HEALTH1 - United Nations · 2010. 1. 8. · Chapter 4 HEALTH1 Healthcare in Ecuador has improved substantially over the last 30 years but spending remains low compared to other countries

Chapter 4 HEALTH1

Healthcare in Ecuador has improved substantially over the last 30 years but spending remains low compared to other countries in the region and there are severe inequalities in the system, particularly in access to healthcare for the indigenous population, the poor and those living in rural areas. A series of reforms in the 1990s remain to be implemented, hindered by a series of economic crises. The main focus of this note, however, is on how to reduce infant mortality rates, particularly for marginalized groups. The recommendations aim to help the country achieve the Millennium Development Goal of reducing infant mortality by two-thirds in the period 1990–2015. For many better-off sectors of the population this target is already in sight. The analysis shows that with a series of initiatives that are almost budget-neutral the MDG is almost within reach for the whole population, suggesting more robust measures may be needed in the near future. The recommendations that follow, would also go some way to addressing the overall problems in health service delivery, including low access to services, inequalities across population groups and geographical areas, and low quality of services. The recommendations, which analysis shows would have the greatest impact on infant mortality, are to protect the immunization program that has suffered recent budget cuts and expand the Free Maternity program (guaranteeing free access to free pre-natal and delivery to all women). In the long-term the government could consider a universal health insurance system as a means of delivering healthcare, especially to the poor. However, current financial constraints make that option impractical at present. The analysis uses a series of econometric and modeling techniques aimed at finding the most cost-effective ways to achieve the desired results.

“Soy indígena, vengo de Tigua ... cuando nos enfermamos, nosotros nos curamos con plantas, pero cuando

es una cosa seria, vamos a ver al doctor.” “Nosotros nunca vamos a ver al médico. Cuando hace falta, compramos remedios en la farmacia.”

“Si nos enfermamos, tomamos té de cedrón. Nunca nos hacemos controlar los dientes.” Quechua children, aged 10 to 12, Ecuador

(UNICEF, 2004: 14)

4.1 Public health spending in Ecuador is among the lowest in Latin America. Spending amounts to a meager 2 percent of GDP (only Haiti spends less). Yet huge strides have been made in improving health conditions. Since 1970, infant mortality rates have been cut by 70 percent to 34 per 1,000 live births. However, important inequalities remain in the health system, with higher mortality rates and limited access to health care for the indigenous population, the poor and those living in rural areas.

4.2 Health provisioning has shifted towards greater emphasis on primary health care in public provisioning and in-patient hospital care by private health providers. Health policy reforms have included increases in user fees for public services, decentralization, special programs providing free health care for the poor (including the free maternity program), and the introduction of 1 This paper was prepared by Rob Vos (Institute of Social Studies, The Hague). Ruth Lucio, Mauricio León, and

José Rosero Secretaría Técnica del Frente Social, Ecuador) and José Cuesta (Institute of Social Studies, The Hague). We are grateful to Karina Lara and Cesar Amores for valuable research inputs and to Arjun Bedi for advice on the modelling of the demand for health care and infant mortality determinants.

67

demand subsidies for health through the conditional cash transfer program Bono de Desarrollo Humano (Human Development Payment). Policies have not been consistent, though, as the health policy framework has changed a number of times during the 1990s and not all reforms have as yet been fully pushed through (such as the decentralization). The reforms were implemented under tightening budget constraints, as the health budget (in real per capita terms) declined for most of the 1990s. Recent upward budget adjustments have mainly gone into salary increases for health workers.

4.3 All of this has had a mixed outcome on health conditions in Ecuador, but principally the policy reforms as implemented have failed to address the fundamental problems in health service delivery, including low degrees of access to services, inequalities across social population groups and geographical areas, and low quality of services. Inequalities in access to health facilities have increased, partly because of the introduction of user fees and partly because health inputs (particularly drug supplies) have fallen well behind requirements. On the other hand, continued expansion of the immunization program and the introduction of the free maternity program have compensated for this effect, at least for young children and pregnant women, and this is likely part of the explanation for the continued decline in infant mortality during the 1990s. However, the remaining large differences between rich and poor, non-indigenous and indigenous, and urban and rural populations, suggest such specific interventions have not been sufficient to ensure greater equity in health conditions. It is even less clear to what extent Ecuador’s health system is sufficiently prepared to take on the challenge of increasing health demands due to the epidemiological transition towards injuries and chronic and degenerative diseases, which already dominate adult disease prevalence. The main focus of this paper will be on the health risks affecting mothers and young children and finding more cost-effective ways to use the health budget to further reduce infant mortality, particularly for those groups with the least favorable health status.

4.4 We find that improving access to maternal and child care (vaccinations, pre-natal controls, and professionally assisted child delivery) is critical in order to reduce infant mortality to the level of the Millennium Development Goal for all population groups (reduce it by two-thirds in the period 1990-2015). The budget implications should not form any impediment to achieving these targets as these are estimated at just 0.01 percent of GDP per annum. In fact, there would not even be an immediate need to increase the health budget to achieve the target for reducing infant mortality, as there is underutilized capacity in the public health system and the system has a distorted personnel structure with an oversupply of doctors and a shortage of nurses, such that cost-saving adjustments could cover the additional expenditures. In other words, all that is required would be a more adequate priority setting in the public health budget, protecting the immunization program to reach full coverage, a targeted expansion of the Free Maternity program and optimizing use of available resources for personnel and infrastructure. The conditional cash transfer program (Bono de Desarrollo Humano) is also aimed at stimulating use of maternal and child health care services, but at present—we estimate—just about 5 percent of the beneficiaries of the program obtain the cash transfer by making use of such services. While close coordination between the two programs is important, the degree of overlap does not seem very big. Over time, such compensatory schemes should be replaced by an adequate health insurance scheme. But with the current low coverage of health insurance and inadequacies in the supply of health services this seems a more appropriate alternative for the long term.

68

4.5 The remainder of this paper is organized as follows. In Section 2, we describe trends in overall health conditions in Ecuador over the past decades through changes in the mortality and epidemiological profiles of the population. We also present analyses of inequalities in health determinants that most impact child mortality. In Section 3, we describe the major changes in health policies during the 1990s, while in Section 4 we analyze how those changes have impacted on health inputs and trends in public health expenditures. Section 5 addresses equity issues, showing who benefits most from health expenditures in Ecuador. Section 6 looks at the efficiency of health spending, particularly that aimed at improving maternal and infant health status. We do this through a two-stage health production function approach, modeling first the determinants of access to health care and choice among public and private providers. Next, we model the determinants of child survival in the first year after birth and assess which policy variables seem most effective in reducing infant mortality. In Section 7 we assess the budget implications of alternative interventions and spell out a budget tracking methodology that links health inputs to outputs.

I. Health Status and Access to Health Services

4.6 Rising Life Expectancy and Declining Infant Mortality. Health conditions for the Ecuadorian population have improved considerably over the past 50 years. Life expectancy increased from 48 to 72 years between 1950 and 2000 (Figure 4.1). This upward trend was sustained during the 1990s with the average Ecuadorian gaining another 5 years of life expectancy. Parallel patterns are found in declining child and infant mortality rates.

.

404550556065707580

Total Males Females

0

5

10

15

20

25

55 60 65 70 75 80 85 90 95 2000

404550556065707580

55 60 65 70 75 80 85 90 95 200055 60 65 70 75 80 85 90 95 2000

Total Males Females

0

5

10

15

20

25

55 60 65 70 75 80 85 90 95 2000

Glo

bal m

orta

lity

rate

(per

1,0

00)

INEC Population projections

INECEstadisticas Vitales

Figure 4.2. Global Mortality Rate, 1960-2001Figure 4.1. Life Expectancy (in years) for Males and Females, 1950-2000

Source: INEC (2003) and INEC Stadísticas VitalesNote: Estimates from administrative records of health centers (Estadísticas Vitales) are for the end of the period. The estimated projections from INEC (2003) are based on mortality tables for 2001 (and CELADE’s life expectancy tables).

Note: Estimates from administrative records of health centers (Estadísticas Vitales) are for the end of the period. The estimated projections from INEC (2003) are based on mortality tables for 2001 (and CELADE’s life expectancy tables).

Source: INEC (2003) and INEC Stadísticas Vitales

1950-1955-1960-1965-1970-1975-1980-1985-1990-19951950-1955-1960-1965-1970-1975-1980-1985-1990-1995

404550556065707580

Total Males Females

0

5

10

15

20

25

55 60 65 70 75 80 85 90 95 2000

404550556065707580

55 60 65 70 75 80 85 90 95 200055 60 65 70 75 80 85 90 95 2000

Total Males Females

0

5

10

15

20

25

55 60 65 70 75 80 85 90 95 2000

Glo

bal m

orta

lity

rate

(per

1,0

00)

INEC Population projections

INECEstadisticas Vitales

Figure 4.2. Global Mortality Rate, 1960-2001Figure 4.1. Life Expectancy (in years) for Males and Females, 1950-2000

Source: INEC (2003) and INEC Stadísticas VitalesNote: Estimates from administrative records of health centers (Estadísticas Vitales) are for the end of the period. The estimated projections from INEC (2003) are based on mortality tables for 2001 (and CELADE’s life expectancy tables).

Note: Estimates from administrative records of health centers (Estadísticas Vitales) are for the end of the period. The estimated projections from INEC (2003) are based on mortality tables for 2001 (and CELADE’s life expectancy tables).

Source: INEC (2003) and INEC Stadísticas Vitales

1950-1955-1960-1965-1970-1975-1980-1985-1990-19951950-1955-1960-1965-1970-1975-1980-1985-1990-1995

4.7 The overall mortality rate dropped from 13.8 in 1960 to 4.5 per 100,000 inhabitants in 2001 (Figure 4.2). This rate has not changed much during the 1990s. In contrast, the infant mortality rate has continued to fall in an almost linear trend since 1950 reaching 34 per 1,000 live births for 1995–2000 according to the population census based life expectancy tables, down

69

from 140 in 1950–55 (INEC 2003). This impressive trend is consistent across various data sources (see Figure 4.3). Child mortality rates follow similar trends as these mainly reflect infant mortality since most child deaths are concentrated in the first year of life (Table 4.1). Since 1970, the infant mortality rate has fallen by 70 percent in Ecuador, which is as impressive as the achievements in the rest of the Americas where the rate has fallen on average at a similar rate (WHO 2003). Achievements in Chile, Costa Rica, and Cuba have been even more impressive with reductions of over 80 percent.

0

20

40

60

80

100

120

140

160

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000

Estadísticas vitales Endesa-Endemain Population censuses

0

1

2

3

4

5

6

7

8

1950-55

1955-60

1960-65

1965-70

1970-75

1975-80

1980-85

1985-90

1990-95

1995-2000

Fert

ility

rat

e

0

20

40

60

80

100

120

140

160

IMR

Fertility rate Infant mortality rate

0

20

40

60

80

100

120

140

160

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000


0

1

2

3

4

5

6

7

8

1950-55

1955-60

1960-65

1965-70

1970-75

1975-80

1980-85

1985-90

1990-95

1995-2000

Fert

ility

rat

e

0

20

40

60

100

120

140

160


Figure 4.3. Infant Mortality, Estimates from Various Sources, 1950-2002- Figure 4.4. Fertility and Infant Mortality, 1950-2000-

Note: See also INEC (2003) and WHO Life Tables in López et al. (2002), for a similar compilation including some additional estimates from the DHS surveys.

Source: INEC (2003); Demographic and Health Surveys (Endesa-Endemain) of 1987, 1989, 1994, and 1999; INEC Estadísticas Vitales.

Source: INEC (2003); Demographic and Health Surveys (Endesa-Endemain) of 1987, 1989, 1994, and 1999; INEC Estadísticas Vitales.Note: See also INEC (2003) and WHO Life Tables inLópez et al. (2002), for a similar compilation including some additional estimates from the DHS surveys.

0

20

40

60

80

100

120

140

160

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000


0

1

2

3

4

5

6

7

8

1950-55

1955-60

1960-65

1965-70

1970-75

1975-80

1980-85

1985-90

1990-95

1995-2000

Fert

ility

rat

e

0

20

40

60

80

100

120

140

160

IMR


0

20

40

60

80

100

120

140

160

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000


0

1

2

3

4

5

6

7

8

1950-55

1955-60

1960-65

1965-70

1970-75

1975-80

1980-85

1985-90

1990-95

1995-2000

Fert

ility

rat

e

0

20

40

60

100

120

140

160


Figure 4.3. Infant Mortality, Estimates from Various Sources, 1950-2002- Figure 4.4. Fertility and Infant Mortality, 1950-2000-

Note: See also INEC (2003) and WHO Life Tables in López et al. (2002), for a similar compilation including some additional estimates from the DHS surveys.

Source: INEC (2003); Demographic and Health Surveys (Endesa-Endemain) of 1987, 1989, 1994, and 1999; INEC Estadísticas Vitales.

Source: INEC (2003); Demographic and Health Surveys (Endesa-Endemain) of 1987, 1989, 1994, and 1999; INEC Estadísticas Vitales.Note: See also INEC (2003) and WHO Life Tables inLópez et al. (2002), for a similar compilation including some additional estimates from the DHS surveys.

Table 4.1. Reaching the MDG for Child Mortality (Projections with Unchanged Trends of 1990s, per 1,000 Live Births)

1990 2001 2015 (p) Year of reaching MDG at unchanged trends

Males Females Males Females Males Females Males Females 0-1 year olds 56.1 44.0 30.8 22.6 14.4 9.7 2010 2008 1-5 year olds 17.1 15.1 6.1 4.6 1.6 1.0 2002 2000 Source: INEC, Population Censuses 1990 and 2001. Note: MDG target is to cut child mortality by two-thirds between 1990 and 2015. Projections assume continued linear trend as observed during 1990s.

4.8 The drop in infant mortality coincides with a long-term decline in fertility rates as Figure 4.4 shows. Fertility dropped from almost 7 births per woman in the 1950s and 1960s to 2.8 in 2000–05. During the 1990s fertility dropped faster in rural than in urban areas, but the rate is still 1.5 times higher for rural women. Fertility and infant mortality may be mutually dependent, as higher fertility raises the risk of early child birth, whereas infant deaths may induce higher fertility as a replacement effect. In Ecuador it is noteworthy that the decline of the fertility rate has been less pronounced and started later than the decrease in infant mortality. This would suggest that improvements in health services and socio-economic conditions influenced

70

conditions of child delivery and care prior to influencing reproductive health practices at the family level. It would also suggest that the replacement effect is not very strong.

4.9 The Millennium Development Goals for health include a target of reducing child mortality by two-thirds between 1990 and 2015. This target was already reached in 2001 for girls aged 1–5 and with unchanged (linear) trends this should also have been reached for boys aged 1–5. The projections in Table 4.1 suggest that the goal for infant mortality (0–1 year olds) could be reached by 2010 for boys and by 2008 for girls. We want to investigate further whether the trends are in fact linear (Figure 4.3 would suggest as much), to what extent health policies are important in determining the trend and whether the observed improvements are equal for all population groups.

4.10 Taking a comparative perspective for Latin America and the Caribbean, a simple correlation confirms the typical negative, log-linear relationship between per capita income and infant mortality often found in cross-country data sets (see e.g. Hanmer and White 1998). The two panels of Figure 4.5 show further that Ecuador tends to be below the trend line, which might suggest health policies have had a positive effect on its position of having a relatively low infant mortality rate compared to its mean income level, and since 1970 the country has moved further away from the trend line.2

2 The findings in Figure 4. 3 are the same when using GDP at PPP values or when using transformed data (logs,

square roots, etc.). The logarithmic specification as shown by the trend line gives the best fit for the data set.

71

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000

1970 GDP pc (constant 95 $)

Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,0002001 GDP pc (constant 1995 US$)

Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

Source: World Bank Development Indicators, 2003 (CD-ROM).

Figure 4.5. Latin America and the Caribbean: Infant Mortality and GDP Per Capita, 1970 and 2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001



y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970 y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001 y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001

y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970 y = -29.556Ln(x) + 298.17R2 = 0.4486

0

20

40

60

80

100

120

140

160

0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000


Infa

nt m

orta

lity

rate

(197

0)

ECU

1970

y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001 y = -16.554Ln(x) + 156.92R2 = 0.6173

0

10

20

30

40

50

60

70

80

90


Infa

nt M

orta

lity

Rat

e (2

001)

ECU

2001



4.11 Looking at child mortality by population groups it appears that (see Table 4.2):

• Boys have a higher probability of dying at an early age than girls, probably largely for biological reasons as such a pattern is found worldwide (WHO, 2003).3

3 Higher child mortality for boys is found with few exceptions around the world. In China, India, Nepal and

Pakistan, mortality in girls exceeds that of boys. This disparity is particularly noticeable in China, where girls have a 33 percent higher risk of dying than their male counterparts. These inequities are thought to arise from the preferential treatment of boys in family health care-seeking behaviour and in nutrition. Throughout Latin America child mortality is higher for boys.

72

Table 4.2. Socio-Demographic Profile of Infant (0-1 years) and Child Mortality (0-5 years) (per 1,000 live births), 1995-2000 1985-90 1995-2000 Age groups Age groups 0 - 1 1 - 5 0 - 5 0 - 1 1 - 5 0 - 5 Ethnicity (language) Non-indigenous 53.3 17.8 70.0 35.0 7.5 38.5 Indigenous 98.8 48.0 141.5 66.0 26.0 90.5 Ethnicity (self-definition) Indigenous n.a. n.a. n.a. 66.0 26.0 89.8 Black (Afro) n.a. n.a. n.a. 38.0 10.0 48.3 Mestizo n.a. n.a. n.a. 31.0 7.0 37.8 Caucasian n.a. n.a. n.a. 26.0 5.3 30.8 Other n.a. n.a. n.a. 29.0 6.5 35.3 Area of residence Urban 40.5 11.0 51.0 27.0 5.5 32.3 Rural 71.3 29.3 98.5 44.0 12.8 56.0 Poverty by UBN Extremely poor 66.3 26.3 90.5 42.0 12.0 53.5 Poor 60.5 22.5 81.8 38.0 10.0 47.5 Non-poor 31.5 7.0 38.5 25.0 4.8 29.3 Total 55.3 19.0 73.3 34.0 8.3 42.3 Source: Estimates based on INEC, Population Censuses of 1990 and 2001, using same estimation method as indicated in note to Figure 4.3.

• Indigenous children have about double the child mortality rate as non-indigenous children (i.e. 90.5 against 38.5 per 1,000 live births). They form the most disadvantaged group in this sense. As we shall try to analyze later on, this is likely caused by the fact that indigenous mothers tend to have less education, less knowledge of (modern) reproductive health care and limited access to professional pre-natal care and birth delivery assistance. As can also be derived from Table 4.2, the gap in the risk of dying at an early age between indigenous and non-indigenous born has not narrowed during the 1990s.

• Child and infant mortality are also significantly higher for Afro-Ecuadorians. The probability of early child death for this population group is 10–15 percent higher than the average, though the difference is much less than that for the indigenous population.

• The risk of child mortality is significantly higher in rural than in urban areas, even though this difference has narrowed slightly from 1.9 times to 1.7 times during the 1990s.4

• Children of poor families are at greater risk than those of non-poor families,5 but here we also observe a reduction in the gap in the 1990s (from 2.1 to 1.6 for child mortality and from 1.9 to 1.5 for infant mortality).

4 For infant mortality this gap fell from 1.75 to 1.63. 5 Poverty is measured here according to the unsatisfied basic needs measure (UBN) as define by SIISE (2003).

73

4.12 The observed differences by socio-demographic groups are more or less expected patterns. The question to be answered later on is to what extent inequalities in access to health services have influenced these patterns.

4.13 Available data also suggest a significant decline in maternal mortality. According to INEC’s health statistics, the maternal mortality rate dropped from 203 per 100,000 live births in 1971 to 117 in 1990 and to 46 in 2002 (SIISE 2003). If these data are correct, a continuation of the declining trend in maternal mortality as observed during the 1990s would be sufficient to reach the MDG target for this health indicator (a cut by 75 percent between 1990 and 2015) by 2008. However, there are strong reasons to believe that the administrative records of maternal deaths as registered by health centers strongly underestimates the actual level. Different estimates are drawn from the Demographic and Health Surveys Encuesta Demografica y de Salud Materno Infantil (ENDEMAIN). Based on this source, Bortman (2003: 301) maternal mortality was an average of 302 per 100,000 in 1981–87 and 159 during 1988–94, while the WHO database records a rate of 210 for the mid-1990s. The latter rate is relatively high by Latin American standards, well above that in Chile, Colombia, Costa Rica, Mexico and the richer countries in the region, but similar to that of Peru and close to that of Honduras and Nicaragua, the latter being much poorer nations than Ecuador. We must emphasize that all these estimates lack precision, such that it is difficult to infer strong conclusions about the trends and at best it is safe to say that the rate likely fell considerably in recent decades.

4.14 A Shifting Epidemiological Profile. The decline in fertility and increase in life expectancy has changed the demographic profile of the Ecuadorian population towards a higher average age. However, the population is still rather young with 54 percent under the age of 25. This segment of the population was still on the rise during the 1950s and 1960s when fertility rates were still very high. The turning point came around 1965. The share of children under age 5 has dropped from 17.6 percent in 1965 to 11.8 percent in 2000.

4.15 The epidemiological profile (causes of mortality and morbidity) has also changed in past decades, away from traditional “child diseases” (malnutrition and communicable respiratory and infectious diseases) towards a greater prevalence of diseases associated with higher levels of

0

10

20

30

40

50

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

Preventable diseases Chronic and degenerative

Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs

Communicable diseases and maternal and nutritional conditionsNon-communicable diseasesInjuries

60

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


0

10

20

30

40

50

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


60

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


Source: INEC, Estadísticas Vitales, 1979-2000 and SIISE (2003).Note: Preventable death causes include: infectious and respiratory diseases, malnutrition, abortions and deaths due to inadequate assistance at birth. Chronic-degenerative diseases include, among other, cancer,cardiovascular and cerebrovascular diseases, mental diseases, and disorders of the nervous system.

Figure: 4.7 Distribution of Adjusted Lost Life Years by Types of Diseases and Age Groups, 1995

Source: CEPAR/PHR (1999), El Peso de la Enfermedad en el Ecuador, Quito (Burden of Disease Study).

Figure 4.6. Main Causes of Mortality, 1979-2000 (share of total deaths)

0

10

20

30

40

50

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


60

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


0

10

20

30

40

50

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


60

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000


Violence Other

Not well de fined

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

< 5 5 a 14 15 - 44 45 - 59 > 60Age groups

% o

f tot

al a

djus

ted

lost

life

yea

rs


Source: INEC, Estadísticas Vitales, 1979-2000 and SIISE (2003).Note: Preventable death causes include: infectious and respiratory diseases, malnutrition, abortions and deaths due to inadequate assistance at birth. Chronic-degenerative diseases include, among other, cancer,cardiovascular and cerebrovascular diseases, mental diseases, and disorders of the nervous system.

Figure: 4.7 Distribution of Adjusted Lost Life Years by Types of Diseases and Age Groups, 1995

Source: CEPAR/PHR (1999), El Peso de la Enfermedad en el Ecuador, Quito (Burden of Disease Study).

Figure 4.6. Main Causes of Mortality, 1979-2000 (share of total deaths)

74

economic well-being and urban lifestyles, such as cancer and cardiovascular health risks. We show this change in health risk patterns for the main causes of death in Figure 4.6. The share of ‘preventable’ diseases declined from over 60 to below 30 percent between 1979 and 2000, while that of ‘chronic-degenerative’ diseases increased from 22.5 to 42.2 percent in the same period. As the administrative records of health services, on which these data are based, tend to understate the prevalence of preventable diseases, the epidemiological transition might have taken place at a somewhat slower pace than suggested by Figure 4.5, but it’s undeniably there.

4.16 The main causes of mortality and morbidity in early childhood (0-5 years of age) remain, however, related to preventable, communicable diseases, as one would expect. The burden of disease study by CEPAR/PHR (1999) for 1995 confirms this, showing that more than 70 percent of years of life lost at an early age are due to infectious and respiratory diseases (plus nutritional deficiencies and maternal and perinatal conditions), while in the age group 5–45 the major causes of lost life years are associated with injuries and chronic-degenerative diseases and in those 45 and older the latter type of disease dominates (see Figure 4.7).

4.17 Malnutrition is reported as a direct cause of infant death in less than 5 percent of the

e

the prevalence of malaria has been closely associated with the El Niño

ants. The indicated differences in the risk of early child

Annex Tables A46-A53. Some of the salient findings are as follows:

cases and, according to INEC’s health statistics, this share has been declining. This figure likely underestimates the impact of malnutrition, as it is a compounding factor in infectious diseases.

4.18 The prevalence of AIDS increased during the 1990s, according to official data from thMinistry of Health.6 Reported cases increased to 764 in 2002, up from 85 in 1990. This implies an increase in the prevalence of AIDS from 0.8 to 6.0 per 100,000 inhabitants. The greatest proportion of cases is clustered in Guayas province (about 60 percent of the total) and heterosexuals account more than half of them. Growth in recent reported cases is especially high among women, particularly sex workers. The impact on child mortality of the rising prevalence of AIDS is not known.

4.19 In recent decadesweather phenomenon. The number of malaria cases increased strongly after the 1982–83 floods, declined after extra malaria eradication efforts were made, resurged in the late 1980s after a light reoccurrence of El Niño, only to decline steeply during the 1990s until the disaster hit strongly again in 1997–98. Typically it takes several years for the pandemic to subside and eradication measures to take effect. Most recently, the number of malaria cases continued to rise until 2001 when it peaked at 883 per 100,000 inhabitants, up from 103 in 1996 and 686 in 1999. Since 2001 malaria prevalence has declined again. Coastal lowlands and the Amazon region are most at risk. Most cases are clustered in the parts of the Costa region most at risk of flooding (see Vos, Velasco and De Labastida 1999).

4.20 Inequalities in Health Determindeath by socio-economic group are also reflected in most indicators of likely determinants of infant mortality. For this we make use of a number of sources, including the 2001 population census, a special module of the Integrated System of Household Surveys of INEC (SIEH) on access to social services of December 2003,7 as well as other recent surveys (the 1999 LSMS and the 2000 EMEDINHO household surveys). Some detailed results are reported in Statistical

6 Ministry of Health, Epidemiological Statistics, 1990-2002. Also reported in SIISE (2003). 7 The special module was sponsored and designed by the Integrated System of Social Indicators of the Social

Cabinet (STFS-SIISE).

75

• Malnutrition. Data limitations do not allow a longer view of trends in malnutrition in Ecuador, but given economic development trends one should expect a significant

and—most particularly— born. Figure 4.8 shows the link between income

•

and 2001, mainly due to increased coverage in rural areas.9

•

as dropped from 19.7 to 10.3 percent

decline till the 1980s. Despite recurrent economic woes ever since, the prevalence of malnutrition also fell in the 1980s and 1990s as suggested by several point estimates. Stunting (low height for age) among children under age 5 fell from 34 to 26 percent between 1985 and 2000.8 Underweight (low weight for age) declined on average from 17 to 12 percent in that samerural households, in the Sierra region, in poor householdsthose that are indigenousdistribution and malnutrition, indicating that stunting still affects more than one third of the extreme poor (first three deciles) and underweight about 17 percent of the poorest. Malnutrition (for stunting) is also between 1.5 and 1.9 times higher for the indigenous population, depending on whether it is identified through language or self-definition (Statistical Annex Table A46).

Water and Sanitation. Access to drinking water and sewerage systems has greatly improved in past decades. Access to safe drinking water increased from 70 to 81 percent between 1990

Figure 4.8 Malnutrition (0-5 years old) and Inequality, 2000

30%

35%

40%

P%

)

period. Malnutrition most affects children that live in

Inequalities in access run parallel to differences in child mortality rates with indigenous, rural, and poor households showing the larger deficiencies in access to water and sanitation. Coverage of safe drinking water is 58 percent for the rural population against 95 percent for the urban population and about 70 percent for the extreme poor against over 90 percent for the richest household groups at the end of 2003. These data consider all forms of access to drinking water. Piped drinking water to people’s homes has less coverage and shows starker inequalities with 65 percent of urban and 19 percent of rural households served in 2001 (up from respectively 56 percent and 8 percent in 1982).

Education. The education of the mother is typically found to be a significant determinant of lower fertility and infant mortality rates. In Ecuador these factors point in the same direction. Female illiteracy h

8 SIISE (2002). 9 The coverage in rural areas increased from 44.5 to 63.5 percent, while that in urban areas reached 92 percent by

2001, up from 91 percent in 1990. Data are based on population census data and refer to all sources of safe drinking water. The coverage of piped drinking water in homes has increased more slowly, i.e. from 30 percent in 1974 to 38 percent in 1990 and 48 percent in 2001.

0%

5%

10%

15%

20%

25%

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Per capita income deciles re

vale

nce

(

Stunting (low height for age)

Underweight (low weight for age)

Source: INEC, EMEDINHO (Special Module of Household Survey), 2000.

76

over the past two decades, and the average level of education of (all) female adults increased from 4.7 to 7.1 years of completed schooling.10 The 2003 SIEH survey estimates female illiteracy at 11 percent for all women over 15 and at 5.7 percent for women of fertile age (15–49 years). However, the latter rate is significantly higher for poor (12 percent), indigenous (21 percent) and rural (13 percent) women. The mean level of education of women of fertile age (15-49 years) is 9.2 years, but for indigenous women in that age group it is 6.5 years, for rural women 6.4 years and poor women (first three quintiles) 6.7 years.

Knowledge of Family Planning and Sexually Transmitted Disease. Knowledge and use of contraceptives has increased. During the 1990s, use of family planning methods increased from 53 to 66 percent of w

•

omen of reproductive age.11 This share

•

d

increased more significantly for women in rural than urban areas, respectively increasing from 40 to 58 percent and from 62 to 71 percent. This no doubt has had a positive impact on declining mortality and fertility rates. Yet again here, remaining differences in use of contraceptives follow the earlier pattern, with women in indigenous, rural, and poorer families showing much less knowledge and use (see Statistical Annex Table A47). Most starkly, 82 percent of indigenous women between 15–45 years report not using any form of contraception. HIV/AIDS does not have a high prevalence in Ecuador, but—as indicated—has been on the rise, particularly among women. Yet, about one third of the population is unaware of the risk that the disease may also be transmitted to the child when giving birth and just over half of the population says it knows how to protect against the risk of HIV/AIDS during sex. Again this awareness varies across population groups showing again the earlier patterns of greater deficiencies among indigenous, rural and poorer families.

Access to Health Services. The population groups showing higher disease risk and infant mortality also appear to find greater difficulty seeking professional medical assistance. As shown by Statistical Annex Table A48, cases of diarrhea anrespiratory disease have a similar incidence across population groups, but those seeking medical assistance during illness are more likely to be among the non-poor, non-indigenous and urban population.12 Differences in access to qualified medical assistance for pre-natal care and child delivery also follow this socio-demographic pattern (see Statistical Annex Table A49). About 84 percent of child delivery takes place with professional care at health centers and hospitals, but this share is only 64 percent for indigenous women, against almost 88 percent for the non-indigenous

10 Based on population census data for, respectively, 1982 and 2001. 11 Data from SIISE (2003), based on data reported in demographic and health surveys (ENDEMAIN) for 1989

and 1999. Please note that the reported data refer to women that currently use contraceptive methods. Including those who have used contraceptives in the past but not currently would increase the coverage for 1999 to 81 percent. The share of women using modern forms of birth control (i.e. excluding methods such as withdrawal before ejaculation, biorhythm, and other less secure practices) is 75 percent, such that the share of women that are currently using birth control methods considered to be secure was 49.5 percent in 1999 (i.e. 75 percent of 0.66).

12 The reported data in Statistical Annex Table A48 are admittedly crude, as no specific questions were raised regarding the severity of the complaints (i.e. a simple cold is reported in the same category as bronchitis or other pulmonary diseases). Nonetheless, assuming the likelihood of a more severe illness is greater among the more vulnerable, the differences in access to health services would be even starker than reported.

77

population. One third of rural births are completed without medical assistance against only 6 percent of urban births. Between 63 and 80 percent of child deliveries of the poorest 30 percent in the income distribution are professionally attended against 90 percent or more for the richest segments, with the latter more typically making use of private health services. Despite the remaining deficiencies, coverage of professionally assisted child delivery has increased in recent decades with likely concomitant effects on falling infant mortality.

Access to Targeted Social Assistance Programs. As discussed in the next section, in recent years a variety of health programs have been introduced targeted at women and early childhood protection through nutrition and

•

health care. The free maternity care

program (Maternidad Gratuita), introduced in 1998, is one of the more important ones and reached by the end of 2003 about 30 percent of pregnant women (Statistical Annex Table A51).13 The impact in terms of coverage of this program will also be included in the aforementioned data on access to professional birth delivery assistance. It is noteworthy, though, that thus far this program is reaching to a much reduced extent the vulnerable groups, as for instance coverage for indigenous and for rural women is only 19 percent and for the poorest 30 percent it is about 20 percent. Other important programs include several nutritional programs like the National Nutritional Program for Children (PANN 2000), the school breakfast program (PAE) and the cash transfer program Bono de Desarrollo Humano (BDH). PANN has two components: nutritional support for children between 6 and 24 months of age (called Mi Papilla) and nutritional support for pregnant and breastfeeding women (called Mi Bebida). The Mi Papilla program reaches (at the end of 2003) about one third of the children in the age group between 6 and 24 months. The program’s coverage is somewhat higher for the vulnerable groups, i.e. indigenous, rural and poorest segments of the population (see Statistical Annex Table A52). Overall, though, the program’s coverage is rather low as just 39 percent of poor children have access to Mi Papilla and 38 percent of poor, pregnant and breastfeeding mothers receive the benefits of Mi Bebida. Thus, exclusion errors in targeting are big, even though inclusion errors are less severe (10-12 percent).14 A recent impact evaluation of the program in one municipality (Santo Domingo) indicated that the program effectively reduced prevalence of underweight (low weight for age) by about 10 percentage points, but could not detect a clear impact on stunting.15 A broader impact evaluation would be required in order to know whether these findings could be generalized. The PAE reaches about 20 percent of the total population (and more of the primary school-age population, see Vos and Ponce 2004), but aims at groups over 5 years old and the nutritional value of the provided snacks is relatively low.16 A specific

13 Please note that the data in Statistical Annex Table A51 refer to “knowledge” of the program, i.e. not actual

coverage. Nonetheless, information from the program itself estimates beneficiaries at about 1 million women, which also implies coverage of about 30 percent.

15

f minimum nutritional requirements for, respectively, energy and protein intake ficient if not supplemented by complementary meals at home.

14 See Statistical Annex Table A54 for more detailed estimates and Ponce (2004) for a further analysis of the program’s targeting efficiency. See MSP/OPS/ICT (2004).

16 According to information from the Secretaría Técnica del Frente Social (STFS), the school snacks represent about one third and two-thirds ofor breakfast or lunch. That is, clearly de

78

nutritional program covering children in age group 2-5 years is lacking at the moment, although some nutritional support is provided through the early childhood programs. However, these have a rather low coverage (about 12 percent of the target group). The BDH is a conditional cash transfer program and is to replace other existing cash transfer programs (Bono Solidario and Beca Escolar). By the end of 2003, the BDH reached 18 percent of the population. Eligibility for this program relates to a poverty indicator and is conditioned on having a child in primary education or accessing a health care center for maternal and child care. Thus not all beneficiaries of this program will necessarily have accessed maternal care. In fact, only 5 percent of BDH beneficiaries are families with no school-going children and only children under age 5. Given the conditions of the program (see Section 3), this gives us to suspect that the program’s impact more likely will be on education than on increasing access to health care. The program overall does benefit our vulnerable groups more than others.

Immunization. Ecuador’s• permanent vaccination campaign has been successful and

• urance in Ecuador is low by any standard.

hampering their access to health services.

in most parts of the country parents from all layers of society demand a vaccination card for their children. Data on immunization coverage can sometimes be confusing as much undercoverage may relate to children of less than one year of age not (yet) having received the full doses. Aggregate estimates from the Demographic Health Survey (DHS) and Encuesta Demográfica y de Salud Materno Infantil (ENDEMAIN) show little increase in coverage (for all main vaccinations) during the 1990s (moving from 75 to 77 percent between 1989 and 1999), but this could be an underestimation. The SIEH’s health module of December 2003 suggests that of the total number of 2–5 year olds, 87 percent have a vaccination card, 99 percent have received the full doses of BCG and 70 percent the DPT vaccination.17 Coverage of immunizations appears to be rather equitable as there are no significant differences in coverage across social groups (see Statistical Annex Table A50).

Health Insurance. Coverage of health insLess than 20 percent of the population has access to any form of health insurance. Those that are insured tend to be part of the public social security system (IESS) and these are mainly urban working families. Private health insurance schemes cover less than 2 percent of the population. The peasant insurance scheme (Seguro Social Campensino) provides some coverage for farmers, but nonetheless reaches just 11 percent of the rural population.18 Access to health insurance is typically lower for the poorer segments and the indigenous population (see Statistical Annex Table A52),

17 Reported coverage of DPT in the SIEH survey of 2003 is much lower than that reported in the 1999 LSMS

(ECV) survey. The reasons for this difference are not very clear, but given the permanence of the immunization program (PAI) we do not believe there has been actual reduction in coverage between 1999 and 2003. Rather, the discrepancy reflects differences in measurement and sampling errors. The reported data thus need to be taken with some caution. The general perception is that vaccination coverage is high (over 80 percent), at least for key vaccinations (BCG and DPT).

18 The figure is based on the SIEH survey of 2003. The LSMS of 1999 reports a somewhat higher coverage: 16.8 percent of the rural population.

79

• y will increase the prevalence of chronic

II. Health Policies in the 1990s

4.21 Ecuador’s public health ralized and supply-driven. The government funds and subsidiz ries and public hospitals from

reforms aiming at greater coverage, efficiency and equity.

ry

4.23 Sup s initiated in this period were aimed at improv reducing inequalities in access through targeted health interventions. A consensus was sought for these reforms with stakeholders within the health sector itself. FASBASE was the stellar program

Life Styles. Smoking and alcohol consumption by pregnant women may affect the health of the unborn child, as much as theand degenerative diseases such as cancer and cardiovascular diseases. We have no comparable evidence on how such life styles may have affected the epidemiological profile over time. The Living Standards Measurement Study (LSMS) and Encuesta de Condicionales de Vida (ECV) survey of 1999 suggests, though, that the prevalence of regular cigarette and alcohol consumption is fairly low in Ecuador, with 8 percent of the population of 15 years and older smoking daily and about 25 percent engaging in alcohol consumption. However, tobacco and alcohol may be characterized as luxury and masculine goods as the use increases with income levels and as women hardly consume them (see Statistical Annex Table A53). In other words, these factors affect the more vulnerable groups to a small degree at best, but are not likely a factor of significance in infant mortality. While engaging more in tobacco and alcohol consumption, males and the better off do compensate (on average at least) by doing more sports, but one should add that less than 30 percent of the adult population is engaging in some sporting activity. This may be a sign that the epidemiological transition as observed earlier and rising problems of obesity are likely to continue in the near future.

system has traditionally been centes public health centers, dispensa

tax revenue. The social security system (IESS) runs its own hospitals and health centers. In recent decades an increasing number of private hospitals and health centers have emerged. In fact, coverage of services per inhabitant has increased for private hospitals and clinics, whereas it has decreased for public in-patient care and remained stable for public primary, outpatient care. The relative reduction in the supply of public health services is partly associated with budgetary restrictions (see next section) and partly with a shift in health policies away from a supply and input-driven system of health provisioning and towards greater incentives to demand for health care and more decentralized delivery. In addition, several targeted health programs have been introduced to repair inequities in the system and put greater emphasis on preventing risk of early child death. Health reform policies have not followed a consistent path in this direction though. Rather, the strategic view underlying reforms has shifted during the decade, such that it is hard to speak of a consistent reform path.

4.22 Ecuador’s health policy reforms since 1990 may be divided into three distinct phases:

• 1991–96: Supply-driven

• 1997–2000: Regulatory and institutional reforms in support of decentralized. deliveof health services and better quality control.

• 2000–04: Decentralization, health insurance reform and demand incentives.

ply-Driven Reforms (1991–96). The reforming coverage of health services, improving the efficiency in delivery and

80

around which the reforms were to be implemented. The program’s name–Strengthening and Expansion of Health Services (Fortalecimiento y Ampliación de Servicios Básicos de Salud)—covers the reform intentions well. FASBASE was financed by the World Bank. Under the program a new system of primary health care provisioning was set up, essentially allowing for a decentralization of planning and decision-making for health providers in newly defined administrative health sector regions. The reform put priority on primary and preventative health care, including an investment program for enhancing the coverage of drinking water and sanitation services and targeted nutritional support for women and young children. The program also established a network of emergency hospital care managed by municipalities and a training program for health workers within the public sector. Finally, the program further aimed at improving the supply of drugs to health centers and strengthened the immunization program (PAI).

4.24 From its inception, the FASBASE program suffered from problems in budget management, encountering tensions between ensuring adequate allocation for the program’s activities and pressures from the Ministry of Health (MoH) to fill gaps in financing its “regular” current expenditures. The progress in enhancing the targeted assistance to young mothers and

d,

made official through decree 502 (DE-502). However, no proper system of

infants, training of local health workers and strengthening of the PAI are probably the main achievements of FASBASE, but overall the program’s results are a far cry from the goals set.

4.25 The strengthening of the immunization program can be singled out as a successful improvement of health services that continued during the subsequent periods. The program goes back to 1977 and has had a consistent presence since then. During the 1980s emphasis was on the fast expansion of immunization coverage, but subsequently other elements have been addeincluding health and childcare education. During the 1990s the program’s capacity to monitor sanitary and environmental conditions for the risk of epidemic diseases and other public health risks has been strengthened. Its success is attributed to its near full emphasis on operations in the field (and reaching beneficiaries often in their homes) and repeated campaigns stressing the importance of vaccinations. Constraints are still found, though, in reaching vulnerable populations in areas of difficult geographical access, and there is some cultural resistance among segments of the indigenous population, particularly in the Amazon region and the Andean highlands. Since 2001, the Law for Immunizations protects the program’s budget. Nonetheless, the draft national government budget for 2004 included a severe cut in the PAI budget, but efforts are being made to revert this. Growth in vaccination coverage and the eradication of poliomyelitis, neonatal tetanus and measles have made an important contribution to the reduction in infant mortality.

4.26 During this period, fees for the use of public health services were introduced and then increased. This started on the initiative of some public health centers that faced problems in meeting health demands. Initially only small contributions were asked for. In 1998, the practice of cost-recovery wascontrol of the level of fees and use of funds was introduced to regulate cost sharing. DE-502 also regulates self-management of hospitals and health centers. Shortly after introducing the decree, the free maternity law was introduced (see below), which guaranteed access free of charge to infant and maternal health care. Given the lack of proper control, application of DE-502 or the free maternity law tends to be mostly at the discretion of those in charge at health centers and hospitals, creating much discontent among users. In subsequent years, more incentives have been given for self-management of hospitals allowing for charging of higher user

81

fees in exchange for lower public transfers. As the former in practice have become more secure and easier forms of financing, the trend has been for health centers and hospitals to extent cost-recovery and increase fees.

4.27 Institutional and Regulatory Reforms (1997–2000). This period is characterized by a series of legal changes to the organization of the health system, for which support was sought or pressure was provided by stakeholders outside the health system. The establishment of a National Health and Social Security System was introduced as an article in the constitution.

rough social participation committees that monitor the performance of health centers.

n irregular and a

). The budget for the Maternidad Gratuita program

More concretely, a new project—Modersa—was to complement the FASBASE program and overcome some of its difficulties through a modernization of the health delivery system. Modersa, also financed by the World Bank, aimed at initiating a process of decentralization of health decision-making to local governments (provinces, municipalities) and encourage community participation to ensure health services meet local demands. The program would also establish contracts with private and public providers, stimulating greater competition and consumer choice. A new regulatory framework was introduced to deal with problems of adverse selection in the market for private health insurers, as well as a regulatory system to standardize quality control and accreditation of hospitals. There was also an attempt to introduce an ambitious new system of human resource management to provide better career perspectives and incentives to public health sector workers, but this plan proved too ambitious for the MoH to handle.

4.28 To provide a legal framework for equity enhancing measures, Modersa managed to get laws and decrees approved establishing, among others, the autonomy of hospitals in decision-making and health management through DE-520 as mentioned above and mechanisms of social control thCongress and the association of municipalities pushed for further decentralization of health care towards municipalities (Ley de Descentralización y Participación Popular of 1999), while the National Council of Women (CONAMU) effectively pressured for free maternity care (through the Ley de Maternidad Gratuita y Atención Infantil of 1998). In addition, a law was introduced regulating the supply of generic medicines for the poorer segments of society.

4.29 Yet much of all this is still to be widely implemented. Only five municipalities (Bucay, Cotacachi, Chordeleg, Loja, and Santo Domingo) had obtained decentralized responsibility for health management by early 2004. The law for generic medicines came with a special budget for the MoH to purchase such drugs, but subsequent budget allocations have beelack of inventory and storage capacity has caused shortages in supply. Therefore, the stark increase in demand for medicines (which increased by 155 percent between 1988 and 1998) has not been satisfied through this mechanism.

4.30 No precise records exist, but the free maternity health service is assumed to have reached between 1.5 and 2 million beneficiaries annually between 2001 and 2003, consisting in equal shares of pregnant women and the recently born. This represents about 30 percent of the target population (see Statistical Annex Table A52was around US$20 million for 2003 and 2004, of which about 80 percent is financed by the Fondo de Solidaridad and the remainder by a fixed share (3 percent) of excise taxes on alcohol and tobacco. In addition, several other public institutions and international organizations (INNFA, MoH, municipalities, PAHO/WHO, UNICEF, GTZ, and CONAMU) have contributed in communicating the program’s benefits to the population. The budget is allocated on a per capita basis to health centers in municipalities (cantons), according to the number of women and

82

young children in the area. An additional 10 percent allocation is given to those municipalities with a poverty incidence higher than 70 percent, with limited access to health services and with above average maternal and infant mortality rates.19 The budget covers only the cost of medicine and basic health equipment for maternal care and no provision is made for health personnel costs or hospital infrastructure. This budget decision is based on the assumption that public health units have such costs already covered and, implicitly, that health units have excess capacity. There is no impact evaluation of the program, but experts in the field say that it has been effective in providing better medical assistance at child delivery and therefore likely has contributed to a reduction in the infant mortality rate. Nonetheless, the program still faces several challenges, particularly in improving access for indigenous and rural women and overcoming implementation problems. Expansion of coverage of the program is hampered by the lack of control of cost-recovery practices at hospitals and health centers as mentioned above and by implementation problems at the municipal level. As to the latter problem, resources for the program are transferred to municipalities, which compensate local health centers on the basis of the number of women and children that have received medical assistance. However, this decentralized operation is very recent and thus far only 42 out of 219 municipalities have opened an account to receive the funds, while most other municipalities still need to overcome administrative hurdles before being able to manage the resource flow.

4.31 Towards Universal Health Insurance, Decentralization and Demand Subsidies (2000–04). During this period, the ambitions of Modersa were toned down in the face of slow progress on several fronts in the modernization of the health sector. New priorities were no longer set with Modersa as the main driving force, but by agents outside the health ministry, including the

ears, partly because of delayed payment of

nitiatives by several municipalities (Guayaquil, Quito, Cotacachi, among )

council for the modernization of the state (CONAM) and the ministry of social affairs (MBS/STFS) that has a lead in the design of social protection programs. Four new priorities were set: (i) renewed priority for expanding coverage of basic infrastructure and equipment of health centers; (ii) health insurance reform; (iii) new impetus to the decentralization process; and (iv) introduction of new forms of demand subsidies.

4.32 Budget allocations during this period were not consistent with the first priority and, as discussed in the next section, most of the health budget increases in this period were for wage adjustments for medical personnel in public health centers, particularly in 2002 and 2003. Strikes of medical personnel were frequent in these ysalaries and in part also because of complaints of a lack of resources for maintenance, equipment and drugs. Scarcity of drug supplies in public health centers increasingly forced most patients (including social security affiliates) to purchase medicines themselves in private pharmacies, increasing out-of-pocket expenses and likely reducing access to health care for the poor (see below, Section 5).

4.33 Universal coverage of health insurance was launched as a strategic objective for the health sector. Concrete and workable proposals in this direction as yet have to be worked out, but the issue has set much of the tone for health sector reform discussions during 2003. This has included separate iothers , aiming to develop their own health insurance schemes for their local constituencies. None of these cases have developed into operational schemes. In Guayaquil’s case part of the plan is to incorporate IESS affiliates into the municipal plan and for beneficiaries of the Bono de

19. See Executive Decree No. 2704 (article 16) of June 5, 2002.

83

Desarrollo Humano (see below) US$1 of their payment would be withheld for insurance coverage of a basic package of services. However, this is not a worked-out plan and there is no clarity on how the insurance plan would lead to universal coverage or what providers it will give access to. The plan of taking US$1 from an already small cash transfer to the poor is being opposed by several central government agencies, as well as international organizations. Quito’s plan has moved to the stage of negotiations with public and private providers about coverage of services and tariffs for treatments. It is far from clear whether the outcomes will be such that the insurance could become universal and affordable and would cover an adequate basic package of services. It is difficult to see how these initiatives will evolve into fully workable schemes without considering simultaneously reform of IESS and public health providers.

4.34 In 2001, Congress approved a new social security law that tries to address some of the fundamental weaknesses of the system and allow for reforms to revert the financial crisis it faces. The law separates the functions of health provisioning and financing and establishes a health inspection to improve regulation and control of the system. Supervision of health financing and

pal and provincial

insurance is put in the hands of the regulators of the banking system (Superintendencia de Bancos y Seguros). The law is held up, however, in the Constitutional Court because of objections to two of its articles. Nonetheless, a start has been made with separating the functions of provisioning and financing and modest progress has been made to allow coverage of family members (spouses and children) of social security affiliates. None of this, however, has brought a solution to the system’s financial crisis much closer. The number of affiliates and their contributions are simply too low to cover rising health demands and the increasing costs of care. The reform of IESS should become part of the proposed move towards a universal health insurance system, but unfortunately to date these are taken as separate issues.

4.35 In 2003, the national commission on decentralization was established to oversee the implementation of the shift of responsibilities and resources to local governments as established in the Ley de Descentralización y Participación Popular (The Law of Decentralization and Popular Participation) of 1997. This should lead to a greater role for municigovernments in the delivery of health services, as well as to greater autonomy of the health institutions themselves. The main actors in the health sector tend to agree the system should move in this direction, although there is still a stumbling block around the precise areas of responsibility of provincial and municipal governments. In addition, there are a range of implementation problems, including the ones already mentioned related to the transfer of resources to municipalities for the Maternidad Gratuita program, problems in hiring qualified personnel by local governments, and a lack of health monitoring systems in support of health management at the local level and the lack of surveillance of attainment of national and centrally established goals. A new budgetary procedure was introduced in 2003 to transfer funds to local governments (municipalities) that have become part of the decentralized system of health delivery. Municipal health directorates are to administer the resources of public hospitals and health centers and allocate these to the service providers. The municipal health administration is to be monitored by local health councils in charge of consensus building around health plans and monitoring their effective implementation. National public health programs will remain funded centrally. Figures 4.9a and 4.9b give a schematic presentation of the change in the budget transfer mechanism in public health. In practice, however, by early 2004 no decentralized transfers had been effectuated through this new system.

84

Figure 4.9a. Current, Centralized Financing of Health Care Provided by Ministry of HealthFigure 4.9a. Current, Centralized Financing of Health Care Provided by Ministry of Health

Central Government

Ministry of Economy and Finance

MOH: Nationwide Special Programs

MOH: Provincial Department

MOH: Regional Health Sections, Hospitals

Health PostHealth Sub-Centers

Health Centers

Central Government





Health PostHealth PostHealth Sub-Centers

Health Sub-Centers

Health CentersHealth Centers

Central Government

85





Health PostHealth PostHealth Sub-Centers

Health Sub-Centers

Health Centers

Central Government



Health Post



Health Centers Health PostHealth Sub-

CentersHealth Sub-

CentersHealth CentersHealth CentersHealth CentersHealth Centers

Central Government




Municipality: Municipal Health

Department

Health PostHealth Sub-

CentersHealth Centers

Regional Health Sections, Decentralized Health

Providers

Municipal Hospital

Municipal Health Center

1. Sectoral consensus2. Health planning3. Monitoring and evaluation of plan4. Accountability

1. Local health policies2. Local regulation and control3. Monitoring and evaluation4. Financing and resource management5. Local public

Central Government





Department

Health PostHealth PostHealth Sub-

CentersHealth Sub-

CentersHealth CentersHealth Centers


Providers

Municipal Hospital

Municipal Hospital





Figure 4.9b. Fiscal Decentralization of Health Care Provided by Ministry of Health

Central Government





Department


CentersHealth Sub-



Providers

Municipal Hospital

Municipal Hospital





Central Government





Department


CentersHealth Sub-



Providers

Municipal Hospital

Municipal Hospital









Figure 4.9b. Fiscal Decentralization of Health Care Provided by Ministry of Health

Box 4.1. Decentralization of Health Care in Cotacachi The municipality of Cotachachi in the province of Imbabura has made the most progress in establishing a decentralized health system. The municipality has 40,000 inhabitants comprising (mostly) indigenous people, as well as an Afro-Ecuadorian and Mestizo population. It comprises 9 parroquías (or communities) of which 8 are rural and 1 urban. According to the poverty map for Ecuador, the poverty incidence as measured by the unsatisfied Basic Need (UBN) index in the urban area of Cotacachi is 64 percent, while that for the rural communities ranges between 70 and 95 percent. In 1996, the first municipal general assembly—a democratic practice based on age-old indigenous traditions—established an intersectoral health committee (along with other such committees for education, tourism, environment, municipal governance and production). This committee (CIS) has been institutionalized and develops a municipal health plan for Cotacachi. Based on an integral diagnosis of health risk, it proposed a complete restructuring of health care provisioning and preventive health care. The municipality has 1 hospital, 7 primary health centers and 2 health posts. Equipment and coverage of services at the hospital was still rather basic by the end of the 1990s and is to be restructured to provide medical services that fit the municipality’s epidemiological profile better. In the same way, a reorganization for the primary health centers is proposed. Most importantly, intercultural medical practices have been designed that combine elements of traditional and modern medicine in an attempt to attract more citizens to the health system and provide care that responds better to their perceived needs and perceptions. The costs of restructuring the health system have been fully detailed according to existing budget procedures. For this, technical teams from the Ministry of Economy and Finance, Ministry of Health, and the municipality of Cotacachi worked together on a joint budget proposal. Local monitoring and evaluation mechanisms are to monitor the proper use of the resources and exercise quality control of health services. On the basis of the draft local health plan and the budget negotiations between the municipality and the Ministry of Health, an agreement was reached—signed by the President—to establish a decentralized health system for Cotacachi according to the locally developed plan in 2003. The flow of the resource has yet to reach Cotacachi’s health system. The effectiveness of this new health care model thus still needs to be proven in practice.

4.36 It seems that matters are stalled at the Ministry of Economy, although as the implementation problems with the Maternidad Gratuita program suggest, many municipalities may not yet be capable of operating the new system. On the other hand, there are some municipalities that have successfully reorganized health care even ahead of the new system, such as Cotacachi (see Box 4.1) conditional cash transfer program, Bono de Desarrollo Humano (BDH), was introduced in 2003, which is to gradually replace existing cash transfer systems, specifically the Beca Escolar (School Payment) and the Bono Solidario (Solidarity Payment). The conditionality consists of the family having either children attending primary school or mothers and young children attending health centers.20 The BDH was set up as part of discussion on improving Ecuador’s social safety net. The idea of introducing direct health demand subsidies was already under earlier consideration as a separate program, but was brought under the BDH as part of an effort to better integrate social policies through the Social Cabinet (Frente Social). While still in the process of absorbing the eligible beneficiaries of the Bono Solidario program, the BDH reached about 1.2 million beneficiaries in the first quarter of 2004. Preliminary evidence also shows that targeting inefficiencies are also still present in the BDH program, indicating that about 31 percent of benefits are leaking to non-poor and undercoverage of eligible poor is about 42 percent.21 Apart from this, two issues emerge regarding health benefits. First, because the BDH has triple objectives (providing social protection, increasing school attendance, and improving access to health care for mothers and young children), it is not immediately clear how much of the program’s impact will serve the health objective directly. It

20 Specifically, the conditions for the target group are that in education: the nuclear family has children in age 6-

15, which are enrolled in school and are registered to attend at least 90 percent of classes; in health: the nuclear family has children in the age group of 0-6 years who should be receiving bi-monthly health controls at designated health centers; for families with children in both age groups, the conditions for education prevail in order to receive the cash transfer.

21 Calculated on the basis of the SIEH survey of December 2003. Eligible poor are defined through the SELBEN composite welfare indicator, which is also used by the program to identify eligible and non-eligible population. (See Statistical Annex Table A54 for details).

86

may be assumed that, families with children of school-going age may be using proof of school attendance, rather than that of visiting health centers to access the cash transfer, as the former condition prevails (see footnote 20). Beneficiaries with only children aged 5 or younger comprise no more than 5 percent of the total. This would suggest that with the current coverage of the BDH, the cash transfer would essentially provide a stimulus to education with a likely small direct impact on access to maternity care. Given the substantially higher infant mortality rates for the poor, rural and indigenous population and given that the BDH in principle is targeted at these groups, it seems worth considering ensuring better coverage of households within these groups with children under 5. Second, the BDH is not implemented in coordination with the PANN 2000 or Maternidad Gratuita programs. The latter is not targeted, but aims at enhancing access for the poorest to maternal health care. The former is targeted towards the poor, but as discussed in Section 2 has rather low coverage. The BDH would provide an extra incentive to access such health care for the poorest in rural areas in particular. From the discussion in Section 2, we conclude that special attention should be given so that the three incentive schemes will effectively help to reach the indigenous poor as well. Mechanisms of social control to ensure that the two programs effectively provide adequate maternal care still need to come into effect, but could be the channel through which the coordination of the three programs takes place.

III. Trends in Health Expenditures 4.37 Social expenditure levels are low in Ecuador compared to other Latin American countries, both as a share of GDP and on a per-capita basis (see Vos et al., 2003). Real expenditures per capita have fallen more or less continuously over the past two decades. Ecuador spent approximately 4–5 percent of GDP (or on average about US$55 per capita) on the social sectors during the 1990s, compared to a Latin American average of 12 percent (or US$550 per capita).22 The per-capita figure for Ecuador improves slightly, to approximately US$130 or around 9 percent of GDP, when social security benefits are included. However, pensions are paid only to those retiring from jobs in the formal sector and mostly do not reach the poorer segments of society.23 Real per capita social expenditure has fallen staggeringly since the early 1980s, and, although there has been a visible recovery since 2000, it currently stands below levels reached a quarter of a century ago (see Figure 4.10).

4.38 The decline in social expenditures has hit education and health spending hardest. During the 1990s, the composition of social spending shifted in favor of targeted social protection programs (including the introduction of the cash transfer program Bono Solidario) and against budgets for universal social services in education and health. Between 2001 and 2003,

Source: Vos et al. (2003); updated for 2001-3 from M inistry of Economy-UN ICEF fiscal data base. Public expenditures refer to central government budget only. Social expenditures include education, health, and social assistance (including cash transfer programs).

Figure 4.10. Trends in Real per Capita Health Expenditures and Total Social Expenditures of Central Government, 1973-2004 (in US$ of 2000)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03

per

capi

ta so

cial

exp

. (U

S$ o

f 200

0)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

per

capi

ta h

ealth

exp

. (U

S$ o

f 200

0)Total social exp.(left-hand scale)

Health (right-hand scale)

Source: Vos et al. (2003); updated for 2001-3 from M inistry of Economy-UN ICEF fiscal data base. Public expenditures refer to central government budget only. Social expenditures include education, health, and social assistance (including cash transfer programs).


0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03

per

capi

ta so

cial

exp

. (U

S$ o

f 200

0)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

per

capi

ta h

ealth

exp

. (U

S$ o

f 200




0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03

per

capi

ta so

cial

exp

. (U

S$ o

f 200

0)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

per

capi

ta h

ealth

exp

. (U

S$ o

f 200



0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03

per

capi

ta so

cial

exp

. (U

S$ o

f 200

0)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

per

capi

ta h

ealth

exp

. (U

S$ o

f 200



22 Figures are in constant US dollars of 1997 as reported in ECLAC (2001) and, for Ecuador, Vos et al. (2003).

Data refer to spending by the central government only and do not include social security. 23 This observation needs some qualification though as, after the 1999 crises, the real value of pension benefits

plunged and financial assets of pensioners were decimated, turning many elderly into ‘new poor’.

87

education and health budgets increased significantly, mainly due to various rounds of salary increases for teachers and medical personnel in public service, as personnel spending carried much of the burden of budget cuts during the 1990s. Figure 4.10 also shows the dramatic decline in real per capita public spending on health services during the 1990s. Per capita health expenditures continued to increase during the 1980s, from US$10 per capita in 1980 to US$17 in 1990. During the 1990s health expenditure fell steeply, to return back to the 1980 level by 2000. During 2001 and 2003, salary adjustments as well as the expansion of special health programs, like the Maternidad Gratuita, led to a significant recovery in spending and back to the level reached in 1990. In any case, the level of public health expenditures is still very low by international standards and in 2003–04 did not surpass a meager 1.2 percent of GDP. No systematic data are available for the health expenditures of the social security system, which not only provides health insurance, but also health care. IESS and the Farmers Insurance scheme (Seguridad Social Campesino, SSC) together add another 0.8 percent of Gross Domestic Product (GDP) (about US$10 per capita) to total public health expenditures. This share appears to have been rather stable since 1995 (Table 4.3). The cost of services provided by the SSC to poor farmers amounts to 0.1 percent of GDP.

Table 4.3. Social Expenditure1 of Central Government as a Percentage of GDP2

1973 1975 1980 1985 1990 1995 2000 2001 2002 2003 20043

Total Social Expenditure 3.5 3.3 5.3 4.7 4.4 3.6 4.0 4.3 5.1 5.0 4.8 Education 2.9 2.5 4.3 3.5 2.8 2.4 1.9 2.2 2.7 2.6 2.6 Health 0.5 0.7 0.9 1.1 1.3 0.9 0.8 0.8 1.3 1.3 1.2 Social Assistance 0.1 0.1 0.1 0.1 0.3 0.3 1.3 1.3 1.1 1.1 1.0 - Cash transfer programs 0.8 0.7 0.5 0.6 0.6 Other 0.1 0.1 0.1 0.1 0.3 0.3 0.5 0.6 0.5 0.5 0.5 Memo: Health-related spending of Social Security System - IESS - Farmers Insurance (SSC)

0.8 0.7 0.1

0.7 0.6 0.1

0.8 0.7 0.1

0.8 0.8

Source: Central Bank and Ministry of Finance data. Updated and adjusted series from Vos, et al. (2003). Notes: (1) Social expenditures refer to central government budget only. Social expenditures include education, health, social welfare and labor, and cash transfer programs. Cash transfer program refers to Bono Solidario for 1999–2002 and includes Beca Escolar and Bono de Desarrollo Humanothereafter. (2) Social expenditure share in GDP calculated on the basis of constant price series in dollars of 2000. The share at current prices is slightly higher on average (0.3 percent points for the 1990s and 0.1 percent for the whole series), but the trends are the same. The difference between the constant and current price shares is explained by the difference in deflators for government spending and GDP, the former being—on average—slightly higher. (3) Numbers for 2004 refer to provisional budget for social expenditure data and Central Bank projection of GDP growth.

88

4.39 The measurement of the recovery in real health spending from 2000 onwards is sensitive to the choice of deflator, but this is not so much the case for the years prior to that. In Figure 4.11, health spending has been deflated using, alternatively, a weighted price for public consumption and investment (weighted for their respective importance in health spending)24 and a nominal wage index for public employees. As much of health spending is on salaries (about 60 percent of the total in 2003) and recent budget increases have been mainly driven by salary adjustments, the wage index should be a good proxy for price changes in health services.25 Figure 4.11 shows that if one takes the wage deflator, there is no actual recovery in real health spending, confirming the hypothesis that most, if not all, budget increases after 2000 went into rising (nominal) salaries of medical personnel and suggesting that the overall rise in real per capita social expenditures during 2000–04 is overstated in Figure 4.10.

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Real spending (wage deflator)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002


Source: Data series used in Vos e al. (2003) and updated for 2001-2 from Ministry of Economy –UNICEF fiscal data base. Deflators are from the Central Bank’s national accounts (NA deflators and the INEC Urban Household Surveys (for the average wage of public employees).

Figure 4.11. Real Health Spending per Capita, Using Altern-ative Deflators, 1982-2002 (in constant price dollars of 2000)

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002


0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002


Source: Data series used in Vos e al. (2003) and updated for 2001-2 from Ministry of Economy –UNICEF fiscal data base. Deflators are from the Central Bank’s national accounts (NA deflators and the INEC Urban Household Surveys (for the average wage of public employees).

Figure 4.11. Real Health Spending per Capita, Using Altern-ative Deflators, 1982-2002 (in constant price dollars of 2000)

4.40 How have health services been affected by the decline in real public spending.? First, there could have been a shift in the structure of health spending, e.g. from more expensive curative care to cheaper preventative care. No sufficiently detailed health budget data are available to check whether there has been such shift, but the discussion of health policies would suggest no major shift in this sense. Yet in practice, some of this has happened due to a shift in provisioning of hospitals towards the private sector and a shift in the public health sector budget towards cheaper outpatient care

0.0

5.0

10.0

15.0

20.0

25.0

1974

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

hosp

ital b

eds p

er 1

00,0

00

total iess msp privados otros

Source: INEC, Anuario de recursos y actividades de salud , various years.

Figure 4.12. Health Input Supplies: Number of Hospital Beds (per 100,000 inhabitants), 1994-2001

0.0

5.0

10.0

15.0

20.0

25.0

1974

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

hosp

ital b

eds p

er 1

00,0

00


0.0

5.0

10.0

15.0

20.0

25.0

1974

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

hosp

ital b

eds p

er 1

00,0

00


Source: INEC, Anuario de recursos y actividades de salud , various years.

Figure 4.12. Health Input Supplies: Number of Hospital Beds (per 100,000 inhabitants), 1994-2001

24 There are no consistent long data series detailing public expenditures by sector and by expenditure items (e.g.

wages, other current expenditures and investment). A national accounts series with disaggregated public sector accounts indicates that historically a bit less than 10 percent of the health budget was spent on investment in health infrastructure but that this increased to about 25 percent in the early 90s, declining to about 10 percent by 1995. We use the latter share for weighting the two deflators.

25 Again, no consistent published breakdown exists for public expenditures, let alone social expenditures by ages and other expenditure components. Available breakdowns from a UNICEF-Ministry of Economy and Finance database show a somewhat erratic pattern for 1995–2000 probably due to classification errors. For 1996–98 the share is about 70 percent jumping to over 80 percent in 1999-2000. The national accounts series quoted in the previous footnote would put the share at around 75 percent during the period 1980–92. This share may be somewhat lower for overall public consumption, but likely the national accounts deflator for government consumption should by and large reflect the trend in the unit cost (wage rate) of public employees. Figure 4.5 shows this appears to be the case until 2000, but not thereafter.

89

through health centers. This may be derived from health input data:

• Coverage of hospital beds fell from 20 to 16 per 100,000 inhabitants between 1974

• utions doubled between 1980 and 2000, but most of

• of medical personnel has increased over the past two decades. Most

and 2001 (see Figure 4.12). Most of this decline is due to reduced coverage by public hospitals. Coverage by private clinics almost doubled during this period from 2.4 to 4.7 beds per 100,000 inhabitants.

The number of public health institthis increase is due to the expansion of out-patient care. The number of public hospitals increased from 144 to 178 in this period, while that of (cheaper) health centers without hospital beds increased from 1,406 to 2,839. The number of private hospitals more than tripled during the period 1980–2000, with 165 new private hospitals emerging in the 1990s. Private agents now run 70 percent of hospitals, though no more than 30 percent of hospital beds, suggesting strong growth of small private clinics.

The availabilitynotably, the number of doctors per 10,000 inhabitants increased from 4.8 to 16.0 between 1975 and 2001 and the availability of nurses increased from 1.4 to 5.1 (see Figure 4.13a). The rising trend continued during the 1990s. However, again the increase is entirely due to the expansion of private health providers. As shown in Figure 4.13b, the availability of doctors in public health institutions stagnated around 8 per 10,000 inhabitants during the 1990s, whereas in private clinics it quadrupled from 1.8 in 1990 to 7.9 in 2001. The supply of nurses increased in both private and public segments, respectively from 0.5 to 0.9 and from 3.0 to 4.2 per 10,000 between 1990 and 2001, but the ratio of nurses per doctor increased in the public sector segment (from 0.4 to 0.5), while it decreased in the private segment (from 0.25 to 0.1). The increase in the share of nurses to doctors in public health care implies a shift from a more expensive to a cheaper composition of medical personnel and thereby (given more or less stagnant coverage) explains in part the decline in public health expenditures. Nonetheless, Ecuador’s composition of health workers seems excessively biased towards doctors. Most countries in the world, particularly more developed countries, tend to have more nurses than doctors. Ecuador has the same ratio of doctors per 10,000 inhabitants as Great Britain, but the availability of nurses

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

1975 1980 1985 1990 1995 1998 1999 2000 2001

med

ical

per

sonn

el p

er 1

0,00

0

Doctors Nurses Auxiliary personnel

.

Nurses-public Nurses-private

Source: INEC, Anuario de recursos y actividades de salud, various years.


0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001

med

ical p

erso

nnel

per 1

0,00

0

Doctors-public Doctors-private

Personnel (per 10,000 Inhabitants) of Medical Personnel (per 10,000 Inhabitants)Figure 4.13a. Health Input Supplies: Total Coverage of Medical Figure 4.13b. Health Input Supplies: Public Versus Private Coverage

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

1975 1980 1985 1990 1995 1998 1999 2000 2001

med

ical

per

sonn

el p

er 1

0,00

0

Doctors Nurses Auxiliary personnel

.

Nurses-public Nurses-private



0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001

med

ical p

erso

nnel

per 1

0,00

0

Doctors-public Doctors-private

Personnel (per 10,000 Inhabitants) of Medical Personnel (per 10,000 Inhabitants)Figure 4.13a. Health Input Supplies: Total Coverage of Medical Figure 4.13b. Health Input Supplies: Public Versus Private Coverage

90

is ten times less. The shift in services towards the private sector (mostly small clinics) has exacerbated this imbalance.

4.41 The shift towards private provisioning eased pressure on the public health budget, but—as discussed further in Section 6—it also reduced access of the poor to health care. The shift in public provisioning towards primary health care is not immediately visible in the health budget structure. The shares of spending for primary health care centers and public hospitals decreased between 1995 and 2004, as Figure 4.14 shows. That for primary health centers decreased from 28 to 23 percent and that for hospitals from 41 to 33 percent. The share of expenditures for special public health programs (drugs supplies, maternal and child care, immunization and other preventive health care programs) dropped from 22 to 17 percent, despite the introduction of the free maternity program (cost about US$21 million in 2004) and a preventive health care program oriented towards the indigenous population (US$6 million). Most of the drop is due to the decline in the budget for drug supplies. The general wage increases of civil servants and health workers explain the increase in the share of general administration cost of the Ministry of Health.

Figure 4.14. Public Health Budget by Main Types of Care, 1995 and 2004 (percentage shares of total budget)

Source : Ministry of Economy and Finance – UNICEF data base.

0%

10%

20%

30%

40%

50%

19952004

Public Health programs (include free maternity law)

Public Hospitals

Primary Health Care

General Administration

MoH



0%

10%

20%

30%

40%

50%

19952004


Public Hospitals

Primary Health Care


MoH



0%

10%

20%

30%

40%

50%

19952004


Public Hospitals

Primary Health Care


MoH

4.42 Until the year 2000, there was a semi-autonomous institute (CEMEIN) with its own budget in charge of providing medical equipment and drugs to health centers and hospitals. This budget has subsequently been absorbed into the general budget of the MoH, but the total available budget declined from US$2.6 to US$1.6 million between 1995 and 2004. In addition, the budget item is frequently used to cover other current expenditures of the ministry. This has led to the under provisioning of public health centers and patients having to purchase their own medicines at private pharmacies, as signaled earlier. This is a strong indication that the quality of public centers has deteriorated (and this is perceived as such). As we shall discuss in Section 6, this is a factor of influence in access to maternal and child health care.

4.43 Irregular salary payments to medical personnel have become another factor affecting the quality of public health provisioning. Strikes have affected health attention with some frequency during recent years, not just to pressure for salary increases, but also because of serious delays in payment. In addition, in the first quarter of 2004, full-time medical personnel in public health centers decided to reduce working hours (and thus hours of attention) from 8 to 4 hours per day, as a form of protest against recent salary increases which benefited equally (in absolute terms) those working full-time as those working only part-time. Poor management of budget cash flow and the low capacity of the government to build social consensus in the health sector is thus seriously jeopardizing the quantity and quality of health provisioning.

4.44 The creation of the free maternity and child care program and the preventive health care program for the indigenous population could help reduce the gap in infant mortality rates for the poor and indigenous population. However, at the same time—as mentioned in Section 3—the

91

source of primary health care, is progressive and about equi-proportional. Benefits received through public hospitals also show a progressive distribution (that is less unequal than per capita consumption), yet 32 percent of those benefits accrue to the richest 20 percent. Vos et al. (2003) provide these findings based on a detailed analysis of social expenditures in Ecuador based on LSMS data for 1999. Figures 4.15a–c summarize the results in the form of Lorenz curves. The results also show that the aggregate public health expenditures are distributed progressively, but are clearly not pro-poor (Figure 4.15c).26

4.48 Special health and nutrition programs (school meals, PANN-Papilla and Bebida), as well as the BDH are clearly pro-poor as the beneficiary incidence estimates based on the 2003 SIEH survey show (see Figure 15b). Around 50 percent of the beneficiaries of these programs belong to the poorest 20 percent by income. Nonetheless, all these programs show important targeting inefficiencies (see Statistical Annex Table A.9).

4.49 Out-of-pocket expenses in health care are four times higher than public expenditures. The inputed benefits of public health spending represent 2.1 percent, while private expenditures represent 8.6 percent of household consumption (see Table 4.4). Significant efficiency and equity gains could be achieved by merging all centralized interventions, and retargeting them toward meeting the Millenium Development Goals in health.

Figure 4.15c. Incidence of Public and Private Expenditures,

Source:Vos et al. (2003).

0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%

cumulative % populatio n

cum

26 These results are obtained using a traditional expenditure incidence analysis imputing the cost of providing

health services as benefits to households depending on the number of beneficiaries of subsidized health care per household. The estimates have to be taken with the usual caution associated with the method. We will not spell these out here, but see e.g. Demery (2003) and Vos et al. (2003) for a discussion. The estimates in Vos et al. (2003) take into account the differences in unit costs per type of provider when assigning benefits. Benefits are assigned to households as follows:

∑=

≡3

1ii

i

ijj S

EE

X

where Xj is the value of the total health subsidy inputed to household j. Eij represents the number of medical consultations and other individualized health benefits of household j at education level i, and Ei the total number of consultations (across all households) at that level. Si is government spending on health by type of provider i, and i (=1, .. , 3) denotes the type of provider (by two types of classifications: institutional – IESS, MoH, Farmers Insurance – and level of care – hospitals, primary health centers, health dispensaries).

ulat

ive

% b

enef

its

Per capita consumption Equity linePublic health exp. Private health exp.

1999Figure 4.15c. Incidence of Public and Private Expenditures,


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e

Per capita consumption Equity linePublic health exp.

0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e




0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e




0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


1999

0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


0%

20%

40%

60%

80%

100%

0% 20% 40% 60% 80% 100%


cum

% b

enef

itsul

ativ

e


1999

93

4.50 The poorer segments of the population both receive relatively more benefits (as a share of their income) from public health services and spend relatively more themselves on health care.

The poorest 20 percent spend 11.5 percent of total household expenditures on health, while the richest 20 percent spend more in absolute terms, but this represents 7.3 percent of their total consumption. More than half (55 percent) of out-of-pocket health spending is on medicines and this share is somewhat higher for the poorest, who spend more than 60 percent of out-of-pocket expenses on drugs (Table 4.5). The poorest spent no less than 7 percent of total expenditures on medicines in 1999 and this share has increased since 1995. This may reflect rising costs of medicine due to the changing disease pattern and the demographic change in the population as

discussed in Section 2, but given the short time period this rise in private spending on medicine is likely more closely associated with inadequacies in the supplies of medicines at clinics and health centers (and, possibly because of that, rising self-medication). The generic medicine program should have helped reduce costs to households in more recent years, but as discussed in Sections 3 and 4, the program’s reduced budget and failing budget execution probably have led to very little relief for poor households.

Table 4.4. Public Benefits and Out-Of-Pocket Expenses in Health, 1999 (Shares of Total P.C. Household Consumption of Each Decile)

P.C. Consumption

Deciles

Benefit of Public Health

Expenditures

Private Health Expenditures

Poorest 10% 5.5% 12.1% 2 4.0% 10.8% 3 2.4% 9.1% 4 4.1% 11.2% 5 3.9% 10.8% 6 3.2% 10.1% 7 2.6% 11.5% 8 2.4% 10.5% 9 1.9% 8.4%

Richest 10% 0.9% 6.2% Total 2.1% 8.6% Source: Vos et al. (2003) and INEC, ECV (LSMS) 1999.

94

Table 4.5. Out-Of-Pocket Health Expenditures of Households by Deciles and Type of Expenditure (Expressed as Percentages)

P.C. Consumption

Deciles

Medical Consults

Medicines X-rays and Laboratory

Tests

Hospitaliz-ation

Glasses and Dental Prostheses

Other Total

Share of Total Private Consumption Expenditures Poorest 10% 1.4 7.5 0.7 0.9 0.4 1.1 12.1 2 1.2 6.7 0.7 0.1 0.3 1.8 10.8 3 1.0 6.3 0.6 0.1 0.2 0.8 9.1 4 1.1 6.8 0.8 0.7 0.2 1.7 11.2 5 1.0 6.9 0.7 0.4 0.5 1.4 10.8 6 1.3 5.9 0.9 0.2 0.4 1.4 10.1 7 1.0 7.9 1.1 0.1 0.5 0.7 11.5 8 1.3 5.1 1.2 0.3 0.4 2.1 10.5 9 1.2 4.4 0.9 0.2 0.8 1.0 8.4 Richest 10% 1.0 2.8 0.8 0.3 0.6 0.6 6.2 Total 1.1 4.8 0.9 0.3 0.5 1.0 8.6

Share of Out-Of-Pocket Health Expenditure Poorest 10% 11.6 62.2 6.1 7.5 3.0 9.5 100.0 2 11.2 62.3 6.8 0.8 2.6 16.2 100.0 3 10.6 69.5 7.1 1.2 2.7 8.9 100.0 4 9.9 60.3 7.1 6.0 1.7 15.0 100.0 5 8.9 63.5 6.8 3.3 4.8 12.7 100.0 6 12.8 58.2 8.7 2.2 4.4 13.7 100.0 7 9.0 69.0 9.8 1.1 4.6 6.4 100.0 8 12.5 48.8 11.2 3.2 4.1 20.1 100.0 9 13.7 52.4 10.5 2.8 9.2 11.4 100.0 Richest 10% 16.5 45.3 13.8 5.3 9.5 9.7 100.0 Total 12.8 55.1 10.4 3.5 6.1 12.1 100.0 Source: INEC, ECV (LSMS), 1999.

4.51 About 20 percent of private health expenditures are for the treatment of children under the age of 5 (Table 4.6). Again, poorer households tend to spend relatively more than rich households, i.e. about 3 percent of total consumption expenditures for the poorest deciles and 1 percent for the richest deciles. This is a pattern that could be expected, as the poorest quintile has about 30 percent of small children against the richest with about 12 percent. Middle-income groups also tend to have fewer small children than the poorest, but their share of expenditures on health care for children is not very different from the poorest, suggesting they have greater access to health services.

95

V. Access to Health Care and Cost-Effectiveness in Health Spending

4.53 In Section 2, we concluded that infant mortality has decreased substantially in recent decades and that Ecuador seems well on track to reach this MDG target well before 2015. However, we also concluded that large differences in the risk of child death persist across socio-economic groups. In order to reduce such differences (and speed up the reduction in infant mortality) it would be useful to know what interventions would be most cost-effective. For that we will first model determinants of access to health care and those of the risk of early child death. Subsequently, we will relate the determinants that can be linked to policy interventions to the cost of such interventions and simulate the budget implications of alternative policy options aimed at reaching the MDG for child mortality. It is important to notice the close interaction among health and non-health variables in determinants found. This points to the need to consider factors that go beyond health sector spending and institutional capacity.

4.54 Joint Modeling of Access to Health Services During Delivery and Infant Mortality. The socioeconomic profile of infant mortality above concludes that wide gaps by socioeconomic categories, geographical location and vulnerability groups prevail in Ecuador. The critical question to ask is what proportion of these differences is explained by unequal access to health provision, specifically among the poor vis-à-vis the non-poor; and the indigenous population vis-à-vis non-indigenous peoples.

4.55 The present analysis connects access to health services during birth with the phenomenon of infant mortality. Traditionally, the health demand literature has focused on separating the effects of price-related and socioeconomic-related variables shaping access to services (see e.g. Gertler, Locay and Sanderson, 1987; Gertler and van der Gaag, 1990). In addition, most previous infant mortality studies have emphasized the likely mutual causation between infant mortality and fertility decisions (Hanmer and White, 1998). However, a combined specification of infant mortality and the choice of health service has not been attempted before.

4.56 The new approach explores first what motivates Ecuadorian individuals when selecting among different health services (either public or private or no service at all). Once these determinants are known, the effects of access to health services on infant mortality are analyzed. As a result, public health policy strategies are linked to the key development goal of infant mortality reduction. This link is two-fold: one shows the effects of private and public interventions on infant mortality directly; the other shows the impact on infant mortality through its induced change to health demand.

4.57 The traditional emphasis in the previous literature is not completely overlooked, though. Price considerations are not directly included although controlling for affiliation (if any) to different health provision categories may well convey price differences faced by health users. The relation between fertility and infant mortality is taken into account implicitly since many of the determinants of fertility are also factors affecting mortality. In addition, interesting behavioral relationships such as replacement and/or resource-competition among household members are controlled for in the proposed infant mortality specification. This is believed to capture the consequences of an intertwined relation between fertility and mortality.

4.58 Data. The database for both models comes from the demographic and health survey (ENDEMAIN) of 1999 that investigated in a exhaustive manner the birth history of every single child born alive in a period of six year before this survey, as well as the use of reproductive

97

health services by mothers and sanitary conditions of the household. In this sense, ENDEMAIN stands out from other surveys such as the LSMS and the population census of 2001 that ask information of mortality just for the last child born and also do not find out the age at which the children died. The sample size of the 1999 ENDEMAIN is representative at the national, regional and provincial levels. The data set for the CPH modeling is composed of 9,391 children born alive since January 1994, of which 301 died before the first year and 9,090 survived. For the health demand model, the number of observations for pregnant women that gave birth is 8,789. See Annex 4.3 for further details on the data sources.

4.59 Modeling Strategy. We apply two techniques to estimate the demand for health services for peri- and post-natal maternal care and the determinants of infant mortality: a multinomial logit model of demand for health care and the Cox Proportional Hazard model to estimate the probability of child survival at early age. The general specifications of these models may be summarized as follows:

Estimation Estimation technique Model function Specification Demand for health services during child delivery

Multinomial Logit (MNL)

∑=

== n

k

z

z

iik

ii

e

ejPSob

0

)(Prβ

β [1]

PSi : health provider ‘i’ (none, public, private); zi: determinants of health care demand for child delivery.

Infant mortality determinants

Cox Proportional Hazard (CPH) survival model )()( 0 tHetH ij

ij x

j

∑=

β

[2]

Hi(t): risk of infant ‘j’ to die in period (t) before reaching one year of age; H0(t): risk of infant of reference group to die in period (t) before reaching one year of age; xi: determinants of infant mortality.

Source: Author’s estimate

4.60 The Multinomial Logit Model (MNL) is an appropriate method to analyze multiple, discrete decisions where the dependent variable may take at least three different values. In the demand for maternal health care we distinguish three options: no access to health care, public and private providers. The MNL specification is further appropriate if we assume that the alternatives for choice of provider are independent. That is, the probability of choosing one or the other option does not depend on the existence of other option(s). In the case of the choice for health care for child delivery, this hypothesis implies that a pregnant woman who is making a choice between receiving maternal health care or not, or between a public and private provider, is not influenced by the presence of an alternative care option. This assumption is likely to be valid as long as the differences in cost or quality between the options are sufficiently large. The assumption may not be true if, for instance, there would be little difference in cost between private and public services, or if the quality of public health care were perceived to be so poor that potential users think they might be equally well off receiving no care at all. Should such conditions hold, the preferred method would be a conditional logit or ordered probit model, which would allow interdependence in the ordering of alternative choices. In Ecuador, there is no a priori reason for any user to first try a public service before soliciting private service. Hence, we assume the MNL assumption is valid for our case. The model specification is based on consumption theory as applicable to health care demand. The theoretical foundations are spelled out in Annex 4.1.

4.61 ‘Hazard.’ models estimate the relative probability of a particular group incurring a particular risk as compared to another reference group. For instance, one could estimate the probability of infant death of girls relative to that of boys. Similarly, survival models estimate the

98

same type of probability but take as the dependent variable the time (number of months) it takes for the event (death) to occur or not to occur within a given time frame (one year, in case of infant mortality). Probit or logit models have often been used to estimate determinants of infant mortality, but this requires defining a binary or discretionary dependent variable (death or no death). In the case of the Cox Proportional Hazard (CPH) or survival model, we are allowed to consider all observed cases of children under age one, whereas a logit or probit specification (when properly estimated) would require considering only relevant cases that have actually reached age one to determine survival or non-survival. As a consequence, the CPH model does not require scrutinizing the sample of infants first and avoids possible biases in the estimators. Such a bias will necessarily be present in probit or logit estimations to the extent that the probability of dying in the first year of life is concentrated during specific months (say the first) after birth.27 Models using a binary or discrete dependent variable would thus considerably overestimate the impact of the given set of determinants on infant mortality, as the number of non-survivors per month lived would be incorrectly weighted.

4.62 Duration models have been extensively used in socioeconomic analysis in issues such as unemployment spells, education enrolment or social benefit schemes (see e.g. Nickell et al., 1991). There is also a rather extensive epidemiological literature on the duration of health conditions including, infant, child and adult mortality (Masset and White, 2003). We spell out the theoretical foundations of the CPH model in greater detail in Annex 4.1.

4.63 Model Specification and Specification Problems. The existing literature suggests a large number of possible determinants of access to health services and infant mortality. Tables 4.7 and 4.8 summarize the expected effects of a wide possible range of determinants, including the ones analyzed in Section 2.

4.64 It is widely recognized that the estimation of health production functions is subject to problems of interdependence and endogeneity. Many of the likely determinants, such as family income and parents’ education are interrelated, such that having both as determinants could give biased estimators. In addition, there tend to exist mutual relationships between health inputs (supply variables) and health outputs (illness incidence) causing endogeneity in the estimation of health production functions (see e.g. Deolalikar, 1996). In particular, there is a potential case for health supply variables at a regional or municipal level to be endogenously determined by the magnitude of infant mortality. If so, higher infant mortality rates would be driving a higher provision of health services.

4.65 The traditional econometric solution to the endogeneity problem is the use of instruments for endogenous variables (Greene, 2001). Obvious candidates for current health supply variables are lagged health supply, based on the assumptions that health policy changes require time both to articulate and to generate expected outcomes. In Ecuador, a practical constraint prevents a satisfactory use of instruments for contemporary supply variables. Statistical sources on supply variables at district (parroquía) level only permit the construction of lagged variables for 1998 and these variables are limited to per capita in-patient hospital services. While this rules out a strategy based on instrumental estimation, the virtually identical significance, size and sign of the infant mortality coefficients with contemporaneous and lagged supply variables confirm that endogeneity may not be a serious problem after all. The hypothesis of no serious endogeneity is

27 In Ecuador, the evidence shows that most child deaths occur within the first five months after birth, thus

reducing considerably the risk of death after having survived the first five months.

99

further confirmed by the very low explanatory power of infant mortality averages on averages of public health expenditures per municipality. In other words, differences in the incidence of infant mortality do not appear to explain much of the distribution of public health provision.

4.66 With respect to the interdependence issue, non-economic variables such as education or geographical location are not entirely independent from economic variables such as consumption levels or poverty status. For instance, education may not only affect infant mortality because it is closely and positively linked to consumption levels (its economic effect) but also because it affects preferences and skills referring to infant care (non-economic effects). All these effects from education are expected to affect infant mortality, and the inclusion of consumption in the infant mortality function acts as a control variable for the economic effects of education. A similar issue may arise with access to drinking water and other sanitation conditions which are likely linked to consumption levels and poverty status, not in the least also because sanitary conditions were used as explanatory variables in the construction of the consumption variable for the ENDEMAIN survey (see Appendix A.3). Apart from checking the relevance of such different dimensions, we also test for multicollinearity and drop variables causing such problems. Table 4.7. The Expected Effects of Possible Determinants on the Demand of Maternal Health Services

Variable Expected effect on demand for

medical services

Hypotheses

Personal characteristics: Mother’s age Ambiguous Effect will depend on the magnitudes of two conflicting effects: increased

demand with age as risk increases vs. decreased demand for professional attention as mother’s experience rises.

Mother’s education Increase in demand for professional services

Partly this effect captures an economic relation (higher education is associated with higher economic status), while it may also capture different preferences between educated and non-educated women. If there are important differences of quality in favor of private health services, one could expect better educated women (with better access to information) to prefer such services.

Mother’s ethnic background

Decrease in demand for professional services

Individuals of ethnic background may prefer traditional delivery practices attended by individuals other than professional medical staff

Household characteristics: Urban location Increase in

demand for professional services

Residence in urban areas typically implies better access to medical facilities for delivery and other health-related practices.

Sierra / Costa location Increase in demand for professional services

Residence in highlands or coastal locations increases access to medical facilities with respect to forest and jungle areas (Amazonía). This variable may well capture therefore differences in public infrastructure that make health services more accessible in certain areas.

Socioeconomic status: (i) per capita consumption level; (ii) per capita consumption separated by poverty status

Increase in demand for professional services

The higher the economic status of the household (proxied by either consumption levels or poverty status), the more likely that a mother is both willing to use and able to afford available medical services offered to her at the time of delivery. Similarly, if there are important perceived differences in quality between public and private services favoring the latter, higher income groups likely will prefer using private health services.

Health inputs and medical conditions: Personal: Affiliation to public social security scheme or private health insurance

Increase in demand for professional services

Affiliation to health systems makes it more likely that mothers use the services covered by their affiliation and those affiliated with the social security system will be expected to use public health services.

Premature delivery Increase in demand for

Complications in the delivery typically increase the demand for professional staff attention.

100

professional services

Prenatal visits Increase in demand for professional services

Prenatal visits may indicate a mother’s higher preference for professional attention throughout pregnancy. These visits indicate also the existence of institutional facilities providing such attentions.

Community health services: District (parroquía) average availability of hospital beds District (parroquía) average availability of medical personnel

Increase in demand for professional services.

Larger public health supply indicators at district level reflect a wider access to public services, which may lead to higher demand for these services in detriment to no-care or private care. There is however a potential problem of endogeneity associated with the fact that the level of public provision can be targeted to communities with lower health supply indicators. Alternatively, the wealthiest communities might also capture more per capita public with political leverage.

Note: These variables refer to zi in the estimation model [1] Table 4.8. The Expected Effects of Possible Determinants on Infant Mortality Variable Expected effect on

infant mortality Comments

Personal and biological characteristics:

Infant’s sex Larger probability of male infant mortality (IM)

Male infants are reported to survive less frequently than females in the first year of life (WHO 2003)

Infant’s ethnic background

Ambiguous Once controlled for education, consumption, behavioral, household composition and community-wide health supply differences among different ethnic groups, it is not clear whether ethnicity-driven aspects have additional impacts on the probability of IM.

Multiple delivery (twins, etc.)

Increase in probability of IM

Multiple delivery increases the probability of less healthy infants.

Premature delivery Increase in probability of IM

Premature delivery increases the probability of less healthy infants.

Birth order Decrease in probability of IM

Under the hypothesis of intra-household competition, higher-ordered infants (first born) are more likely to capture available resources (food, care) in the household to ensure their survival.

Behavioral characteristics of mother:

Type of attention received during delivery

Professional attention decreases probability of IM

Professional attention should reduce the probability of delivery problems causing infant mortality. It is, however, an open question whether the impact on infant mortality is significantly different between private and public health care and most likely will be associated with differences in quality of service.

Prenatal visits Ambiguous More prenatal control visits may indicate a higher preference and larger opportunities for professional attention, thus reducing health risks for the fetus and the infant. It may also indicate, however, that fetus complications exist and that the infant when born is more prone to illness or physical deformation.

Breastfeeding Decreased probability of IM

Biologically, it has been shown that breast feeding improves the nutritional and immunological conditions of the infant. Also, it may indicate the mother is around to personally take care of the infant. Further, breastfeeding provides a natural form of contraceptive, which may help spread the interval between births and reduce risk of early child death.

Contraceptive use Ambiguous Use of contraceptives may indicate that mothers are more aware of – and more willing to use – available medical services regarding infant care as well as other health issues. It will also reduce the number of children born to the mother (fertility) and may help spread the interval between births, reducing the risk of early child death. However, though less likely, it may also increase health risks for mothers if their use is uncontrolled or protracted. It is, however, unlikely that this variable properly captures mother’s health status.

Household characteristics:

Socioeconomic:

101

Variable Expected effect on infant mortality

Comments

Mother’s age Ambiguous Effect will depend on the interaction of two conflicting effects: older age at birth increases health risks for newborn infants but also increases the experience of maternal and infant health care. There might also be a non-linear relation between mother’s age and IM. This kind of argument (hypothesis) is tested by introducing the square of the variable of mother’s age.

Mother’s education Decreased probability of IM

Again, this variable captures an economic relation that implies that better educated parents devote more resources to the benefit of infant’s health. It also accounts for a presumably better use of available child caring technology, which should also positively affect infant health status. Finally, higher education of the mother implies a higher opportunity cost of personally taking care of her infant. This should lead to less time devoted to infant health care. There is, therefore, a possible trade-off between the quality and quantity of time mothers devote to infant care.

Household size Ambiguous Two potential conflicting effects: more individuals in the household may mean higher fertility rates (thus higher health risk) and tougher competition for resources available to infants. But, it may indicate that other household members may take care of infant welfare in case of need.

Socioeconomic status: (i) Poverty status (ii) Per capita consumption

Decreased probability of IM

A higher economic status (captured either through levels of consumption or poverty conditions) is related to a lower probability of IM. However, the socioeconomic effect may be already captured by other variables such as parental education, basic infrastructure in the household or geographical location.

Basic social infrastructure: Water Sanitation


Better sanitation and drinking water access for the household should improve hygienic and health conditions for all members. However, this effect may be already captured by socioeconomic status or education variables, as the distribution of social infrastructure within households is correlated with socioeconomic status (and geographical location).

Geographical: Urban residence Decreased

probability of IM The effects of urban location on IM that are not already captured by economic or educational variables may be due to wider access to health facilities in urban areas with respect to rural areas.

Sierra / Costa residence


Residence in highlands or coastal areas implies more favorable living conditions for infants (and other individuals) with respect to forest and jungle areas (Amazonía), so it should be associated with decreases in IM

Community characteristics:

District (parroquía) immunization rate

Likely decreased probability of IM (might be ambiguous)

On the one hand, higher immunization rates at the community level indicate wider access and use of public health facilities that should lead to improved health status of infants. But, they may also indicate that unfavorable community conditions attract targeted policy intervention. The underlying health performance at community level should then be responsible for higher individual IM rates.

Note: These variables refer to xi in the estimation model [2]

4.67 A final specification problem refers to omitted variables. While trying to test for as many likely determinants as possible, data limitations did not allow for the inclusion of some. For instance, it was not possible to include indicators of mother’s health status during delivery, as these are not reported in the demographic and health survey (ENDEMAIN). Instruments for mothers’ health status (and life style) such as alcoholic or smoking habits are not reported either. The use (abuse) of contraceptives or pre-natal visits can be considered to be imperfect instruments only. The expected relation between mothers’ health and socioeconomic conditions may partially offset this omission. Further, the model does not consider the price of health services, as these are not reported in the ENDEMAIN survey. The consumption variable is also not part of the survey, but could be imputed using an indirect method (see Annex 4.2 for the estimation method). Per capita consumption is used to pick up both the income and cost differential effect on demand for health services, which is of course an imperfect approximation.

102

4.68 Results of the Access to Health Care Model. Tables 4.9a and 4.9b present the results of the demand for health services model in the form of elasticities. More detailed estimation results are in Statistical Annex Table A54a and A54b.

4.69 Personal. characteristics of the mother seem not to have a consistent significant effect on the choice of medical attention. However:

• Indigenous mothers seem not to prefer deliveries attended by professional medical personnel. Being an indigenous mother increases the probability of unattended delivery by 27 percent (marginal effect; Table A.10a). Being non-indigenous increases both the probability of demanding public and private delivery services, roughly by the same amount each (around 13-14 percent).

• Mother’s education is highly significant and the simultaneous significance of per capita consumption indicates that there are non-economic effects underlying this relationship. In fact, the magnitude of the impact of each additional year of education is relatively modest. Each 1 percent increase in the education level decreases the probability of child delivery without professional care by 0.7 percent, while it is expected to increase private service demand by 0.4 percent and public service demand by 0.15 percent.

Table 4.9a. Effects of Determinants on Access to Health Provision Services (Elasticities) Effect on the

probability of no professional

maternity care during delivery

(percent)

Effect on the probability of

public maternity care during

delivery (percent)

Effect on the probability of

private maternity care during delivery

(percent) Personal characteristics Mother’s age motherage 1.162 -0.111 -0.877* Mother’s age squared motheragesq -0.636 0.049 0.504** Mother’s ethnic background ethnic 0.043*** -0.010*** -0.020*** Mother’s education (years of school) motherschool -0.745*** 0.147*** 0.404*** Medical variables (reference: no affiliation to health system) Affiliation to IESS Segiess -0.013*** 0.025*** -0.040** Affiliation to other health insurance system Segother 0.025*** -0.015*** 0.007 Premature delivery Premature -0.020*** 0.005*** 0.010*** Pre-natal visits Prenatal -0.675*** 0.242*** 0.139*** District (parroquía) average supply of inpatient hospital services

tasin99 -0.065 0.092*** -0.130***

District (parroquía) average supply of medical personnel

tasper99 -0.010 0.008*** -0.006***

Household characteristics Urban residence Urban -0.264*** 0.005*** 0.241*** (reference: Amazonía) Costa Coast -0.356*** -0.131*** 0.613*** Sierra Sierra -0.291*** 0.053*** 0.167*** Per capita consumption (log) Logcons -0.832*** 0.010*** 0.773*** Source: Multinomial logit model estimates based on 1999 Demographic and Health Survey for Ecuador (ENDEMAIN). See Statistical Annex Table A54a for detailed estimation results. Notes: * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent

4.70 Estimates for health inputs and medical condition variables confirm the hypotheses of Table 4.7 with some qualifications:

103

• Given the difficulty of using the instruments, the magnitude of estimated coefficients for average availability of in-patient hospital services at the parroquía level confirms the existence of some degree of endogeneity bias.

Table 4.9b. Combined Effect of Poverty and Consumption

Effect on the probability of no professional care during child delivery

(%)

Effect on the probability of public

care during child delivery (%)

Effect on the probability of private

care during child delivery (%)

Per capita consumption of the poor

Povertycons -0.143 *** -0.059 0.258***

Per capita consumption of the non poor

Nopovertycons -0.192 *** -0.050 0.284***

Source: Multinomial Logit Model (MNL) estimates based on 1999 Demographic and Health Survey for Ecuador (ENDEMAIN). See Statistical Annex Table A54b for detailed estimation results. Note: *significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

Table 4.10. Impact Elasticities of Infant Mortality Determinants Variable Effect (percent) Personal and biological characteristics Sex (male=1) sexh 0.168 ** Ethnic background etnia 0.003 Multiple delivery multi 0.006 Premature birth prematu 0.133 *** First born primo -0.149 *** Behavioral factors (reference: no medical assistance) Private health care private -0.175 *** Public health care public -0.116 * Pre-natal controls prenatal -0.461 *** Breastfeeding breast -18.496 *** Use of contraceptives anticon -0.002 Household characteristics Mother’s age motherage -0.714 Mother’s age squared motheragesq 0.352 Mother’s education (years of schooling) motherschool -0.460 ** Household size hhsize -0.993 *** Poor household (dummy times per capita consumption) poverty 0.130 Access to safe drinking water water 0.134 Access to sewerage sanitation 0.225 Urban residence urban 0.042 Costa location coast -0.140 Sierra location sierra -0.188 * (reference: Amazonía) District (parroquía) has a satisfactory immunization record immune -1.102 *** Birth in 1994 b1994 -0.050 Birth in 1995 b1995 -0.060 Birth in 1996 b1996 0.016 Birth in 1997 b1997 -0.044 Birth in 1998 b1998 -0.002 (reference: 1999) Source: Cox Proportional Hazard model of infant survival. Estimates based on 1999 Demographic and Health Survey for Ecuador (ENDEMAIN). See Statistical Annex Table A55 for detailed estimation results. Notes: * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

104

• In contrast, average availability of medical personnel at public health services may capture quality considerations of health provisioning well. The quantitative effect is rather small though. Each 1 percent increase in the supply of health workers (per 100,000) increases the demand for public provisioning by a meager 0.01 percent, while it decreases the demand for private services by 0.01 percent and that of no use of services equally by 0.01 percent. These particular estimates need to be taken with some caution, but it seems safe to conclude that improvements in the supply of health workers will have a very limited impact on the demand for maternal health care.

• Affiliation to the social security system (IESS, SSS) increases, as expected, the

probability of choosing public provision. The effect is rather small and in addition it appears to be (partly) offset by a lower demand for private services. Hence expanding public health insurance per se will not have a big impact on the demand for professional maternity care.

• Premature delivery decreases the probability of unattended delivery, resulting in a

very similar increase in private and public intervention. As a premature case appears, mothers may want to take no chances and ask for professional assistance.

• Pre-natal controls decrease the probability of unattended delivery by almost 20

percent (marginal effect). As a result, there is an increase in the demand for public provision of 15 percent and 4 percent for private provision, reflecting that most prenatal control schemes are public and targeted to the poor.

4.71 Effects on the choice of delivery services from household characteristics are as expected:

• Residence in Urban Areas and Regions. Other than the Amazonía decreases the probability of unattended delivery (by 11 percent in the case of urban residence; 18 percent for Costa; and 14 percent for Sierra), and results typically in increases in public and private provisions. Interestingly, the resulting increase in demand for professional care is stronger for private than public institutions.

• Each 1 percent increase in per capita consumption increases the demand for

private services nearly proportionally (0.7 percent) to the detriment of unattended deliveries, and no substantial change in demand for publicly provided care. This key result may indicate that Ecuadorians are willing to immediately shift to private rather than to public services. This might be attributed to a lack of confidence in the perceived quality of public health care. Also, it is worth noting that the impact of a higher consumption level only differs slightly between poor and non-poor, as Table 9b indicates. An important policy implication would be that an unconditional cash transfer program (such as the Bono Solidario) would be expected to enhance demand for private health care. In contrast, a conditional cash transfer (such as BDH) would increase access to public health care services because of its conditionality and not because the additional income would induce such demand.

105

4.72 Results of the Infant Mortality Model. Estimates from the survival model are reported in Tables A.10 and A.11 and mostly confirm the expected determinants of infant mortality.

4.73 Regarding Personal and Biological Factors:

• Being born as a male or prematurely both increase the probability of infant mortality. Although the impacts are statistically quite significant, their magnitudes are rather low (elasticities are 0.17 and 0.13, respectively).

• Being a first born (with or without other siblings) implies having a lower probability of dying in the first year of age. This is taken as evidence that if there is competition for resources within the household, older siblings have a better chance of capturing resources for survival. Higher risk associated with multiple delivery does not seem to have significant effects.

• More interestingly, indigenous infants are not more prone to infant mortality than non-indigenous for reasons unrelated to socio-economic, household or behavioral factors (controlled for by a different set of variables). What matters more, it seems, are the factors associated with a lack of access to medical assistance for birth delivery.

4.74 Behavioral and Health Policy. Variables seem in general to have significant bearings on infant mortality.

• Professional Medical Care reduces the probability of infant mortality, but the significance and magnitude of this effect varies somewhat by category of service. That is, the effect on infant mortality of private care is larger than the effect of public health care. The elasticities do not differ widely, but it is perhaps of importance that the statistical significance of public attention only passes confidence intervals of 10 percent while the significance of private attention is accepted at 1 percent. Prenatal visits—when conducted—reduce infant mortality, indicating the effectiveness of this intervention in preventing future infant deaths. These findings point out the rising demand for and effectiveness of private over public services, which should be a guiding criteria in future sector reforms.

• Breastfeeding also significantly reduces infant mortality, its effect being the largest among both behavioral and non-behavioral factors. If all women giving birth breastfeed their children, this is expected to reduce infant mortality by 18 percent.

• Coverage of immunization is also very significantly related to lower prevalence of infant mortality. For each 1 percent increase in the coverage of immunization, infant mortality goes down by about 1.1 percent. Once again, the effectiveness of this preventative action stands out.

• Access to safe drinking water and sewerage is not found to have a significant impact on infant mortality. As indicated, this result may well be due to problems of multicollinearity with other determinants, such as education and area of residence. However, we also do not find a very high simple correlation between infant mortality and sanitary conditions, such that the latter might be a conditioning or compounding factor rather than a direct determinant.

106

4.75 Socio-economic characteristics present a mixed combination of effects, substantiating the relations between some of these variables.

• Mother’s age does not affect infant’s survival possibilities, probably indicating the offsetting effects of more experience and more biological risk as age rises.

• However, higher levels of maternal education reduce infant mortality: each percentage increase in the average level of maternal education reduces the probability of infant mortality by 0.5 percent.28

• In contrast, the poverty status of the household does not seem to affect infant mortality. This is not so surprising as maternal education is related to consumption levels and socioeconomic status of the household. The same explanation may underlie the statistical insignificance of water and sanitation services. Although it is difficult to disentangle the possible different effects of education on infant mortality, the fact that this variable strips the significance of other socioeconomic variables may point to the importance of the economic determinants of infant mortality.

• Of the remaining significant determinants of infant mortality, household size affects the probability of infant mortality. As discussed in Table 4.8, we might expect an ambiguous effect, as a larger household size could reflect higher fertility and thus greater health risk, as well as reflecting greater competition for household resources. The alternative hypothesis is that it increases the likelihood that other household members take care of infants. Apparently, this latter effect outweighs the former.

• Geographical Location does not seem to have a substantial impact on infant mortality on its own. Urban residence per se does not reduce infant mortality having already controlled for factors related to geography, such as access to services, or socioeconomic patterns. Residence in the Costa or Sierra does not affect significantly the probability of infant mortality, although there is some weak effect attributed to living in the highlands (Sierra). This may be related to better transport and health infrastructure as compared to Amazonian areas, but this kind of effect is neither substantial nor systematic across non-Amazonian areas.

• Finally, the regression equation includes an additional variable for the year in which the child was born. The latter allows us to check for an unexplained progress over time in reducing infant mortality which remains after controlling for all other microeconomic and non-economic factors. However, the time dummies are not significantly different from zero (not even at a 10 percent confidence interval). Therefore, we can conclude there is no apparent unexplained trend in infant mortality.

28 Several specifications including paternal education and differences in parental education did not prove

significant, though.

107

VI. Health Budget Training

4.76 Ecuador’s government budget system has long been highly centralized, rule oriented and input based. The health budget has been no exception. The input-based characteristic has led to incremental allocations (in nominal terms) according to the cost changes of health supply components. Government budget rules in Ecuador since the 1970s have included a host of fixed allocations from specified revenues (including the repartition of oil revenues) and fixed-share or growth rules, such as the rule that health spending should increase at least as fast as aggregate public spending. Even though this has not been adhered to in practice and has not been able to prevent real health expenditures from falling for a prolonged period of time, it has imposed rigidities in the spending pattern and made the health budget adjust pro-cyclically at least until 1995, as analyzed in a study by Vos et al. (2003). In recent years, as discussed in Section 4, expenditure trends have been dominated by salary adjustments and the expansion of special health programs, leaving investment in health infrastructure and maintenance as well as resources for drugs supplies as the first items cut when budget adjustments are required. Attempts at a system of more decentralized budget allocation and performance-based budgets have been part of the health policy reforms of the 1990s. Implementation of these reforms has been difficult, as explained in Section 4, leaving much of the traditional health budgeting system still in place.

4.77 A first step towards a more comprehensive result-oriented budgeting system would be to put together all (central and decentralized) programs and interventions aimed at improving health outcomes and monitor each for their cost-effectiveness. In line with the MDGs, further reduction of infant mortality is a priority target for the Ecuadorian government. In that vein, the results of the health demand and child survival models of Section 6 may serve as a starting point for the development of a result-oriented expenditure tracking methodology.

4.78 The two models serve to establish input-output relationships in health, i.e. between policy interventions, access to health services and expected health outcomes measured by infant mortality. The relative importance of each determinant discussed in Section 6 may be expressed as an elasticity expressing the impact of a 1 percent change of a given determinant on, respectively, the probability of professionally assisted child delivery and child survival (see Tables 9a/b and 10). After linking these to unit costs we obtain a basis for making budget projections for alternative resource allocations. We perform this analysis in three related steps:

4.79 The Impact on Access to Health Services. In our case the specification is for the probability of professionally assisted childbirth and pre-natal controls. For this we use the relevant elasticities from Tables 4.9a and 4.9b for the likelihood of receiving no medical assistance during child delivery. The key policy variables are access to health insurance and availability of health services and medical personnel. In addition, we assume that expansion of the free maternity program will increase professionally assisted childbirth commensurately. Household determinants include the educational level of the mother and per capita household consumption. The averages of the latter two variables are assumed to change at fixed rates of 1.5 percent per annum, thus imposing a trend in rising access to health services due to improving

108

socio-economic conditions.29 We assume further that use of pre-natal controls has similar determinants and impacts as the probability of medically assisted child delivery.

4.80 Impact on Infant Mortality. Results of policy simulations on the demand for maternal care (at child delivery and pre-natal controls) are subsequently used in the health “production function” for child survival. The survival model further suggested that breastfeeding has a strong positive effect on avoiding early child death. We did not model the determinants of breastfeeding practices. We checked whether having access to pre-natal maternal care increases the likelihood of breastfeeding, but this does not appear to be the case as there is no significant difference in such practice between those that do and those that do not have access. Moreover, since breastfeeding practices are already widespread (over 90 percent) we do not consider increases in the budget simulations. According to the model findings, this then leaves the coverage of the immunization program as the policy variable with the most important expected effect on reduction of infant mortality. Ethnicity and female education variables are considered as the most relevant household variables in determining infant mortality.

4.81 Budget Implications. We add unit cost estimates for the relevant public health input variables, such as salaries of health workers, construction and maintenance of hospitals and health centers, as well as of the special programs, specifically the immunization and free maternity programs.30 In addition, we considered the possibility of increasing access to health services through a subsidy to the health insurance premium and thereby broaden coverage of health insurance, but this has been dropped from the results reported below as increasing health insurance coverage has very little direct impact on infant mortality. We did not consider expansion of the BDH to enlarge its impact on the use of health services. As discussed in Section 4, this program currently mainly has a (direct) impact on access to education. By changing the relevant health input variables, the required health budget will adjust according to the change in coverage of services times the corresponding unit cost parameter. In the simulation results presented below we use two alternative scenarios. One assumes that if demand for maternal health care increases there will have to be a related increase in the supply of doctors, nurses, and health infrastructure to meet that demand.31 The second assumes, as the free maternity program does, that the additional demand for health care can be met using existing resources, assuming slack capacity.32 We assume further that the nominal health budget (for all items) is adjusted for a given inflation rate (3 percent per year) to account for changes in input costs. Also nominal salaries of health workers are adjusted over time this way. In the budget simulations presented below, we assume constant real wages for doctors and nurses.

4.82 The models presented in Section 6, did not detect structural differences across population groups in the impact of the mentioned determinants, except for the income (consumption) variable. The gaps in health status (as measured by infant mortality and access to health 29 The growth in years of schooling of females thereby is assumed follow the trend of the 1990s, while economic

growth is—albeit modest—more optimistic (and clearly more stable) than that achieved during the 1990s. 30 See Statistical Annex Table A56 for a summary of the unit cost estimates as used in the budget projections. 31 As we make a projection of the total public health budget and focus only on demand for maternal and child care,

we assume in the reported simulations this affects 20 percent of the overall health budget. So a 10 percent increase in spending on this area would lead to a 2 percent increase in the overall budget.

32 This has not been investigated systematically, but there is some evidence suggesting substantial underutilization of public health resources. Using health input indicators provided by INEC’s Vital Statistics and Data on Human Resources in Hospitals and Health Centers, the average number of medical consults (of any kind) per hour in public health services is less than one.

109

services) between poor and non-poor and indigenous and non-indigenous thus are explained essentially by differences in initial conditions. The objective of the budget simulations is to see what would be cost-effective ways to achieve both the MDG in child mortality and reduce inequality in health status across population groups. The latter can be achieved by targeting interventions at the (extreme) poor and indigenous population.33

VII. Projections for Access to Maternal Care and Infant Mortality

4.83 Table 4.11 shows the results for three budget scenarios under the assumption that there is a 0.2 percent endogenous increase in medical personnel and infrastructure for each 1 percent increase in demand for public maternity care:

• Reaching full coverage of the vaccination program for all. This would require an annual growth in the number of beneficiaries of the program of about 3 percent per year until 2015 (particularly to cope with the current undercoverage of DPT vaccinations).

• Expansion of the Free Maternity program targeted at the poor and indigenous population, such that these target groups get unrestrained access to maternal and childcare by 2015. For this, the program’s expansion would amount to about 2 percent per annum for the (non-indigenous) poor and 3 percent for the indigenous population.

• Combination of the first two scenarios.

4.84 Our target is to reduce overall infant mortality from 34 to 18 per 1,000 live births between 2004 and 2015. For the poor the reduction should be from 42 to 22 and for the indigenous population from 66 to 33. Our baseline simulation projects improvements in education and per capita consumption forward to 2015 under the assumptions indicated above, but assumes that health programs show no further expansion from the coverage reached in 2003. The baseline then projects that without health input improvements infant mortality would reach 30.5 per 1,000 live births by 2015.

4.85 The simulations presented in Table 4.11 show that neither reaching full coverage of the immunization program nor the expansion of the free maternity program targeted at the poor are sufficient to reach the MDG targets for infant mortality. The expansion of the immunization program would reduce infant mortality to 20.1 per 1,000 live births by 2015. As the immunization program is universal and initial coverage does not differ much across population groups, this policy would reduce infant mortality for all, but would not narrow differences between poor and non-poor or indigenous and non-indigenous.

33 In the reported budget simulations, the “poor” refers to the “extreme poor” as identified by the UBN index.

110

Table 4.11. Achieving MDG for Child Mortality and Budget Implications

Baseline projection for

IMR

MDG for

IMR

Baseline projection for assisted

child delivery

Simulated impact

Additional budget cost

(over baseline) (annual

average 2004-2015)

IMR Access to

maternal care

US$ million

% of GDP

2000 2015 2015 2003 2015 2015 2015 Full coverage immunization program by 2015 3.5 0.01% Total population 34.1 30.5 18.2 84% 82% 20.1 81% Poor 42.0 37.4 21.9 78% 76% 24.9 74% Non-Poor 28.9 25.8 10.4 87% 86% 16.8 85% Indigenous 66.0 58.7 32.6 65% 63% 39.6 62% Non-indigenous 30.2 27.0 15.2 86% 83% 17.6 83% Full coverage free maternity program for poor and indigenous by 2015 3.8 0.01% Total population 34.1 30.5 18.2 84% 81%

which would seem quite affordable. Combining the two would lead to an annual cost of US$7.2 million over the baseline budget (or 0.02 percent of GDP). The combination of these two policies would be sufficient to reach the MDG targets for the poor and indigenous population groups, but would still fall slightly short for the non-poor.

4.88 As discussed in Section 3 and under the budget scenario assumptions (see footnote 32), there is some reason to believe that there is slack capacity in the public health system (at least in relation to executing its present tasks). In fact, the design of the Free Maternity Program assumes the existence of such slack capacity. Dropping the assumption that supply of medical personnel and infrastructure adjusts automatically in response to increased demand for maternity care reduces the budgetary requirements of the expansion of both programs to US$5.7 million per year.

4.89 If the health ministry decides to change the somewhat distorted proportion of doctors to nurses, it could “finance” the indicated extra budget requirements through cost-savings, if gradually (over the 11-year time span of the simulations) it reduces the number of doctors (by 40 a year) and increases the number of nurses (by about 50 a year). This would slightly raise the share of nurses to doctors in public health centers from 0.5 to 0.56. Bringing this proportion to 1:1 could imply an annual cost saving of US$29 million or about 8 percent of the total public health budget.34

4.90 Further fine-tuning of these budget scenarios will be needed to guide actual policy decision-making. Of course, the relatively low cost estimates are influenced by the assumption of a stable macroeconomic environment in the coming decade with low inflation and a steady rise in real per capita incomes. Actual conditions may be more volatile, for which the budget tracking methodology may help keep a focus on the health outcomes and make the health budget much more anti-cyclical than now. Such a budget strategy would comprise the protection of full immunization coverage and possibly a much stronger move towards wider coverage of health insurance. The latter is where the future of health financing should lie, but the country is still a long way from universal health insurance. A targeted expansion of the Free Maternity Program would be an effective option for the short to medium term. At present, as analyzed, this program does not overlap much with the benefits of the Bono de Desarrollo Humano cash transfer program, even though potentially such overlap may become important with the possible expansion of the BDH. All of this would require appropriate coordination for which the health budget-tracking model might provide meaningful input. “These proposed measures would increase the demand for public health services. The infant mortality model suggested that birth delivery attended by private health providers would increase the probability of child survival in the first year more than if delivery takes place in a public health center. As indicated, the difference in effectiveness is not very large though. Nonetheless, if the BDH and Free Maternity programs would give the poor equal access to private health centers, the impact on infant mortality could be enhanced by more. Yet, availability of private clinics typically is at greater distance from the homes of poor families, such that the expected additional impact likely will be very small indeed. The additional consideration is that the quality of public health care is in need of improvement to meet at least the apparently better standards at private health centers.

34. Such a scenario, leaving all other things equal, would require firing some 2,500 medical doctors and hiring

some 2,600 nurses in the period 2004-2015.

112

4.91 These policy directions can be fully consistent with the envisaged decentralized budget allocation mechanisms. The main requirement is that decentralized budgeting incorporates a focus on health outcomes, rather than on health inputs.

4.92 In conclusion, we found the somewhat paradoxical outcome that infant mortality rates have come down steadily, despite low public health expenditures, an apparently poorly functioning health system and a persisting high incidence of malnutrition in Ecuador. Much of the answer is to be found in the continued improvement of education levels and the urbanization of the population with associated improvements in sanitary conditions and knowledge of adequate reproductive health conditions. The same factors likely also have contributed to the strong reduction in fertility rates that has further pushed down infant mortality in a mutually reinforcing trend. Health interventions have mattered, particularly the expanded coverage of the immunization program and the program’s related contribution to improving environmental conditions for disease control. During the 1990s this role of the program has been strengthened along with specific interventions trying to ensure sufficient access for the poor to maternity care (among others through the Free Maternity program). Such policies seem effective at low cost. Obviously, more specific implementation issues will require additional attention, such as reaching the indigenous population with (culturally) adequate health services, quality control of health delivery and avoiding absenteeism of health workers. Overall rising health costs need to be taken into account as well since the country’s epidemiological profile is shifting towards a predominance of cancer and cardio-vascular diseases. Ecuador’s challenges in improving health delivery systems are large, but resource constraints cannot be an excuse for not achieving the MDG targets for infant mortality.

113

References

Bortman, Martin. 2003. ‘Health,’ Vicente Fretes-Cibils, Marcelo M. Giugale, José Roberto López-Cálix (ed.). “Ecuador: An Economic and Social Agenda in the New Millennium.” Washington D.C.: The World Bank.

CEPAR/PHR. 1999. “El Peso de la Enfermedad en el Ecuador.”

Collett, D. 1994. “Modelling Survival Time in Medical Research.” London: Chapman and Hall.

Cox, D. R. and D. Oakes. 1984. “Analysis of Survival Data. London.”: Chapman and Hall.

Demery, Lionel. 2003. “Analyzing the incidence of public spending”. François Bourguignon and Luiz Pereira da Silva (eds.) The Impact of Economic Policies on Poverty and Income Distribution. Evaluation Techniques and Tools, Oxford and Washington D.C.: Oxford University Press and World Bank, pp.41-68.

Deolalikar, Anil B. 1996. “Government health spending in Indonesia: Impact on children in different economic groups.” in Dominique van der Walle and Kimberly Nead (eds.) Public Spending and the Poor. Theory and Evidence, Baltimore: Johns Hopkins University Press (for World Bank), pp. 259-289.

ECLAC. 2001. “Panorama Social de América Latina.” Santiago: United Nations - Economic Commission for Latin America and the Caribbean.

Gertler, P., L. Locay and W. Sanderson. 1987. "Are User Fees Regressive? The Welfare Implications of Health Care Financing Proposals in Peru.": Journal of Econometrics 36(1/2) pp 67-80.

Gertler, P. and J. van der Gaag. 1990. “The Willingness to Pay for Medical Care: Evidence from two Developing Countries.” Baltimore: Johns Hopkins University Press.

Greene, W.H. 2001. “Econometric Analysis.” 3rd Edition, Englewood Cliffs, N.J.: Prentice-Hall.

Hanmer, Lucia and Howard White. 1998. “Under-five mortality in Sub-Saharan Africa.”: ISS, The Hague.

INEC. 2003. “Ecuador: Estimaciones y proyecciones de población 1950-2025.”: INEC.

Jenkins, Stephen P. 1995. “Easy Estimation Methods for Discrete-Time Duration Models.” Oxford Bulletin of Economics and Statistics, Vol. 57 (1): 129-38.

López, A.D., O.B. Ahmed, M. Guillot, B.D. Ferguson, J.A. Salomon, C.J.L. Murray, and K.H. Hill. 2002. World Mortality in 2000 : Life Tables for 191 countries, Geneva: World Health Organization.

Masset, E. & H. White. 2003. “Infant and Child Mortality in Andhra Pradesh. Analysing Changes Over Time and Between States.” Working Paper No. 8. London: Young Lives.

MSP/OPS/ICT. 2004. “Evaluación de impacto del programa PANN 2000.” Quito (mimeo, March).

Mwabu, G. and M. Ainsworth, and A. Nyamete. 1993. "Quality of Medical Care and Choice of Medical Treatment in Kenya: An Empirical Analysis." Journal of Human Resources 28(4): 838-862.

114

Nickell, S., R. Layard, and R. Jackman. 1991. “Unemployment, Macroeconomic Performance and the Labour Market.” Oxford, London: Oxford University Press.

Ponce, Juan. 2004. “Focalización del Programa Nacional de Nutrición y Atención a la Niñez.” Quito: Secretaría Técnica del Frente Social (mimeo, April).

SIISE. 2002. “La desnutrición infantil en el Ecuador.” Boletín Indice No. 2, Quito.

SIISE. 2003. “Sistema Integrado de Indicadores Sociales.” Versión 3.5, CD-ROM database, Quito: STFS.

UNICEF. 2004. “Ensuring the Rights of Indigenous Children.” Unicef Innocenti Research Centre, Innocenti Digest 11, Florence: UNICEF.

Vos, Rob, Margarita Velasco, and Edgar de Labastida. 1999. ‘Los efectos económicos y sociales del fenómeno de El Niño, 1997-98’, in: Enrico Gasparri, Carlo Tassara and Margarita Velasco (eds.) “El fenómeno de El Niño en el Ecuador, 1997-1999. Del desastre a la prevención.” Quito: CISP-SEDEH-SIISE.

Vos, Rob (ed.) et al. 2003. ¿Quién se beneficia del Gasto Social en el Ecuador?, Estudios e Informes del SIISE-STFS No. 4, Quito: Secretaría Técnica del Frente Social.

WHO 2003. World Health Report 2003, Geneva: World Health Organization.

115

Annex 4.1

Theoretical Underpinnings of Health Demand and Infant Mortality Models Demand for Health Care Given data limitations, the use of health care is not measured in terms of the quantity of health care consumed but in terms of choices among alternative health care providers or the probability of choosing a particular health care provider. The framework that we propose to use is a standard framework that has been employed in several papers that have estimated the use of health care (see Gertler, Locay and Sanderson, 1987; Mwabu, Ainsworth and Nyamete, 1993). The framework is a short-run static model with a utility function defined by health status and the consumption of all other goods. Consider an individual confronted with an illness. This individual has to choose among alternative health care providers (including self-care). Health care providers offer different levels of service at varying costs. An individual has to make a discrete choice amongst these providers. Conditional on an individual's health status, the type of illness, availability of information, and income an individual chooses the alternative that yields the highest utility. This description of the manner in which an individual may make a choice concerning health care provision may be formalized by considering utility conditional on receiving care from health care provider (HCP) j. Utility conditional on choosing provider j is given by,

( , ,ij ij ij ijU U H C T= )

)

(1) Where Hij is the expected health status of the individual conditional on receiving treatment from provider j, Cij is the consumption of all other goods except those associated with health care, Tij represents the non-monetary costs of access to provider j. The expected improvement in health care status is unobservable but is assumed to depend on the characteristics of an individual (health status, habits, etc.) and the quality of health care received by the individual. This allows us to write a health production function defined over Xi, the attributes of an individual and Zj, the attributes of the provider j. Hence,

( ,ij i jH H X Z= (2)

Turning to the second argument in the utility function, the level of consumption that is possible depends on the income of the individual and the costs associated with buying health care. If the user fee associated with provider j is Pj and Y is an individual's income then,

Cij = Yi - Pj. (3)

Substituting 3 into 1 yields,

( , ,ij ij i j ijU U H Y P T= − ) (4) i.e., utility is a function of the expected health status of an individual, the level of consumption and the non-monetary costs associated with using provider j. To guide empirical work it is suitable to substitute 3 and 2 into 1. This yields a function where utility is given as,

( , , , , )ij i j i j ijU U X Z Y P T= (5)

Thus, the benefits from visiting a particular health care provider depend on individual characteristics, the attributes of the provider, individual income, user fees faced with provider j and non-monetary costs associated with visiting provider j.

116

In order to empirically determine the probability of choosing a provider we need to choose a particular form for the conditional utility function and to introduce a stochastic disturbance. There are several possible choices for the form of the utility function. What is required is a form for the utility function that is consistent with well-ordered preferences. As shown in Gertler, Locay and Sanderson (1987) a suitable form for the utility function is the semi-translog where health and non-price access costs enter in log form and consumption enters in both log and log-squared forms. Other suitable forms for the utility function include parameterizations that are log-linear in heath status and consumption or a utility function that is linear in health status but log-linear in consumption (see Mwabu et al., 1993 for a discussion). The exact form of the utility function that will form the basis for our analysis will be determined at a later stage.

For the time being consider a utility function which may be written as follows: Uij = Vij+ εij (6)

Where, Vij is the systematic part of the utility function and depends on individual characteristics and provider attributes. The idiosyncratic part is represented by εij. An explicit form of the empirical utility function may be written as,

j i jUij W K ijα β′ ′= + +ε (7)

Where, Wi = [Xi,Yi] and Kj = [Zj, Pj, Tj]. Estimates of α differ across alternatives while estimates of β are equal across the various alternatives.35

An individual's health care provider choice may now be expressed as:

HCPi = j if Uij > max{Uik}, j =1… J, k ≠ j, (8)

Where HCPi is a health care provider indicator.

The parameters of (7) and the probability that individual i chooses health care provider j may be obtained by estimating a multinomial discrete choice model. The selection rule (8), combined with the assumption that the stochastic error term follows a Weibull distribution, defines a multinomial logit model where:

1Pr( ) exp( ) / exp( )

J

ij i j i j k i kk

P HCP j W K W Kα β α β=

′ ′ ′ ′= = = + +∑ (9)

Estimates of the required parameters may be obtained by maximum-likelihood estimation of (9). The results of the multinomial logit model may be used to compute the welfare costs associated with the imposition of user fees. These welfare costs may be measured in terms of compensating variations. Finally, as discussed above, choosing a particular health care provider depends on individual characteristics as well as the attributes of the choice and, accordingly, it is better to view (9) as a reduced form relationship rather than as a demand function. Survival Analysis

Duration models have been extensively used in socioeconomic analysis in issues such as unemployment periods, education enrollment and social benefit schemes (Nickell et al, 1991, Meyer, 1990). There is also a rather extensive epidemiological literature on the duration of health conditions, among then infant, child and adult mortality (Masset and White, 2003). Theoretically, a survival function is first constructed as a reduced-form equation relating a set of determinants, xi, with the probability of observing an event, for example the death of an infant in a period or interval ‘t’. Interestingly, a survival function estimates

35 For simplicity the utility function is depicted as a linear function of the explanatory variables. At the time of

empirical implementation appropriate alternative forms will be examined.

117

conditional probabilities that a particular event does not take place in a given period given that it has not occurred in previous periods. Also, survival models (such as the Cox Proportional Hazard model used in this study) typically assume that determinants of the survival process are time invariant throughout the analyzed period.36 Following Greene’s (2001) notation, let ‘T’ be a random continuous variable with a probability function f. This probability function indicates the number of periods elapsed until the studied event takes place in a period ‘t’. This probability function is dependant on a set of variables, xi, capturing anything from socioeconomic conditions to individual characteristics to any other factor that affects the duration of the studied event. Let F( ) be the cumulative probability of the duration variable, T. The probability that an event takes place in a period ‘t’ is given by:

∫ ≤==t

tTobdxxftF0

)(Pr)()( (10)

Conversely, an underlying survival function indicates the probability that the duration of the process unfolding in the observed event takes ‘t’ periods to materialize is given by:

)(Pr)(1)( tTobtFtSu ≥=−= (11) The survival function in (11) indicates that each period is independent of the previous as far as the probability of observing the event is concerned. This is typically not true in the kind of socioeconomic events like infant mortality or unemployment periods. More appropriately, a survival function can be expressed as a process of intertwined relations of xi upon F(t) conditional on survival in previous periods:

)()( tTtTtSc ≥== (12) In the case of infant mortality, ‘t’ typically represents the number of successive months within the first year after birth that the infant remains alive. Thus, the survival probability of remaining alive in the fourth month of life for an infant is the conditional probability that the infant survived the first, second and third month after birth. This (conditional) survival function in (12) can be expressed conveniently in the form of a hazard rate, that is, the ratio between the probability of failure (death), and success (survival) of an event taking place. As Jenkins (1995) shows, (12) can be re-arranged in the following way:

∏ ∏−

= =

−−

=−⋅=>==1

1 1

))(1()(1

)())(1()()(Pr)(t

k

t

ku kh

ththkhthtTtTobtS (13)

Cox and Oakes (1984), parameterized this conditional probability in the form of proportional hazards with respect to a baseline individual leading to the Cox Proportional Hazard model (CPH) explained below. Using maximum likelihood estimation, the CPH can estimate the unknown coefficients, βi, of a set of determinants, xi, on infant mortality.

36 This assumption can be easily relaxed, however.

118

Annex 4.2 Methodology of Estimating Per Capita Household Consumption in the ENDAMAIN (DHS)

Survey

The procedure to calculate per capita household consumption was developed by SIISE as part of the derivation of a Human Development Index (HDI) at the provincial level (UNDP 2001). The income variable of the HDI was proxied by per capita consumption as the consumption measure is considered to be more reliable than income estimates in household surveys. No existing survey provides representative data at the provincial level for this variable. For this reason two sets of survey data were combined: the 1998 LSMS, Encuesta de Condiciones de Vida (ECV) and the 1999 Demographic and Health Survey (DHS). The former reports consumption data, but not at the provincial level; the latter has representative socio-economic data at the provincial level, but does not report consumption. The 1998 LSMS was used instead of the more recent 1999 LSMS, because the latter did not cover the Amazon region of the country.

Observed consumption data from the 1998 LSMS survey were imputed to the households of the ENDEMAIN simple using a simulation exercise. The procedure is as follows:

• A set of categorical variables (X1, X2, ... , Xn ) related to the quality of housing and other living conditions indicators was selected using an optimal scaling technique (principal components analysis). These variables were subsequently used as explanatory variables in a multiple regression analysis.

• Two regression models were specified, one for urban areas and one for rural areas using data from the 1998 LSMS survey. Monthly per capita consumption was taken as a dependent variable in a linear regression in the following general format: y = ao + a1*X1 + a2*X2 +... + an*Xn, where y is per capita consumption, ai are the regression coefficients and Xi the selected qualitative variables.

• After validating and testing the regression models, the corresponding parameters were used against the corresponding values for the explanatory variables as measured in the 1999 ENDEMAIN survey to impute per capita consumption for each of the household members.

Principal Components Analysis Valuation of categorical variables through the optimal scaling method is obtained by maximizing linear correlations between a given set of variables. This way qualitative variables are quantified and the optimal combination of variables is sought for use in regression models (that assume linear relationships). Figure A4.1, demonstrates the advantage of constructing an index of this sort, using statistical procedures. The arbitrary assignation of scores or weights to categorical variables (e.g. through policy evaluators) would not adequately discriminate between the different elements (Part I in the Figure). For instance, if we had categorized the quality of housing with values 1, 2 and 3 to identify homes constructed with bricks and cement, wood, and “adobe,” respectively, we would not value the distances in quality as such. Instead, using optimal scaling, we obtain statistically derived weights that consider the distance between the units of observation based on maximum linear correlations (see part II of Figure A2).

The first step in the procedure implies selecting and comparing common variables in both surveys. Variables were selected if they showed similar frequency distributions and those

119

showing very different frequencies were excluded. Valuation of the selected categorical variables was done on the basis of the 1998 LSMS survey using (non-linear) principal components analysis, which is part of the family of multivariate factor analysis. The objective of the procedure is to maximize linear correlation between two variables, to increase in this way the distance between observation units (in our case: households). It also allows one to distinguish ambiguous cases better.

Figure A4.1 Assigning optimal scales to components of categorical variables.

Variable A

IIRelationship between variables after applying optimal scaling model.

Relationship between variables without applying optimal scaling model.

Element 1 after optimization I

Element 2 after optimization

Element 1 before optimization

Element 2 before optimization

Variable B

120

Table A4.1. Common Variables of 1998 LSMS and 1999 ENDEMAIN Surveys and Optimal Scaling of Categorical Variables.

Variables Weights (score)


Type of housing (CEP_VIVI) Education household head (NIV_JEF) 1 Other -3,35 1 None -1,342 Shed/”choza” -3,35 2 Literacy campaign -1,063 Mediagua -0,89 3 Primary -0,64 Rented room 0,39 4 Secondary 0,585 Apartment 0,39 5 Superior / university 1,916 House 0,39 6 Post-graduate 2,95Tenancy of housing (CEP_TENE) Education spouse (NIV_CON) 1 Other -1,35 1 None -1,512 Tenancy in exchange for services -1,35 2 Literacy campaign -1,193 “Cedida” -1,35 3 Primary -0,74 Rented -1,35 4 Secondary 0,935 Own housing 0,74 5 Superior / university 2,44Roof construction (CEP_TECH) 6 Post-graduate 2,781 Other -1,71 7 No spouse -0,112 Palm leaves / “paja” -1,71 Área of residence (CEP_REG) 3 Teja -0,76 1 Rural Sierra -1,124 Zinc -0,57 2 Rural Costa -1,125 Eternit 0,85 3 Rural Amazonía -1,126 Hormigón / bloque / ladrillo 1,62 4 Urban Amazonía 0,67Floor construction of housing (CEP_TECH)

5 Urban Costa 0,67

1 Other -1,41 6 Urban Sierra 1,132 Dirt -1,41 Persons per bedroom (HACINA2) 3 Wooden -0,73 1 More than 4 persons -1,444 Cement / stones 0,19 2 3-4 persons -0,795 Parquet / ceramic tiles 1,55 3 0-2 persons 0,95Wall construction of housing (CEP_PARE)

Sound equipment (EQUIPO)

1 Other -1,74 1 Does not have -0,882 Bahareque -1,57 2 Has 1,143 Wood -1,29 Refrigerator (REFR) 4 Adobe -0,98 1 No -1,145 Bricks 0,73 2 Yes 0,87Type of toilet (CEP_SSHH) Telephone connection (TELEFO) 1 None -1,64 1 No -0,542 Latrine -1,35 2 Yes 1,853 Toilet and septic tank -0,20 Washing machine (LAVADO) 4 Toilet and sewerage connection 0,96 1 No -0,29Electricity (CEP_LUZ) 2 Yes 3,491 No -3,23 Car (CARRO) 2 Yes 0,31 1 No -0,42Energy for cooking CEP_COMB) 2 Yes 2,41 Other -2,36 Computer (COMP) 2 Electricity -2,36 1 No -0,22

121



4,524 Gas 0,43 Television (TELEVI) Drinking water supply (CEP_AGUA) 1 No -2,031 River / lake / stream -1,94 2 Yes 0,492 Open well -1,04 3 Piped water (shared access) -0,87 3 Wood / carbon -2,36 2 Yes -0,48 6 Drinking water supply by truck 0,06 7 Piped drinking water in home 1,08 Sources: INEC, ECV, 1998; CEPAR, ENDEMAIN, 1999. Note: In parentheses is the label of the variable as used in the linear regression models. This table lists the selected variables with their corresponding values. The order and magnitude of the derived weights are as expected, that is higher scores (weights) reflect better living conditions. Regression Models Alter valuating the categorical variables, these were subsequently inserted into the consumption model, taking logarithms of observed per capita household consumption. The estimation results for urban and rural areas are, respectively:

Rural areas (Adjusted R2 = 0.832):

Per capita consumption = exp(13.57013682)*exp(A_AGUA*0.0191595822347977 + A_CARRO*0.204389371615797+ A_COMB*0.1545491562636985 + A_CONY*0.0497950918830602 + A_EQUIPO*0.0426059807807566 + A_HACI*0.0771703896456176 + A_JEFE*0.116035513892694 + A_PISO*0.106651022931832 + A_REFR*0.223143443359268 + A_SSHH*0.0373208677801488 + A_TECH*0.0517787473764033 + A_TELEFO*0.0517069259231406 + A_VIVI*0.0148335861089794 -A_MIEM*0.369007006249754)

Urban areas (Adjusted R2 = 0.765):

Per capita consumption = exp(13.8124792990525)* exp(A_AGUA*0.0469161683673776 + A_CARRO*0.178953644054217 + A_CONY*0.0351708061398845 + A_EQUIPO*0.0634251606620911 + A_HACI*0.0725197379069467 + A_JEFE*0.0660413752052221 + A_PISO*0.0397410910875448 + A_REFR*0.0829802496104904 + A_SSHH*0.0583510687159971 + A_TELEFO*0.0997169614546676 + A_VIVI*0.032614743431203 - A_MIEM*0.526590305558633 + A_LAVA*0.0778544004268161 + COSTA*0.0581064991569207 + A_COMB*0.0682568172266199)

122

The results were subsequently validated by reproducing the poverty incidence and average consumption levels between observed values of the 1998 LSMS and the imputed values of the 1999 ENDEMAIN surveys (see Table A4.2). Actual values are reproduced with great precision. Only in the case of Urban Costa is there a discrepancy of about 5 percent between the two estimates. In all other cases the difference is less than 2 percent.

Region

Observed

Imputed (ENDEMAIN

1999) Observed

Imputed (ENDEMA

IN 1999)Observed

(ECV 1998)

Imputed (ENDEMA

IN 1999)

Costa 34.9 30 70 71.6 46.7 41.7Sierra 22 20 69.1 69.8 45.5 46.6Amazonía 27.7 27 58.6 59.1 52.6 49.6Country 30.2 26 68.8 69.5 46.4 44.4

Costa 455,610 416,919 246,183 215,711 384,878 358,842Sierra 640,430 522,848 233,261 251,724 442,005 379,081Amazonía 447,511 413,795 277,594 284,947 310,928 322,573Country 522,906 458,574 241,752 242,774 406,465 366,917Sources: INEC, ECV, 1998; CEPAR, ENDEMAIN, 1999.

Poverty Incidence (consumption) ( percent)

Per capita Household Consumption (1998 sucres)

Table A4.2. Poverty Incidence and per Capita Consumption, According to Observed Values of the 1998 LSMS Survey and Imputed Values of ENDEMAIN Survey of 1999

Urban Areas Rural Areas Total

In the demand for health services model, we use the consumption estimates for each household and individual.

123

Annex 4.3

Description of Primary Data Sources Used in this Study

Unit of

Analysis Population

Census LSMS (ECV) SIEH – Survey ENDEMAIN (DHS)

Year 2001 1999 December 2003 1999 Dwellings 21,462 Households 2,887,087 5,824 18,959 Persons 12,156,608 25,980 82,317

Sample size

Women in reproductive age in sample

6,204,056 6,531 22,940 14,285

4. Defined as: Women 12 years and older Women aged 15 – 49 years Women aged 15 – 49

years Women ages 15 – 49 years

Population (universe)

All dwellings, households and individuals in Ecuadorian territory

All dwellings, households and individuals in Costa and Sierra regions (Amazonia and Galápagos Islands are not included)

All dwellings, households and individuals in country (except Galápagos)

Dwellings, one household per dwelling, one woman of reproductive age per dwelling (covers all four regions of the country)

Sample frame Pre-census 2001 Population Census of 1990 Population Census of 2001

Population Census of 1990

Representativity of data

National, regional, provincial, cantons, parroquias, urban, rural

Costa, Sierra, Quito, Guayaquil, urban and rural

National, Costa, Sierra, Amazonía, Urban, Rural, Provinces except Galápagos

National, Urban, Rural, Sierra, Costa, Amazonía, Galápagos, Quito, Guayaquil, 15 provinces of Costa and Sierra

Key variables of health modules

Fertility Infant mortality (indirect method)

Child care Immunizations Breastfeeding Disease prevalence Access to health services Insurance and health expenditures Fertility and maternal health Infant mortality (indirect method) Health practices and attitudes Nutrition (Anthropometry)

Epidemiological control: malaria and dengue Immunization Birth control and delivery Infant mortality (indirect method)

Child care Immunizations Breastfeeding Maternal,, infant and child health Health insurance Fertility Maternal and infant mortality (direct method) Sexual and reproductive health STD and HIV/AIDS Domestic violence Birth rate

124

Documents

HEALTH1 - United Nations · 2010. 1. 8. · Chapter 4 HEALTH1 Healthcare in Ecuador has improved substantially over the last 30 years but spending remains low compared to other countries