Print version ISSN 1405-7425
Pap. poblac vol.16 no.63 Toluca Jan./Mar. 2010
What do we know about adult mortality and data quality in Peru? Mortality coverage levels and trends from recent decades
¿Qué sabemos acerca de la mortalidad y calidad de datos en Perú? Niveles de cobertura de mortalidad y tendencias en décadas recientes
Mario PiscoyaDíaz & Bernardo L. Queiroz
Accurate knowledge of adult mortality levels and trends in the developing world is hampered by its widespread lack of complete vital registration systems. Although knowledge of infant and child mortality was once affect by the same problem, surveybased techniques have been more successful in estimating child and infant than adult mortality. The main objective of this paper is to estimate mortality rates for the population aged 5 and above, in Peru by sex. The study evaluates the degree of coverage, and corrects the level of mortality, when necessary, using different methodologies. The literature does not indicate the best method to investigate mortality data problems. Thus, the implementation of alternative methods will improve the understanding of the mortality levels and trends in Peru in recent decades.
Key words: Peru, adult mortality, demographic methods.
El conocimiento preciso sobre los niveles de mortalidad en adultos y sus tendencias en países en desarrollo ha sido detenido por una falta generalizada de sistemas de registros vitales. A pesar que el conocimiento sobre mortalidad infantil se vio alguna vez afectado por el mismo problema, técnicas basadas en encuestas han sido más efectivas al estimar la mortalidad infantil y adulta. El principal objetivo de este artículo es estimar las tasas de mortalidad para la población de cinco años y más, por sexo, en Perú. El estudio evalúa el grado de cobertura y corrige el nivel de mortalidad, cuando es necesario, usando diferentes metodologías. La literatura no indica el mejor método para investigar problemas de datos sobre mortalidad. De esta forma, la implementación de métodos alternativos mejorará la comprensión de los niveles de mortalidad y sus tendencias en Perú en décadas recientes.
Palabras clave: Perú, mortalidad adulta, métodos demográficos.
The knowledge of the levels and trends of adult mortality in developing countries is limited by the quality of available data, for both death counts and population. The most common problems faced by developing countries are incomplete coverage of the vital registration system, errors in age declaration for both population and death (United Nations, 1983; Preston, Elo, Stewart, 1999).
Latin America, and Peru specifically, have undergone major demographic changes (Chackiel and Plaut,1996; Chackiel, 1999). Although mortality levels are still high, in comparison to the developed world, Peru observed a rapidly decline in child and infant mortality because of medical and other technological changes. In addition, economic modernization and changes in social and cultural norms are related to a rapid decline in fertility and important changes on how parents relate and invest in their children. Those two elements (fertility and infant mortality) of the demographic transition have been widely studied in the area (INEI, 1986; Guzman,1989; Chackiel,2004; Celade,2004; INEI, 2005)
Adult mortality, however, compared to fertility and infant and child mortality is much less understood in the large majority of developing countries, and Peru is no exception (Timeaus, 1996). Nevertheless, it is important and relevant to study and understand levels and patterns of adult mortality in a scenario of rapid population aging and major changes in the socioeconomic scenario.
The main objective of this paper is to estimate mortality rates for the population aged 5 and above, in Peru by sex. The study evaluates the degree of coverage, and corrects the level of mortality, when necessary, using different methodologies. The literature does not indicate the best method to investigate mortality data problems. Thus, the implementation of alternative methods will improve the understanding of the mortality levels and trends in Peru in recent decades.
Estimating adult mortality using census and vital registration data requires the application of a variety of evaluation methods, adjust them, if necessary, and some key assumptions that are discussed in the following section. The data needed are two population age and sex distributions from successive censuses, and an age and sex distribution of deaths reported either in the population census or some sort administrative data record (vital registration).
We obtained population distribution by age and sex from the population censuses for 1993 and 2005. The National Statistics Office run the census in 1836, 1850, 1862, 1876, 1940, 1961, 1972, 1981, 1993 and 2005. The first 4 censuses listed only provide population totals by political division and sex without an age breakdown. From 1940, censuses provide age group distributions of the population, nevertheless age declaration in the 1940 census is very inaccurate and an improvement was observed in 1961 census (Arriaga, 1966). It is important to remark that the age declaration has been improving with time (Arriaga, 1966; Moser 1985).
An examination of 19611981 census data shows a slight deficit of males for almost all age groups, except the youngest (Moser, 1985). For 1993 and 2005 censuses we find the same pattern observed by Moser (1985). A closer comparison of both single age distributions suggest an under enumeration of population aged 02 in 1993 census, the number of survivors of those born from 1991 to 1993 in 2005 census was greater than the observed in 1993 census.
We obtain mortality data from vital registration, the data correspond to the years 20002004. Historically, vital system statistics are considered of bad quality (Arriaga, 1966; laspers and Orellana, 1996). A detailed revision of the coverage of death registration systems for Latin America was performed by laspers and Orellana (1996). In the specific case of Peru, they showed that the omission rate was around 45 per cent in the period between 1960 and 1985 and it appears to have not dropped since then.
We also used mortality estimates from the AMDC project (reference here). Hill & Choi (2004) estimated vital registration coverage and analyzed data quality for a series of developing countries in the world. They produced estimates for Peru from the early 1960 to the 1990s. We use their estimates to analyze the recent trend in Peruvian mortality.
The death distribution methods are widely used to estimate adult mortality (Timaeus, 1991). They compare the distribution of deaths by age with the age distribution of the living and provide age patterns of mortality in a defined reference period. There are two major approaches: the General Growth Balance Methods (Hill, 1987), and the Synthetic Extinct Generation method (Benneth & Horiuchi, 1981). The death distribution methods make several strong assumptions: that the population is closed to migration that the completeness of recording of deaths is constant by age, that the completeness of recording of population is constant by age, and that ages of the living and the dead are reported without error.
The Bennett and Horiuchi (1981) method is a synthetic cohort analog of Vincent's (1951) method of extinct generations, known as Synthetic Extinct Generations (SEG) method. In these methods, agespecific growth rates are used to convert an observed distribution of deaths by age into the corresponding stationary population age distribution. Since in a stationary population the deaths above each age x are equal to the population aged x, the deaths in the stationary population above age x provide an estimate of the population of age x. The completeness of death registration relative to population is estimated by the ratio of the deathbased estimate of population aged x to the observed population aged x. The synthetic estimate of the population aged x is given by:
where N(x) is the estimated population aged x, D(y) is the observed number of deaths at age y, and r(z) is the agespecific growth rate of the population at age z.
The General Growth Balance (GGB) method (Hill 1987) is a generalization to all closed populations of the Brass (1975) Growth Balance method. The Demographic Balancing Equation expresses the identity that the growth rate of the population is equal to the difference between the entry rate and the exit rate. This identity holds for openended age segments x+. The entry rate x+ minus the growth rate x+ provides a residual estimate of the death rate x+. We can then estimate the residual from population data from two population censuses and compared to a direct estimate using the recorded deaths (from the census or vital registration), and comparing these two records we can estimate the completeness of death recording relative to population (Hill, 2003).
Hill (1987) shows that:
Where N1 and N2 are population counts at two time points separated by t years, D are recorded intercensal deaths, and k1, k2, and c are respectively the completeness of the first and second population counts and the intercensal deaths.
Hill and Choi (2004) proposed that the combination of SEG and GGB might be more robust than either one individually. The combined method consists of first applying GGB to estimate any changes in census coverage (k1/k2), using the estimate to adjust one or other census to make the two consistent, and then applying SEG using the adjusted population data in place of the reported.
The death distribution method has also an important limitation. Since they compare a distribution of deaths to an intercensal population, they estimate intercensal completeness of mortality estimates, and not the completeness at the beginning or end of the intercensal period. This matter is of particular importance when a distribution of deaths comes from data from the latest census or when vital registration is available for recent years (Hill et al, 2009). For example, in the case of Peru we obtained mortality data from 2000 to 2004. Thus, the age pattern of mortality reflects the age distribution of the population around the time of the the second census, not to the average age distribution of the population over the intercensal period. In order to take this into account, we should first used the deaths and the population from the second census to calculate agespecific death rates, and then estimated average annual deaths for the intercensal period by applying the death rates to an estimate of the age distribution of the intercensal population.
The assumption of both the SEG and GGB methods that the population is closed to migration is also of importance to Peru, since it has experienced recent net emigration (INEI, 2009). Both the SEG and GGB methods use information on deaths and growth rates cumulated above a series of ages x. If there is some age x above which net migration is negligible, the performance of the methods above that age will be unaffected. The intercept and slope of the GGB method were obtained by orthogonal regression to points for the ages 35+ to 75+. The coverage of deaths for the SEG method was obtained by averaging the estimates for the age ranges 35+ to 75+. Finally the combined method was applied by using GGB estimates to adjust the data for the estimated census coverage change and then applying the SEG method.
Measures of adult mortality
We also show summary measures of adult mortality to illustrate our analysis. We assume that adult mortality can be defined as starting at age 15, about the inflection point at which declining mortality risk in childhood are replaced by rising risks in adulthood. The primary measure of adult mortality used in this paper is the probability of dying between the age 15 and 60, 45q15. This measure covers a substantial age range, but avoids the problems inherent in the measurement of old age mortality (Hill, 2003).
The performance of the evaluation methods are better seem graphically, Figure 1 to 3 show the results for GGB, SEG and SEGadjusted. Figure 1 shows the dignostic plot for the General Growth Balance method. For the GGB method, the xaxis shows the observed death rate for ages x+, and the yaxis represents the death rates for ages x+ obtained by residual. The results from the SEG method are presented in Figure 2 and Figure 3. For the SEG method, the figure show the ratio of the deathbased estimate of population aged x to the observed population aged x.
Following Bhat (1990), we fitted a orthogonal regression to estimate a straight line to the points for the age segments considered in the GGB estimation. The slope of the fitted to the points estimates the adjustment factor necessary to correct observed death rates to the death rates estimated by residual. The intercept of the line provides an estimate of the relative coverage between the two censuses used in the analysis. The analysis of the graph confirms our concern about the assumption of closed population in Peru. The points at young ages or both males and females are very irregular and off the fitted line, and the estimate of census coverage indicates better coverage in the first census what is consistent with problems arising from net emigration (Hill & Choi, 2004). Overall, the fit of the observations (death rates) are quite good, and improve over time and for age groups 35 and above the results imply that age reporting is surprisingly good and the assumptions of the GGB method are met.
The results for the slope, for both males and females, and census coverage are show in Table 3. The results indicate that death registration in Peru is almost complete, the slope of the line indicates a small adjustment of 1,07 to the death rates for males and 1.02 for females, indicating coverage of 93 per cent for males and 98 per cent for females. The results hold also when we fitted the method to the openended interval 35 and over to avoid possible distortions caused by migration. The probability of dying between ages 15 and 60, considering the adjustment factors, are 0.1248 and 0,085 for males and females, respectively.
Table 2 and Figure 2 shows the results obtained from application of SEG Method. The results are very different from the application of the General Growth Balance. SEG estimates death coverage of about 80 per cent for both sexes. The results obtained in the GGB estimation indicated census coverage changes in Peru for the period. The ascending completeness of mortality record by age, shown in Figure 2, is a result of the change in census coverage discussed before Hill & Choi (2004).
We used the procedure developed by Hill and Choi (2004) to overcome this limitation. Figure 3 shows the death coverage by age without the ascending pattern observed before. By applying a correction to census coverage, we estimated completeness of death registered at 86 per cent for men and 93 per cent for females, much closer to what was obtained using the GGB method.
Figure 3 shows the results for men and female, and in both cases it can be observed that there is a better consistency between the completeness of death registration and the age. The ascending pattern in coverage by age, that was cause by changes in census coverage, is not observed and results are very consistent with GGB estimation.
The quality of mortality data in Peru improved steadily over the last halfcentury. Death registration coverage in the 1960s was about 84 per cent for males and 90 per cent for females (estimation not shown), reaching over 93 per cent for both sexes in the most recent period. Overall, the results indicate that the death registration coverage for Peru have improved significantly over time.
Figure 4 shows the probabilities of dying between the ages of 15 and 60 by sex by decade for Peru. The probability of dying increases from the 1960s to the 1980s, especially for males, and starts a steadly decline since them. The increase in the probability of dying for males, from about 0,14 to about 0,17 can be explained by the impressive increase in the external causes of deaths observed during that time in Peru (Guibovich & Gonzales, 2008).
Regional levels and differentials
The data obtained from the National Statistics office allows us to study regional differentials in registration coverage in Peru. Since the country is marked by high income inequality and regional differences we now concentrate state level results. Death coverage varies widely across regions in Peru from 45 per cent in Huacavelicato 105 per cent in lea. The methods adjustment also shows a wide variation, since small areas are less suitable for the assumption of closed population.
In order to summarize the situation in Peru, we adopt the classification system proposed by laspers and Orellana (1996). According with that system, coverage of registered deaths are rated as follows (see Table 4 ):
• 'Good', when coverage is greater than 90 per cent.
• 'Satisfactory', when it is between 80 and 89 per cent.
• 'Fair', when it is between 70 and 79 per cent.
• 'Deficient', when it is lower than 70 per cent.
The distribution of quality of death reporting follows very closely the level of development and distance to Lima. The regions located near to the ocean show a much higher coverage than the regions in the interior of the country. In general, poverty levels seem to have strong correlation with data quality and death recording. The more developed regions such as Lima, Callao, Arequipa and Tacna have death coverage above 90 per cent whereas the poorest areas of the country, such as Ayacucho, Cajamarca and Loreto, have coverage levels below 70 per cent.
Discussion and conclusion
The knowledge of the levels and trends of adult mortality in developing countries is limited by the quality of available data, for both death counts and population. The most common problems faced by developing countries are incomplete coverage of the vital registration system, errors in age declaration for both population and death. However, it is extremely important to produce estimates of data quality and death rates for they are important to elaborate public policies and to evaluate the impact of socioeconomic changes and policies. The purpose of this paper has been to apply the methods to Peru and produce an systematic analysis of mortality data quality, record coverage, and produce estimates of adult mortality by sex and region of Peru.
The quality of the diagnostic plots and the consistency of the adult mortality estimates in these applications suggest that the methods are working well. Nevertheless, it is important to discuss some important issues when applying the Death Distribution methods to developing nations: a) the methods seem to work well, but with considerable uncertainty, thus it is important to produce more work on how to deal with these uncertainty; b) as discussed by Hill & Choi (2004), different data limitations affect GGB and SEG methods differently and results should be carefully analyzed. For example, SEG is strongly affect by changes in census coverage whereas GGB shows some advantage by allowing for systematic errors.
We compared our probabilities of adult mortality to estimates from the United Nations (tables not shown) as a plausibility test and find very similar results. It is important to point that they use very different methodological approach from us. We observed that after a period of increase in adult mortality, probably related to external causes, adult mortality in Peru is declining. We also find that female adult mortality is lower during all time and probability of dying between ages 15 and 60 are declining faster for females than males. (Figure 5)
We find that death registration records vary widely across regions in Peru and are high correlated with development indicators. The regions located near to the ocean show a much higher coverage than the regions in the interior of the country. In general, poverty levels seem to have strong correlation with data quality and death recording. The more developed regions have death coverage above 90 per cent whereas the poorest areas of the country have coverage levels below 70 per cent. The results for small areas, however, should be carefully analyzed since they are less suitable for the assumption of closed population.
The exercise performed here suggests that the Death Distribution Methods (GGB and SEG) are a feasible way to study and understand adult mortality in developing countries. Although, new data sources, such as DHS, and o indirect methods, such as orphanhood, have been used for a long time and have improved our knowledge of adult mortality and provide information on a more regular basis. It seems from the findings that data collected from vital registration offices combine to censuses data provide a unique opportunity to study regional and, in some cases, socioeconomic differentials in adult mortality.
Nevertheless, all efforts should be directed to produce high quality vital registration records, including cause of death. In the case of Peru, the least developed regions of the countries are the one that should received largest investments, since the results indicate that less than 60 per cent of deaths are recorded in these areas. Advances in these are would produce better and more reliable estimates of mortality and improve our understanding in trends and differentials for a series of developing countries.
AGOSTINHO, C. & B. QUEIROZ, 2008, Estimativas de mortalidade adulta para o Brasil no periodo 19802000: uma abordagem metodológica comparativa, in Anais do XVI Encontro Nacional da ABEP, Caxambu. [ Links ]
BENNETT, N. G. and S. HORIUCHI, 1981, "Estimating the completeness of death registration in a closed population", in Population Index 47(2). [ Links ]
BHAT M. "Estimating transition probabilities of age misstatement", in Demography 1990, 27. [ Links ]
BRASS, William, 1964, Uses of census or survey data for the estimation of vital rates, (E/CN.14/CAS.4/V57), Paper prepared for the African Seminar on Vital Statistics, Addis Abbaba, 1419 december. [ Links ]
BRASS, William., 1975, Methods for estimating fertility and mortality from limited and defective data, Chapel Hill, Carolina Population Center, University of North Carolina. [ Links ]
BRASS, William, 1975, Methods for estimating fertility and mortalityfrom limited and defective data, Chapel Hill, N.C., International Programme of Laboratories for Population Statistics, North Carolina Population Center. [ Links ]
CEPAL, 2001, "América Latina: tablas de mortalidad 19502025", in Boletín Demográfico, núm. 67, Celade, Santiago de Chile. [ Links ]
CELADE, 2004, Boletín Demografico, núm.74, Santiago de Chile. [ Links ]
CHACKIEL, J.& R. PLAUT, 1996, "Demographic trends", in I. TIMAEUS, J. [ Links ]
CHACKIEL, L. RUZIKA, Adult mortality in Latin America, Oxford University Press. [ Links ]
CHACKIEL, J., 1999, Mortality in Latin America, in Health and Mortality Issues of Global Concern. United Nations y Population and Family Study Centre (CBGS). [ Links ]
CHACKIEL, J. 2004, "La transición de la fecundidad en América Latina 19502000", in Papeles de Población, num. 41, Universidad Autonoma de Mexico. [ Links ]
GUZMAN, J., 1989, "Trends in socioeconomic differentials in infant mortality in selected latin american countries", in L. RUZICKA, G. WUNSCH, P. KANE, Diferential mortality: methodological issues and biosocial factors, Oxford University Press. [ Links ]
GUIBOVICH, G. & F. GONZÁLES, 2008, "Mortalidad en el Peru, 1987 versus 1997", in Revista Peruana de Epidemiologia, vol. 12, num 2. [ Links ]
HILL, K., B. QUEIROZ, L. WONG, F. POPOLO, J. PLATA, J. ROSALES & C. STANTON, 2009, "Estimating pregnancyrelated mortality from census data: experience from Latin America", in Bulletin of the World Health Organization, vol. 87. [ Links ]
HILL K, Y. CHOI & I. TIMAEUS, 2005, "Unconventional approaches to mortality estimation", in Demographic Research, nov. 13(12). [ Links ]
HILL, K., Y. CHOI, 2004, "Death distribution methods for estimating adult mortality: sensitivity analysis with simulated data errors", in Adult mortality in developing countries workshop, The Marconi Center, Marin County, California. [ Links ]
HILL, K. 2003, "Metodos para estimar la mortalidad adulta en los paises en desarollo: una revision comparativa", in Notas de Poblacion, num. 76. [ Links ]
HILL, Kenneth, 2000, Methods for measuring adult mortality in developing countries: a comparative review, The Global Burden of Disease in Aging Populations/Research Paper núm. 01.13. [ Links ]
HILL, K, 1987, "Estimating census and death registration completeness", in Asian and Pacific Census Forum, 1(3): 813. [ Links ]
HUYNEN, M. L. VOLLEBREGT, P. MARTENS & BENAVIDES, 2005, "The epidemiologic transition in Peru", in Revista Panamericana de Salud Publica, vol. 17, num. 1. [ Links ]
INEI, 1986, Peru. Encuesta Demográfica y de Salud Familiar, Instituto Nacional de Estadística e Informática (ENDES). [ Links ]
INEI, 2005, Peru. Encuesta Demográfica y de Salud Familiar, Instituto Nacional de Estadística e Informática (ENDES). [ Links ]
INEI, 2009, Peru. Evolución del movimiento migratorio peruano, Instituto Nacional de Estadística e Informática, diciembre 2008. Informe Técnico núm. 2, febrero. [ Links ]
PRESTON, Samuel, Patrick HEUVELINE and Michel GUILLOT, 2000, Demography: measuring and modeling population processes, Blackwell, Oxford. [ Links ]
STANTON, C., J. HOBCRAFT, K. HILL, M. NAGHAVI, F. MUNENE, W. MAPETA, N. KODJOGBÉ, V. RABEZA, B. SISOUPHANTONG & O. [ Links ]
CAMPBELL, 2001, "Every death counts: measuring maternal mortality via a census", in Bulletin of the World Health Organization, 79(7). [ Links ]
TIMAEUS, Ian et al., 2001, The impact of HIV/AIDS on adult mortality in South Africa, Technical Report, Burden of Disease Research Unity/South Africa Medical Council Center. [ Links ]
UNITED NATIONS, 1997, Manual X: indirect techniques for demographic estimation, United Nations, New York. [ Links ]
UNITED NATIONS, 2002, Population Division. Methods for estimating adult mortality, New York. [ Links ]
VINCENT, P., 1951, "La mortalite des viellards", in Population, 6(2). [ Links ]
ZLOTNIK H & K. HILL, 1981, "The use of hypothetical cohorts in estimating demographic parameters under conditions of changing fertility and mortality", in Demography 18(1). [ Links ]
Information about author(s)
Mario Ernesto PiscoyaDíaz. Es bachiller en Estadística por la Universidad Nacional Mayor de San Marcos, Lima, y maestro en Estadística por la Universidad Federal de Minas Gerais. Es alumno del Programa de Doctorado en Demografía del Centro de Desenvolvimento e Planejamento Regional (Cedeplar) de la Universidad Federal de Minas Gerais. Sus áreas de intereses son salud sexual y reproductiva, y mortalidad adulta en América Latina. Actualmente se desempeña como docente en el Instituto de Matemática y Estadística de la Universidad Federal de Goiás, Brasil.
Bernardo L. Queiroz. Ostenta una licenciatura en Ciencias Económicas por la Universidade Federal de Minas Gerais, una maestría en Demografía por la misma institución y un doctorado en Demografía por la University of California at Berkeley. Actualmente es profesor del Departamento de Demografía de la Universidade Federal de Minas Gerais e investigador del Centro de Desenvolvimento e Planejamento Regional (Cedeplar). Entre sus publicaciones más recientes están: "Estimating pregnancyrelated mortality from census data: experience from Latin América", en Bulletin of the World Health Organization, 2009 (en colaboración con Kenneth Hill, Laura Wong, Fabiana del Popolo and Cindy Stanton); "Compressão da mortalidade: um estudo da variabilidade da idade à morte na população do Estado de São Paulo, Brasil, 19802005", en la Revista de Saúde Pública (en coautoría con Marcos Gonzaga y Carla Machado). Sus líneas de investigación son: mortalidad adulta, demografía formal, transferencias intergeneracionales y seguridad social.