Child access to health services during the economic crisis: An Indonesian experience of the safety net program

ARTICLE IN PRESS

Social Science & Medicine 63 (2006) 2912–2925 www.elsevier.com/locate/socscimed

Child access to health services during the economic crisis: An Indonesian experience of the safety net program Eunike Suci Atma Jaya Catholic University, Jakarta, DKI, Indonesia Available online 11 September 2006

Abstract Child health has been a serious problem in Indonesia for several decades. The prolonged Indonesian economic crisis in 1997 had a tremendous impact on poor children who suffered due to malnutrition. In 1998, the Indonesian government launched a broad social safety net program to protect the poor from becoming poorer. In the health sector this took the form of Jaring Pengaman Sosial Bidang Kesehatan (JPS-BK) or the Social Safety Net in Health Sector program. Adopting the model of health services utilization of Andersen and Newman, I examine the extent to which JPS-BK contributed to better health services for poor children in four provinces, by using a simplified version of Andersen and Newman’s model of health services utilization which emphasizes the importance of contextual determinants. Variables used in the study included child outpatient visits, health card possession, household income, and poverty status. Using data sets from the JPS-BK longitudinal study, I compared utilization of health services between baseline data collection at Rounds One and Three, which was taken a year afterward. In addition, I used the Village Potentials data set from the Indonesian Bureau of Statistics and employed factor analysis to raise one variable representing the village/ neighborhood developmental level. Basic statistics were used to examine possible changes between study rounds and logistic regression was used to examine the effect of health card possession on child health services utilization. Two significant improvements occurred during the first year of the program: (i) more sick children visited outpatient facilities and (ii) more children lived in households possessing health cards. The JPS-BK increased the ‘‘potential access’’ that was demonstrated by the significant increase in health card possession regardless of the visit, and ‘‘realized access’’ that was demonstrated by the significant increase in child outpatient visits regardless of health card possession. Further research needs to be undertaken to explore the dynamics of outpatient visits and the actual use of health cards. r 2006 Elsevier Ltd. All rights reserved. Keywords: Indonesia; Safety net program; Child health; Developing countries; Health seeking behavior; Health services Utilization

Background Child health has been a serious problem in Indonesia for several decades. The trend of child health indicators shows that child health in Indonesia has improved significantly over the last Tel.: +62 21 7074 0597; fax: +62 21 570 8830.

E-mail address: [email protected]. 0277-9536/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2006.07.014

forty years. Infant mortality rate (IMR) has declined from 145 deaths per 1000 live births in 1971 to 52 deaths per 1000 live births in 1995. The under-five mortality rate also declined from 218 deaths per 1000 live births in 1971 to 71 deaths per 1000 live births in 1995 (Central Bureau of Statistics, 1998). However, the indicators remain at levels well below those of neighboring countries. For example, in 1997 the IMR of Indonesia was 41

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

deaths per live births, while that of Malaysia, the Philippines, Thailand, and Vietnam were 11, 37, 30, and 38 deaths per 1000 live births, respectively. The under-five mortality rate showed a similar pattern (Ministry of Health, 1999a). Because the country is an archipelago, Indonesia is particularly challenged by the problems of providing equal health services, and in fact, child health status varies broadly across provinces. Based on the 1997 Indonesian Demographic and Health Survey, the Central Bureau of Statistics (1998) reported that the IMR in Yogyakarta was 23 per 1000 live births. In contrast, the IMRs in West Nusa Tenggara, Central Sulawesi, and Southeast Sulawesi were 111, 95, and 78 per 1000 live births, respectively. The pattern of the under-five mortality rates is similar to the IMR. As happen in many other countries, the IMR of rural children was higher that that of urban ones. The long monetary and economic crises started in mid-1997 ruined almost all the health indicator improvements Indonesia had achieved previously. In addition, it had a devastating impact on household purchasing power (Ananta, 2003; Junaidi, 2001). In the health sector, for example, drug prices skyrocketed and health services became less affordable. Rapid surveys in Jakarta as well as several other cities and rural areas in Java suggested that drug prices doubled or even tripled after the crisis (World Bank, 1999). The economic crisis, coupled with subsequent political crises and natural disasters (e.g., earthquakes, floods, El Nin˜o), prohibited many of the rural and urban poor from achieving a subsistence standard of living (BAPPENAS (Badan Perencanaan Pembangunan Nasional (National Development Planning Agency)), 2004). The number of poor people increased from 34.5 millions in February 1996 to 49.5 millions in December 1998 (Junaidi, 2001), and the existing poor became poorer as many of them lost jobs. Among those affected by the crisis, poor children suffered most and they are most likely to suffer in the future owing to the lasting effects of malnutrition. Utomo (2003) reported that in East Java severe malnutrition has increased among infant and young children since 1997. The 1998 SUSENAS data, quoted by Utomo, affirms that 10.5% of under-fives were severely underweight, implying that 2.3 million under-five children were severely malnourished. The younger they are, the more vulnerable to illness they become because they are at the beginning stage of physical development.

2913

In 1998 more than half of children under the age of two on the populous island of Java suffered from malnutrition (Bresnan, 1999). Unlike neighboring countries that were able to recover from the economic crisis relatively quickly, Indonesia suffered a prolonged depression owing to severe drought, political turmoil, internal conflicts, and the lack of political will to solve economic problems. The World Bank argued that ‘‘weak institutions and endemic corruption, at a time of high political uncertainty’’ made the situation in Indonesia much worse than the situations of other East Asian countries (World Bank, 1999). Starting in October 1998 (the 1998/1999 fiscal year), the Indonesian government, with the support of the International Monetary Fund and the World Bank, launched a broad social safety net known as Jaring Pengaman Sosial (JPS)—the social safety net—to cover the economic, education, and health sectors. During the first year of implementation, the government had spent about 30% of its expenditure for the program (Ananta & Siregar, 1999). The program’s aim was to protect the poor from the negative impact of the structural adjustment programs during the economic crisis which began in July 1997. Since the safety net was a rescue program which was one part of a broader structural adjustment package, the program was temporary and not intended to replace the long-run programs on poverty alleviation (Ananta, 1999). When the crisis passed, it could be integrated into the national development program which is more sustainable. In 2001 the government reduced its international loan so as to be more independent. At the same time a program called Impact Control Program for Energy Subsidy Reduction in Social Safety and Health sector (PPDPSE) was launched to reduce gasoline subsidy because the subsidy was enjoyed by the rich. The subsidy, which was initially intended to reduce the gasoline price, was used directly for various activities needed by the poor, such as providing health services, clean water, public transportation, education, and other social services (BAPPENAS, 2004). It is expected that the PPDPSE sustained the safety net programs. The Jaring Pengaman Sosial Bidang Kesehatan (JPS-BK) or the Social Safety Net Program in Health Sector aimed to develop and maintain the health and nutritional status of poor families. The program was expected to provide modern medical services for the poor, irrespective of their income

ARTICLE IN PRESS 2914

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

and expenditure fluctuations (Suryahadi, Suharso, & Sumarto, 2001). Some of the specific objectives of the program were to provide subsidies for medicines and imported medical equipment, free health services and referrals, immunizations, and food supplementation for both children and pregnant/ lactating mothers. With the budget of about 1.7 billion rupiah for the 1998–1999 fiscal year, the government subsidized generic drugs and imported medical equipment, and free health services for poor mother and children (Muhammad, 2000). The funds for health services were distributed straight to health centers called puskesmas and other services under the puskesmas system as a block grant. The puskesmas directors and the village midwives were responsible for the program management. This direct distribution was a new strategy to empower health providers who would communicate directly with the lay people, so that poor families would be able to access health services appropriately (Kusnanto, 1999). Furthermore, direct distribution also sought to reduce the loss of funds due to the inefficient bureaucratic system, which is very costly because of the widespread corruption in the country.1 To identify the target groups and update the list of poor families, each village had a team consisting of village leaders, village midwives, women’s organizations, non-governmental organizations, and other important people. The ratio of government and non-government members was to be 1:1 (Ministry of Health, 1999c). Since the lists of poor families provided by the National Family Planning Coordinating Board (BKKBN) were outdated and inaccurate, it was important that the village team update the list.2 Any family not already on the list and meeting at least one of the program criteria was added to it. On the other hand, any family included on the BKKBN list not meeting at least one of the eligible criteria was removed from the JPS-BK program. Eligible families were defined as those families who did not eat twice daily or did not bring their sick members to health centers, families whose head-of-household lost his/her job due to a mass dismissal, and families with children who dropped out of school due to financial problems (Ministry of Health, 1999b). 1

Schwarz notes that corruption is endemic in Indonesia and, in many cases, is socially tolerated (1994). 2 The list of poor families provided by BKKBN is commonly used as a national reference.

After the poor families were identified, puskesmas directors sent the list of the poor to the local government at the regency level to request the funds. In order to access free health services, puskesmas directors and village leaders distributed health cards to poor families based on the evaluation reported by the village team. The JPS-BK program distributed health cards to all poor families no matter where they resided. The program also has a well-planned budget and guideline for the target groups (Suwandono, 1999). Since JPS-BK is a new safety net program that was administered across the nation, its implementation may have led to many problems. For example, the front line health providers had never before received a large block grant due to the top-down health system in Indonesia. This situation created a problem in fund management, as many health providers did not know how to spend the funding. Also, the poor infrastructure, especially in remote areas outside Java, made it difficult to reach some villages on small islands directly. In addition to these problems, not all puskesmas heads and village midwives had or were able to open a bank account, a prerequisite for receiving grant funds; the standard definition of ‘‘poor’’ may not have been appropriate given the local situation, and homeless persons who were unable to obtain ID cards were then also unable to obtain health cards. Other problems accessing free health services through health cards related to the lack of information on how to use the cards and inability to show the card at the assigned health facilities (e.g., forgot to bring the cards or lost them). In order to evaluate the JPS-BK program, the Ministry of Health and BAPPENAS initiated a multi-centered longitudinal study in five provinces to measure the extent to which JPS-BK actually benefited poor families, especially pregnant/lactating mothers and children, by increasing their access to health services and improving their health status with supplemental foods (Suwandono, 1999). The longitudinal study began in December 1998 and was comprised of three rounds: (i) December 1998–February 1999; (ii) April–June 1999; and (iii) September–November 1999. Fig. 1 shows that the longitudinal study was conducted after the monetary crises during which the Indonesian currency fell to almost threefold per US dollar than before the crisis. The monetary data show a decline of the Indonesian purchasing power.

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

2915

JPSBK 18000 2

1

15000

3

Rp./$

12000

9000

6000

3000

Jul-02

Apr-02

Jan-02

Oct-01

Jul-01

Apr-01

Jan-01

Oct-00

Jul-00

Apr-00

Jan-00

Oct-99

Jul-99

Apr-99

Jan-99

Oct-98

Jul-98

Apr-98

Jan-98

Oct-97

Jul-97

Apr-97

Jan-97

0

Fig. 1. The currency rate before, during and after the economic crisis; and the time schedule of the JPS-BK program and the three-panel study (1998–1999). Note: The JPS-BK programs started in October 1998. Bars 1, 2, and 3 show the time and duration of the longitudinal study for the first, second, and third rounds.

Using the data from the longitudinal study, I attempted to measure the extent to which the JPSBK program increased child access to health care services. I concentrated on children under five years old because they were likely to be the primary victims of the economic crisis. This was important because the first five years of a child’s life are the most crucial to individual development. Moreover, Indonesian child health programs, including the JPS-BK, usually focus on children five years of age and younger.

Research questions My primary study objective was to measure the extent to which the JPS-BK contributed to improved child access to health services in selected provinces (i.e., Central Java, Yogyakarta, East Java, and South Sulawesi). A measure of the JPSBK program used in the study was health card possession, and I attempted to evaluate the extent to which health cards were distributed and the effects of the distribution on child outpatient visits. Secondly, I evaluated the determinants (i.e., predisposing, enabling, and need factors) that influenced the utilization of child health services during the first year of the JPS-BK program.

Theoretical background Ronald Andersen has developed a framework for health services utilization based on a socio-behavioral model. His initial framework, introduced in his 1968 monograph, A Behavioral Model of Families’ use of Health Services, emphasized three major factors: (1) predisposing characteristics; (2) enabling factors of access to care; and (3) actual illness or ‘‘need’’ for care. The framework has since been adopted and further developed by many other scholars, who have demonstrated its usefulness (e.g., Beckman, 1984; Wan, 1989; Wolinsky, 1978; Wolinsky & Johnson, 1991). I adopted and simplified a model of health services utilization developed by Ronald Andersen and John Newman in 1973 which emphasized evaluation at the individual level and the importance of societal determinants of the health care system. According to Andersen and Newman (1973), some individuals have a greater propensity to use health services than others, and their propensity can be predicted by individual and social characteristics. The characteristics per se are not the reasons an individual seeks care, but are factors which predispose the individual to seek care. Age, for example, is not the reason an individual seeks care. However, elderly persons are more likely to

ARTICLE IN PRESS 2916

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

seek care than younger adults because the elderly are more vulnerable to degenerative diseases. In addition, certain enabling factors also determine an individual’s use of health services. Andersen defines an enabling factor as a ‘‘condition which permits a family to act on a value or satisfy a need regarding health service use’’ (Andersen & Newman, 1973, p. 109). Individuals, although they are predisposed to use health services, need some means available for them to do so. Conditions enabling a family to use health care services could be a higher household income or socio-economic status and possessing health insurance or third party payer. The presence of predisposing and enabling conditions might not be enough to lead individuals to use modern health care. Individuals need to perceive illness or the probability of its occurrence if they are to use health services. Some indicators of perceived illness are number of disability days, symptoms the individual experiences in a given time period. Perceived illness can also be measured by selfreports that generally state an individual health condition (e.g., excellent, good, poor). Since Andersen’s model was developed in the US when the country established new federal socialwelfare programs in 1960s, the application of the model to developing countries needs to be modified. During that time, American scholars debated issues of access to care and equitable treatment. Traditional medicine, self-help treatments, and other alternative healings were not popular in the US and were therefore non-issues. Benyoussef and Wessen (1974) were among the first to point out the differences in health services utilization between developed and developing countries. Using Tunisia as the focus of their study, they found that modernization (e.g., rural/urban differential) is the key predictor of health services utilization in developing countries. Not only does the rural– urban differential reflect the availability of services (Thind & Andersen, 2003), it also reflects the differences between the populations and the technical sophistication of the available services (Benyoussef & Wessen, 1974). The study of Thind and Andersen (2003) found that, among others, urbanization, time to reach a health facility, and economic status were strong predictors of child health care utilization in the Dominican Republic. A second feature that distinguishes the contexts of the US and Indonesian healthcare crises is how the people in these countries pay for health services. In the early 1970s, US citizens relied on voluntary

family health insurance to reimburse them for many medical expenses. In Indonesia, most people do not have health insurance; instead they pay for health services out-of-pocket. Finally, the Andersen–Newman model does not incorporate detailed characteristics of the environment. Though Andersen and Newman mention rural–urban natures of communities, there are many more environmental factors that affect health services utilization. For example, an individual who lives in a rural village where a health center is available tends to visit the center more often than one who lives in a similar village but must travel a distance to reach a health center. Different from the US, health centers in Indonesia are available only at the district level. For my study, I simplified the model to emphasize the importance of contextual determinants appropriate to Indonesia. They were village/neighborhood developmental level, provincial origin, distance to district office, number of different types of epidemics, and pharmaceutical supply. In addition to health card possession, I used household income, poverty status, and puskesmas medical expenditure as enabling factors. Variables representing predisposing factors were child age, sex, family size, and mother’s educational level. Illness level was represented by the number of symptoms. The above behavioral model of health services utilization is fundamental, given that a major goal of the model is to provide measures of access to medical care (Andersen, 1995). Andersen defines two measures of access to health care: (i) potential access, which is the presence of enabling sources and (ii) realized access, which is the actual use of health services. In addition, he determines the equity of access: an equitable access occurs when demographic and need variables account for most of the variance in utilization, while inequitable access occurs when social structure, health beliefs, and enabling resources determine who gets medical care (Andersen, 1995). Methods The study employed a time series non-experimental ‘‘ex post facto’’ design, as the JPS-BK study collected information in three rounds. Although the JPS-BK study was longitudinal, my study used a time series method owing to the fact that the JPSBK data lacked unique ID codes for children. Baseline data were collected in Round One and

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

follow up data the later two rounds four and twelve months after the baseline. My study examined the change in health services utilization between Rounds One (December 1998–February 1999) and Three (September 1999–November 1999) in four provinces: Central Java, Yogyakarta, East Java, and South Sulawesi. Two JPS-BK data sets were used in the study: the individual and the puskesmas data sets. In addition to these, I used the Village Potentials (podes) data set that were provided by the Indonesian Bureau of Statistics to obtain contextual variables. Since child health services were obtained only for those who were sick, the analyses were performed only for sick children (n ¼ 4861). I performed a factor analysis to simplify the description of village characteristics taken from podes data. The analysis yielded one variable called community developmental level that represented the characteristics and modernization level of the community. This variable highly correlated with rural/urban status with Pearson correlation coefficient of 0.63, and therefore I combined both variables and used a single variable called village/ neighborhood developmental level. Using the median as the cut-off point, I categorized community developmental level into two groups: lower and higher developmental levels. I then made a crosstabulation between dichotomous variables of developmental level and rural–urban status, and distinguished three levels of village type: (i) urban neighborhoods, which mostly have high developmental level; (ii) higher-development rural villages; and (iii) lower-development rural villages. In addition to this variable, I used distance to the district

2917

office and number of different epidemics in the study, which refers to the number of different epidemic types that occurred in a village at least once during the year prior to survey. As for the puskesmas data, I used three variables representing the puskesmas performance: (i) the score of puskesmas’ staff personnel; (ii) the score of puskesmas’ pharmaceutical supply; and (iii) the puskesmas JPS-BK medical expenditure. One dichotomous-dependent variable used in the study was child outpatient visits. The major independent variables included number of illnesses, age, sex, household income, mother’s educational levels, health card possession, health center’s performance, and level of village development. I employed basic statistics to examine whether the variables varied with rounds. I found that almost all demographic and contextual variables (i.e., child’s age, child’s sex, poverty status, mother’s educational level, mother’s age, family size, number of different epidemics in village/neighborhood, and distance to district office) in Rounds One and Three were comparable. This means that I could examine the change in the child health services utilization with minimal confounding factors. Only one variable, village/neighborhood developmental level, varied with rounds and therefore analyses using this variable need to be evaluated carefully. For the analyses, I employed basic statistics to examine whether the child outpatient visits varied with the individual, household, and contextual variables. To examine the effect of the JPS-BK program (i.e., health card possession) on child health services utilization with a number of control variables, I used logistic regressions.

Table 1 Frequency distribution of sick children visiting an outpatient care facility Variable

Total sick children Province East Java Yogyakarta Central Java South Sulawesi Village/neighborhood developmental level Lower-development rural village Higher-development rural village Urban neighborhood

Visited

Did not visit

Total

n

Row (%)

n

Row (%)

n

Row (%)

3245

55.2

1616

27.5

4861

100

1679 121 960 485

68.2 67.6 65.1 65.1

783 58 515 260

31.8 32.4 34.9 34.9

2,462 179 1,475 745

100 100 100 100

1206 833 1206

65.2 65.9 69.0

644 431 541

34.8 34.1 31

1,850 1,264 1,747

100 100 100

ARTICLE IN PRESS 2918

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

Findings Table 1 shows the percentages of child outpatient visits in the four provinces. It shows that the percentages of the visits in Yogyakarta and East Java provinces were higher than those in Central Java and South Sulawesi provinces. However, they did not differ significantly. In addition, urban sick children were more likely to visit an outpatient facility than their rural counterparts. They did not differ significantly. Both findings indicated that contextual factors did not affect individual health seeking behavior. The next analyses aimed to evaluate whether child outpatient visits varied with household and individual backgrounds. Table 2 shows that sick children who visited and did not visit an outpatient facility did not differ with regard to poverty status and household income, but did differ significantly with regard to health card possession (Chi-square ¼

51.36, po.0001). More specifically, sick children living in households possessing a health card were more likely than others to visit an outpatient facility. Table 2 presents the percentages of child outpatient visits for mothers of different educational levels. It shows clearly that as their mother’s educational level increased, the proportion of sick children visiting an outpatient facility increased (Chi-square ¼ 16.45, po.001). The table also shows that the average age of sick children who sought outpatient care differed significantly from sick children who did not seek it. Younger sick children were more likely than older sick children to visit an outpatient facility (po.0001). In regards with the mother’s age and the average numbers of symptoms, the statistical differences between visiting and non-visiting children are too small to have practical implication. Table 3 shows that about 63% of sick children in Round One and 71% in Round Three sought

Table 2 Frequency distribution of outpatient visits of sick children by poverty status and health card possession (pooled data) Variable

Visited

Did not visit

Total

n

Row (%)

n

Row (%)

n

Row (%)

Total sick children Possessed health card (Chi-square ¼ 51.36, po.0001) 0—No 1—Yes Poverty statusa (Chi-squares ¼ 2.0, p ¼ 0.15) 0—Poor 1—Very poor Missing value Mother educational level (Chi-square ¼ 16.45, po.001) Never went to school Not finished elementary S Finished elementary school Finished junior high school+ Child status: grandchild Missing value

3245

66.8

1616

33.2

4861

100

632 2613

57.8 69.4

462 1154

42.2 30.6

1094 3767

100 100

1171 2068 6

65.5 67.5 66.7

617 996 3

34.5 32.5 33.3

1788 3064 9

100 100 100

258 673 1272 459 534 49

59.7 65.2 68.8 69.5 66.1 62.8

174 360 578 201 274 29

40.3 34.8 31.2 30.5 33.9 37.2

432 1033 1850 660 808 78

100 100 100 100 100 100

Variable

Visited

Household incomeb (in thousand rupiah per day) (t-test ¼
7.18, p ¼ .0918) Age of mother (po.001) Number of symptoms (po.0001) Child age (po.0001) a

Did not visit

n

Mean

Std. dev.

N

Mean

Std. dev.

2977 2660 3060 3245

6.2 29.3 2.6 1.6

2.6 6.3 1.2 1.3

1486 1309 1505 1616

6.0 30.0 2.3 1.8

2.7 6.7 1.2 1.3

Poverty status is based on percent of total expenditure on food. Very poor families spent more than 80% of their total expenditure on food. b Household income was based on daily household income in thousand rupiahs and excluded outliers of greater than 12,500 and lower then 300 rupiahs per day.

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

2919

Table 3 Frequency distribution of child outpatient visits and health card possession, by round Variable

Round One

Total sick children Outpatient visits (Chi-square ¼ 53.9, po.001) 1—Visited 2—Did not visit Health card possession (Chi-square ¼ 601.61, po.001) 0—No 1—Yes

Round Three

Total

n

Col. (%)

n

Col. (%)

n

2351

100.0

2510

100.0

4861

1471 880

62.6 37.4

1774 736

70.7 29.3

3245 1616

886 1465

37.7 62.3

208 2302

8.3 91.7

1094 3767

Table 4 Frequency distribution of sick children visiting an outpatient care by round and health card possession children Variable

Round One

Round Three

Visited

Total sick children (n ¼ 4861) Row (%) Health card possession 0—No Row (%) 1—Yes Row (%)

n

Col. (%)

Did not visit n

1471

100

880

63 (Chi-square ¼ 24.568, po.001) 498 33.9 56 973 66.2 66

388 492

outpatient care. The increase of outpatient visits in Round Three was significant (Chi-square ¼ 53.9, po.001). One possible factor contributing to the significant increase in outpatient visits was the increase in the number of children living in households that possessed health cards, given that poor families can obtain free health services by showing the cards. The table shows that 62% of sick children in Round One and 92% in Round Three lived in households possessing health cards. The 30% difference in health card possession among sick children shows a significant increase in the health card possession from Rounds One to Three (Chisquare ¼ 601.61, po.001). The above significant increase in health card possession, however, does not describe the association between health card possession and outpatient visits. In addition, this association needs to be evaluated to determine whether outpatient visits increased only for children living in families possessing a health card. The row percentages in

Tot. Col. (%)

Visited n

Col. (%) 100

Did not visit n

100

2351

1774

736

37

100

44.1 44 55.9 34

886 100 1465 100

71 (Chi-square ¼ 4.280, p ¼ .039) 134 7.6 74 64 1640 92.5 662 71

Tot. Col. (%) 100

2510

29

100

10.1 36 90.0 29

208 100 2302 100

Table 4 show some interesting findings. First, there were more sick children in Round Three (n ¼ 2510) than Round One (n ¼ 2351). Second, regardless of health card possession, outpatient visits of sick children increased by 8% from 63% in Round One to 71% in Round Three. Among those possessing health cards, the visits increased by 4% from 66% in Round One to 71% in Round Three. Interestingly, those not possessing health cards show a significant increase in out patient visits by 8% from 56% in Round One to 64% in Round Three. One reason for the increase number of sick children is the fact that baseline data in Round One was collected two months after JPS-BK was launched. Indonesia had no experience in implementing social safety net programs, and the initial months of the program were used for preparation. During the first two months of the program, not all of the eligible mothers and children received health cards and understood how to use the cards.

ARTICLE IN PRESS 2920

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

Increasing the number of sick children might not reflect the real increase of sick children. The finding of the increase in child outpatient visits regardless health card possession implies the ‘‘realized access’’ (Andersen, 1995) of the JPS-BK program. One possible explanation of the significant increase among sick children not possessing health cards is that health cards were distributed by the puskesmas director. It is possible that poor families who had not received health cards expected to get one on site when they brought their sick children to puskesmas. Looking at the column percentages, Table 4 shows that in Round One, about 66% of sick children visiting outpatient facilities lived in families possessing a health card. In Round Three, about 93% of sick children visiting outpatient facilities did. Interestingly, the table shows that health card possession increased not only among sick children visiting outpatient facilities, but also among sick children not visiting outpatient facilities. In Round One, about 56% of sick children not visiting an outpatient facility lived in families possessing a health card, while in Round Three about 90% of sick children did. This finding implies that the health card possession increased significantly during the first year of JPS-BK program regardless of child outpatient visits. Since health card possession is an enabling factor in the study, the finding also indicates that the JPS-BK program significantly increased ‘‘potential access’’ (Andersen, 1995) to health services for the poor. Looking at the detailed relationship between health card possession and outpatient visits in each round, Table 4 reveals an interesting finding. Namely, in Round One, the percentage of health card possession of sick children visiting outpatient facilities and those not visiting them differed significantly (Chi-square ¼ 24.57, po.001). In Round Three, however, the percentage of health card possession of the two groups did not differ significantly, given that almost all households possessed health cards. The findings indicate that within a year of the program, the JPS-BK had successfully distributed health cards to almost all poor families. The overall results of the evaluation on child outpatient visits show that health card possession, number of different symptoms, and children’s age had the strongest relationships to child outpatient visits (po.0001). Since these descriptive analyses

were performed independently, I performed multivariate analyses. Logistic regressions were chosen because the dependent variable is dummy. Four models were used in Round One and five models were used in Round Three. For both Rounds One and Three, the first four models are the same. The first model evaluated the effect of JPS-BK program (e.g., health card possession) on child outpatient visits. The second model added controls for the individual determinants: age, sex, and number of symptoms. The third model added controls for household determinants: poverty status, household income, family size, and mother’s educational level. The fourth model added controls for the contextual determinants: provincial origin, village/neighborhood developmental level, distance to the district office, number of different epidemics, and puskesmas medical supply. The fifth model used for Round Three data only added controls for the other two puskesmas determinants: puskesmas expenditure and staff personnel. Tables 5 and 6 show the results of the regression models in Rounds One and Three; from these, four findings can be derived. First, Table 5 shows that in all the Round One models, controlling for individual, household, and contextual variables, health card possession performed consistently as a significant predictor of child outpatient visits. Sick children living in households possessing health cards were more likely to visit an outpatient facility than those living in households not possessing health cards (OR between 1.48 and 1.54, po.0001). This finding shows that, through health card distribution, the JPS-BK program significantly increased child outpatient visits. In contrast, Table 6 shows that in all Round Three regression models, health card no longer acted as a significant predictor of child outpatient visits. Again, this was due to the fact that after a year if the JPS-BK program, most poor households possessed health cards. Second, in both Rounds One and Three, all models, excepting the fifth, show child’s age significantly influencing child outpatient utilization. Models 2, 3, and 4 in Tables 5 and 6 show consistently that younger sick children were more likely to visit an outpatient facility than their older counterparts, controlling for other individual, household, and contextual factors, as well as number of symptoms, round and health card possession (OR from 0.89 to 0.90, po.005). The significance of child age was consistent with the general epidemiological hypothesis that younger

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

2921

Table 5 Logistic regression of child outpatient visits, Round One Parameter

JPS-BK Health card possession (not possessed) Individual determinants Child age (year) Child sex (female) Number of symptoms Household determinants Poverty status (poor) Household income (thousand Rp.) Family size Mother’s educational level (no school/not finish elementary) Finished elementary + No informationa Contextual determinants Province (East Java) Central Java Yogyakarta South Sulawesi Village/neighborhood developmental level (Lower-development rural) Higher-development rural Urban Distance to district office Number of different epidemics Puskesmas pharmaceutical supply N Intercept
2 Log L

Model 1

Model 2

Model 3

Model 4

Est.

OR

Est.

OR

Est.

OR

Est.

OR

****0.43

1.54

****0.42

1.52

****0.40

1.49

****0.39

1.48

**
0.11 0.08 ****0.19

0.90 1.09 1.20

*
0.11 0.07 ****0.19

0.90 1.07 1.20

*
0.11 0.06 ****0.19

0.90 1.06 1.21


0.06 0.01
0.00

0.94 1.02 0.99


0.05 0.02
0.03

0.95 1.03 0.97

*0.31 0.10

1.36 1.1

0.28 0.11

1.32 1.12


0.23 0.42
0.12

0.79 1.52 0.89


0.03
0.03 0.00
0.03
0.19 1935 0.10 77.33

0.97 0.97 1.00 0.98 *0.83

2351 0.25 24.57

2200 -0.06 58.88

2031 -0.25 67.24

Note: * o.005, **o.001, ***.0005, ****o.0001. a Information on mother’s educational level was unavailable because the children were the grand-children of the household heads. Income higher than 12,500 rupiahs and distance higher than 300 km were excluded from the analyses.

children are more vulnerable to infectious diseases and therefore more likely to obtain health care services than their older counterparts. Third, the results from Models 2 to 5 in both rounds also show consistently that sick children who were reported to have experienced more symptoms were more likely to visit an outpatient facility than those reported to have experienced fewer symptoms, controlling for age, health card possession, household and contextual factors (OR from 1.20 to 1.37, po.0001). This is not surprising given that higher number of symptoms reported by caretakers may indicate the seriousness of the illness, and therefore these children were more likely to visit an outpatient facility. Fourth, Tables 5 and 6 reveal an interesting finding that household and contextual variables did

not influence child outpatient visits significantly. Only few variables showed significant effects on child outpatient visits. However, their significance was either in only a single model in one round or with a very low significance level (po.005), and therefore uninterpretable. Interestingly, the number of different epidemics shows as a significant predictor in Round Three (Models 4 and 5). Namely, sick children living in villages/neighborhoods with fewer epidemic types were more likely to visit an outpatient facility than those living in villages/ neighborhoods with more epidemic types. This finding needs careful evaluation because podes data did not provide information on the frequency of an epidemic occurred in a village. A village with fewer epidemic types, if the epidemics occurred intensively, might result in more sickness cases. It is

ARTICLE IN PRESS 2922

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

Table 6 Logistic Regression of child outpatient visits, Round Three Parameter

JPS-BK Health card possession (not possessed) Individual determinants Child age (year) Child sex (female) Number of symptoms Household determinants Poverty status (poor) Household income (thousand Rp.) Family size Mother’s educational level (no school/not finish elementary) Finished elementary + No informationa Contextual determinants Province (East Java) Central Java Yogyakarta South Sulawesi Village/neighborhood developmental level (Lower-development rural) Higher-development rural Urban Distance to district office Number of different epidemics Puskesmas pharmaceutical supply Other puskesmas determinants Puskesmas expenditure (million Rp.) Puskesmas staff N Intercept
2 Log L

Model 1 Est.

OR

Model 2 Est.

OR

Model 3 Est.

OR

Model 4 Est.

OR

Model 5 Est.

OR

0.31

1.37

0.23

1.26

0.24

1.27

0.27

1.31

0.37

1.45

*
0.11 0.15 ****0.29

0.89 1.17 1.33

*
0.11 0.11 ****0.26

0.89 1.12 1.30

*
0.11 0.09 ****0.27

0.90 1.10 1.31


0.11 0.12 ****0.32

0.90 1.13 1.38

0.17 0.00
0.07

1.19 1.00 0.93

0.13 0.02 *
0.08

1.14 1.02 0.92

0.19 0.03
0.08

1.20 1.03 0.92

0.08 0.06

1.09 1.06

0.09 0.08

1.10 1.08

0.10 0.11

1.11 1.12


0.19
0.38 0.30

0.82 0.68 1.35

*-0.41
0.52 0.35

0.67 0.60 1.42


0.09 0.13 0.00 *
0.17 0.01

0.91 1.14 1.00 0.85 1.01


0.20 0.01 0.01 ****
0.28 0.07

0.82 1.01 1.01 0.76 1.07

0.00
0.01 1768 0.11 96.71

1.00 0.99

2510 0.59 4.14

2365 0.13 59.17

2151 0.42 59.06

2115 0.39 77.00

Note: * o.005, **o.001, ***.0005, ****o.0001. a Information on mother’s educational level was unavailable because the children were the grand-children of the household heads. Income higher than 12,500 rupiahs and distance higher than 300 km were excluded from the analyses.

possible that some areas were hot spots for a certain type of epidemic. Discussion From the above analyses on the overall effects of the JPS-BK program on child health status, we learned that two significant improvements occurred during the first year of the JPS-BK program: (i) more sick children visited an outpatient facility and (ii) more children lived in households possessing health cards. From these findings we could derive that through health card distribution, the JPS-BK program significantly increased the ‘‘potential

access’’ and the ‘‘realized access’’ of poor children to health services. The increase in the ‘‘potential access’’ was demonstrated by the significant increase in health card possession regardless of the visits, while the increase in the ‘‘realized access’’ was demonstrated by the significant increase in child outpatient visits. The new strategy of health card distribution empowered local people to identify poor families more effectively. The finding shows that by the end of the first year of the program, almost all poor sick children had access to health services. It is worthwhile to note that the health card system was not new. Previously, this system received aid from the

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925

existing program for poverty alleviation, called Inpres Desa Tertinggal (IDT), or the ‘‘presidential instruction on the backward village’’ program. Under this program, all families—poor or not— within ‘‘backward villages,’’ or villages not yet modernized, received health cards. Although privileged families could take advantage of the IDT card system, the social stigma of living in the backward villages discouraged many of them from using it. On the other hand, poor families living outside the backward villages did not receive health cards to obtain free services. Because of their limited budget, unclear target, and socio-cultural barriers, the IDT programs failed to achieve the expected purposes (Kusnanto, 1999). The logistic regressions on child outpatient visits show clearly that there were only three strong predictors of child outpatient visits: (i) level of susceptibility to illness (i.e., child’s age); (ii) household ability to obtain health services (i.e., health card possession); and (iii) children’s health condition (i.e., number of symptoms). These three predictors represent the predisposing, enabling, and illness need determinants, respectively. Although predisposing and enabling determinants contain many other less direct factors, child age, health card possession, and number of symptoms were the only significant determinants. This implies that, for the poor, child health care services utilization during the first year of the JPS-BK program were influenced more directly by individual and enabling factors than by demographic, household, or contextual ones. The policy implication of this finding is that village team should update the list of poor people in the area regularly to make the correct target people received health cards. To prevent health card distribution from ‘moral hazard’ (giving the cards to relatives) as indicated by Utomo (2003), village team should include people from different institution and be monitored by local non-government organization. Evaluating the equity of access, furthermore, the child’s age (demographic factor) and the number of symptoms (need) acted as the measures of ‘‘equitable access’’, while the health card possession (enabling resource) acted as a measure of ‘‘inequitable access.’’ The significance of these factors on child health services utilization made it difficult to examine the equity of access to health care clearly. Realizing that enabling resource determines who gets medical care, however, we may conclude that child’s access to health care in the study was

2923

equitable because: (i) the study focused on poor children whose households were eligible for health cards and (ii) in Round Three, almost all households possessed health cards and the logistic regressions demonstrated that health card possession was no longer a significant predictor of child health care utilization. The results of analyses in this study have theoretical implications for the Andersen’s framework of health care services utilization in developing countries, particularly how the framework could be applied for poor children. Previously, I argued that applications of the framework in developing countries should consider contextual characteristics, such as level of modernization and rural/urban status, given that Andersen developed the framework based on the health circumstances in the US in 1960s. The contextual characteristics commonly distinguish developing countries from the developed ones. However, the results of the study showed that these characteristics did not influence child health services utilization in the four provinces in Indonesia. It is possible that specific conditions of my study, which was generally carried out during the economic crises among children who were under the poverty level, made those factors trivial. The significance of contextual determinants could be found in more diverse circumstances. The most similar and recent study under such circumstance was carried out by Thind and Andersen (2003). They evaluated the predictors for health services utilization of Dominican children under five years old who had respiratory illnesses. Using the 1991 Demography and Health Survey data, they found that sex, location (rural/urban), and possession index quartile (to represent household economic status) significantly influenced the parent’s decision to seek care for their sick children. In contrast to the results of their study, I found that age, number of symptoms, and health card possession were the predictors for child health services utilization. Although sex in their study and age in my study are individual predisposing determinants, sex relates more to local norms on gender expectations and preferences, while age relates more to the susceptibility to illnesses. The significance of the number of symptoms in my study also relates directly to the factors of illness severity and need. Indeed, both studies found that enabling factors (possession index quartile and health card possession) influenced child

ARTICLE IN PRESS 2924

E. Suci / Social Science & Medicine 63 (2006) 2912–2925

health services utilization significantly. Since my study focused on poor children, it may be that, for poor households, factors that are more direct have stronger influence than less direct factors. In addition, because they are poor, norms and other contextual factors contribute very little to the decision to visit a health facility. The policy implication of this finding is that the government needs to firstly address the issues of financial barriers as direct enabling factor to increase the access of primary health services for the poor. Since the JPS-BK program was temporary, this factor should be evaluated later under normal economic condition. Conclusion The results of this study have direct implications for developing a better safety net program in Indonesia. The study demonstrated that health card possession was a strong enabling factor for the poor to obtain health care services. We also learned that the JPS-BK program had successfully distributed health cards to the poor. The percentage of children living in households possessing health cards in Round Three was 92% by the end of 1999. Merely possessing a health card, however, may not enable poor families to obtain health care services, because health card holders need information on the procedures for using it. Therefore, any health policy attempting to improve access to health care services for the poor should not only be concerned with the distribution of health cards, but also with providing the proper information to the poor on how to use them. Realizing that JPSBK was a rescue program to cushion the poor from becoming poorer during the economic crisis, Indonesian policy makers should also consider developing a more established program to improve the health of poor children. Acknowledgment I would like to thank Albert Wessen and Roger Avery for their time and comment which helped develop this report References Ananta, A. (2003). What do we learn from the crisis?: Insights on human development in Indonesia during 1997–99. In A. Ananta (Ed). The Indonesian crisis: A human develop-

ment perspective. Singapore:Institute of Southeast Asian Studies. Ananta, A., & Siregar, R. (1999). Social safety net policies in Indonesia: Objectives and shortcomings. ASEAN Economic Bulletin, 16(3), 344–360. Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36(March), 1–10. Andersen, R. M., & Newman, J. E. (1973). Societal and individual determinants of medical care utilization in the United States. Milbank Quarterly, 51, 95–124. BAPPENAS (2004). Kebijakan dan Strategi Pemenuhan Kebutuhan Sosial Dasar Bagi Masyarakat Miskin. Jakarta. BAPPENAS Beckman, L. J. (1984). Analysis of the suitability of alcohol treatment resources for women. Substance and Alcohol Actions Misuse, 5(1), 21–27. Benyoussef, A., & Wessen, A. (1974). Utilization of health service in developing countries—Tunisia. Social Science & Medicine, 8, 287–304. Bresnan, J. (1999). The United States, the IMF, and the Indonesian financial crisis. In A. Schwarz, & J. Paris (Eds.), The politics of post-Suharto Indonesia (pp. 87–112). New York: Council on Foreign Relation Press. Central Bureau of Statistics (1998). Indonesia Demographic and Health Survey (IDHS) 1997. Calverton, MD: Macro International Inc. Junaidi, P. (2001). Access of health care for poor people during crisis: Indonesian case. Journal of the Indonesian Medical Association (JIMA), II(VI), 1–11. Kusnanto, H. (1999). Ketepatan Sasaran Keluarga Miskin dalam JPS-BK: Survey di Jateng, Jatim Bagian Selatan DIY. Medika, Special Edition, September 24–26. Ministry of Health (1999a). Indonesian Health Profile 1999. Jakarta: Health Department. Ministry of Health (1999b). Pedoman Pelaksanaan Program Jaring Pengaman Sosial Bidang Kesehatan. Jakarta: Health Department. Ministry of Health (1999c). Pedoman Pelaksanaan Program Jaring Pengaman Sosial Bidang Kesehatan Bagi Puskesmas. Jakarta: Health Department. Muhammad, M. (2000). Laporan Tim Pengendali Gugus Tugas peningkatan Jaring Pengaman Sosial Tahun Anggaran 1999/ 2000. Jakarta: April, 2000, unpublished. Suryahadi, A., Suharso, Y., & Sumarto, S. (2001). Coverage and targetting in the Indonesian social safety net programmes. ASEAN Economic Bulletin, 18(2), 161–176. Suwandono, A. (1999). Hasil Studi longitudinal mutisenter JPSBK. Medika, (Suppl), 8–13. Thind, A., & Andersen, R. (2003). Respiratory illness in the Dominican Republic: What are the predictors for health services utilization of young children. Social Science & Medicine, 56, 1173–1182. Utomo, B. (2003). Health status of Indonesians during the economic crisis. In A. Ananta (Ed.), The Indonesian crisis: A human development perspective. Singapore: Institute of Southeast Asian Studies. Wan, T. T. (1989). The behavioral model of health care utilization and older people. In M. Ori, & K. Bon (Eds.), Aging and health care. New York: Rouledge.

ARTICLE IN PRESS E. Suci / Social Science & Medicine 63 (2006) 2912–2925 Wolinsky, F. D. (1978). Assessing the effects of presdisposing, enabling, and illness-morbidity characteristics on health service utilization. Journal of Health and Social Behavior, 19(December), 384–396. Wolinsky, F. D., & Johnson, R. J. (1991). The use of health services by older adults. Journal of Gerontology: Social Sciences, 46(6), S345–S357.

2925

World Bank (1999). Report and recommendations of the president of the international bank for reconstruction and development to the executive directors on a proposed social safety net adjustment loan in the amount of US $600 million to the Republic of Indonesia. Environment and Social Development Unit, East Asia and Pacific Region.