T im aeu s, I. M ., Chackiel, J. & Ruzicka, L. (Editors). A dult M ortalin- in Latin A m ^  
Oxford, U.K.: Oxford U niversity Press, 1996.
9 Adult Mortality Decline in Costa Rica
LUIS ROSERO-BIXBY
Princeton University, Princeton, USA
1. Introduction
Costa Rica is, along with Cuba, the country with the best health indicators in 
Latin America. Life expectancy at birth in Costa Rica was 77.9 years for 
women and 72.7 years for men in 1990, figures which are comparable with 
those for Western Europe and the USA. Study of the epidemiologic transition 
in Costa Rica has usually focused on children, reflecting the fact that the 
main component of mortality decline has been the prevention and control of 
premature deaths (Behm, 1976; Rosero-Bixby, 1986; CELADE et al., 1987; 
Cervantes and Raabe, 1991). The greater availability of information on child 
health and the young age-structure of the population (in 1960, 47 per cent of 
the population were children under 15 years of age) have reinforced the 
emphasis on studying the young.
This chapter shifts the previous emphasis on childhood to focus on adult 
mortality. The purpose is to describe the mortality transition at adult ages, to 
identify its key components, and to make inferences about its likely determi­
nants. The chapter has five sections: socio-economic and public-health back­
ground of Costa Rica; the data and methods used; decline in risks of dying in 
two age intervals (20-49 and 50-79 years); analysis of risks of dying by cause 
of death; and areal analysis of adult mortality and its correlate across 100 small 
geographical units.
2. Background Information on Costa Rica
2.1. Socio-Economic Background
Costa Rica is a small Central American country with a population of about 3 
million. Its economy is dependent on export of tropical crops, predominantly 
coffee and bananas. Per capita income in 1990 was about US $1900, slightly
This paper is part of a collaborative research project into health policy of the University of Costa 
Rica and the Nordic School of Public Health. The project was partially supported by a grant of the 
Swedish Agency for Research C ooperation with Developing Countries (SA REC). D r Leonardo 
M ata contributed to this paper with helpful comments.
Adult Mortality Decline in Costa Rica 167
lower than the Latin American average. Approximately one half of the popu­
lation resides in rural areas and a third of the labour force is employed in 
agriculture. The country has had greater social than economic development: 
only 7 per cent of adults are illiterate, almost all children attend school, 79 per 
cent of the population is covered by the social security system, and 94 per cent 
enjoy a piped water supply (Table 9.1).
After World War II, Costa Rica enjoyed more than three decades of vigor­
ous economic growth. The real growth in gross domestic product (GDP) was 
greater than 6 per cent per annum (more than 3 per cent in per capita terms) 
between 1950 and 1980. Moreover, welfare-oriented institutions helped to 
distribute the product of such progress comparatively evenly among the vari­
ous social strata (González-Vega, 1985). In the early 1980s, however, inter­
national and domestic circumstances brought this continuous progress to an 
end. The country experienced an acute economic crisis in the 1980s, from 
which it has not yet recovered. Consequently, GDP and expenditure on public 
health per capita in 1990 were lower than in 1980. Although the main indica­
tors of health status of the population did not deteriorate during this decade, 
other social indicators, such as school attendance, have decayed (Table 9.1).
2.2. The Health Sector
Significant public health programmes began in Costa Rica in the 1920s, with 
the creation of the Sub-Secretariat of Hygiene and Public Health in 1922; the 
Sub-Secretariat was promoted to ministeriaHevel in 1929 (Meza-Lago, 1985). 
An ambitious social security system started in 1941 which provided, among 
other benefits, medical and hospital care to workers in the formal sector and, 
since 1955, to their families. The health sector was substantially reorganized in 
the early 1970s. All hospitals were transferred to the social security system, 
which, in turn, expanded its coverage from 39 to 70 per cent of the population 
over a ten-year period (Table 9.1). The Ministry of Health established a 
programme of primary health care to reach rural areas and urban slums 
(Saenz, 1985). By 1980, this programme covered 60 per cent of the Costa Rican 
population with services that included quarterly visits to every household by a 
health worker.
At present, Costa Rican medicine is highly socialized. Only 1.2 per cent of 
the hospital discharges come from the private sector. Coverage of public 
medical services is high: for example, 97 per cent of births occurred in hospitals 
in 1990. However, the quality of the service may be questioned. The budget 
outlay for health in the public sector peaked at 7.6 per cent of the GDP in 1980 
and today represent 5.6 per cent of GDP (Table 9.1).
2.3. Demographic Trends
Mortality probably started to decline in Costa Rica before the end of the 
nineteenth century, after the cholera epidemic of 1856. In contrast, the birth
168 Luis Rosero-Bixby
Table 9.1. Demographic, socio-economic and health characteristics of Costa Rica, 
1950-90
Indicators 1950 1960 1970 1980 1990
Demographic
Populations (1000s) 862 1236 1731 2284 3015
Natural growth (%) 3.2 3.8 2.6 2.7 2.3
Total Fertility Rate 6.7 7.3 4.9 3.7 3.2
Under age 15 (%) 43 47 46 38 35
Socioeconomic
Per capita GDP in 1990-USS 808 1103 1527 1999 1937
Male work force in agriculture (%) 63 59 49 35 33
Illiteracy (% in ages 15+) 21 16 13 10 7
Enrolment in primary and secondary 39 52 61 64 60
school (% ages 5-19) 
Health (general)
Life expectancy at birth 55.6 62.6 65.4 72.6 75.2
Infant mortality rate (1000s) 95 80 67 2D 16
Public health expenditure
Per capita in 1990-USS 19 33 78 152 109
As % of GDP 2.2 3.0 5.1 7.6 5.6
Hospital care
Beds per 1000 population 5.1 4.6 4.1 3.3 2.5
Discharges per 1000 population 95 101 111 117 105
Births in institutions (%) 20 49 70 91 97
Medical care
Physicians per 10,000 people 3.1 2.8 5.6 7.8 8.9
Medically certified deaths (%) 60 65 71 84 96
Health insurance coverage (%) 8 15 39 70 79
Outpatient consultations per capita
(public health services) — 1.1 2.0 2.9 2.5
Primary health care
Administrative coverage (%) 0 0 0 60 57
Sanitation
Population with piped water (%) 53 65 75 83 94
Population with faeces disposal (%) 48 69 86 94 97
Source: U pdated from Rosero-Bixby (19856).
rate remained high until 1960 and even increased briefly in the 1950s. Conse­
quently, population growth accelerated, reaching 3.8 per cent in 1960—one of 
the highest growth rates in the world. By 1960 the birth rate began to decline, 
which resulted in a reduced population growth (Table 9.1).
Adult Mortality Decline in Costa Rica 169
In contrast with its present very low mortality, Costa Rica’s population 
growth (2.3 per cent) and total fertility rate (3.2 births per woman) remained 
moderately high in 1990. The low mortality level and a transitional age- 
structure with relatively large population at ages at which the risk of death is 
minimal, result in a crude death rate below 4 per 1000, one of the lowest in the 
world.
2.4. The Epidemiologic Transition
Early data suggest that at the beginning of this century mortality was lower in 
Costa Rica than in Latin America as a whole (Rosero-Bixby, 1985a). Social 
homogeneity inherited from colonial times, a more egalitarian distribution of 
land, the absence of militarism, and an emphasis on education by governments 
of all parties are some factors that seem to account for this initial advantage of 
Costa Rica (Mata and Rosero-Bixby, 1988). As in the rest of the continent, the 
most rapid improvement in life expectancy took place after World War II. 
Costa Rica’s life expectancy rose sharply from 46 to 63 years between 1940 and 
1960. This progress has been linked with the adoption of cost-effective tech­
nologies, such as antibiotics, DDT, and vaccines, as well as to government 
initiatives, such as the social security system established in 1941. The standard 
of living in Costa Rica also increased substantially during this period (Rosero- 
Bixby, 1991a).
After modest gains in life expectancy in the 1960s, a breakthrough in the 
trend took place in the 1970s (Caldwell, 1986). Life expectancy rose from 65 to 
73 years between 1970 and 1980, owing to a dramatic decline in the infant 
mortality rate from 67 to 21 per 1000. This decline has been linked mainly 
to the implementation of cost-effective primary health care programmes 
among rural populations. Conventional health interventions, favourable 
socio-economic circumstances, and a substantial fertility reduction have also 
been identified as significant factors in the improvement of child health in the 
1970s (Rosero-Bixby, 1986).
In spite of the economic recession, mortality continued falling during the 
1980s, but at a slower pace. Life expectancy increased from 72.6 to 75.2 years 
and infant mortality declined from 21 to 16 per 1000 between 1980 and 1990 
(Table 9.1). This progress is intriguing since it occurred under difficult socio­
economic circumstances and declining public health expenditure and services 
(Table 9.1).
3. Data and Methods
Most analyses in this paper are based on the risk of dying in the age groups 20 
to 49 years and 50 to 79 years. These risks measure the probability of dying in 
those age intervals by a person that has reached the initial age of the interval.
170 Luis Rosero-Bixby
The risks were derived from the age-specific mortality rates (5m J and the 
following approximate relations (Kleinbaum et al., 1982: 107):
= 1 -  exp - 5 - 1 : m .
i j
= 1 -  exp - 5 - 1 : m .
The risk of dying was also computed by specific causes of death. In this case, 
the risk represents the probability of dying of a particular cause (or group of 
causes) in the corresponding age interval in the absence of other causes of 
death. The following relation aggregates the risks of dying by k different 
causes i:
e=i-n(i-e)
>=i
The proportional contribution of the i-th cause of death to the decline in the 
risk of dying by all causes—the ‘attributable risk decline’ D ,—was estimated 
using the following relation:
— = •  Y D  =1 
' A OQ 1 --00,, ’ T  '
where Q is the mean risk in the period and the operator A indicates the 
amount of change (the first difference) during the period.
The data source from which the risk of dying by all causes is drawn is a series 
of life tables covering the period 1920-80 (Rosero-Bixby and Caamaño, 1984). 
These tables include corrections for under-registration of deaths. The series 
was updated with life tables computed for 1985 and 1990 (not shown) employ­
ing the same procedures and adjustments as for the 1980 life table.
Seventeen groups of causes of death were defined, based on the classifi­
cation into twelve groups developed by Preston et al. (1972). Table 9.2 presents 
the definitions of these groups according to the International Classification of 
Diseases, 5th to 9th Revisions. The age-specific death rates by cause in 1951— 
2 and 1961-2 come from statistical yearbooks (‘Anuario Estadístico’) pub­
lished by the Directorate of Statistics and Censuses. For 1971-2,1981-2 and 
1989-90 the data come from computer files obtained from the Directorate of 
Statistics and Censuses.
The areal analysis of adult mortality is based on two cross-sections centred 
in 1973 and 1984 for 100 counties. Mortality rates were computed with three- 
year averages of deaths (i.e. 1972-4 and 1983-5) from vital statistics computer 
files using the census population in the age intervals 20-49 years and 50-79
Adult Mortality Decline in Costa Rica 171
Table 9.2. Definition of the groups of causes of death. International Classification of 
Diseases, 5th to 9th Revisions
G roups o f causes o f death  International Classification of D iseases Revisions
5th 6th 7th 8th 9th
(1938) (1948) (1955) (1965) (1975)
Respiratory tuberculosis 13 1 1 -8 10-12 10-12
M alaria 28 37 111-116 84 84
Diarrhoea] disease 119-120 104 571-572 8 - 9 8 - 9
Acute respiratory 33. 1 0 6 -9 8 8 -9 3 480-502 466 -  491 466.480 -  491
infections
O ther infectious and 1 -1 2 . 14 -27 . 2 -3 6 . 3 8 -4 3 9 -1 1 1 , 0 - 7 .  13 -83 , 0 - 7 .  13 -8 3 .
parasitic diseases 30. 3 4 -4 4 , 177 117-138 85-1 3 6 8 5-1 3 9
M alnutrition 6 7 -7 1 . 73 6 4 -6 5 286-293 2 6 0 -9 , 2 8 0 -5 2 6 0 -9 . 2 8 0 -5
M aternal 141-150 115-120 640 -  689 630 -678 630-678
Digestive cancer 46 4 5 -4 8 150-159 150-159 150-159
Respiratory cancer 47 4 9 -5 0 160-163 160-163 160-163
U terus cancer 48 5 2 -5 3 171-174 180-182 179-182
O ther cancer 45. 4 9 -5 7 . 74 4 4 .5 1 ,5 4 -6 0 140 -  9, 164-170. 14 0 -9 , 170 -9 . 1 4 0 -9 .1 7 0 -8 , 
175-239 183 -  239 183-239
Cardiovascular 58, 8 3 .9 0 -1 0 3 70. 7 9 -8 6 330 -  34 .4 0 0 -6 8 3 9 0 -4 5 8 390-459
Diabetes 61 63 260 250 250
Cirrhosis 77.124 105 581 571 571
A utom obile accidents 170 138 810-835 810-825 810-825
O th e r accidents and 163 -9 . 171-6 . 139-150 800 -  9 .840  -  999 800 -  7. 830 -  999 800 -  7, 8 3 0 - 999
violence 178-195
Ill-defined, senility 162. 199. 200 136-137 790-795 790-795 790 -  795
years as denominators. Child mortality rates were estimated from the census 
data on the proportions of surviving children.
The analysis considers the following explanatory variables:
(a) Socio-economic development, as estimated by an index computed from 
a linear combination of census data on seven indicators. These indicators with 
their weights indicated in parentheses are household income (1), proportion of 
non-agricultural activities (1), urbanization (1), women in the labour force (3), 
literacy (3), school attendance (2), and school attainment (2). The weights 
were determined with exploratory factor analysis. Detailed information from 
this index and the definition of the 100 counties is presented elsewhere 
(Rosero-Bixby, 1991b: ch. V).
(b) Social security, an indicator of access to medical services, is the pro­
portion of population covered by the social security system according to the 
1973 and 1984 censuses.
(c) Medically assisted deaths, a proportion computed from vital statistics.
(d) Primary health care services, as estimated by the proportion of popu­
lation living in areas administratively covered by the rural and community 
health programs in 1984 (these programmes started in 1972).
172 Luis Rosero-Bixby
(e) Improvement in access to secondary health care, as estimated by the 
proportion of population living in the catchment areas of health centres or 
clinics opened between 1970 and 1983.
(f) Travel time to San José in 1970 and 1984, an indicator of access to health 
and other services only available in the capital, estimated as the average of the 
times for each census tract, as assessed by supervisors of the 1984 census 
(unpublished information).
Multiple regression models for this areal data set are estimated with Gener­
alized Least Squares, using as a weighting variable the square root of the 
population (young or older adults) in the county. This procedure purges the 
distortion caused by the different variance of residuals due to the different 
demographic size of counties (Hanushek and Jackson, 1977: 150-68). The 
regression models were estimated using the computer package SHAZAM.
4. Level and Trend in Adult Mortality
In 1920, a 20-year-old Costa Rican had a 40 per cent chance of dying before 
reaching age 50; in 1990 the risk was only 6.1 per cent for men and 3.3 per cent 
for women, a fall of more than 80 per cent (Table 9.3). This decline is of similar 
magnitude to that observed in infant mortality, which fell from 200 to 16 per 
1000 during this period (Rosero-Bixby and Caamaño, 1984). Young adult 
mortality diminished steadily during these 70 years except in the period 1960-
5, when men’s mortality increased slightly. The fastest decline occurred in the 
1950s when Costa Rica benefited from discoveries associated with World War 
II, notably antibiotics and DDT. The risk of dying for young adults diminished 
by about 5 per cent a year during that decade. The second fastest decline took 
place in the 1940s, and was probably conditioned by factors similar to those 
operating in the 1950s. In addition, the 1970s were highly advantageous for 
women, probably due to improved reproductive health conditions.
The decline in the risk of dying at older ages (50 to 79 years) was not as 
dramatic as that at younger ages, but was still substantial. The probability of 
dying in this age interval was 54 per cent among men and 40 per cent among 
women in 1990, about 40 per cent lower than 70 years earlier (Table 9.3). The 
fastest decline in mortality at older ages took place in the late 1980s, which was 
a period characterized by adverse socio-economic conditions and reductions in 
expenditure on public health. The early 1970s is another period of accelerated 
reduction in the mortality of this age group (Table 9.3). The temporal se­
quence of mortality decline among older adults is thus different from that 
among young adults. There is, however, some coincidence in the cohort se­
quence of the transition. The beneficiaries of the accelerated decline at older 
ages in the 1980s were the same cohorts that experienced a fast decline at 
young adult ages in the 1950s.
The widening of the sex differential in adult mortality, especially among
Adult Mortality Decline in Costa Rica 173
Table 9.3. Risk of dying among adults in the age groups 20-49 and 50-79 
years, Costa Rica 1920-90
Year Mortality per 1000 Sex ratio Annual change (%)
Male Female Male Female
Young adults (20-49 years)
1920 402 404 0.99 -4.2 -4.2
1930 264 265 0.99 -1.0 -1.1
1940 239 238 1.00 -4.3 -4.8
1950 155 148 1.05 -4.5 -5.1
1960 99 89 1.11 0.7 -1.8
1965 102 81 1.25 -1.1 -3.9
1970 96 67 1.43 -1.8 -4.3
1975 88 54 1.62 -2.9 -4.8
1980 76 43 1.78 -3.6 -3.6
1985 64 36 1.79 -0.9 -1.4
1990 61 33 1.84
Old adults (50-79 years)
1920 858 853 1.01 -0.7 -0.9
1930 797 782 1.02 0.0 -0.2
1940 793 771 1.03 -0.5 -0 .9
1950 751 704 1.07 -0.8 -1.6
1960 690 598 1.15 -0.2 -0.3
1965 684 588 1.16 -0.7 -1.0
1970 660 560 1.18 -1.2 -1.9
1975 621 511 1.22 -0.7 -1.4
1980 601 475 1.27 -0.7 -1.3
1985 579 446 1.30 -1.5 -2.1
1990 538 401 1.34
Source: updated from Rosero and Caam ano (1984).
young adults, is a noteworthy feature of the mortality transition in Costa Rica 
and elsewhere. Until 1940 there was almost no difference in the risk of dying 
by sex. Since then the pace of mortality decline has been slower for men, 
raising their relative risk of dying in respect to women. In 1990, men had a risk 
of dying 84 per cent higher than women at young adult ages and 34 per cent 
higher at older adult ages.
How does the adult mortality transition in Costa Rica compare with other 
countries? The contrast with the United States suggests that the decline in 
Costa Rica has been extraordinary, particularly among men (Figure 9.1). In 
1920, the risk of dying of young Costa Rican adults, men and women, was 
double that of US citizens. It was about 16 per cent higher at older adult ages. 
Costa Rican men closed this gap by 1960 and nowadays they have a 21 per cent
174 Luis Rosero-Bixby
AGES 20-49
AGES 50-79
Year
Fig. 9.1. Risk of dying in the age groups 20-49 and 50-79. Costa Rica and the 
USA, 1920-90
Source: Costa Rica: Table 9.2; US: National Centers for Health Statistics.
lower risk of dying than their US counterparts at young adult ages and 7 per 
cent lower risk at older ages (Figure 9.1). The mortality decline for Costa 
Rican women was relatively less steep (in spite of being faster than that of 
men), but it was fast enough for women to catch up with their US counterparts 
in the 1980s. At present, adult women in Costa Rica and the US experience 
approximately the same risks of dying.
A comparison with two European populations—France and the Czech Re­
public—suggests once again that the adult mortality decline in Costa Rica, 
especially among men, has been exceptional (Figure 9.2). In 1985, men’s life 
expectancy at age 40 in Costa Rica was five years longer than in the Czech 
Republic and one year longer than in France, whereas that of women was two 
years longer than in the Czech Republic but about two years lower than in
Adult Mortality Decline in Costa Rica 175
Fig. 9.2. Life expectancy at age 40 in Costa Rica, the Czech Republic and 
France, 1950-85
Source: Costa Rica: updated from Rosero-Bixby and Caamano (1984); France and 
the Czech Republic: Rytchtarikova et al. (1990).
France. Figure 9.2 also illustrates that continuous progress in contemporary 
adult mortality is by no means the rule. In the Czech Republic, as in the rest 
of Eastern Europe and the former Soviet Union, adult mortality has stalled or 
deteriorated since the 1960s (Bourgeois-Pichat, 1985).
Another feature of adult mortality trends in Costa Rica is that the widening 
of the sex-gap has been less severe than in industrialized countries. Although 
progress in reducing adult male mortality in Costa Rica has been slower than 
for adult women, it was substantially faster than in industrialized countries. 
Conversely, although mortality decline among Costa Rican women has been 
substantially faster than among men, it has not been fast enough to catch up 
with the levels prevailing in industrialized countries.
5. Causes of Death
A salient aspect of the epidemiologic transition is the shift in cause of death 
patterns from infection and malnutrition to degenerative and man-made con­
ditions as the leading causes of death (Omran, 1982). Mohs (1991) contrasts
176 Luis Rosero-Bixby
two paradigms in the explanation of this shift, the “malnutrition paradigm 
which emphasizes the improvement in living conditions, and the ‘infectious 
diseases paradigm’ which underscores the role of health interventions, es­
pecially in the lagged transitions of the less developed countries. How well 
does the Costa Rican cause-of-death pattern of decline in adult ages fit this 
picture of the epidemiologic transition?
Data available for the period since 1951 show that the cause of death profile 
has changed in Costa Rica in the expected direction: the importance of infec­
tions and malnutrition has declined, whereas that of degenerative and man- 
made conditions has increased (Figure 9.3). More specifically, the place left 
vacant by the decline in infectious and deficiency diseases was taken largely by 
accidents and violence at young adult ages (38 per cent of deaths in 1989-90) 
and by cardiovascular diseases at older adult ages (39 per cent of deaths in 
1989-90). It is important to point out, however, that the contribution of 
infectious and deficiency diseases to adult mortality was somewhat limited in 
the past: 40 and 19 per cent of deaths in the age groups 20 to 49 years and 50 
to 79 years, respectively, in 1951-2.
The proportion of ill-defined causes of death (including those classified as 
‘senility’) gives an idea of the quality of cause of death data. This proportion 
was 9 and 12 per cent at young and older ages respectively in 1951-2, which is 
unusually low for a developing country four decades ago. Reflecting an im­
provement in data quality, the group of ill-defined causes declined to 2 per 
cent of deaths in 1989-90 in both age groups, a figure that is consistent with the 
fact that 96 per cent of deaths in Costa Rica were medically certified in 1990. 
The reduction in the proportion of ill-defined causes of death, in spite of being
Accidents and 
Cancer violence □  Ill-defined 
Infection and Cardiovascular and 
nutrition diabetes Other
100 r
1951/2 1971/2 1989/90 1951/2 1971/2 1989/90 
Young adults (20-49) Old adults (50-79)
Fig. 9.3. Distribution of deaths by groups of causes: Costa Rica, adults, 1951-90
Adult Mortality Decline in Costa Rica 177
a favourable trend, is a nuisance for the study of cause-specific mortality over 
time. It is possible, for example, that a part of the decline in infections and 
malnutrition has been blurred by improved diagnosis, or that a part of the 
increase in the contribution of cardiovascular diseases has been an artifact of 
better diagnoses.
Figure 9.4 shows the evolution in the risks of dying from selected groups of 
causes of death, as defined in Table 9.2. The upper part of the figure presents 
conditions of infectious or nutritional origin. The dramatic declines in respir­
atory tuberculosis at young adult ages during the 1950s and in acute respira­
tory infections during the 1970s stand out. The virtual elimination of malaria 
in the 1950s was a notable achievement in both age groups. The reduction in 
maternal mortality in the 1960s, probably a byproduct of the fertility transi­
tion, was another noteworthy trend. The risk of dying from diarrhoeal diseases 
and nutritional conditions exhibits a sustained decline during the period, al­
though it is not as fast as the aforementioned reductions. The only clear 
increase in mortality from diseases of infectious origin is for acute respiratory 
infections at older ages between 1951 and 1971. The risk of dying from these 
conditions rose from 33 to 60 per 1000 in these twenty years. Part of this 
increase could be an artifact of changes in diagnostic practices.
The lower part of Figure 9.4 shows the evolution in the risk of dying from 
degenerative and man-made conditions. There are several negative trends. 
The most serious are the increase in automobile accidents, cardiovascular 
diseases, and diabetes from 1951-2 to 1971-2. Mortality from automobile 
accidents at young adult ages underwent a seven-fold increase in this period, 
whereas that caused by diabetes at older adult ages underwent a three-fold 
increase. Mortality from respiratory cancer (chiefly lung cancer) also increases 
substantially, particularly for young adults in the 1960s and for older adults in 
the 1970s. The risk of dying from respiratory cancer in 1989-90 is three and 
two times greater than in 1951-2 at young and older adult ages, respectively.
The most notable trend in this group of causes of death was the sharp 
decline in cardiovascular mortality in the 1970s (Figure 9.4). The decline was 
one of 38 per cent and 25 per cent at young and older adult ages respectively 
during the decade. Major declines in diabetes, accidents and violence (includ­
ing automobile accidents) also took place during this decade. The large re­
duction in mortality from cancer of the uterus (chiefly cervical neoplasms) in 
the 1960s has been linked to screening programmes implemented, in part, 
along with family planning services (Rosero-Bixby and Grimaldo, 1987). In 
turn, the decline in digestive cancer (chiefly stomach cancer) has a special 
significance, due to its extraordinarily high incidence in Costa Rica. The risk of 
dying from digestive cancer was 46 per cent lower in 1989-90 than in 1951-2 
at young adult ages, and 18 per cent lower than in 1961-2 at older adult ages.
Table 9.4 shows the estimated contributions of each cause of death to the 
overall mortality decline. This ‘attributable decline’ depends on both the pace 
of change in a particular cause and its baseline level. About three-quarters of
178 Luis Rosero-Bixby
YOUNG ADULTS (20-49) OLD ADULTS (50-79)
1950 1960 1970 1980 1990
20 r
15 Cardiovascular^--'^
OOO ■------- *  \
Accidents & violence 'y .
GQ).
05 10
*o
o
Diabetes
Accidents S
^Uterine Cancer 
—̂  Respir*10̂  Canea*
«Respira*03 ------- -------- Ul Uterine Cancer 
1950 1960 1970 1980 1990 1950 1960 1970 1980 1990
Fig. 9.4. Risk of dying by selected causes of death: Costa Rican adults, 1951-90 
Source: Table 9.2 and 9.A1.
Adult Mortality Decline in Costa Rica 179
Table 9.4. Attributable decline in adults’ risk of dying in 1951-71 and 1971-89
Cause of death Percent attributable decline
Ages 20 -49 Ages 50-79  
1951-71 1971-89 1951-71 1971-89
Respiratory TB 37.0 4.4 16.5 4.1
Malaria 7.9 0.1 29.7 0.1
Diarrhoeal diseases 3.0 1.1 20.0 0.6
Acute respiratory infections -0.1 8.7 -44.8 22.1
Other infections and parasitic 12.9 7.0 34.0 3.0
Malnutrition 6.3 2.3 17.2 3.4
Maternal 9.4 5.5 0.0 0.0
Subtotal: infection-nutrition 76.4 29.1 72.7 33.3
Digestive cancer 4.4 3.5 -19.4 1.2
Respiratory cancer -1.2 0.7 -3.8 -3.5
Uterus cancer 3.1 1.9 2.3 1.7
Other cancers 3.9 2.7 11.5 4.6
Subtotal: cancer 10.2 8.8 -9.3 4.1
Cardiovascular -1.5 23.4 -108.6 45.5
Diabetes -1.7 0.7 -38.4 7.4
Subtotal: cardiovascular-diabetes -3 .2 24.1 -147.0 52.9
Automobile accidents -12.1 4.7 -12.9 1.1
Accidents and violence -0.5 8.4 12.3 -2.8
Subtotal: accidents-violence -12.6 13.1 -0.6 -1.7
Cirrhosis 1.5 -2.4 -1.7 -2.4
Ill-defined, senility 15.0 11.0 92.1 19.7
Residual 12.7 16.4 93.8 -5.9
Total 100.0 100.0 100.0 100.0
Source: Table 9.A1.
the mortality reduction at both young and older ages from 1951-2 to 1971-2 
can be attributed to the control of infectious diseases and malnutrition. This 
contribution declined to about 30 per cent in the 1970s and 1980s. In this later 
period, cardiovascular diseases accounted for approximately one-quarter of 
the decline at young adult ages and one-half at older adult ages. An increase 
in accidents and violence counteracted the adult mortality decline in the 1950s 
and 1960s by 13 per cent, whereas in the 1970s and 1980s external causes 
contributed 13 per cent to the decline. The contribution of tuberculosis and 
malaria to the early mortality transition is outstanding. The control of these 
two diseases alone explains about 45 per cent of the adult-mortality decline 
from 1951-2 to 1971-2 in the two age groups studied. A 9 per cent contri-
180 Luis Rosero-Bixby
bution of maternal deaths to the decline in young adult mortality in this early 
period is also noteworthy.
A large decline of ill-defined causes of death, which is almost as large as the 
decline in all other causes, confounds analysis of the components of the early 
mortality transition in the older age group (Table 9.4). Part of the negative 
contributions of cardiovascular diseases (-109 per cent), acute respiratory 
infections (-45 per cent), diabetes (-38 per cent), and digestive cancer (-19 
per cent) is probably due to the reduction in ill-defined causes of death. 
However, this artifact could only explain a fraction of these negative contri­
butions (44 per cent at the most). It seems that the increases in these causes of 
death and automobile accidents diminished by more than one-half the poten­
tial mortality decline brought about by other causes. This adverse effect may 
be ascribable to the negative influence of economic development and afflu­
ence: the increase in automobile accidents is a clear example of this. Part of the 
increase in diabetes, cardiovascular diseases, respiratory cancer and, perhaps, 
respiratory infections can be linked to modem lifestyles, including obesity, 
smoking, sedentary habits, and consumption of animal fats. It seems, however, 
that these deleterious consequences of progress were neutralized in the 1970s 
and 1980s.
Data by sex on the cause-specific mortality (Table 9.5), are only available 
for the period since 1971. They corroborate the aforementioned trends, such 
as the large contribution of cardiovascular disease to the mortality decline. At 
young adult ages, the most important factor in the mortality decline among 
men was the control of accidents and violence (23 per cent contribution). 
These causes of death made almost no contribution to the decline among 
women. Cardiovascular diseases played a more important role in women (29 
per cent) than in men (18 per cent). Maternal mortality was the second most 
important component (12 per cent contribution) among women. At older 
adult ages, the contribution of digestive and respiratory cancer exhibits major 
sex-differences. Mortality from these tumours increased among men and made 
a negative contribution to the transition. Among women, there was almost no 
increase in respiratory cancer and digestive cancer decreased. Therefore, the 
widening of the sex differential in mortality at young ages was a result of the 
reduction in maternal mortality and a more rapid decline of cardiovascular 
mortality among women. At older adult ages, the increasing differential re­
sulted largely from an increase in deaths from respiratory and digestive cancer 
among men.
How does the cause of death profile of Costa Rican adults compare with that 
of other populations? A comparison with Argentina and Chile has shown that 
infectious and parasitic diseases mortality (including pneumonia and 
bronchitis) is relatively low in Costa Rica, whereas men’s mortality from 
accidents is higher (Arriaga, 1991; Table 9.4). Other international comparisons 
single out Costa Rica for its high mortality from stomach cancer and accidents 
and its relatively low mortality from lung cancer and cardiovascular diseases 
(Brouard and Lopez, 1985; Mesle, 1985).
Adult Mortality Decline in Costa Rica 181
Table 9.5. Risk of dying by sex for 17 groups of causes of death, young and older 
adults, Costa Rica, 1971-90
Cause of death Male risk per 1000 Fem ale risk per 1000 A ttributable 
decline 
1971-2 1981-2 1989-90 1971-2 1981-2 1989 -  90 1971-89 (% )
Males Females
Young adults (20 -4 9 )
Respiratory 1.7 1.1 0.5 1.7 0.6 0.2 3.6 5.3
tuberculosis
D iarrhoeal 0.5 0.1 0.1 0.3 0.1 0.1 1.5 0.6
diseases
Acute respiratory 3.6 1.0 0.7 2.8 0.7 0.5 8.9 8.6
infections
O ther infections 2.8 1.1 1.0 3.1 0.8 0.7 5.4 8.8
and parasitic
M alnutrition 1.1 0.2 0.3 0.8 0.4 0.2 2.6 2.0
M aternal 0.0 0.0 0.0 4.0 1.2 0.7 0.0 11.9
Digestive cancer 6.2 4.3 4.5 3.4 3.8 3.0 5.3 1.4
Respiratory cancer 1.0 1.2 0.5 0.6 0.5 0.6 1.5 -0.2
U terus cancer 0.0 0.0 0.0 3.7 3.0 2.6 0.0 3.9
O ther cancer 4.8 4.5 4.4 7.2 5.9 5.9 1.3 4.5
Cardiovascular 15.8 10.8 9.8 14.3 7.9 6.3 18.5 29.0
D iabetes 0.9 0.9 1.0 1.8 1.1 1.3 -0.5 2.0
Cirrhosis 2.4 3.3 3.5 1.0 0.9 1.3 -3.4 -1.2
A utom obile 12.0 8.4 9.5 1.7 1.5 1.5 7.9 1.0
accidents
O ther accidents. 22.5 18.3 17.7 3.2 2.4 3.1 15.1 0.7
and violence
Ill-defined, senility 5.8 1.9 1.4 2.7 2.2 0.4 13.4 8.1
Residual 13.9 7.4 7.8 10.2 7.0 6.5 18.7 13.7
Total 91.4 62.7 61.2 60.8 39.3 34.3 100.0 100.0
Old adults (50 -79 )
Respiratory 18.4 8.4 7.9 9.3 3.6 2.6 5.7 3.0
tuberculosis
D iarrhoeal diseases 4.4 3.4 3.6 4.8 1.8 3.3 0.4 0.6
A cute respiratory 61.4 22.6 17.2 59.0 16.9 13.3 24.6 20.6
infections
O ther infections. 15.9 8.4 9.0 9.9 7.1 4.2 3.8 2.5
and parasitic
M alnutrition 11.3 4.7 3.9 10.1 2.1 3.0 4.0 3.1
Digestive cancer 99.5 102.1 108.9 67.5 52.4 55.2 -5.6 5.7
Respiratory cancer 17.1 30.0 30.3 8.3 11.6 10.3 -7.2 -0.9
U terus cancer 0.0 0.0 0.0 21.2 16.5 14.4 0.0 3.0
O ther cancer 72.0 63.4 63.0 52.4 42.7 43.3 5.1 4.2
Cardiovascular 299.0 233.8 231.4 245.6 174.5 166.0 49.0 43.4
Diabetes 27.5 13.7 14.1 41.9 22.5 25.1 7.3 7.5
Cirrhosis 12.6 12.7 18.5 5.5 9.5 9.4 -3.2 -1.7
Autom obile 16.8 15.7 13.2 4.8 1.9 3.6 2.0 0.5
accidents
O ther accidents. 20.4 25.9 28.7 6.2 9.9 9.6 —4.6 -1.5
and violence
Ill-defined, senility 60.8 53.3 14.4 42.4 42.4 7.7 25.7 15.4
Residual 98.6 101.0 110.3 71.9 72.3 82.9 -7 .0 -5.3
Total 597.6 526.5 514.6 507.3 400.5 378.9 100.0 100.0
182 Luis Rosero-Bixby
Table 9.6 compares the risks of dying from selected causes in Costa Rica and 
the USA in the late 1980s. Heart disease and respiratory cancer are the key to 
the comparatively low mortality of Costa Rica (see risk-differences in Table 
9.6). Consumption of cigarettes is substantially lower in Costa Rica than in the 
USA and is a likely explanation of some of this differential (Ravenholt, 1990). 
Other plausible explanations of lower heart disease mortality in Costa Rica 
are a less stressful lifestyle, less sedentary habits, particularly in rural areas, 
less fat and protein in the diet, and a lower prevalence of obesity among men. 
The sharp decline in cardiovascular mortality in both Costa Rica and the USA
Table 9.6. Risk of dying at ages 25-74 years for selected causes of death, 
comparison between Costa Rica, 1988 and the USA, 1987
Cause of death (ICD Codes8) Mortality per 1000
CR-1988 US-1987 Difference Ratio
CR-US CR/US
Males
Heart disease (390-429) 117.8 204.5 -86.7 0.58
Respiratory cancer (162) 18.8 68.5 -49.7 0.27
Other cancer (140-208, 72.9 101.1 -28.2 0.72
exc. 151, 162)
Diabetes (250) 11.3 10.7 0.6 1.05
Other accidents and violence 38.5 35.2 3.3 1.09
(800-9, 820-999)
Cirrhosis (571) 17.7 14.1 3.6 1.25
Automobile accidents (810-819) 18.4 12.7 5.7 1.45
Stroke (430-438) 34.0 26.6 7.4 1.28
Stomach cancer (151) 47.8 5.8 42.0 8.30
Females
Heart disease (390-429) 71.7 99.9 -28.2 0.72
Respiratory cancer (162) 5.7 28.5 -22.9 0.20
Other cancer (140-208, exc. 67.6 86.5 -19.0 0.78
151.162.180)
Other accidents and violence 9.2 10.9 -1.7 0.85
(800-9, 820-999)
Automobile accidents (810-819) 3.9 5.3 -1.5 0.72
Cirrhosis (571) 6.8 6.2 0.6 1.09
Stroke (430-438) 27.0 21.1 5.9 1.28
Cervical cancer (180) 11.0 2.8 8.3 3.96
Diabetes (250) 20.0 9.7 10.3 2.06
Stomach cancer (151) 18.5 2.4 16.1 7.56
"International Classification of Diseases (1CD). 9th Revision. 
Source: PA H O , 1990. Vol. 1. Tables III-9.
Adult Mortality Decline in Costa Rica 183
in the 1970s and 1980s is an encouraging development, suggesting that 
progress need not necessarily result in an increase in such mortality. The 
future of mortality from respiratory cancer in Costa Rica is less appealing. An 
increase in smoking in young generations in the 1960s, which probably took 
place because of the boom in television and communications, will have an 
impact on mortality from respiratory cancer thirty to fifty years later, that is 
after 1990.
In the group ‘other cancer’ Costa Ricans are again fortunate (Table
9.6). The differences from the USA under this rubric come mostly from 
bladder and prostate cancer in men and breast cancer in women. The future 
evolution of bladder and prostate cancer in Costa Rica is uncertain, but 
breast cancer will probably increase substantially. A slight increase of 
breast cancer mortality has already been observed and a substantial 
increase (32 per cent in twenty years) in the incidence of this cancer has been 
projected just as a consequence of past fertility reductions (Rosero-Bixby et 
al., 1987).
Mortality from stomach cancer is eight times higher in Costa Rica than in 
the USA. Nevertheless, an encouraging decline has already occurred in Costa 
Rica and will probably continue in the future. Stomach cancer mortality has 
been decreasing dramatically in the USA (more in women than in men) for 
several decades: in 1973-4, this rate was one-fifth its level in 1935 among white 
females (Devesa and Silverman, 1978; Table 9.4). Although the causes of this 
decline are still unknown, it seems reasonable to expect a similar trend in 
Costa Rica.
The comparison with the USA also suggests that Costa Rica could achieve 
important mortality reductions from the death rates due to stroke in both 
sexes, automobile accidents in men, and diabetes and cervical cancer in 
women. Several technologies are already available to intervene against these 
causes of death.
6. Areal Variation in Adult Mortality
The data examined so far have shown puzzling relationships between 
socio-economic development, expenditure on health services and adult-mor- 
tality transition in Costa Rica. National indicators of adult mortality, socio­
economic progress, and health interventions follow disparate trends. 
Moreover, adults in Costa Rica face similar, or even lesser, mortality risks than 
their counterparts in the USA, despite the enormous differences between 
them in income, social organization, and availability of health facilities. The 
analysis of causes of death offers hints about the forces that reduced adult 
mortality in Costa Rica but also raises questions regarding the possible deter­
minants of cardiovascular mortality trends. An examination of areal variation 
in adult mortality and its correlates might provide additional clues about the
184 Luis Rosero-Bixby
role that socio-economic development and health interventions play in deter­
mining adult mortality in Costa Rica.
The areal (or ecological) analysis uses a territorial division of Costa Rica 
into 100 ‘counties’ defined for other purposes (Rosero-Bixby, 1991¿>). Four- 
fifths of the counties are in the range of 6000 to 60.000 population. The analysis 
covers two three-year periods centred on the census years 1973 and 1984. The 
adult mortality rates at young ages (20 to 49 years) and older ages (50 to 79 
years) are analysed separately. Three mortality rates are studied in each age 
group: all causes of death; infectious diseases and malnutrition; and cardio­
vascular diseases and diabetes. It is important to keep in mind the considerable 
random variation existing in these county-level rates because of the small 
numbers involved, particularly in the young adult age group.
Are there geographical regularities in adult mortality in Costa Rica? The 
choropleth map of cardiovascular mortality at older ages in 1973 (Figure 9.5) 
suggests geographical clustering: counties with higher cardiovascular mortality 
tend to line up along an interoceanic axis between the ports of Puntarenas and 
Limón, including the Central Valley and the capital city, San José. It is pre­
cisely along this axis that Costa Rica traditionally concentrated its socio­
economic and demographic development (Hall, 1985). The map therefore 
suggests a correspondence between cardiovascular mortality and develop­
ment. Other choropleth maps (1984, all causes of death, young adult ages) 
present less clear geographical clustering. Maps (not shown) for the young 
adult ages and infection-malnutrition conditions, in particular, resemble a 
chess board, without spatial regularities.
Figure 9.5 also includes a map describing the pace of decline in cardio­
vascular mortality between 1973 and 1984. The geographical pattern is less 
clear in this than in the 1973 mortality level map. Nevertheless, a ‘regression to 
the mean’ phenomenon is clearly present: those counties with the lowest rates 
in 1973 tend to be the ones where the rate increases most in the following 
decade.
Pearson correlation coefficients (r) can be used to check the existence of 
ecological covariations between adult mortality in Costa Rican counties 
and their socio-economic and health status (Table 9.7). To illustrate, correla­
tion coefficients smaller than 0.20 in absolute value indicate lack of, or non­
significant, association; coefficients between 0.20 and 0.39 indicate a modest 
but significant association; and coefficients of 0.40 or larger indicate a close 
association. Mortality rates at young and older adult ages are weakly corre­
lated (Table 9.7). The largest correlation coefficient between young and adult 
mortality is a modest 0.31 for infections and malnutrition in 1973. In addition, 
neither young nor older adult mortality rates are correlated meaningfully with 
infant mortality. The only two correlations with infant mortality of a signifi­
cant magnitude are negative in sign. Mortality from cardiovascular disease 
tends to be high in counties with low infant mortality in 1973, a pattern 
consistent with that shown in Figure 9.5. The lack of a positive association
(a) Mortality rate in 1973 (b) Proportional decline 1973-84
30 Km
Fig. 9.5. Level and change in the mortality rate from cardiovascular diseases and diabetes at ages 50-79
Adult Mortality Decline in Costa Rica
186 Luis Rosero-Bixby
Table 9.7. Relationship between adult mortality and selected variables, 100 Costa 
Rican counties, 1973 and 1984
Variable3 All causes Infection- Cardiovascular-
nutrition
1973 1984
1973 1984 1973 1984
Correlations between young adult mortality and: 
Mortality among:
Older adults 0.18 -0.06 0.31 -0.08 0.21 0.22
(50-79 years)
Infants 0.06 0.04 0.14 0.08 -0.10 -0.02
Socioeconomic
development:
10 years earlier 0.12 0.02 0.16 0.04 0.29 0.22
contemporary 0.11 0.01 0.16 0.04 0.32 0.23
Social security 0.14 -0.06 0.10 0.03 0.33 0.06
Medically assisted 0.08 -0.18 0.20 -0 .08 0.38 -0.09
deaths
Primary care — -0 .05 — -0 .03 — -0.12
Travel time to -0.3  0.08 -0.02 0.17 -0.27 -0.14
San José 
Correlations between older adult mortality and:
Mortality among:
Young adults 0.18 -0.06 0.31 -0.08 0.21 0.22
(20-49 years)
Infants -0.31 0.01 0.15 -0.01 -0.35 -0.06
Socioeconomic
development:
10 years earlier 0.52 0.48 0.19 0.28 0.60 0.59
contemporary 0.49 0.46 0.22 0.26 0.58 0.59
Social security 0.53 0.43 0.21 0.29 0.59 0.57
Medically assisted 0.48 0.42 0.17 0.11 0.67 0.54
deaths
Primary care — -0.24 — -0.17 — -0.24
Travel time to -0.48 -0.44 -0.17 -0.21 -0.51 -0.53
San José
"For definitions see text.
between adult and infant mortality suggests that conclusions from previous 
studies about the determinants of the breakthrough in the Costa Rican infant 
mortality in the 1970s cannot be extrapolated to explain the adult mortality 
transition.
Adult Mortality Decline in Costa Rica 187
Young adult mortality shows almost no association with indicators of de­
velopment and health services (Table 9.7). In contrast, older adult mortality, 
especially cardiovascular deaths, correlates closely with socio-economic and 
health indicators, with mortality tending to be lower in backward counties and 
areas farther away from San José. These inverted correlations are modest for 
death rates linked to infection and nutrition (about 0.20 to 0.30). whereas they 
are substantial (0.50 to 0.60 at older ages) for cardiovascular diseases and 
diabetes.
Data errors, such as differential integrity in the registry of deaths or biased 
reports of the place of residence of the dead, are unlikely explanations for the 
odd behaviour of adult mortality rates, since the problem is not present in 
infant mortality rates based on the same data source. For example, an earlier 
study shows that the infant mortality rate in 1972-5 and 1982-4 was negatively 
and significantly correlated with such characteristics as the proportion of 
deaths medically-certified, coverage of social security, and proportion of 
dwellings with plumbing (Cervantes and Raabe, 1991; Tables 9.1 and 9.2).
One plausible explanation of the inverse association is reverse causality 
between adult mortality and place of residence: sick adults may tend to mi­
grate to places with better health facilities and higher levels of development, 
artificially increasing mortality rates in the better-off counties. The targeting of 
health interventions on backward areas could also generate reverse causality 
since high mortality counties would appear as being better-off in terms of 
health services. Another explanatory mechanism is the existence of ‘frailty’, or 
selection effects: backward counties may have low adult mortality rates be­
cause historically high child mortality eliminated the frail children in each 
cohort, leaving alive a select group of more resistant individuals (Vaupel et al., 
1979). A more straightforward explanation is that economic progress and 
modern lifestyles indeed raise adult mortality, particularly that of cardiovascu­
lar origin.
A further step in the analysis is modelling mortality decline, rather than its 
level, as a function of changes in the potential explanatory variables. An 
analysis based on changes rather than levels has advantages like controlling 
the effect of unmeasured variables (e.g. the ‘targeting’ problem); in addition, 
some data errors neutralize each other by the computation of changes (Liker 
et al., 1985). Random and other errors, however, can be magnified by the 
computation of changes, and the ‘regression to the mean’ phenomenon can 
introduce considerable noise into the variance of changes (Bohmstedt, 1969). 
Due to these problems, the magnitude of correlations for changes is usually 
lower than for levels (Freedman and Takeshita, 1969).
Table 9.8 shows the results of estimating, with generalized least squares, 
multiple regression models of the percentage decline in mortality during 1973- 
84 for: young adults, all causes of death; older adults, all causes; older adults, 
infection and malnutrition; and older adults, cardiovascular diseases and
188 Luis Rosero-Bixby
Table 9.8. Standardized regression coefficients (t-ratios) on the relative decline in 
adult mortality from 1972-4 to 1983-5, 100 Costa Rican counties
Explanatory variables Young adults Older adults (50-79 years)
(20-49 years)
All causes Infection- Cardio­
nutrition vascular
Initial mortality level (1972-74) 0.159 -0.059 0.158 0.007
( 1.45“) (-0.55) (1.54a) (0.07)
Pace of socioeconomic 
development 
A decade earlier (1963-73) 0.102 -0.040 0.219 -0.028
(0.81) (-0.32) (1.80b) (-0.23)
Contemporary (1973-84) -0.002 -0.014 0.027 -0.163
(-0.02) (-0-11) (0.22) (-1.32a)
Relative decline in travel -0.107 -0.091 -0.146 0.113
time to San José 1970-84 (-0.86) (-0.74) (-1.24) (0.93)
Increase in Social Security -0.099 -0.233 -0.158 -0.047
System coverage 1973-84 (-0.73) (-1.78») (-1.25) (-0.36)
Increase in the proportion -0.144 -0.086 -0.106 0.126
of MD assisted deaths (-0.96) (-0.74) (-0.94) (-1.10)
Increase in primary health 0.077 0.009 0.061 -0.153
care services 1972-84 (0.51) (0.06) (0.42) (-1.04)
Improvement in access to -0.055 -0.022 -0.055 -0.030
secondary' health care 1970-84 (-0.41) (-0.17) (-0.43) (-0.23)
Adjusted R2 0.033 0.006 0.059 0.014
“Significant at 0.90. 
h Significant at 0.95.
diabetes. Cause-specific mortality decline at young adult ages cannot be ana­
lysed because the small number of deaths involved yields highly unreliable 
estimates.
In the four multiple regression models in Table 9.8 only a few regression 
coefficients appear statistically significant. Initial mortality levels exert a posi­
tive and significant influence on the declines in young adult and in older adult 
mortality from infections and malnutrition. These significant effects are mani­
festations of the regression to the mean trend: mortality is easier to lower 
where it is high initially. The speed of socio-economic progress ten years 
earlier appears to be a significant factor in the speed of decline in older age 
mortality from diseases of infectious and nutritional origin, whereas contem­
porary progress exerts a negative influence on the decline in mortality from
Adult Mortality Decline in Costa Rica 189
cardiovascular disease. Social security coverage also has a significant negative 
effect on older age mortality from all causes.
A perturbing result in Table 9.8 is that the data fail to show any meaningful 
impact of interventions like the development of new clinics or the expansion of 
primary health coverage upon adult mortality. This is in sharp contrast with 
the findings of earlier studies that showed a clear linkage between these 
interventions and the infant mortality decline in the 1970s (Rosero-Bixby, 
1986).
7. Conclusions
Mortality at young adult ages has declined during the present century to a 
similar degree to infant mortality (80 per cent from 1920 to 1990). The pro­
portional decline at older adult ages was also substantial but represented only 
about one half of that at younger ages. Women’s mortality decreased more 
than men’s, giving rise to a widening sex differential. The most fruitful period 
in the control of young adult mortality was after the Second World War; at 
older adult ages it was the late 1980s. This exceptionally fast decline has 
allowed Costa Rica to close the gap with industrialized countries. At present, 
adult men in Costa Rica face lower risks of dying than men in countries such 
as the USA and France (women face similar or slightly higher risks than their 
counterparts of these nations).
It is hard to trace an association between the pace of adult mortality decline 
and trends in socio-economic development and in health interventions in 
Costa Rica. The decade of fastest economic growth and expansion of expendi­
ture on health was the 1960s. Most of the Costa Rican fertility transition also 
occurred in these years. It was precisely in this decade, however, when the 
mortality transition presented signs of stagnation. Conversely, it was in the 
decade of stalling, or even deteriorating, socio-economic conditions and public 
health interventions—the 1980s—that the greatest reductions in older adult 
mortality occurred. On the other hand, the most fruitful period for young adult 
mortality—the 1950s—was the decade with the fastest progress in educational 
enrolment and sanitation, as well as in the expansion of basic hospital care like 
institutional childbirth. The existence of cohort effects, long latency periods, 
and complex lagged reactions might be the cause for these puzzling temporal 
relationships. It is also possible that some degree of independence actually 
exists between adult mortality and the socio-economic and programmatic 
environment. In this case, the key factors explaining adult mortality trends 
might be the diffusion of innovations among health professionals and of life 
styles (good and bad) and health practices among the general population.
Areal analyses of adult mortality suggests that cardiovascular and diabetes 
mortality is higher in more prosperous communities. Moreover, the decline in 
such mortality between 1973 and 1984 correlated negatively—and signifi­
190 Luis Rosero-Bixby
cantly—with socio-economic improvement. In contrast, the decline in mor­
tality from infections and malnutrition was positively associated with ten-vear 
lagged improvements in well-being. This, of course, is not a conclusive proof 
that development has a negative effect on adult mortality. Bad data, frailty, 
and the reversed causal linkages discussed already could explain all or part of 
this correlation. A longitudinal study in selected communities would give more 
conclusive answers about the impact of progress and health interventions 
upon the components of adult mortality and, in particular, cardiovascular 
deaths. Nevertheless, the results of this somewhat simplistic exercise match 
other pieces of evidence, such as the higher mortality rates in the USA and the 
adverse trends in some causes of death in the 1950s and 1960s. They suggest a 
picture in which socio-economic development does not necessarily enhance 
survival chances at adult ages.
The analysis of mortality trends by cause of death gives additional insights 
into the existence of a dark side to the development of the epidemiologic 
transition. The period from 1951-2 to 1971-2 saw substantial mortality in­
creases from cardiovascular diseases, diabetes, automobile accidents, respira­
tory cancer, and respiratory infections. These probably offset more than one 
half the mortality decline generated by the control of other conditions. Im­
proved diagnoses alone can hardly explain these mortality increases. Some 
increases were probably linked to affluence and features of modem lifestyles 
like smoking, sedentary habits, obesity, consumption of animal fats, and an 
increase in the use of motor-vehicles. It seems, however, that these deleterious 
consequences of progress were largely neutralized during the 1970s and 1980s.
A comparison with causes of death in the USA showed that Costa Rica’s 
adult mortality is comparatively low because of substantially lower risks of 
dying from heart disease and respiratory cancer. Cigarette smoking, a habit 
associated with modernity, seems likely to play a central role in this difference.
The classic shift in the cause of death profile from infectious diseases and 
malnutrition to degenerative and man-made conditions is observed in the 
adult mortality transition in Costa Rica. At present, the leading causes of 
death are accidents and violence at young adult ages and cardiovascular dis­
eases at older adult ages (each represents almost 40 per cent of deaths in the 
corresponding age group). In the early stages of Costa Rica’s epidemiologic 
transition, the control of infections and malnutrition was the key factor of 
mortality decline. About three-quarters of the decline between 1951 and 1971 
at both young and older adult ages can be ascribed to infectious and nu­
tritional diseases. In particular, respiratory tuberculosis and malaria account 
for about 45 per cent of the decline in the two age groups, an outstanding 
contribution linked to the importation of cost-effective technological advances 
(chiefly streptomycin and DDT). The contribution of infections and malnu­
trition to the adult-mortality decline shrank to about 30 per cent during the 
period from 1971-2 to 1989-90. During this second stage, the control of 
cardiovascular diseases became the key factor in the decline at older ages,
Adult Mortality Decline in Costa Rica 191
whereas both cardiovascular disease and accidents and violence fuelled the 
decline at young ages.
Trends under way and a comparison with the epidemiological profile of the 
USA suggest that substantial adult reductions in mortality might come in the 
future from control of stomach cancer, as well as from mortality from stroke, 
automobile accidents, diabetes (particularly in women), and cervical cancer. 
In addition, it is an important challenge for Costa Rica to keep its comparative 
advantage in heart disease mortality. Increases must be expected in lung and 
breast cancer mortality as consequence of past rises in smoking and reduced 
fertility respectively.
Shifting the focus from infant to adult mortality has exposed new facets of 
the epidemiologic transition in Costa Rica. Trends in the pace of adult mor­
tality decline differ from those in infant mortality. Geographical patterns and 
ecological correlates are also different. The determinants of these disparate 
trends and geographical patterns are thus probably different as well. In par­
ticular, while the breakthrough in child survival in the 1970s can be ascribed 
largely to programmatic interventions, the explanation for the recent decline 
in adult mortality is more elusive.
Appendix 9.1
Table 9.A1. Risk of dying for 17 groups of causes of death, young and old adults, 
Costa Rica 1951-90
Causes of death Risk of dying per 1000
1951-2 1961-2 1971-2 1981-2 1989-91
Young adults (20-49)
Respiratory tuberculosis 19.94 3.37 1.70 0.86 0.38
Malaria 3.95 0.17 0.03 0.00 0.00
Diarrhoeal diseases 1.91 0.80 0.40 0.10 0.07
Acute respiratory infections 3.19 2.16 3.24 0.86 0.61
Other infections and parasitic 9.35 2.80 2.94 0.95 0.84
Malnutrition 4.06 1.01 0.95 0.28 0.26
Maternal 6.68 5.74 1.99 0.61 0.34
Digestive cancer 6.99 6.07 4.82 4.07 3.76
Respiratory cancer 0.20 0.29 0.79 0.85 0.58
Uterus cancer 3.42 4.06 1.88 1.49 1.31
Other cancer 7.87 6.37 5.94 5.20 5.14
Cardiovascular 14.31 14.08 15.06 9.37 8.09
Diabetes 0.51 0.67 1.36 0.97 1.15
Cirrhosis 2.43 1.02 1.69 2.10 2.41
Automobile accidents 0.92 2.79 6.94 4.97 5.52
Other accidents, and violence 12.74 12.99 12.98 10.44 10.47
Ill-defined, senility 11.67 4.92 4.23 2.03 0.92
Residual 18.29 12.92 12.05 7.20 7.14
Total 121.21 79.28 76.26 51.15 47.96
Old adults (50-79)
Respiratory tuberculosis 23.9 14.0 13.6 5.8 5.0
Malaria 18.7 0.5 0.2 0.0 0.0
Diarrhoeal diseases 17.1 15.1 4.6 2.6 3.4
Acute respiratory infections 33.1 47.4 60.0 19.6 15.1
Other infections and parasitic 33.7 17.5 12.7 7.7 6.4
Malnutrition 21.4 16.1 10.7 3.3 3.4
Digestive cancer 71.6 98.6 82.9 76.1 80.5
Respiratory cancer 10.1 10.4 12.5 20.3 19.7
Uterus cancer 12.6 15.8 11.2 8.7 7.6
Adult Mortality Decline in Costa Rica 193
Table 9.A1. Continued
Causes of death Risk of dying per 1000
1951-2 1961-2 1971-2 1981-2 1989-90
Other cancer 68.2 60.7 61.4 52.3 52.2
Cardiovascular 219.4 234.1 271.1 202.9 197.0
Diabetes 11.5 19.8 35.1 18.3 20.0
Cirrhosis 7.8 10.4 8.9 11.1 13.8
Automobile accidents 2.5 5.1 10.6 8.5 8.2
Other accidents, and violence 20.7 19.2 13.1 17.7 18.9
Ill-defined, senility 104.7 63.3 51.1 47.4 10.7
Residual 137.3 140.3 84.6 85.8 95.8
Total 579.6 569.7 552.2 463.3 446.3
References
Arriaga, E. E. (1991), ‘Similarities and differences in the adult mortality transition in 
developing countries’, IUSSP Seminar on Causes and Prevention of Adult Mortality 
in Developing Countries, Santiago, Chile, 7-11 October.
Behm, H. (1976), La Mortalidad en los Primeros Años de Vida en Países de la América 
Latina. Costa Rica 1968-69, CELADE, Series AN. 1024, San José, Costa Rica.
Bohmstedt, G. W. (1969), ‘Observations on the measurement of change’, in: E. F. 
Borgatta (ed.) Sociological Methodology, Jossey-Bass, San Francisco.
Bourgeois-Pichat, J. (1985), ‘Recent changes in mortality in industrialized countries’, 
in: J. Vallin and A. Lopez (eds.) Health Policy, Social Policy and Mortality Prospects, 
Ordina Editions, Liège, Belgium.
Brouard, N., and A. Lopez (1985), ‘Cause of death patterns in low mortality countries: 
a classification analysis’, in: International Population Conference, Florence, 1985. 
Vol. 2, International Union for the Scientific Study of Population, Liège, Belgium.
Caldwell, J. C. (1986), ‘Routes to low mortality in poor countries’, Population and 
Development Review, 12:171-220.
Centro Latinoamericano de Demografía (CELADE), Ministerio de Salud & 
Universidad de Costa Rica (1987), Los Grupos Sociales de Riesgo para la Sobrevida 
Infantil 1960-1984, Series A  1049, San José, Costa Rica.
Cervantes, S., and C. Raabe (1991), ‘Determinants of the evolution of the health 
situation of the population’, Scandinavian Journal o f  Social Medicine, Supplement 
46: 43-52.
Devesa, S. S., and D. T. Silverman (1978), ‘Cancer incidence and mortality trends in the 
United States: 1935-74', Journal o f  the National Cancer Institute, 60: 545-71.
Freedman, R., and J. Y. Takeshita (1969), Family Planning in Taiwan: An Experiment 
in Social Change, Princeton University Press, Princeton.
González-Vega, C. (1985), ‘Health improvements in Costa Rica: the socio-economic 
background’, in: S. Halstead, J. Walsh, and K. Warren (eds.) G ood Health at Low
194 Luis Rosero-Bixby
Cost: Proceedings o f  a Conference Held at the Bellagio Conference Center. The 
Rockefeller Foundation, New York.
Hall, C. (1985). Costa Rica, a Geographic Interpretation in Historical Perspective, 
West view Press. Boulder, Colo.
Hanushek, E. A., and J. E. Jackson (1977), Statistical Methods for Social Scientists, 
Academic Press. Orlando, FI.
Kleinbaum. D. G., D. G. Kupper, and H. Morgenstem (1982). Epidemiologic Research, 
Van Nostrand Reinhold, New York.
Liker, J. K., S. Augustyniak, and G. J. Duncan (1985). ‘Panel data and models of 
change: a comparison of first difference and conventional two-wave models’. Social 
Science Research, 14: 80-101.
Mata, L„ and L. Rosero-Bixby (1988), National Health and Social Development in 
Costa Rica: A Case Study o f  Intersectorial Action, PAHO, Technical Paper No. 13, 
Washington, DC.
Meslé, F. (1985), ‘Evolution des causes de décés dans quelques pays á faible mortalité’, 
in: International Population Conference, Florence, 1985, Vol. 2, International Union 
for the Scientific Study of Population, Liège, Belgium.
Meza-Lago. C. (1985), ‘Health care in Costa Rica: boom and crisis’. Social Science and 
Medicine, 21: 13-21.
Mohs, E. (1991), ‘General theory of paradigms in health’, Scandinavian Journal o f  
Social Medicine, Supplement 46: 14-24.
Omran, A. R. (1982), ‘Epidemiological transition: theory’, in: International Encyclo­
pedia o f  Population, The Free Press, New York, 172-83.
Pan American Health Organization (PAHO) (1990), Health Conditions in the 
Americas, 1990 Edition, Vol. I, Scientific Publications 524, Washington, DC.
Preston. S., N. Keyfitz, and R. Schoen (1972), Causes o f  Death: Life Tables for National 
Populations, Seminar Press, New York.
Ravenholt, R. T. (1990), ‘Tobacco’s global death march’, Population and Development 
Review, 16: 213-40.
Rosero-Bixby, L. (1985a), ‘The case of Costa Rica’, in: J. Vallin and A. Lopez (eds.) 
Health Policy, Social Policy and Mortality Prospects, Ordina Editions, Liège, 
Belgium.
------ (1985ft), ‘Infant mortality decline in Costa Rica’, in: S. Halstead, J. Walsh, and
K. Warren (eds.) G ood Health at Low Cost: Proceedings o f  a Conference held at 
the Bellagio Conference Center, The Rockefeller Foundation, New York.
------(1986), ‘Infant mortality in Costa Rica: explaining the recent decline’, Studies in
Family Planning, 17: 57-65.
------ (1991a), ‘Socioeconomic development, health interventions, and mortality decline
in Costa Rica’, Scandinavian Journal o f  Social Medicine, Supplement No. 46: 33-42.
------ (19916), Interaction Diffusion and Fertility Transition in Costa Rica, Doctoral
dissertation. University Microfilms International, University of Michigan, Ann 
Arbor.
------and H. Caamaño (1984), ‘Tablas de vida de Costa Rica 1900-1980’, in: Mortalidad
y Fecundidad en Costa Rica (pp. 7 -19), Asociación Demográfica Costarricense, San 
José, Costa Rica.
------ and C. Grimaldo (1987), ‘Descriptive epidemiology of cancer of the breast and
uterine cervix in Costa Rica’, PAH O Bulletin, 21: 250-60.
------M. W. Oberle, and N. C. Lee (1987), ‘Reproductive history and breast cancer in a
Adult Mortality Decline in Costa Rica 195
population of high fertility, Costa Rica, 1984-5’, International Journal o f  Cancer, 40: 
747-54.
Rytchtarikova, J., J. Vallin, and F. Mesle (1989), ‘Comparative study of mortality 
trends in France and the Czech Republic since 1950’, Population (English Selection 
No. 1), 44: 291-321.
Saenz, L. (1985), ‘Health changes during a decade: the Costa Rican case’, in: S. 
Halstead, J. Walsh, and K. Warren (eds.) G ood Health at Low  Cost: Proceedings of 
a Conference held at the Bellagio Conference Center, The Rockefeller Foundation, 
New York.
Vaupel, J. W., K. G. Manton, and E. Stallard (1979), ‘The impact of heterogeneity in 
individual frailty on the dynamics of mortality', Demography, 16: 439-54.