rates of Thais in their 20s and 30s by two to three times. Using the indirect method to adjust for undercount of deaths, these mortality rates and their impact on life-expectancy are dramatically greater. If these indirect adjustments are correct, these represent very substantial increases in adult mortality over a very short period and the magnitude of this increase may not be adequately reflected in the life tables for Thailand published by international agencies such as the United Nations and the World Bank.
These observations are consistent with other studies of the Thai HIV epidemic which suggest a rapid increase in new HIV infections largely due to commercial sex, peaking in the late 1980s, and early 1990s, which was relatively well-controlled with an aggressive national campaign by the mid-1990s.22–26 While we have emphasized the role of the HIV epidemic on young adult mortality in the 1990s, mortality increases at these ages are also likely to have been partly driven by a rise in road traffic deaths, due to increased motorization that accompanied economic development over this period,27 and inadequate prevention measures.
Thailand is a good example of a country where health policy is responsive, flexible and effective. This is very well-illustrated by the HIV/AIDS epidemic, where preventive measures have limited the spread of the epidemic.22–26 These policy responses have led to a generalized decline in young adult mortality from HIV since the late 1990s, and this is expected to continue. In part, the urgency of policy responses to control the HIV epidemic in Thailand was highlighted by mortality trends evident from the vital registration system. Reliable, relevant and timely data on the level and cause pattern of mortality are crucial to guide policy responses to protect and promote population health. Careful application and interpretation of demographic principles and methods permits correction of registered mortality data for undercount of deaths, and thus greatly increases their utility for public health purposes. The mid-decade SPCs proved particularly useful in our analysis for determining levels of infant and child mortality; the evidence above suggests, however, that the SPC should return to its ‘dual record’ system of evaluating coverage of deaths, used in the 1980s, rather than the simpler assessment of vital registration coverage based on household reports of registration used in the 1990s. This is being implemented in the current SPC for 2005–06.
While these methodological ‘fixes’ can help to provide a better understanding, conflicting estimates of vital registration coverage in the 1990s preclude a more definitive assessment of mortality changes and the actual impact of the HIV epidemic in Thailand during this period. If there has been a real decline in adult mortality registration, as suggested by the indirect methods, then this is of major significance for the Thai health system and the reasons for this need to be carefully understood and addressed. Better registration of infant and child deaths is also clearly needed.
In addition to improving coverage of mortality registration, the system for assigning cause of death also requires further development if public policy is to more appropriately respond to the emerging public health challenges, particularly from chronic diseases and injuries.27–31 With the significant gains in reducing child mortality in Thailand over the past few decades, public health efforts are increasingly being orientated towards controlling diseases such as ischaemic heart disease, stroke, cancer and chronic lung disease, which largely affect adults and which, if unchecked, can lead to reversals in mortality decline similar to what has been observed from HIV/AIDS.2–4 A key component of the evidence base to inform policy responses to these public health challenges is a reliable, timely information system on the causes of all deaths. In parallel with increased efforts to improve the coverage of mortality registration, research is urgently required to evaluate the quality of cause of death data and to suggest measures to strengthen systems for correctly diagnosing deaths, and thus more appropriately informing policy action to hasten the epidemiological transition in Thailand.
Contributors: K.H. carried out the adult mortality analysis and drafted the initial article. P.V. carried out the child mortality analysis, compiled the data and contributed to draft revisions. P.P. provided technical advice on the analysis, draft revisions and interpretation of results. Y.P. assisted with data compilation, draft revisions and coordination of the research. S.S.L. contributed to the analysis, detailed revisions of the article and coordination of the research. A.D.L. conceived the study, contributed to the analysis, and detailed revisions of the article. All authors approved the final product.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
We thank Jesse Klafter for research assistance with the revision of the GGB, SEG and combined GGB/SEG estimates, and Mollie Hogan with research assistance with referencing. This research is conducted under the Setting Priorities using Information on Cost-Effectiveness (SPICE) project and is supported by an international collaborative research grant from the Wellcome Trust, United Kingdom (071842/Z/03/Z) and the National Health and Medical Research Council of Australia (301199).
Conflict of interest: None declared.
KEY MESSAGES
|
|
References |
|---|
|
TopAbstractIntroductionData sourcesMethodsResultsDiscussion and implications for...AcknowledgementsReferences |
|---|
1 Omran AR. The epidemiologic transition: a theory of the epidemiology of population change. Milbank Memorial Fund Quarterly (1971) 49:509–38.[CrossRef][Web of Science][Medline]
2 Leon D, Chenet L, Shkolnikov V, et al. Huge variation in Russian mortality 1984–94: artefact, alcohol, or what? Lancet (1997) 350:383–88.[CrossRef][Web of Science][Medline]
3 Moser K, Shkolnikov V, Leon D. World mortality 1950–2000: divergence replaces convergence from the late 1980s. Bulletin of the World Health Organization (2005) 83:202–09.[Web of Science][Medline]
4 McMichael A, McKee M, Shkolnikov V, Valkonen T. Mortality trends and setbacks: global convergence or divergence? Lancet (2004) 363:1155–59.[CrossRef][Web of Science][Medline]
5 Hosegood V, Vanneste A-M, Timæus I. Levels and causes of adult mortality in rural South Africa: the impact of AIDS. AIDS (2004) 18:663–71.[CrossRef][Web of Science][Medline]
6 Suwee W, Santipaporn S. Utilization of the 2000 Population and Housing Census of Thailand. (2002) 20th Population Census Conference Ulaanbaatar: Mongolia.
7 Tangcharoensathien V, Faramnuayphol P, Teokul W, Bundhamcharoen K, Wibulpholprasert S. A critical assessment of mortality statistics in Thailand: potential for improvements. Bulletin of the World Health Organization (2006) 84:233–38.[CrossRef][Web of Science][Medline]
8 National Research Council. Fertility and Mortality Changes in Thailand, 1950–1975 (1980) Washington DC: National Academy of Sciences.
9 Choprapawon C, Porapakkham Y, Sablon MA, Panjajaru R, Jhantharatat B. Thailand's national death registration reform: verifying the causes of death between July 1997 and December 1999. Asia Pacific Journal of Public Health (2005) 17:110–16.
10 United Nations. Manual X: Indirect techniques for demographic estimation (1983) New York: United Nations.
11 Chandrasekar C, Deming ED. On a method of estimating birth and death rates and the extent of registration. Journal of the American Statistical Association (1949) 44:101–15.[CrossRef][Web of Science]
12 Coale AJ, Demeny P. Regional Model Life Tables and Stable Populations (1983) 2nd edn. Princeton, New Jersey: Princeton University Press.
13 Hill K. Age patterns of child mortality in the developing world. Population Bulletin of The United Nations (1995) 39:112–32.[Medline]
14 Hill K, Pande R, Jones G. Trends in Child Mortality in the Developing World 1960–1990 (1997) New York: UNICEF.
15 Bennett NG, Horiuchi S. Estimating the completeness of death registration in a closed population. Population Index (1981) 47:207–21.[CrossRef][Web of Science]
16 Brass W. Methods for Estimating Fertility and Mortality from Limited and Defective Data (1975) Chapel Hill: University of North Carolina Press.
17 Hill K. Estimating census and death registration completeness. Asian & Pacific Population Forum (1987) 1:8–13.
18 United Nations. Methods for Estimating Adult Mortality (2002) New York, NY: United Nations.
19 Hill K. Methods for Measuring Adult Mortality in Developing Countries: a Comparative Review. (2001) International Population Conference, 2001: Salvador, Brazil.
20 Prasartkul P, Vapattanawong P. The completeness of death registration in Thailand: Evidence from Demographic Surveillance System of the Kanchanaburi Project. (2006) World Health and Population.
21 National Statistical Office Thailand. Expectation of life at birth from the survey of population change [cited August 2006]. Available from: http://web.nso.go.th/eng/stat/popchang/popchgt5.htm.
22 Ruxrungtham K, Brown T, Phanuphak P. HIV/AIDS in Asia. Lancet (2004) 364:69–82.[CrossRef][Web of Science][Medline]
23 UNAIDS. Relationship of HIV and STF declines in Thailand to behavior change - A synthesis of existing studies: UNAIDS Key Materials Publication. (1998).
24 Hanenberg R, Rojanapithayakorn W, Kunasol P, Sokal D. Impact of Thailand's HIV-control programme as indicated by the decline of sexually transmitted diseases. Lancet (1994) 344:243–45.[CrossRef][Web of Science][Medline]
25 Hanenberg R, Rojanapithayakorn W. Changes in prostitution and the AIDS epidemic in Thailand. AIDS Care (1998) 10:69–79.[Web of Science][Medline]
26 The Thai Working Group on HIV/AIDS Projection. Projections for HIV/AIDS in Thailand: 2000–2020 (2001) Bangkok: Karnsana Pringtin Press.
27 Suriyawongpaisal P, Kanchanasut S. Road traffic injuries in Thailand: trends, selected underlying determinants and status of intervention. Injury Control and Safety Promotion (2003) 10:95–104.[CrossRef][Medline]
28 Sritara P, Cheepudomwit S, Chapman N, et al. Twelve-year changes in vascular risk factors and their associations with mortality in a cohort of 3499 Thais: the Electricity Generating Authority of Thailand Study. Int J Epidemiol (2003) 32:461–68.
29 Aekplakorn W, Stolk R, Neal B, et al. The Prevalence and Management of Diabetes in Thai Adults: The International Collaborative Study of Cardiovascular Disease in Asia. Diabetes Care (2003) 26:2758–63.
30 The InterASIA Collaborative Group. Cardiovascular risk factor levels in urban and rural Thailand - The International Collaborative Study of Cardiovascular Disease in Asia (InterASIA). European Journal of Cardiovascular Prevention & Rehabilitation (2003) 10:249–57.[CrossRef][Web of Science]
31 Lapham S, Skipper B, Brown P, Chadbunchachai W, Suriyawongpaisal P, Paisarnsilp S. Prevalence of alcohol problems among emergency room patients in Thailand. Addiction (1998) 93:1231–39.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Ebrahim Prostate cancer, Cochrane, Durkheim and Beer Lao Int. J. Epidemiol., April 1, 2007; 36(2): 263 - 264. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||