Abstract
Access to skilled birth attendants and emergency obstetric care are thought to prevent early neonatal deaths. This study aims to examine the association between the type of delivery attendant and place of delivery and early neonatal mortality in Indonesia.
Four Indonesia Demographic and Health Surveys from 1994, 1997, 2002/2003 and 2007 were used, including survival information from 52 917 singleton live-born infants of the most recent birth of a mother within a 5-year period before each survey. Cox proportional hazards regression models were used to obtain the hazard ratio for univariable and multivariable analyses.
Our study found no significant reduction in the risk of early neonatal death for home deliveries assisted by the trained attendants compared with those assisted by untrained attendants. In rural areas, the risk of early neonatal death was higher for home deliveries assisted by trained attendants than home deliveries assisted by untrained attendants. In urban areas, a protective role of institutional deliveries was found if mothers had delivery complications. However, an increased risk was associated with deliveries in public hospitals in rural areas. Infants of mothers attending antenatal care services were significantly protected against early neonatal deaths, irrespective of the urban or rural setting. An increased risk of early neonatal death was also associated with male infants, infants whose size at birth was smaller than average and/or infants reported to be born early. A reduced risk was observed amongst mothers with high levels of education.
Continuous improvement in the skills and the quality of the village midwives might benefit maternal and newborn survival. Efforts to strengthen the referral system and to improve the quality of delivery and newborn care services in health facilities are important, particularly in public hospitals and in rural areas.
No significant reduction was observed in the risk of early neonatal death for home deliveries assisted by trained delivery attendants compared with those assisted by untrained delivery attendants.
In rural areas, the risk of early neonatal death was significantly higher amongst home deliveries assisted by trained delivery attendants compared with untrained attendants.
In urban areas, infants of mothers with delivery complications who were delivered at health care facilities had a reduced risk of death.
Efforts to improve the quality of village midwife services may benefit women who prefer a home delivery.
Introduction
According to the most recent estimate of child and neonatal mortality, of the 7.7 million under-five deaths globally, 3.1 million occur in the neonatal period (neonatal deaths) (Rajaratnam et al. 2010). Of all neonatal deaths, three-quarters occur in the early neonatal period, i.e. 0–6 days after delivery (WHO 2006). The reduction in neonatal mortality rates has been slower than for post-neonatal and under-five mortality rates, particularly in low- and middle-income countries (Murray et al. 2007; Lawn et al. 2009; Rajaratnam et al. 2010). As a result, the proportion of under-five deaths due to neonatal deaths has been increasing (Lawn et al. 2005; Lawn et al. 2009). Interventions that address neonatal mortality, including early neonatal deaths, are important to help countries achieve the fourth Millennium Development Goals target of reducing under-five mortality by two-thirds by 2015 (United Nations 2001).
Early neonatal death reflects the quality of care received by the mother during the antenatal period and at childbirth (Ngoc et al. 2006; WHO 2006). The availability, accessibility and quality of skilled birth attendance and emergency obstetric care are considered important for reducing the burden of early neonatal deaths (WHO 2006; Lawn et al. 2009). However, a recently published trial conducted in rural communities of six developing countries found no significant reduction in rates of early neonatal deaths after the implementation of essential newborn care training for birth attendants, including nurses, physicians, midwives and traditional birth attendants (Carlo et al. 2010). This finding has raised some doubts about the impact of skilled birth attendants on neonatal survival.
In developing countries, including Indonesia, access to safe delivery services remains a great challenge (Darmstadt et al. 2009). Delivery care services are underutilized, and in some resource-poor settings the services of traditional birth attendants and home deliveries are preferred by many women (Titaley et al. 2010a). The long-standing commitment of the Indonesian Government to improve maternal and newborn health is demonstrated by the implementation of the village midwife programme in 1989 (Geefhuysen 1999; Hatt et al. 2007). By 1996, more than 50 000 midwives had been assigned to villages across Indonesia to improve the quality of delivery services for women who mainly gave birth at home assisted by traditional birth attendants (Geefhuysen 1999). A number of investigators have examined whether the village midwife programme has had a favourable impact on maternal and newborn health (Hatt et al. 2007; Hatt et al. 2009). Using the pooled data of the 1991 to 2002/2003 Indonesia Demographic and Health Survey (IDHS), one study demonstrated that professional attendance at home was not associated with a reduction of early neonatal deaths (Hatt et al. 2009). Another study showed that although the village midwife programme has improved access to trained delivery attendants amongst poor communities, the gap in access to emergency obstetric care has widened (Hatt et al. 2007).
The IDHS 2007 reported that 79% of mothers in Indonesia were assisted by trained attendants at childbirth; however, only 46% of mothers had an institutional delivery (Badan Pusat Statistik-Statistics Indonesia et al. 2008). The disparity in the use of delivery care services was evident across provinces. The use of trained attendants varied from 33% in Maluku to 97% in DKI Jakarta province, and the proportion of institutional deliveries varied from only 8% in Southeast Sulawesi to 91% in Bali Province (Badan Pusat Statistik-Statistics Indonesia et al. 2008). Mothers reported institutional deliveries in only 29% of rural births compared with 70% of urban births (Badan Pusat Statistik-Statistics Indonesia et al. 2008). A study from two districts of West Java Province also reported a relatively low percentage of professional care at birth (33%), despite the investments made in the village midwife programme (Achadi et al. 2007).
This study explores the association between the type of delivery attendant, place of delivery and the risk of early neonatal mortality in Indonesia. Specifically, it examines whether the use of trained birth attendants at home protects against early neonatal deaths. This study also looks at whether trained delivery attendants or institutional deliveries are associated with reduced risk of early neonatal deaths amongst women in urban and rural areas, and whether or not the mothers report delivery complications.
Methods
Data sources
Pooled data from the 1994, 1997, 2002/2003 and 2007 IDHS were used in this study (Macro International Inc. 2009). The IDHS collects demographic and health information by interviewing ever married women aged 15–49 years and ever married men aged 15–54 years, using the Household, Women and Men Questionnaire (Badan Pusat Statistik-Statistics Indonesia et al. 1995; Badan Pusat Statistik-Statistics Indonesia et al. 1998; Badan Pusat Statistik-Statistics Indonesia et al. 2003; Badan Pusat Statistik-Statistics Indonesia et al. 2008). The probability sampling methods of the IDHS have been described elsewhere (Macro International Inc. 1996). All information obtained in these surveys was self-reported.
We examined information of 52 917 singleton live-born infants of the most recent birth of a mother within the period of 5 years before each survey. This included 506 early neonatal deaths. Only the most recent births have detailed information about use of maternal and child care services, so our analyses were restricted to these births. Also, by using only the most recent delivery of a mother for the analyses, we had limited potential recall bias about the mothers’ pregnancy, delivery and post-partum histories.
Variables
Outcome and main exposure variable
The outcome variable was early neonatal mortality, defined as deaths occurring within the first 7 days of life (0–6 days) (WHO 2006). To examine the association that early neonatal death has with delivery attendance and place of delivery, a composite variable was constructed as the main exposure, consisting of type of delivery attendants (trained and untrained attendants) and place of delivery (home, public hospital, private hospital, other public birthing centres and other private birthing centres).
Trained delivery attendants include doctors, midwives and village midwives, based on the World Health Organization (WHO) definition as ‘an accredited health professional who has been educated and trained to proficiency in the skills needed to manage normal (uncomplicated) pregnancies, childbirth and the immediate postnatal period, and in the identification, management and referral of complications in women and newborns’ (WHO 2004). Untrained delivery attendants included traditional birth attendants, family members and other relatives.
All deliveries in health facilities, i.e. public hospitals, private hospitals and birthing centres, were assisted by trained delivery attendants. Other public birthing centres included health centres, sub-health centres, village maternity posts and village health posts. The health centre is the primary care unit at the sub-district level and is usually headed by a medical doctor. Only some health centres have inpatient care used mostly for delivery care services (WHO Regional Office for South-East Asia 2008). Each health centre has around three to five sub-health centres available at the village level that serve groups of villages (WHO Regional Office for South-East Asia 2008). Village maternity and health posts are available at the village level and are managed by a village midwife (WHO Regional Office for South-East Asia 2008). Other private birthing centres included private clinics and private practices of family physicians, obstetrics and gynaecology specialists, midwives, nurses or village midwives.
Other potential risk factors
Using a framework adapted from Mosley and Chen (1984) for child survival in developing countries, we identified 15 potential confounders from the pooled data. These variables were categorized into three main groups: community-level characteristics, socio-economic factors and proximate determinants (Figure 1). Proximate determinants were divided into two sub-groups: maternal and neonatal, and perinatal health care service factors.
Conceptual framework of factors associated with early neonatal deaths in Indonesia. Note: Adapted from the Mosley and Chen framework (Mosley and Chen 1984).
Conceptual framework of factors associated with early neonatal deaths in Indonesia. Note: Adapted from the Mosley and Chen framework (Mosley and Chen 1984).
A wealth index variable was constructed to rank households across all IDHS surveys using principal component analysis (Filmer and Pritchett 2001) of 11 housing characteristics and assets, i.e. source of drinking water, type of toilet, main material of floor, main material of wall, availability of electricity, possession of radio, television, fridge, and bicycle, motorcycle and car ownership.
A composite variable of birth rank and birth interval as a five-category variable was constructed to examine the impact of combined birth rank and birth interval on early neonatal mortality. These categories were first birth rank infants, second and third birth rank infants with previous birth interval of more than 2 years, second and third rank infants with previous birth interval of less than or equal to 2 years, fourth birth rank infants with interval of more than 2 years, and fourth birth rank infants with interval of less than or equal to 2 years.
Another composite variable consisted of the combined birth size and timing of delivery. Information about child size at birth was based on mothers’ subjective assessments and was categorized into smaller than average-size infants, average-size infants and larger than average-size infants. This variable was used as a proxy for birth weight, due to the high percentage of missing values (31%) for the birth weight variable. The timing of delivery consisted of two categories, infants born early and infants born on time. The information about timing of delivery in the 1994 and 1997 IDHS was based on mothers’ reports of whether or not their infant was born on time or prematurely (Badan Pusat Statistik-Statistics Indonesia et al. 1995; Badan Pusat Statistik-Statistics Indonesia et al. 1998). In the 2002/2003 and 2007 IDHS, early born infants were identified based on mothers’ reports that their labour occurred before the ninth month of pregnancy (Badan Pusat Statistik-Statistics Indonesia et al. 2003; Badan Pusat Statistik-Statistics Indonesia et al. 2008).
The delivery complication variable was based on five types of obstetric complications reported by mothers, i.e. prolonged labour referring to strong and regular contractions lasting more than 1 day and 1 night; excessive vaginal bleeding that soaked more than three pieces of cloth; high fever and foul smelling vaginal discharge; convulsions with loss of consciousness; and ‘waters breaking’ more than 6 hours before delivery.
Statistical analysis
To describe the characteristics of infants included in this study, we performed frequency tabulations of all variables included in our analysis. To examine the association between the study outcome and the main exposure variable, Cox proportional hazards regression models were used to obtain the crude hazard ratio (HR) from univariable analyses and adjusted HR from multivariable analyses.
In the multivariable analyses, we employed a hierarchical modelling strategy in which factors potentially associated with the outcome were entered progressively into the model, from the most distal to the most proximate variables (Victora et al. 1997). In the first stage, the more distal factors, i.e. community-level characteristics, were entered to assess their associations with the study outcome. A backward elimination procedure was performed and only factors significantly associated with the study outcome, using a significance level of 0.05, were retained in the model (Model 1).
In the second stage, socio-economic factors were entered into Model 1. This was followed by the backward elimination technique to remove socio-economic factors not significantly associated with early neonatal deaths (Model 2). In the third stage, only maternal and neonatal factors of the proximate determinants were included. A similar approach was applied to retain maternal and neonatal factors significantly associated with early neonatal deaths (Model 3). In the final model (Model 4), the perinatal and health care services factors, including the main exposure variable (combined birth attendance and place of delivery), were entered. Three variables, year of survey, combined region and type of residence (urban/rural), and maternal age at childbirth were selected a priori and retained in the models irrespective of their level of significance.
On a priori grounds we examined the interaction between the main exposure variable (type of delivery attendant and place of delivery) and the type of residence (urban/rural), since studies have reported the influence of urban and rural differences in the use of delivery care services (Shakya and McMurray 2001; Thind and Banerjee 2004). When a significant interaction was found, the final model (Model 4) was stratified based on the type of residence variable. Previous literature has also shown a relationship between delivery complications and the use of delivery care services (Paul and Rumsey 2002; Thind and Banerjee 2004; Titaley et al. 2010b), therefore this interaction was also examined in this study in the stratified models described previously.
All statistical analyses used the STATA/MP version 10.0 (2007) (Stata Corporation, College Station, TX, USA) and survey commands were used for Cox proportional hazards to adjust for the sampling weights and cluster sampling design. All of the HR and 95% confidence intervals (CI) were weighted for the sampling probabilities.
Results
Our study examined data of 52 917 singleton live-born infants, most recently born to a mother within the 5 years preceding each survey. This included 506 early neonatal deaths. The proportion of early neonatal deaths to all neonatal deaths increased from 66% (95% CI: 55.8–75.3%) in the 1994 IDHS to 78% (95% CI: 66.0–86.0%) in the 2007 IDHS. Of all home deliveries assisted by untrained delivery attendants, 95% (95% CI: 94.0–95.1%) were assisted by traditional birth attendants. Of home deliveries assisted by trained attendants, 99% (95% CI: 98.5–99.1%) were assisted by nurses and midwives, including village midwives.
Figure 2 shows the trends of utilization of delivery attendance and place of delivery by survey year. Overall, there was a decreasing trend of home deliveries assisted by untrained attendants, from 59% (95% CI: 56.5–62.0%) in the 1994 IDHS to 25% (95% CI: 23.3–27.5%) in the 2007 IDHS. A substantial increase in the use of private birthing centres was observed, from 5% (95% CI: 4.2–5.9%) in the 1994 IDHS to 29% (95% CI: 27.0–31.3%) in the 2007 IDHS. This was mainly attributable to an increase in the use of private clinics, including private midwife practices. In urban areas, our analyses showed that 18% (95% CI: 16.3–20.3%) of women delivered at home assisted by untrained attendants and 11% (95% CI: 9.8–11.4%) delivered in public hospitals. In rural areas, 55% (95% CI: 53.3–56.6%) of women delivered at home assisted by untrained attendants and only 4% (95% CI: 3.5–4.2%) delivered in public hospitals.
Trends in the use of delivery attendants and place of delivery in Indonesia, IDHS 1994–2007. Notes: All values are weighted for the sampling probability. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
Trends in the use of delivery attendants and place of delivery in Indonesia, IDHS 1994–2007. Notes: All values are weighted for the sampling probability. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
Table 1 shows the baseline characteristics of the study population, along with the results of the univariable and multivariable analyses. There was no significant difference in the risk of early neonatal death between home deliveries assisted by trained attendants and those assisted by untrained attendants (HR = 1.24, P = 0.31) (Table 1). However, deliveries at public hospitals were associated with an increased risk (HR = 2.65, P < 0.01). Amongst socio-economic determinants, higher maternal education had a protective effect against early neonatal mortality, whereas infants born to parents who were both employed had a significantly increased risk compared with infants whose father alone was employed. The median household wealth index was 0.17 where only the father alone was employed, but only -0.25 where both the mother and father were employed, suggesting that in poorer households both parents need to work. However, when household wealth index was entered to replace the parental occupation variable, no significant association between household wealth index and early neonatal death was observed. The distribution of delivery care services across quintiles of household wealth index is shown in Supplementary Web Table 1. Amongst the proximate determinants, male infants, smaller than average-size infants and/or infants reported to be born early, as well as infants whose mother reported delivery complications, exhibited an increased risk of early neonatal deaths.
Hazard ratio (HR) for early neonatal mortality for community, socio-economic and some proximate determinants: the results of multivariable analysis, IDHS 1994–2007
| Variable | N | (%) | Unadjusted1 | Adjusted1 | ||||
|---|---|---|---|---|---|---|---|---|
| HR | (95% CI)2 | P3 | HR | (95% CI)2 | P3 | |||
| Year of survey | ||||||||
| 1994 (Ref)4 | 13 282 | (25.1) | 1.00 | 1.00 | ||||
| 1997 | 13 043 | (24.7) | 0.92 | (0.65–1.30) | 0.65 | 0.93 | (0.66–1.32) | 0.70 |
| 2002/2003 | 12 646 | (23.9) | 0.70 | (0.46–1.06) | 0.09 | 0.71 | (0.47–1.09) | 0.12 |
| 2007 | 13 945 | (26.4) | 0.58 | (0.39–0.86) | 0.01 | 0.54 | (0.36–0.81) | <0.01 |
| COMMUNITY FACTORS | ||||||||
| Region and type of residence | ||||||||
| Java-Bali region: urban | 13 200 | (24.9) | 1.00 | 1.00 | ||||
| Java-Bali region: rural | 17 937 | (33.9) | 0.99 | (0.63–1.55) | 0.98 | 0.77 | (0.48–1.25) | 0.29 |
| Sumatera region: urban | 3238 | (6.1) | 1.13 | (0.68–1.87) | 0.65 | 1.13 | (0.69–1.84) | 0.64 |
| Sumatera region: rural | 8070 | (15.3) | 1.05 | (0.69–1.59) | 0.82 | 0.85 | (0.54–1.33) | 0.48 |
| Eastern Indonesia region: urban | 2642 | (5.0) | 1.11 | (0.69–1.80) | 0.66 | 1.09 | (0.69–1.73) | 0.71 |
| Eastern Indonesia region: rural | 7830 | (14.8) | 1.35 | (0.92–1.99) | 0.13 | 1.03 | (0.67–1.58) | 0.89 |
| Average paternal years of schooling | Mean ± SE = 7.67 ± 0.06 | 0.94 | (0.90–0.99) | 0.02 | ||||
| SOCIO-ECONOMIC FACTORS | ||||||||
| Maternal marital status | ||||||||
| Currently married | 51 538 | (97.4) | 1.00 | |||||
| Formerly married | 1379 | (2.6) | 1.68 | (0.81–3.51) | 0.16 | |||
| Maternal education | ||||||||
| No education | 3671 | (6.9) | 1.00 | 1.00 | ||||
| Incomplete primary | 10 735 | (20.3) | 0.78 | (0.45–1.37) | 0.39 | 0.86 | (0.49–1.51) | 0.61 |
| Complete primary | 16 809 | (31.8) | 0.73 | (0.42–1.25) | 0.25 | 0.93 | (0.53–1.63) | 0.80 |
| Incomplete secondary | 9957 | (18.8) | 0.63 | (0.37–1.08) | 0.09 | 0.83 | (0.46–1.47) | 0.52 |
| Complete secondary | 9057 | (17.1) | 0.50 | (0.27–0.93) | 0.03 | 0.59 | (0.29–1.19) | 0.14 |
| Tertiary | 2685 | (5.1) | 0.37 | (0.16–0.82) | 0.01 | 0.34 | (0.15–0.80) | 0.01 |
| Parental occupation | ||||||||
| Not working mother + working father | 28 857 | (54.5) | 1.00 | 1.00 | ||||
| Working mother + working father | 22 925 | (43.3) | 1.53 | (1.15–2.02) | <0.01 | 1.54 | (1.15–2.07) | <0.01 |
| Unemployed father | 999 | (1.9) | 1.33 | (0.43–4.15) | 0.62 | 1.26 | (0.40–3.98) | 0.69 |
| Household wealth index | Mean ± SE = 0.29 ± 0.03 | 0.93 | (0.86–0.99) | 0.03 | ||||
| PROXIMATE DETERMINANTS | ||||||||
| Maternal and neonatal factors | ||||||||
| Maternal age at childbirth | Mean ± SE = 26.94 ± 0.05 | 1.02 | (1.00–1.05) | 0.10 | 1.02 | (1.00–1.05) | 0.07 | |
| Child sex | ||||||||
| Female | 25 568 | (48.3) | 1.00 | 1.00 | ||||
| Male | 27 348 | (51.7) | 1.55 | (1.17–2.07) | <0.01 | 1.65 | (1.23–2.22) | <0.01 |
| Birth rank and interval | ||||||||
| 1st rank | 17 053 | (32.2) | 1.00 | |||||
| 2nd/3rd rank, interval ≤2 yrs | 3351 | (6.3) | 0.97 | (0.56–1.68) | 0.92 | |||
| 2nd/3rd rank, interval >2 yrs | 19 816 | (37.5) | 0.84 | (0.57–1.23) | 0.36 | |||
| ≥4th rank, interval >2 yrs | 10 775 | (20.4) | 1.22 | (0.84–1.78) | 0.29 | |||
| ≥4th rank, interval ≤2 yrs | 1922 | (3.6) | 1.51 | (0.94–2.42) | 0.09 | |||
| Desire for pregnancy | ||||||||
| Wanted then | 43 175 | (81.6) | 1.00 | |||||
| Wanted later | 5257 | (9.9) | 1.06 | (0.62–1.80) | 0.83 | |||
| Wanted no more | 4384 | (8.3) | 0.94 | (0.57–1.56) | 0.81 | |||
| Delivery complication | ||||||||
| No | 34 467 | (65.1) | 1.00 | 1.00 | ||||
| Yes | 18 080 | (34.2) | 1.81 | (1.36–2.41) | <0.001 | 1.82 | (1.34–2.46) | <0.001 |
| Birth size and timing of delivery | ||||||||
| Average-sized + born on time | 26 507 | (50.1) | 1.00 | 1.00 | ||||
| Average-sized + early born | 435 | (0.8) | 5.70 | (2.63–12.35) | <0.001 | 5.78 | (2.62–12.77) | <0.001 |
| Smaller than average + born on time | 6568 | (12.4) | 2.50 | (1.76–3.56) | <0.001 | 2.49 | (1.74–3.56) | <0.001 |
| Smaller than average + early born | 486 | (0.9) | 16.56 | (10.83–25.32) | <0.001 | 14.85 | (9.50–23.21) | <0.001 |
| Larger than average | 17 475 | (33.0) | 0.99 | (0.67–1.46) | 0.96 | 0.92 | (0.62–1.37) | 0.69 |
| Perinatal health care services | ||||||||
| Use of antenatal care service | ||||||||
| No | 5116 | (9.7) | 1.00 | 1.00 | ||||
| Yes | 47 480 | (89.7) | 0.45 | (0.31–0.64) | <0.001 | 0.49 | (0.33–0.73) | <0.001 |
| Mode of delivery | ||||||||
| Non-caesarean section | 50 383 | (95.2) | 1.00 | |||||
| Caesarean section | 2458 | (4.6) | 1.46 | (0.83–2.56) | 0.19 | |||
| Type of delivery attendants and place of delivery | ||||||||
| Untrained attendants at home | 21 993 | (41.6) | 1.00 | 1.00 | ||||
| Trained attendants at home | 13 886 | (26.2) | 0.91 | (0.63–1.30) | 0.60 | 1.24 | (0.82–1.87) | 0.31 |
| Public hospital | 3319 | (6.3) | 2.40 | (1.52–3.79) | <0.001 | 2.65 | (1.48–4.75) | <0.01 |
| Private hospital | 3164 | (6.0) | 0.61 | (0.33–1.13) | 0.11 | 0.83 | (0.41–1.68) | 0.60 |
| Other public birthing centres5 | 1553 | (2.9) | 0.45 | (0.21–0.99) | 0.05 | 0.65 | (0.29–1.46) | 0.30 |
| Other private birthing centres6 | 8820 | (16.7) | 0.68 | (0.41–1.14) | 0.14 | 1.18 | (0.67–2.08) | 0.57 |
| Variable | N | (%) | Unadjusted1 | Adjusted1 | ||||
|---|---|---|---|---|---|---|---|---|
| HR | (95% CI)2 | P3 | HR | (95% CI)2 | P3 | |||
| Year of survey | ||||||||
| 1994 (Ref)4 | 13 282 | (25.1) | 1.00 | 1.00 | ||||
| 1997 | 13 043 | (24.7) | 0.92 | (0.65–1.30) | 0.65 | 0.93 | (0.66–1.32) | 0.70 |
| 2002/2003 | 12 646 | (23.9) | 0.70 | (0.46–1.06) | 0.09 | 0.71 | (0.47–1.09) | 0.12 |
| 2007 | 13 945 | (26.4) | 0.58 | (0.39–0.86) | 0.01 | 0.54 | (0.36–0.81) | <0.01 |
| COMMUNITY FACTORS | ||||||||
| Region and type of residence | ||||||||
| Java-Bali region: urban | 13 200 | (24.9) | 1.00 | 1.00 | ||||
| Java-Bali region: rural | 17 937 | (33.9) | 0.99 | (0.63–1.55) | 0.98 | 0.77 | (0.48–1.25) | 0.29 |
| Sumatera region: urban | 3238 | (6.1) | 1.13 | (0.68–1.87) | 0.65 | 1.13 | (0.69–1.84) | 0.64 |
| Sumatera region: rural | 8070 | (15.3) | 1.05 | (0.69–1.59) | 0.82 | 0.85 | (0.54–1.33) | 0.48 |
| Eastern Indonesia region: urban | 2642 | (5.0) | 1.11 | (0.69–1.80) | 0.66 | 1.09 | (0.69–1.73) | 0.71 |
| Eastern Indonesia region: rural | 7830 | (14.8) | 1.35 | (0.92–1.99) | 0.13 | 1.03 | (0.67–1.58) | 0.89 |
| Average paternal years of schooling | Mean ± SE = 7.67 ± 0.06 | 0.94 | (0.90–0.99) | 0.02 | ||||
| SOCIO-ECONOMIC FACTORS | ||||||||
| Maternal marital status | ||||||||
| Currently married | 51 538 | (97.4) | 1.00 | |||||
| Formerly married | 1379 | (2.6) | 1.68 | (0.81–3.51) | 0.16 | |||
| Maternal education | ||||||||
| No education | 3671 | (6.9) | 1.00 | 1.00 | ||||
| Incomplete primary | 10 735 | (20.3) | 0.78 | (0.45–1.37) | 0.39 | 0.86 | (0.49–1.51) | 0.61 |
| Complete primary | 16 809 | (31.8) | 0.73 | (0.42–1.25) | 0.25 | 0.93 | (0.53–1.63) | 0.80 |
| Incomplete secondary | 9957 | (18.8) | 0.63 | (0.37–1.08) | 0.09 | 0.83 | (0.46–1.47) | 0.52 |
| Complete secondary | 9057 | (17.1) | 0.50 | (0.27–0.93) | 0.03 | 0.59 | (0.29–1.19) | 0.14 |
| Tertiary | 2685 | (5.1) | 0.37 | (0.16–0.82) | 0.01 | 0.34 | (0.15–0.80) | 0.01 |
| Parental occupation | ||||||||
| Not working mother + working father | 28 857 | (54.5) | 1.00 | 1.00 | ||||
| Working mother + working father | 22 925 | (43.3) | 1.53 | (1.15–2.02) | <0.01 | 1.54 | (1.15–2.07) | <0.01 |
| Unemployed father | 999 | (1.9) | 1.33 | (0.43–4.15) | 0.62 | 1.26 | (0.40–3.98) | 0.69 |
| Household wealth index | Mean ± SE = 0.29 ± 0.03 | 0.93 | (0.86–0.99) | 0.03 | ||||
| PROXIMATE DETERMINANTS | ||||||||
| Maternal and neonatal factors | ||||||||
| Maternal age at childbirth | Mean ± SE = 26.94 ± 0.05 | 1.02 | (1.00–1.05) | 0.10 | 1.02 | (1.00–1.05) | 0.07 | |
| Child sex | ||||||||
| Female | 25 568 | (48.3) | 1.00 | 1.00 | ||||
| Male | 27 348 | (51.7) | 1.55 | (1.17–2.07) | <0.01 | 1.65 | (1.23–2.22) | <0.01 |
| Birth rank and interval | ||||||||
| 1st rank | 17 053 | (32.2) | 1.00 | |||||
| 2nd/3rd rank, interval ≤2 yrs | 3351 | (6.3) | 0.97 | (0.56–1.68) | 0.92 | |||
| 2nd/3rd rank, interval >2 yrs | 19 816 | (37.5) | 0.84 | (0.57–1.23) | 0.36 | |||
| ≥4th rank, interval >2 yrs | 10 775 | (20.4) | 1.22 | (0.84–1.78) | 0.29 | |||
| ≥4th rank, interval ≤2 yrs | 1922 | (3.6) | 1.51 | (0.94–2.42) | 0.09 | |||
| Desire for pregnancy | ||||||||
| Wanted then | 43 175 | (81.6) | 1.00 | |||||
| Wanted later | 5257 | (9.9) | 1.06 | (0.62–1.80) | 0.83 | |||
| Wanted no more | 4384 | (8.3) | 0.94 | (0.57–1.56) | 0.81 | |||
| Delivery complication | ||||||||
| No | 34 467 | (65.1) | 1.00 | 1.00 | ||||
| Yes | 18 080 | (34.2) | 1.81 | (1.36–2.41) | <0.001 | 1.82 | (1.34–2.46) | <0.001 |
| Birth size and timing of delivery | ||||||||
| Average-sized + born on time | 26 507 | (50.1) | 1.00 | 1.00 | ||||
| Average-sized + early born | 435 | (0.8) | 5.70 | (2.63–12.35) | <0.001 | 5.78 | (2.62–12.77) | <0.001 |
| Smaller than average + born on time | 6568 | (12.4) | 2.50 | (1.76–3.56) | <0.001 | 2.49 | (1.74–3.56) | <0.001 |
| Smaller than average + early born | 486 | (0.9) | 16.56 | (10.83–25.32) | <0.001 | 14.85 | (9.50–23.21) | <0.001 |
| Larger than average | 17 475 | (33.0) | 0.99 | (0.67–1.46) | 0.96 | 0.92 | (0.62–1.37) | 0.69 |
| Perinatal health care services | ||||||||
| Use of antenatal care service | ||||||||
| No | 5116 | (9.7) | 1.00 | 1.00 | ||||
| Yes | 47 480 | (89.7) | 0.45 | (0.31–0.64) | <0.001 | 0.49 | (0.33–0.73) | <0.001 |
| Mode of delivery | ||||||||
| Non-caesarean section | 50 383 | (95.2) | 1.00 | |||||
| Caesarean section | 2458 | (4.6) | 1.46 | (0.83–2.56) | 0.19 | |||
| Type of delivery attendants and place of delivery | ||||||||
| Untrained attendants at home | 21 993 | (41.6) | 1.00 | 1.00 | ||||
| Trained attendants at home | 13 886 | (26.2) | 0.91 | (0.63–1.30) | 0.60 | 1.24 | (0.82–1.87) | 0.31 |
| Public hospital | 3319 | (6.3) | 2.40 | (1.52–3.79) | <0.001 | 2.65 | (1.48–4.75) | <0.01 |
| Private hospital | 3164 | (6.0) | 0.61 | (0.33–1.13) | 0.11 | 0.83 | (0.41–1.68) | 0.60 |
| Other public birthing centres5 | 1553 | (2.9) | 0.45 | (0.21–0.99) | 0.05 | 0.65 | (0.29–1.46) | 0.30 |
| Other private birthing centres6 | 8820 | (16.7) | 0.68 | (0.41–1.14) | 0.14 | 1.18 | (0.67–2.08) | 0.57 |
Note:
13200 missing cases were excluded from the analysis.
295% CI = 95% confidence interval.
3P = P value.
4(Ref) = reference group.
5Includes health centres, sub-health centres, village maternity posts, village health posts.
6Includes private clinics and the private practices of family physicians, nurses or village midwives.
All values are weighted for the sampling probability.
The analysis found a significant interaction (P = 0.03) between the type of delivery attendant, place of delivery and the type of residence (urban/rural). The results of the urban/rural stratified analyses are presented in Figures 3 and 4. In urban areas (Figure 3) home deliveries assisted by trained delivery attendants, compared with untrained attendants, tended to be protective against early neonatal deaths, although the difference was not significant, whereas infants delivered in other public birthing centres, including health centres and sub-health centres, were significantly protected against early neonatal deaths. We also found a significant interaction between the occurrence of delivery complications and the type of delivery attendants and place of delivery amongst mothers in urban areas (P < 0.001). Only a borderline significant protective effect was associated with trained delivery attendants for home deliveries amongst mothers in urban areas who had delivery complications (Figure 5). However, the risk of early neonatal deaths reduced significantly for institutional deliveries, except for deliveries in public hospitals which showed no significant reduction (HR = 0.91, 95% CI: 0.38–2.18, P = 0.84). Amongst mothers from urban areas who did not have any delivery complications, we found no significant difference between home deliveries assisted by untrained delivery attendants and home deliveries attended by trained attendants or institutional deliveries (data not shown).
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: urban residents, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: urban residents, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: rural residents, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: rural residents, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives.
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: urban women reporting obstetric complications at delivery, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives. Urban women with delivery complications delivered in ‘other public birthing centres’ were omitted because the sample was too small to interpret.
The association between type of delivery attendant, place of delivery and early neonatal mortality in Indonesia: urban women reporting obstetric complications at delivery, IDHS 1994–2007. Notes: (Ref) = reference group. All values are weighted for the sampling probability. This model was adjusted for year of survey, region, maternal education, parental occupation, maternal age at childbirth, child sex, delivery complications, birth size and timing of delivery and use of antenatal care service. Other public birthing centres include health centres, sub-health centres, village maternity posts, village health posts. Other private birthing centres include private clinics and the private practices of family physicians, nurses or village midwives. Urban women with delivery complications delivered in ‘other public birthing centres’ were omitted because the sample was too small to interpret.
In rural areas, a significantly higher risk of early neonatal death was found amongst home deliveries assisted by trained attendants compared with untrained attendants (HR = 1.62, 95% CI: 1.01–2.57, P = 0.04) (Figure 4). The risk also increased significantly if mothers gave birth in public hospitals (HR = 3.24, 95% CI: 1.59–6.62, P < 0.01). Although not significant, the results showed that deliveries at other types of health facilities were associated with an increased risk of early neonatal death. No significant interaction was observed between the type of delivery attendants and place of delivery and the reported delivery complications for infants from rural areas (P = 0.66). Figures 3 and 4 should be interpreted cautiously since they have different reference groups for which the risk of early neonatal deaths may vary, due to the differences between urban and rural environments and health services.
The protective role of antenatal care services was shown amongst infants from both urban and rural areas. The risk of early neonatal death amongst infants whose mothers attended antenatal care services reduced by 68% (HR = 0.32, 95% CI: 0.14–0.71, P = 0.01) and 48% (HR = 0.52, 95% CI: 0.34–0.80, P < 0.01) in urban and rural areas, respectively, compared with infants whose mothers did not attend any antenatal services.
Discussion
Main findings
Our results failed to show protection against early neonatal deaths from having trained delivery attendants for home deliveries in both urban and rural areas. Only in urban areas, and where the mothers had delivery complications, was there borderline significant protection for home deliveries assisted by trained delivery attendants. In contrast, the risk of early neonatal death in rural areas was significantly higher for home deliveries assisted by trained attendants compared with untrained delivery attendants.
The protective effect of having an institutional delivery was only found in urban areas, particularly for infants whose mothers had delivery complications. In rural areas, a significant increased risk for early neonatal death was associated with deliveries in public hospitals, while no significant difference in the risk was shown between deliveries in other type of health care facilities and home deliveries assisted by untrained delivery attendants. Our results demonstrate the importance of antenatal care services, which significantly protected infants against early neonatal deaths irrespective of the differences between urban and rural settings.
Consistent with previous reports, our study also showed that the risk of early neonatal death was significantly higher amongst male infants, smaller than average-size infants and/or infants reported to be born early, and infants whose parents were both employed (Green 1992; Alonso et al. 2006; Ngoc et al. 2006; Arokiasamy and Gautam 2007). A protective effect of higher maternal education on newborn survival was also confirmed in this study.
Significance of the findings
Although efforts to improve maternal and newborn survival have been made by the Indonesian government over the last 20 years, Indonesia still has relatively high neonatal mortality rates compared with other countries in Southeast Asia, including Malaysia and Thailand (WHO 2006). Our findings indicate that improvement in the quality of health professionals and facilities remains a priority in both urban and rural areas. However, public health policy should take into account the disparities between urban and rural settings. Limited access and availability of emergency obstetric care, especially in rural areas, might prevent women from receiving timely medical attention. Moreover, monitoring programmes to improve the performance of village midwives, who are usually the solo health care provider in the village, are important to ensure the high quality of services. Adequate pre-service and in-service midwifery training for health professionals, including village midwives, will be beneficial. Our findings highlight the need for effective and efficient patient referral system in the event of obstetric emergencies, particularly in remote and rural areas where the availability and accessibility of health services remains a major challenge (Geefhuysen 1999; Titaley et al. 2010a; Titaley et al. 2010c).
Strengths and limitations of this study
In this analysis, we used data from four nationwide surveys covering mothers’ most recent births from 1989 to 2007. Pooling these surveys provided an adequate sample size to estimate the effects of different types of birth attendants and places of delivery on risk of early neonatal death with adequate precision, after adjusting for a wide range of variables from the community, household and individual levels. Compared with a previous study, which also used the IDHS data (Hatt et al. 2009) and examined similar research questions, our study has included the most recent survey (IDHS 2007) and has also taken into account the role of major covariates, including urban/rural differences and the presence of obstetric complication at delivery, along with their interactions with utilization of delivery care services.
The following limitations of our study should be noted when interpreting the findings. Information obtained from the IDHS was all self-reported and not validated. It relies on the mothers’ recall about their pregnancy, delivery and post-delivery conditions and might be subject to recall bias. However, by restricting our study population to the most recent delivery within 5 years prior to each survey, this bias was minimized. Variables included in our data also relied on the availability of the information in the IDHS datasets. Several potential predictors, such as maternal anthropometric measurements or traditional cultural practices, which were not available in IDHS datasets, might have affected newborn survival but could not be examined in this study. In our analysis, a birth weight variable was replaced by mothers’ subjective assessment on infants’ size at birth which relies on mothers’ recollection and knowledge about the size of their infant.
Delivery care services and early neonatal mortality
Our study did not find a protective effect of having trained delivery attendants for home deliveries in Indonesia, particularly in rural areas where the risk of early neonatal death was significantly higher for mothers delivering at home assisted by trained attendants compared with untrained delivery attendants. This finding is consistent with the results from a trial conducted in rural areas of six developing countries, which showed no significant improvement in newborn survival amongst deliveries assisted by trained delivery attendants (Carlo et al. 2010). An earlier study analysing pooled IDHS data also found no association between uptake of professional care at home and early neonatal deaths, although no adjustment was made for the occurrence of delivery complications (Hatt et al. 2009). Similarly, an Indonesian study from two districts in West Java Province reported high maternal mortality ratios even amongst women receiving professional care at deliveries (Ronsmans et al. 2009).
The increased risk of early neonatal deaths amongst home deliveries assisted by trained delivery attendants in rural areas might be due to their lack of clinical experience and skills. The main providers of professional care for home deliveries in rural areas in Indonesia are village midwives (Geefhuysen 1999; Ronsmans et al. 2001; Makowiecka et al. 2008; D'Ambruoso et al. 2009; Titaley et al. 2010b). This finding indicates a continued need to improve the skills and performance of village midwives considering the enormous investment made by the government in deploying a village midwife in almost every village in Indonesia. Efforts to enhance the quality of the village midwives started as early as 1997 through revisions to the training programmes for village midwives (Geefhuysen 1999). These changes aimed to provide more opportunities for hands-on experience to increase their practical skills to manage deliveries and deal with complications (Geefhuysen 1999). Nevertheless, our results suggest that continuous monitoring and evaluation programmes are still required to assess the effectiveness of village midwives in reducing mortality, particularly for home deliveries in rural areas.
Studies have indicated that low midwife density aggravated by poor infrastructure in rural areas might hinder the ability of village midwives or other health professionals to provide health services (Thaddeus and Maine 1994; Makowiecka et al. 2008; Titaley et al. 2010a). Furthermore, the influence of traditional values means traditional birth attendants are often preferred over trained birth attendants, especially in rural and remote settings (Geefhuysen 1999; Titaley et al. 2010a). This ambivalence about the use of trained birth attendants can lead to delays in providing adequate care or referral to health facilities, given that professional care is often only sought when an emergency occurs (Achadi et al. 2007; Makowiecka et al. 2008; Ronsmans et al. 2009). Strengthening the partnership programme (United Nations Children's Fund 2008) between village midwives, traditional birth attendants and community health workers might help to increase the coverage of deliveries by trained attendants and reduce delays in seeking trained health care in emergencies.
Our findings also demonstrate the advantage of institutional deliveries when delivery complications occur. Professional care available at health facilities, along with access to life-saving equipment, is beneficial for both mothers and newborns. Obstetric complications can be better managed in health care facilities and potential delays can be minimized. Moreover, the availability of a team of providers in health care facilities offers another advantage (Koblinsky et al. 2006; Ronsmans et al. 2009). Unfortunately, the protective effect of institutional deliveries on early neonatal deaths in our analysis was only found in urban areas, which might reflect the poor quality and limited access to health services in rural areas. Interventions directed at improving the availability, accessibility and quality of health services, including emergency obstetric and newborn care in rural areas, remain a priority in Indonesia.
The increased risk of early neonatal death amongst infants delivered in public hospitals might also be due to selection bias as a result of the referral of more severe cases to public hospitals (Arokiasamy and Gautam 2007; Hatt et al. 2007). However, of concern was that increased risk was only observed in rural areas. In addition to poor quality of health services in rural settings, families’ reluctance to act on advice to seek care in a public hospital, or delays in referral by village-level providers, might aggravate any delivery complications (Thaddeus and Maine 1994; Arokiasamy and Gautam 2007). Therefore, public health interventions to increase community awareness about the importance of safe delivery care services are essential, along with efforts to strengthen the referral system so that women with obstetric complications at the early stage of labour can receive treatment promptly (Paul and Rumsey 2002). Community-based interventions to avoid delays in reaching health care facilities should be promoted, such as organizing community transportation for pregnant women with a village ambulance as part of the ‘Village Alert’ programme (Learner 2010). The involvement of local community members, for example through women's group meetings (Manandhar et al. 2004) or traditional birth attendants (Jokhio et al. 2005), in encouraging women to use maternal and child health services would also be important supportive interventions.
Other factors significantly associated with early neonatal mortality
The role of maternal education on child health and survival, including neonatal health and survival, has long been established (Caldwell 1979; Ware 1984; Mellington and Cameron 1999; Basu and Stephenson 2005). Maternal education increases mothers’ knowledge about child health and health care services (Caldwell 1979) and improves mothers’ health-care-seeking behaviours (Ware 1984; Mellington and Cameron 1999; Basu and Stephenson 2005), which benefit the health of both the mother and newborn. In this study the association between economic status and early neonatal mortality is also reflected by the significant effect of parental occupation status. A low household wealth index was observed where both the mother and father were employed, and amongst these households we observed an increased risk of early neonatal death. This increased risk of neonatal death might be due to lack of time to attend antenatal care services because of the mother's employment (Simkhada et al. 2008), which in turn might adversely affect the survival of the newborns.
The importance of antenatal care services for preventing early neonatal deaths was confirmed by our study. In addition to health counselling (Chen et al. 2007), early detection of pregnancy complications (Chen et al. 2007), the provision of iron/folic acid supplements (Titaley et al. 2010d) and tetanus toxoid vaccinations (Titaley et al. 2010d) have been reported to protect newborns against early neonatal deaths. Mothers and newborns would benefit from outreach health care services, such as Pusat Pelayanan Terpadu or Posyandu (WHO Regional Office for South-East Asia 2008), that reach community members in rural and remote areas and promote the uptake of antenatal services. These are the communities who are often in the greatest need but are more likely to underutilize maternal and child health services.
Although skilled birth attendance has been claimed as a key intervention to improve neonatal survival (Darmstadt et al. 2005), our results suggest that having trained attendants for home deliveries might not be a critical intervention to reduce neonatal mortality in Indonesia. This was even observed amongst women with delivery complications. Institutional deliveries, or a substantial improvement in skills of home birth attendants and a better referral system, might be required to reduce the risk of neonatal deaths in Indonesia. In contrast, clear evidence of the protective effect of antenatal care services for neonatal deaths was found in this study. This finding supports our previous research using a similar IDHS dataset that demonstrated the protective role of antenatal services, particularly iron and folic acid supplementation, in reducing the risk of neonatal deaths in Indonesia (Titaley et al. 2010d).
Conclusions
The protective role of trained attendants for home deliveries against early neonatal deaths was not confirmed in our study. However, a reduced risk was associated with institutional deliveries particularly among women in urban areas with delivery complications. These findings indicate the need for continuous improvement in the quality of village midwives, who are the main care provider for home deliveries. Efforts to strengthen the referral system and to increase the quality of delivery and newborn care services in health facilities, especially in public hospitals in rural areas, are important.
Health promotion programmes that deliver community-based education about the importance of maternal and newborn health services, including safe-delivery care by using trained delivery attendants or institutional deliveries, remain a priority. Community-based interventions to prevent delays in getting medical attention for complicated deliveries will be beneficial, along with strengthening the partnership between health professionals, traditional birth attendants and community health workers.
Supplementary Data
Supplementary Data are available at Health Policy and Planning Online.
Funding
We are grateful to the Australian Agency for International Development (AusAID) for funding CRT's PhD scholarship in International Public Health at the Sydney School of Public Health, University of Sydney, Australia. This analysis is part of CRT's thesis. CLR is supported by a NHMRC Research Fellowship.
Conflict of interest
None declared.
