Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/wiley/measuring-progress-towards-achieving-millennium-development-goals-in-VYQIxORZG1
References (35)
-
K. Carter, M. Cornelius, Richard Taylor, Shareen Ali, C. Rao, Alan Lopez, Vasemaca Lewai, Ramneek Goundar, C. Mowry (2011)
Mortality trends in FijiAustralian and New Zealand Journal of Public Health, 35
-
Achieving the Millennium Development Goals in the Pacific Islands -
World Health Organization. Under-five Mortality Rate -
World Health Organization. Infant Mortality Rate. In: Indicator and Measurement Registry Version -
(2002)
Probabliity. In: Statistical Methods in Medical Research -
(2011)
Tuvalu National Population Policy 2010–2015 -
S. Hufanga, K. Carter, C. Rao, Alan Lopez, Richard Taylor (2012)
Mortality trends in Tonga: an assessment based on a synthesis of local dataPopulation Health Metrics, 10
-
(2002)
United Nations Common Country Assessment Tuvalu: Final Draft. Suva (FJI): United Nations -
(2011)
Indicator and Measurement Registry Version 1.7.0 -
(2005)
Tuvalu 2002 Population and Housing Census, Volume 2: Demographic Profile, 1991–2002 -
(2007)
Bank and UN Population Division. Levels and Trends of Child Mortality in 2006: Estimates Developed by the Inter-agency Group for Child and Mortality Estimation -
(2010)
Department of Planning and Budget, Ministry of Finance and Economic Development -
(2009)
Tuvalu Demographic and Health Survey 2007: Final Report -
(2011)
Regional Office for the Pacific. WHO Western Pacific Country Health Information Profiles -
K. Carter, C. Rao, Alan Lopez, Richard Taylor (2012)
Mortality and cause-of-death reporting and analysis systems in seven pacific island countriesBMC Public Health, 12
-
Vilimaina Rakaseta, Demography Programme (1996)
Tuvalu population profile : a guide for planners and policy-makers -
(2011)
Geneva (CHE): WHO -
(2008)
Policies and Strategies in Population and Reproductive Health -Suva -
(2009)
2008 Joint Annual Report -
(2012)
Tuvalu 2007 Demographic and Health Survey Facts and Figures at your Fingertips, Infant and Child Mortality [Internet] -
(1991)
Noumea (FJI): Secretariat of the Pacific Community -
(2007)
Demographic and Health Survey Facts and Figures at your Fingertips, Infant and Child Mortality -
An Assessment of the Credibility of Child Mortality Declines Estimated from DHS Mortality Rates -Working Draft -
(2002)
United Nations Common Country Assessment Tuvalu: Final Draft -
P. Coombs (1966)
The International Institute for Educational PlanningInternational Review of Education, 12
-
V. Preedy, R. Watson (2010)
Children Ever Born and Children Surviving -
(1983)
Manual X: Indirect Techniques for Demographic Estimation -
(2012)
Child Mortality Estimates – Tuvalu [Internet] -
(2011)
National MDG Taskforce and UNDP Multi‐Country Office: Tuvalu Millennium Development Goals Progress Report 2010/2011 -
(2008)
Funafuti (YUV): Government of Tuvalu -
Data on reported births and deaths aggregated by age groups. Received through coauthor (SH) in 2011 -
(2010)
Millennium Development Goals Project Officer with Assistance from the Department of Planning and Budget. National MDG Taskforce and UNDP Multi-Country Office -
(1992)
Tuvalu 1991 Population and Housing Census, Volume 2: Analytical Report -
(2011)
WHO Western Pacific Country Health Information Profiles, 2011 Revision -
(2013)
Child Info Monitoring the Situation of Children and Women. Statistics by Area/Child Survival and Health [Internet]
- Publisher
- Wiley
- Copyright
- © 2014 Public Health Association of Australia
- ISSN
- 1326-0200
- eISSN
- 1753-6405
- DOI
- 10.1111/1753-6405.12169
- pmid
- 24750434
- Publisher site
- See Article on Publisher Site
Abstract
probability of dying between birth and age Objectives: Infant mortality rates (IMR) and under-five mortality rates (U5MR) in Tuvalu (2010 I<1year per 1,000 live births, and under- five mortality rate (U5MR) the probability population 11,149) for 1990–2011 were evaluated to determine best estimates of levels and of dying between birth and age <5 years trends. per 1,000 live births. The World Health Methods: Estimates were graphed over time to identify trends/inconsistencies, and censored 1,2 Organization (WHO) states that these for reliability/plausibility. Where possible, 95% confidence intervals (CIs) and tests for linear rates are often used to identify vulnerable trend were calculated. populations as they reflect the social, Results: Ministry of Health (MoH) data indicates IMR and U5MR (per 1,000 live births) declined economic and environmental conditions in over 1990–2008: IMR 62 (95%CI 46-81) for 1991–93 (51 deaths) to 19 (95%CI 10-33) for 2006–08 which children live. The value assigned to (12 deaths); U5MR 67 (95%CI 50-87) for 1991–93 (55 deaths) to 19 (95%CI 10-33) for 2006–08 IMR and U5MR in evaluating child health is (12 deaths). The 2007 Demographic and Health Survey (DHS) suggests recent trends are indicated by their inclusion as Millennium increasing: IMR 24 for 1998–2002 to 31 (95%CI 20-42) for 2003–07; U5MR 29 for 1998–2002 to 1,2 Development Goal (MDG) indicators. The 36 (95%CI 30-43) for 2003–07 (deaths not provided). Tests for linear trend and 95%CIs indicate importance of accurate and appropriate MoH declines are statistically significant, but recent increased estimates from DHS are not, and reporting of these mortality estimates, could be affected by recall bias. and their trends over time, are essential in Conclusions: Small populations provide challenges in interpretation of IMR/U5MR trends. To monitoring child health status, and have ensure the correct interpretation of rates, CIs (95%) and tests for trend should be calculated. implications for policy and resource allocation 3,4 Tuvalu has experienced steady decline in IMR/U5MR over the past 20 years. decisions. This article demonstrates to policy-makers, especially those working Key words: child mortality, infant mortality, Millennium Development Goals, Tuvalu in small populations, that calculation and application of statistical confidence intervals (95%CIs) to mortality estimates facilitates islands staffed by trained nurses. Mortality Where possible, 95%CIs are calculated accurate determination of current status and estimates for Tuvalu, including IMR and U5MR, to assist with evaluation of differences in trends in child health and formulation of estimates and trends of IMR and U5MR. are published by government sources and a evidence-based policy. range of international agencies. This paper identifies and evaluates available data on Tuvalu is a Polynesian island nation, formerly Methods IMR and U5MR in Tuvalu for 1990–2008 to known as the Ellice Islands, in the South 5 Estimates of Tuvalu IMR and U5MR for determine best estimates of levels and trends. Pacific Ocean. It is the fourth smallest country 1990-2008 were obtained from the following This information is relevant to evaluation of in the world, covering a total land area of 26 5 sources: square kilometres across nine coral islands. Tuvalu’s progress towards achieving MDG The population was estimated to have 4 of reducing child mortality by two thirds • Calculations performed by the present reached 11,149 in 2010, making Tuvalu the (between 1990 and 2015) from an IMR of 57.3 study based on previously unpublished smallest member of the United Nations (UN) deaths per 1,000 live births in 1992 to 19.1 by data from Tuvalu Ministry of Health (MoH) by population. Health service delivery occurs 2015; and U5MR from 68.7 deaths per 1,000 Health Information System of annual births through one hospital located on the main live births in 1991 to 22.9 by 2015; according and deaths aggregated by age group for 6 7 island of Funafuti and clinics on the outer to the 2011 Tuvalu MDG target score card. 1991-2008. 1. School of Public Health and Community Medicine, University of New South Wales 2. Ministry of Health, Tuvalu Correspondence to: Professor Richard Taylor, School of Public Health and Community Medicine (SPHCM), Faculty of Medicine, University of New South Wales, Samuels Bldg, Level 2/Rm223, Botany Street, Gate 11, Randwick, NSW 2031; e-mail: r.taylor@unsw.edu.au Submitted: March 2013; Revision requested: July 2013; Accepted: October 2013 The authors have stated they have no conflict of interest. Aust NZ J Public Health. 2014; 38:390-4; doi: 10.1111/1753-6405.12169 390 Australian and New Zealand Journal of Public Health 2014 vol . 38 no . 4 © 2014 Public Health Association of Australia Epidemiology Measuring Millennium Development Goals progress in small populations • Estimates published in reports by However, deaths may be occasionally missed were censored from further analysis, and government departments, WHO, UN or inadvertently double registered, which the remaining three graphed with 95%CIs agencies, the World Bank, Secretariat would inordinately affect rates based on small where possible (Figure 3). Estimates based of the Pacific Community (SPC) and the numbers. on previously unpublished MoH data of European Commission, based on a variety births and deaths <1 year indicate that For estimates of IMR and U5MR based on of methods, obtained through internet and the IMR trend in Tuvalu is decreasing from MoH recorded births and deaths, the number 6,8-12 other searches. 61.7 (95%CI 45.9-81.1) deaths per 1,000 live of deaths <1 year and < 5 years in each births for 1991–1993 (51 deaths) to 18.9 Estimates of IMR and U5MR were graphed triennial period was divided by the number of (95%CI 9.8-33.0) for 2006–2008 (12 deaths), over time to identify trends and inconsistency live birth in the same periods, and the Poisson which is statistically significantly different between sources (Figure 1 and 2). Neonatal method was employed to calculate 95%CIs. (Figure 3, Table 1). The linear trend of IMR and post-neonatal mortality from the 2007 The Poisson method is used for proportions using triennial MoH data for 1991–2008 was Demographic and Health Survey (DHS) are characterised by small numerators; 95%CIs highly statistically significant χ = 18.6 (d.f. 1); also included. Estimates were assessed for for counts are derived from a Poisson table p<0.001, and the R for a linear fit was 0.93. reliability and plausibility based on the source which are then translated into 95%CIs for the 8,9 Estimates contained in Census reports of data and method of estimation. Sources rate. Statistical test for linear trend was carried indicate an IMR of 41 deaths per 1,000 live were censored from further analysis if they out on the MoH triennial IMR and U5MR from 6,11 births for 1990–91 (CEBCS), 51 for 1992–97 were obtained from a secondary source deaths and births over 1991–2008 using a and 35 for 1997–2002 (vital registration (VR) or included implausible figures (such as chi-square and its p-value. The proportion 2 data, deaths not provided) however, the an U5MR lower that the IMR for the same of variance explained (R ) for a linear fit was 12 absence of data on standard errors in Census period). also calculated where one 1.00 is a perfect reports means 95%CIs are unable to be linear correspondence. For estimates based Three sources of estimates for 1990–2008 calculated and this impedes interpretation on retrospective maternal history from the remained following censorship: (1) MoH of these data. Statistical CIs (95%) of IMR 2007 DHS, the 95%CIs for 1997–2007 were estimates based on previously unpublished from MoH data overlap the 2007 DHS point 7 calculated using the normal approximation of population vital registration data; (2) 2007 estimates for comparable periods. Estimates the binomial based on standard errors of the DHS estimates based on household survey contained in the 2007 DHS report suggest 10 rates, provided on request from SPC. data (retrospective maternal history); and (3) a decline in IMR from 37 deaths per 1,000 live Tuvalu Census estimates based on children births for 1993–97 to 24 for 1998–2002, then ever born and children surviving (CEBCS) Results an increase to 31 (95%CI 20-42) for 2003–07 method for 1991 Census, and population (deaths not provided), however, the point Infant mortality rate vital registration for the 2002 Census. Case estimate for the 1998–2002 is within the ascertainment of births and deaths through Extensive variation was found in uncensored 95%CI of the 2003-07 period indicating no the Health Department is usually of high estimates of Tuvalu IMR for 1990–2008 (Figure statistically significant difference (Figure 3). standard in small countries, especially in 1). From six sources of IMR estimates, three children, where these events are infrequent. Figure 1: Tuvalu infant mortality rates 1990-2011 (uncensored). Figure 2: Tuvalu under-five mortality rates 1990-2011 (uncensored). MoH MoH MDG 1992 Baseline DHS U5MR69 DHS Census Reports MDG 60 Census Reports UNIGME 1992 Baseline European Commission UNIGME IMR 57 MDG Report European Commission MDG Report MDG2015 MDG 2015 Target Goal Target Goal U5MR 23 IMR 19 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Year Year MoH:7 Ministry of Health: unpublished vital registration data. Time periods – annually 1991-2008. MoH:7 Ministry of Health: unpublished vital registration data. Time periods – annually 1991-2008. DHS 2007:9 Demographic and Health Survey: retrospective birth history. Time periods – 1993-97 (1995); 1998-2002 DHS 2007:10 Demographic and Health Survey: retrospective birth history. Time periods – 1993-97 (1995); 1998-2002 (2000); 2003-07 (2005). (2000); 2003-07 (2005). Census Reports:7,8 children ever born children surviving – 1992 Census; vital registration – 2002 Census. Time periods Census Reports:8,9 Children ever born children surviving – 1992 Census; vital registration – 2002 Census. Time periods – 1990-91 (1991); 1992-97 (1994.5); 1997-2002 (1999.5). – 1990-91 (1991); 1992-97 (1994.5); 1997-2002 (1999.5). UNIGME:10 United Nations inter-agency group for child mortality estimates. Time periods – annually 1990-2010. UNIGME:11 United Nations inter-agency group for child mortality estimates. Time periods – annually 1975-2010. European Commission:11 secondary source. Time periods – annually. European Commission:12 secondary source. Time periods – annually. MDG Report:6 Millennium Development Goals Progress Report. Time periods – annually 1991-2009. MDG Report:6 Millennium Development Goals Progress Report. Time periods- annually 1991-2009. 2014 vol . 38 no . 4 Australian and New Zealand Journal of Public Health 391 © 2014 Public Health Association of Australia Deaths (per 1,000 live births) Deaths (per 1,000 live births) Taylor et al. Article 30-43) for 2003–07 (deaths not provided). Table 1: Live births and deaths of children under-one and under-five years 1990-2008. However, the point estimate for 1998–2002 Ministry of Health is at the lower bound of the 95%CI of the Period Births Deaths <1 Deaths <5 IMR (95%CI) U5MR (95%CI) 2003–07 period, indicating there would be 1991-93 826 51 55 62 (46-81) 67 (50-87) no statistically significant difference between 1994-96 682 34 39 50 (35-70) 57 (41-78) them (Figure 4). 1997-99 588 28 41 48 (32-69) 70 (50-95) Neonatal and post-neonatal 2000-02 527 17 17 32 (19-52) 32 (19-52) mortality 2003-05 599 21 26 35 (22-54) 43 (28-64) Neonatal and post-neonatal mortality rates 2006-08 634 12 12 19 (10-33) 19 (10-33) from the 2007 Tuvalu DHS for the previous Source: previously unpublished data from Tuvalu Ministry of Health (Health Information System) five and ten-year periods are given in Table 2 Deaths <1: deaths under-one year. Deaths <5: deaths under-five years. (deaths not provided). The majority of infant IMR: infant mortality rate per 1,000 live births. U5MR: under-five mortality rate per 1,000 live births. 95%CI 95% confidence intervals. mortality occurs in the neonatal period (92% for 2003–07) and the neonatal and post- estimate and the 1997–99 U5MR (Figure 4). Under-five mortality rate neonatal mortalities are statistically different. The linear trend of U5MR using triennial MoH As with IMR, extensive variation was found However, wide 95%CIs prevent conclusions data for 1991–2008 was highly statistically in uncensored estimates of Tuvalu U5MR for concerning possible time trends. significant χ = 20.3 (d.f. 1); p<0.001, and the 1990–2008 (Figure 2). From six sources of R for a linear fit is 0.71. Estimates contained in U5MR estimates, three were censored from 8,9 Census reports indicate a decreasing trend further analysis, and the remaining three Table 2: Neonatal and post-neonatal mortality for U5MR from 59 deaths per 1,000 live births graphed with 95%CIs where possible (Figure Tuvalu 1998-2007. for 1990–91 (CEBCS) to 41 for 1997–2002 4). Estimates based on unpublished MoH Neonatal Post-neonatal (VR, deaths not provided) but this change is data of births and deaths <5 years indicate Period NNMR % IMR PNMR % IMR unable to be properly evaluated since there that the U5MR trend in Tuvalu is decreasing (95%CI) (95%CI) are no 95%CIs available. Statistical CIs (95%) from 67 (95%CI 50-87) deaths per 1,000 live 2003-07 29 (13-44) 92.0 3 (0-8) 8.0 of U5MR from MoH estimates contain the births for 1991–93 (55 deaths) to 18.9 (95%CI 2007 DHS estimates for comparable periods 1998-2007 22 (15-28) 77.9 6 (1-11) 22 10-33) for 2006–08 (12 deaths), see Table 1. indicating no statistical difference between Source: Tuvalu Demographic and Health Survey 2007: Final Report.10 Recent estimates for IMR and U5MR by MoH them. Estimates contained in the 2007 DHS Deaths <1: deaths under-one year. Deaths <5: deaths under-five years. are the same – as no deaths ≥1 and <5 were NNMR: neonatal mortality rate per 1,000 live births. PNMR: post- report suggest a decline in U5MR from 44 reported for the 2006–08 period. 95%CIs neonatal mortality rate per 1,000 live births. 95%CI 95% confidence deaths per 1,000 live births for 1993–97 to 29 intervals. Provided on request from the Secretariat of the Pacific calculated for MoH data indicate a statistical Community. for 1998–2002, then an increase to 36 (95%CI difference between the latest 2006–08 Figure 3: Tuvalu infant mortality rates 1990-2008 (censored for reliability and Figure 4: Tuvalu under-five mortality rates 1990-2008 (censored for reliability and plausibility). plausibility). MoH MoH DHS 2007 (5yr estimates) DHS 2007 (5yr estimates) Census Reports Census Reports 80 DHS 2007 (10yr estimate) DHS 2007 (10yr estimate) 70 Estimates and 95% confidence intervals Estimates and 95% confidence intervals 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 Year Year MoH:7 Ministry of Health: unpublished vital registration data. Time periods: 1991-93 (1992); 1994-96 (1995); 1997- MoH:7 Ministry of Health: unpublished vital registration (VR) data. Time periods: 1991-93 (1992); 1994-96 (1995); 99 (1998); 2000-02 (2001); 2006-08 (2007). 1997-99 (1998); 2000-02 (2001); 2006-08 (2007). DHS 2007 (5 yr estimates):10 Demographic and Health Survey 2007: retrospective birth history. Time periods: 1993- DHS 2007 (5 yr estimates):10 Demographic and Health Survey 2007: retrospective birth history. Time periods: 1993- 97 (1995); 1998-2002 (2000); 2003-07 (2005). Statistical confidence interval (95%) provided on request from the 97 (1995); 1998-2002 (2000); 2003-07 (2005). Statistical confidence interval (95%) provided on request from the Secretariat of the Pacific Community (SPC) through personal communications. Secretariat of the Pacific Community (SPC) through personal communications. DHS 2007 (10 yr estimates):10 Demographic and Health Survey 2007: retrospective birth history. Time period: 1998- DHS 2007 (10 yr estimates):10 Demographic and Health Survey 2007: retrospective birth history. Time period: 1998- 2007 (2002). Statistical confidence interval (95%) provided on request from SPC through personal communications. 2007 (2002). Statistical confidence interval (95%) provided on request from SPC through personal communications. Census Reports: Census 19928 children ever born children surviving method. Time period: 1990-91 (1991); Census Census Reports: Census 19928 children ever born children surviving method. Time period: 1990-91 (1991); Census 20029 VR. Time periods: 1992-97 (1994); 1997-2002 (1999). 20029 VR. Time period: 1997-2002 (1999). Statistical confidence interval (95%) calculations: MoH (VR) – Poisson; DHS 2007 – normal approximation of Statistical confidence interval (95%) calculations: MoH (VR) – Poisson; DHS 2007 – normal approximation of binomial. binomial. 392 Australian and New Zealand Journal of Public Health 2014 vol . 38 no . 4 © 2014 Public Health Association of Australia Deaths (per 1,000 live births) Deaths (per 1,000 live births) Epidemiology Measuring Millennium Development Goals progress in small populations attached confidence intervals that need to be The main problem with vital registration Discussion considered when comparing values over time is usually under-enumeration rather 1,13 Extensive variation exists between sources and across countries. Confidence intervals than over-enumeration, and therefore reporting estimates and trends of IMR and for IMR or U5MR estimates cannot be located higher rates from vital registration in the U5MR in Tuvalu for 1990–2008. Discrepancies in the published report of the 2007 Tuvalu past compared to indirect methods are greater than 30 deaths per 1,000 live births DHS, and in the present article 95%CIs plausible. Vital registration can be subject between estimates reported for the same were calculated for DHS estimates using the to over-enumeration if different sources of time period in both IMR and U5MR were normal approximation of the binomial based registration are added together without 20,21 identified (Figures 1 and 2). Compared with on standard errors provided on request from de-duplication, as could happen in Tonga 21,22 credible primary sources of data for IMR SPC through personal communication. or Fiji, but in this instance a single source and U5MR in Tuvalu, many of the estimates (MoH) was used. Case ascertainment of Also obtained from SPC were 95%CIs for the in Figures 1 and 2 are clearly untenable births and deaths, especially in children, neonatal and post-neonatal mortality rates. – methods and sources are often not through the Health Department is usually These suggest that for 2003–07 more than provided, and some estimates are obviously of high standard in small countries where 90% of the infant mortality occurred in the derived from modelling based on simplistic these events are infrequent. However, deaths neonatal period, which implicates antenatal assumptions. The cause for concern in these may be occasionally missed or inadvertently and intra-partum influences rather than discrepancies is that misleading data and double registered, which would inordinately environmental factors during infancy as trends in IMR or U5MR can have significant affect rates based on small numbers. Further, contributing to the IMR. These estimates are implications on subsequent health policy vital registration usually improves with time, also subject to wide 95%CIs. decisions. especially under pressure for monitoring of The 2007 Tuvalu DHS report states that MDG indicators, and thus it is likely that the Three sources remained after censorship of “The trends in childhood mortality show a <1 and <5 declines recorded based on VR IMR and U5MR estimates, and 95%CIs were worsening situation for 2003–2007 compared are real. Evaluation of the plausibility of child applied where possible (Figures 3 and 4). with 1998–2003...”; however, following mortality estimates in Tuvalu from different Comparisons of DHS estimates for 1997–2007 application of 95%CIs as recommended by sources indicates that the declines in IMR and indicate no time trend when 95%CIs are 1,2 13 WHO and UNICEF, the lack of statistical U5MR based on MoH vital registration data taken into account. Further, there were no significance of the difference is clearly are most likely to be accurate. significant differences between estimates identified (Figures 3 and 4). The 2007 Tuvalu from MoH and DHS data for 1997–2007. This report has identified and illustrated that DHS states that great care must be taken Comparison of differences over time using extensive variation in estimations of Tuvalu when interpreting the survey results given 95%CIs and statistical analysis of linear trend IMR and U5MR for 1990–2008 exist (Figures the low number of respondents and the for MoH estimates indicate a decline in IMR 15 1 and 2). Typographical errors perpetuated scope for sampling errors. In addition, the and U5MR in Tuvalu over 1991–2008. 6 in successive reports are one cause of this Tuvalu MDG Progress Report 2010/2011, extensive variation, such as the implausible While the preferred data source for IMR and the Tuvalu National Population Policy 16 reporting of an U5MR of 5 deaths per 1,000 and U5MR estimates is vital registration 2010–2015 indicate that the lower estimates 1,2 live births for the same time period that has with complete coverage, numerous of IMR and U5MR from the 2007 DHS for prior 12,23 an IMR of 41 deaths per 1,000 live births. reports acknowledge the challenges periods (1993–1997, 1998–2002) compared Although the extent of the variation in posed in attempting to generate accurate to the MoH data could arise from recall bias in estimates of IMR and U5MR was shown to estimates in developing countries that lack that many mothers did not recall accurately 1,4,13 6 decrease once estimates were censored for fully functioning registration systems. <5 deaths during the DHS interview, leading reliability and plausibility, the continued Censuses and surveys are used in the absence to subsequent under-enumeration of deaths 16 existence of discrepancies between sources of accurate vital registration to estimate for earlier retrospective periods. A 2008 deemed credible resulted in conflicting IMR and U5MR from: questions on deaths in report on the credibility of child mortality reports of whether trends in IMR and U5MR the household over a reference period; or rates produced by DHS adds that errors in in Tuvalu were decreasing or increasing. a retrospective complete maternal history the recorded dates of birth of children and The 2008 United Nations Population Fund (both direct methods); or questions on misreporting of age at death are further report on achieving MDGs in the Pacific children ever born and children still surviving important sources of bias concerning the 17 Islands outlines that caution must be taken – an indirect demographic technique useful collection of DHS mortality data. Following in evaluating progress towards MDGs in in less literate and numerate populations. assessment of IMR and U5MR estimates Pacific nations, as statistical methods that The use of household surveys in the absence from MoH and 2007 DHS sources, the Tuvalu 6 apply to large populations cannot be applied of adequate vital registration is endorsed by MDG Progress Report 2010/2011 and Tuvalu 1,2 16 without modification to countries with small organisations such as WHO and the United Population Policy 2010–2015 both report populations, particularly for ‘rare’ events such Nations Children’s Fund (UNICEF), with both that trends in these MDG indicators have as child mortality. Furthermore, the report sources indicating that surveys like DHS have decreased in Tuvalu since the early 1990s, suggests that “Pacific Island statisticians become the primary source of data on child with Tuvalu deemed to be on track to achieve may need to undergo further training in mortality in developing countries. These the MDG of reducing child mortality by two- 6,18,19 the measurement of population and health publications state that estimates obtained thirds between 1990 and 2015. trends at ‘microscale. ’” from household surveys should have 2014 vol . 38 no . 4 Australian and New Zealand Journal of Public Health 393 © 2014 Public Health Association of Australia Taylor et al. Article 11. Statistics and Monitoring Division of Policy and for assessment of methods and likely bias Conclusions Strategy. Child Mortality Estimates – Tuvalu [Internet]. and sampling variation. This report concludes New York (NY): UNICEF; 2012 [cited 2012 Nov 13]. Health planners rely on statistical measures that small populations provide challenges Available from: http://www.childmortality.org 12. Tuvalu Government, European Commission. 2008 Joint and trends to assess the effectiveness of in interpretation of IMR and U5MR trends. Annual Report. Funafuti (YUV): Government of Tuvalu; policy approaches. In the case of Tuvalu, the To ensure the correct interpretation of 13. Statistics and Monitoring Division of Policy and preceding analysis suggests that the policy differences in rates, CIs (95%) and tests for Strategy. Child Info Monitoring the Situation of Children measures that have been implemented trend should be calculated. Tuvalu has been and Women. Statistics by Area/Child Survival and Health to achieve the MDG pertaining to infant [Internet]. New York (NY): UNICEF; 2013 [cited 2012 Nov beset with many inaccurate and misleading 2]. Available from: http://www.childinfo.org/mortality. and child mortality have been effective. In estimates of IMR and U5MR; the best html the absence of detailed statistical analysis estimates of IMR and U5MR in Tuvalu indicate 14. Department of Economic and Social Affairs Population Divisions. Estimation of child mortality from information employing confidence intervals, the opposite a decreasing trend in recent periods for both on children ever born and children surviving. In: Manual conclusion may have been drawn. rates. X: Indirect Techniques for Demographic Estimation. New York (NY): United Nations; 1983. p. 73. Through calculation and application of 15. Secretariat of the Pacific Community Central Statistics 95%CIs to mortality estimates where possible, Division, and Macro International Inc. Tuvalu 2007 References Demographic and Health Survey Facts and Figures at their importance in determining statistical your Fingertips, Infant and Child Mortality [Internet]. 1. World Health Organization. Infant Mortality Rate. significance of differences in IMR and U5MR Noumea (NCL): SPC; 2012 [cited 2012 Nov 2]. Available In: Indicator and Measurement Registry Version 1.7.0. from: http://www.spc.int/sdp/index.php?option=com_ [Internet]. Geneva (CHE): WHO; 2011 [cited 2012 estimates and the validity of trends has been docman&task=doc_view&gid=214 Nov 11]. Available from: http://apps.who.int/gho/ demonstrated (Figures 3 and 4). Standard 16. Department of Planning and Budget, Ministry indicatorregistry/App_Main/view_indicator.aspx?iid=1 of Finance and Economic Development. Tuvalu practice in small populations should include 2. World Health Organization. Under-five Mortality National Population Policy 2010-2015. Funafuti (YUV): Rate. In: Indicator and Measurement Registry Version calculation and application of statistical CIs to Government of Tuvalu; 2011. 1.7.0. [Internet]. Geneva (CHE): WHO; 2011 [cited 2012 IMR and U5MR estimations, especially when 17. Sullivan J. An Assessment of the Credibility of Child Nov 11]. Available from: http://apps.who.int/gho/ Mortality Declines Estimated from DHS Mortality Rates indicatorregistry/App_Main/view_indicator.aspx?iid=7 sampling of the population is employed. It is – Working Draft; Revision 1 [Internet]. New York (NY) 3. Rakaseta V, Haberkorn G, Demmke A, Lepers C. Tuvalu standard statistical practice to calculate CIs for UNICEF Statistics and Monitoring Division of Policy Population Profile; A Guide for Planners and Policy- and Strategy; 2012 [cited 2012 Nov 13]. Available makers. Noumèa (NCL): Secretariat of the Pacific estimates based on a sample of a population from: http://www.childinfo.org/files/overall_Results_ Community; 1998. and the appropriate approach involves the of_Analysis.pdf 4. Office of the United Nations Resident Coordinator. 18. UNICEF, WHO, the World Bank and UN Population binomial (exact or an approximation). United Nations Common Country Assessment Tuvalu: Division. Levels and Trends of Child Mortality in 2006: Final Draft. Suva (FJI): United Nations; 2002. CIs are also commonly calculated for rates Estimates Developed by the Inter-agency Group for Child 5. Regional Office for the Pacific. WHO Western Pacific derived from aggregate data based on and Mortality Estimation. New York (NY): UNICEF; 2007. Country Health Information Profiles, 2011 Revision. 19. United Nations Population Fund. Achieving the Manila (PHL): World Health Organisation; 2011. numerators and denominators from different Millennium Development Goals in the Pacific Islands; 6. Millennium Development Goals Project Oc ffi er with sources (in this case recorded deaths Policies and Strategies in Population and Reproductive Assistance from the Department of Planning and Health – Suva. New York (NY): UNFPA; 2008. Budget. National MDG Taskforce and UNDP Multi- and births) since such estimates can be 20. Carter K, Cornelius M, Taylor R, et al. Mortality trends in Country Office: Tuvalu Millennium Development Goals considered as samples in time, in which case Fiji. Aust N Z J Public Health. 2011;35(5):412-20. Progress Report 2010/2011. Funafuti (YUV): Government 21. Hufanga S, Carter KL, Rao C, Lopez AD, Taylor R. the Poisson distribution is more appropriate. of Tuvalu; 2011. Mortality trends in Tonga: An assessment based on a 7. Ministry of Health. Data on reported births and deaths Furthermore, approaches to produce synthesis of local data. Popul Health Metr. 2012;10:14. aggregated by age groups. Received through co- 22. Carter K, Rao C, Lopez AD, Taylor R. Mortality and cause- author (SH) in 2011. Unpublished observations. estimates of IMR and U5MR require of-death reporting and analysis systems in seven pacific 8. Central Statistics Division. Tuvalu 1991 Population and examination concerning the possibility of island countries. BMC Public Health. 2012;12:436. Housing Census, Volume 2: Analytical Report. Funafuti 23. Government of Tuvalu, United Nations Development under-enumeration bias for continuous (YUV): Government of Tuvalu; 1992. Programme. Tuvalu Millennium Development Goals 9. Secretariat of the Pacific Community. Tuvalu 2002 recording (vital registration) and recall bias for Report 2006 [Internet]. Paris (FRA): UN International Population and Housing Census, Volume 2: Demographic 6,17 retrospective methods. Finally, the results Institute for Educational Planning; 2006 [cited 2012 Profile, 1991-2002. Noumea (FJI): Secretariat of the Nov 13]. Available from: http://planipolis.iiep.unesco. Pacific Community; 2005. of all approaches to the production of IMR org/upload/Tuvalu/Tuvalu%20MDG%202006.pdf 10. Central Statistics Division, The Secretariat of the Pacific and U5MR estimates should be compared 24. Armitage P, Berry G, Matthews JNS. Probabliity. In: Community, and Macro International Inc. Tuvalu th Statistical Methods in Medical Research. 4 ed. Boston Demographic and Health Survey 2007: Final Report. – as a matter of course – with each other; (MA): Blackwell Scientific; 2002. p. 65-76. Noumea (FJI): Secretariat of the Pacific Community; confidence in estimates is enhanced when different methods produce similar estimates and trends, allowing for their particular defects. Conflicting data indicates the need 394 Australian and New Zealand Journal of Public Health 2014 vol . 38 no . 4 © 2014 Public Health Association of Australia
Journal
Australian and New Zealand Journal of Public Health – Wiley
Published: Aug 1, 2014
Keywords: ; ; ;