Data for this analysis come from a Health Facilities Survey (HFS), which was conducted April-June 2018 in Java, Indonesia. The HFS was conducted in a face-to-face interview with staff members knowledgeable about PAC provision in their facility, usually the medical director or head midwife. In eight sampled hospitals, multiple wards treated PAC patients. In these cases, interviewers conducted separate surveys within each ward. In seven hospitals, two wards were interviewed, and three wards were interviewed in one hospital. The Guttmacher Institute’s and University of Indonesia Faculty of Public Health’s respective Institutional Review Boards granted ethical approval for this study.
We aimed to include all facilities with the potential to offer PAC in our sampling frame. For hospitals, this was defined as having either an obstetric care ward or an operating theater. We included all public hospitals in the sampling frame, which are classified into four types denoted by the letters A through D. Type A hospitals are the largest and most comprehensive facilities, whereas Type D hospitals are the smallest, with no more than four specialty care wards. We also included private, often religiously-affiliated maternal and neonatal specialty hospitals called Rumah Sakit Ibu dan Anak (RSIA, Mother and Child Hospitals), and Rumah Sakit Anak dan Bunda (RSAB, Child and Mother Hospitals). Finally, health centers with PONED (basic obstetric and neonatal emergency service) registration were included in the sampling frame. PONED health centers, staffed mainly by midwives and GPs, were established in 2008 to combat Indonesia’s high level of maternal and infant mortality by expanding access to basic emergency obstetric and neonatal care.
We extracted information on public and private hospitals from the Ministry of Health Hospital Management Information System website in June 2017, and on PONED health centers from the Ministry of Health 2016 report, a national census of all health facilities. After adjusting for closures and misclassification, the sampling frame consisted of 2,239 health facilities (Table 1). We used stratified random sampling to obtain a sample representative of Java and each of its six provinces. Within each province, we selected 100% of Type A hospitals, 40% each of Type B, C, D, and RSIA/RSAB hospitals, and 20% of PONED health centers. This resulted in a sample of 717 facilities (32%) (Table 1).
Table 1. Java Health Facilities Survey 2018 sample
Type of facility
|
Universe with potential to offer PAC (corrected)
|
HFS sampling fraction (based on corrected universe)
|
Number selected
|
Number of completed interviews
|
Response rate
|
Hospital Type A
|
13
|
100%
|
13
|
11
|
85%
|
Hospital Type B
|
205
|
40%
|
84
|
73
|
87%
|
Hospital Type C
|
446
|
40%
|
192
|
171
|
89%
|
Hospital Type D
|
356
|
40%
|
134
|
123
|
92%
|
RSIA/RSAB (private maternity hospital)
|
233
|
40%
|
97
|
82
|
85%
|
PONED (BEmOC-registered health center)
|
986
|
20%
|
197
|
197
|
100%
|
Total
|
2,239
|
32%
|
717
|
657
|
92%
|
Within each stratum (facility type and province), we first calculated a base weight equal to the inverse probability of selection, and a non-response weight equal to the inverse probability of participation. We weighted all facilities by the composite weight equal to the product of the base weight and the non-response weight. A total of 657 facilities (92% response rate) completed the HFS. For the purposes of this analysis, we collapsed hospitals into three groups: Type A/B, Type C/D, and RSIA/RSAB, as the hospital types within each grouping offer similar levels of obstetric service provision.
We created a composite indicator for each facility summarizing its capacity to treat the most common complications from miscarriages and unsafe abortions: infection, hemorrhage, and internal injury. This indicator classifies capacity to treat PAC patients into two categories, based on the essential services framework:
- Basic PAC capacity is defined as the ability to offer round-the-clock access to a minimum level of PAC service. To meet this standard, a facility must be open 24/7 with at least three appropriate providers on staff, and have a means of contact with and transport to a higher-level facility for referral. The facility must offer services necessary to prevent and treat infection and manage early-gestation pregnancy loss: parenteral antibiotics, IV fluid, uterotonic oxytocics, uterine evacuation for early-gestation pregnancies, and provision of short-acting contraceptives. Facilities are considered to have full basic PAC capacity if they meet all of these criteria, but do not provide the full set of comprehensive service indicators, described next.
- Comprehensive PAC capacity is defined as service provision that can accommodate both basic PAC treatment (defined above) as well as complete care for more advanced interventions: surgery capability (laparotomy), stocks of blood for transfusion, second-trimester uterine evacuation, and provision of long-acting contraceptives (IUD or implant). Primary-level health facilities are typically excluded from this measure, since only hospitals are expected to have the potential to provide this more advanced treatment. Hospitals with all services listed in Table 1 are classified as having comprehensive PAC capacity.
We adjusted these definitions to account for Indonesia’s unique regulations and practices. First, in most other settings where the essential services framework has been applied, midwives and GPs typically count towards the three appropriate providers needed to meet the standard for full PAC capacity. In Indonesia, midwives and GPs have neither authorization nor, generally speaking, training to provide the full set of services for basic PAC. Therefore, for this analysis, only Ob/Gyns count towards the three staff minimum. Secondly, all health centers and hospitals in Indonesia are required to have a means of communication and transport, so we excluded questions about this from the HFS and assume that all facilities in our sample have referral capacity. Thirdly, because Indonesia’s national health insurance does not reimburse hospitals for provision of short-acting contraceptive methods, these methods are less routinely offered in hospital settings. Including this service as an essential indicator caused many hospitals to be classified as lacking the full set of services needed for basic PAC provision, even if they otherwise met the criteria for comprehensive PAC capacity. To account for this, we removed short-term contraceptive method provision as a criterion. Finally, Indonesia’s blood supply chain relies primarily on local International Red Cross facilities, which coordinate with hospitals to ensure a sufficient supply of blood products; hospitals are generally not expected to routinely keep blood products onsite (15). For this reason, we did not consider stocks of blood products a requirement for comprehensive PAC capacity. Table 2 summarizes these definitions.
Table 2. Essential services for basic and comprehensive PAC
Indicator
|
Basic
|
Comprehensive
|
Open 24/7
|
X
|
X
|
>= 3 Ob/Gyn doctors on staff
|
X
|
X
|
IV fluids
|
X
|
X
|
IV antibiotics
|
X
|
X
|
Uterotonic oxytocics
|
X
|
X
|
First trimester uterine evacuation
|
X
|
X
|
Second trimester uterine evacuation
|
|
X
|
Surgical capacity
|
|
X
|
Long-acting contraceptive method (IUD or implant)
|
|
X
|
Respondents reported whether the facility is open 24/7 and the number and type of providers on staff. To determine whether facilities had IV fluids, parenteral antibiotics, uterotonic drugs, short-acting contraceptive methods (pill or injectable), long-acting contraceptive methods (IUD or implant) and blood for transfusion, respondents reported whether the facility offered each service and whether the facility had experienced stock-outs of each commodity at any point in the past three months. Facilities that offered the service and experienced no stock-outs in the past three months were coded as having that equipment or drug. Respondents also reported whether the facility had the ability to perform removal of retained products of conception in both the first and the second trimester, separately.
For the eight hospitals in which multiple wards were interviewed, we coded the entire hospital as having a given service or piece of equipment if any one of its wards did, under the assumption that hospital departments can share supplies or transfer patients to a better-equipped unit when necessary. The number of providers for these hospitals was calculated as the sum of providers reported in all surveyed wards.
Using the Indonesia-adjusted essential services definition and accounting for all of the above indicators, facilities were then coded into one of three categories: incomplete PAC capacity (lacking one or more of the basic PAC criteria), basic PAC capacity, or comprehensive PAC capacity. We then calculated the weighted proportion of facilities in each PAC capacity category, overall and by facility type.
To better understand which components contribute to facilities lacking the full set of PAC capability indicators, we calculated the proportion of facilities reporting each individual service or equipment. Although it was excluded from the list of indicators for this analysis, we also calculated the proportion of facilities that offer a short-term contraceptive method, since this service is appropriate for many PAC patients and has been used in most other published analyses of health system PAC capacity.
We also constructed a hypothetical indicator to summarize PAC capacity under three conditions that would account for potential changes in healthcare policies. First, we investigate what PAC capacity would be if first trimester uterine evacuation were offered in all PONED health centers, which currently are not authorized to provide this service. The second scenario portrays what would happen if authorization to perform PAC were expanded to all GPs. The third scenario investigates the change in capacity if all midwives (but not GPs) could perform PAC. We assessed the impact of each change in authorized provider type individually and, finally, in combination with each other.
Finally, we also calculated the proportion of PAC patients treated with each of four methods. Respondents at each facility that treated PAC patients estimated the percentage of PAC patients treated with different methods, responses for which were grouped into four categories: dilation and curettage (D&C), manual or electric vacuum aspiration (MVA/EVA), misoprostol, and surgery/other methods. At each of the 438 facilities that provided PAC patient caseloads and estimates of the proportion treated with each procedure, we applied the distribution of PAC methods to the number of PAC patients treated at the facility in the year 2018, resulting in an estimate of the number of PAC patients receiving each type of treatment. In the eight facilities where multiple departments were interviewed, we did this in each department and then summed the number of patients treated with each method for a facility-level total. We applied facility weights to the number of patients treated with each method and divided by the total number of PAC patients to calculate the proportion of all PAC patients treated with each method in the year 2018. All analyses were conducted in Stata 15.0.