DOI: 10.1111/j.1471-0528.2008.01913.x www.blackwellpublishing.

com/bjog

Epidemiology

Estimation of population-based incidence of pregnancy-related illness and mortality (PRIAM) in two districts in West Java, Indonesia
C Ronsmans,a S Scott,a A Adisasmita,b P Deviany,b F Nandiatyc
a Infectious Disease Epidemiology Unit, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, UK b Faculty of Public Health, University of Indonesia, West Java, Indonesia c Payakumbuh District Health Ofce, West Sumatra, Indonesia Correspondence: Prof C Ronsmans, Infectious Disease Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London, WC1E 7HT, UK. Email [email protected]

Accepted 6 July 2008.

Objective We introduce a new and untested approach for the

measurement of life-threatening maternal morbidity in populations where not all women give birth in a health facility. By dening complications at the very extreme end of the severity spectrum, we postulate that its count in hospitals can be used to represent the incidence in the general population.
Design We counted all cases of life-threatening obstetric

(dened using the concepts of near miss and met need for life-saving surgery) and maternal mortality.
Results The incidence of maternal mortality and life-threatening complications at the population level was 421 and 1416 per 100 000 births, respectively, resulting in an overall ratio of PRIAM of 1837 per 100 000. The overall incidence of PRIAM was much lower in rural than in urban areas (1529 and 2880 per 100 000, respectively, P < 0.001), and it was lowest in rural Serang (1304 per 100 000). Conclusions The approach tested in this studyrelying on

morbidity in hospitals and all maternal deaths in the population. Using these data, we describe the incidence of life-threatening morbidity in the total population, examine its variation across geographical areas and investigate its relationship with maternal mortality.
Setting Serang and Pandeglang district in West Java, Indonesia. Population or sample All women residing in the two districts. Methods Cross-sectional study of maternal morbidity and

mortality.
Main outcome measures Pregnancy-related illness and mortality

conditions that are absolutely life-threatening such that their count in hospitals can be used to represent the incidence in the general populationis promising but needs further testing in populations with varied disease epidemiology and access to care. Continued investments in hospital-based audits of life-threatening morbidity may ultimately improve the quality and reliability of information on obstetric complications and facilitate the development of rigorous and standard criteria for the definition of life-threatening morbidity.
Keywords Epidemiology, Indonesia, life-threatening maternal

(PRIAM), consisting of life-threatening maternal morbidity

morbidity, maternal mortality, near miss.

Please cite this paper as: Ronsmans C, Scott S, Adisasmita A, Deviany P, Nandiaty F. Estimation of population-based incidence of pregnancy-related illness and mortality (PRIAM) in two districts in West Java, Indonesia. BJOG 2009;116:8290.

Introduction
Falling numbers of maternal deaths in developed countries, difculties in accurately measuring maternal mortality and a growing concern for understanding the factors that contribute to maternal death have stimulated an interest in investigating the levels and causes of life-threatening maternal morbidity.14 Life-threatening obstetric events are thought to be more common than maternal deaths, and populationbased estimates of the incidence of life-threatening maternal

morbidity are a potentially powerful indicator for the evaluation of the health impact of maternal health interventions.14 The larger numbers of life-threatening morbidity may also allow a further disaggregation of data across geographical or other relevant subgroups than is possible for maternal deaths and help identify areas in greatest need of maternal health interventions.5 Measuring the incidence of life-threatening obstetric morbidity remains difcult, however, particularly in populations where not all women give birth in a health facility. Research

82

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

Pregnancy-related illness and mortality (PRIAM)

has shown that the reliability of estimates of morbidity based on womens recall of their birth experience is poor,69 and observation by a trained provider is necessary to measure maternal morbidity reliably. In a systematic review of the incidence of severe acute maternal morbidity,1 only one study was found reporting the incidence of severe morbidity at the population level in developing countries. In this West African study,10 labour or delivery was observed by health professionals in nearly two-thirds of urban births, but the investigators relied on the traditional birth attendants or the womans report of morbidity in the remaining third of all births. In a study in India, Bang et al. trained village health workers to identify morbidity during labour or in the postpartum in women giving birth at home, but the severity of the reported conditions was difcult to ascertain, even among those who were deemed to require emergency obstetric care.11 Given these constraints, estimates of the overall burden of maternal ill-health are incomplete and are largely based on literature from industrialised countries and expert consensus.12 In this paper, we introduce a new and untested approach for the measurement of life-threatening maternal morbidity. By dening complications at the very extreme end of the severity spectrum, we postulate that their count in hospitals can be used to represent the incidence in the general population. We dene complications that are absolutely life threatening such that women who experience these problems are unlikely to survive if they do not receive care in a hospital. Using this approach, population estimates of morbidity could be obtained from hospital data, assuming that women with such complications who do not reach the hospital will die. The aim of this paper was to introduce this new denition of life-threatening complications and to test this approach in two districts in Java, Indonesia. This study was conducted as part of an international research initiative to assess the effectiveness of safe motherhood strategies, called Immpact.13 As part of this research, a comprehensive evaluation was undertaken in two districts in Java, Indonesia, including the measurement of maternal mortality at the population level and a count of all pregnancy-related admissions in the four hospitals serving the two districts.14,15 In this paper, we dene life-threatening obstetric morbidity using the concepts of near miss and met need for life-saving surgery,1,2,5,16 describe the incidence of life-threatening morbidity in the two districts, examine the variation across geographical areas and investigate its relationship with maternal mortality.

West and has a population of 1.1 million.17 Most women (83%) in the two districts give birth at home.14

Denitions of life-threatening morbidity, maternal deaths and pregnancy-related illness and mortality
Our aim was to dene an exhaustive list of complications during pregnancy, labour or delivery that are absolutely life threatening, that are thought to have a high probability of dying if the woman fails to obtain hospital care. We base our denitions on two distinct concepts of severe morbidity reported in the literature. The rst concept is that of near miss, referring to any pregnant or recently delivered woman in whom immediate survival is threatened and who survives because of chance or because of the good care received.2,18,19 Women qualify as near miss when they are acutely ill and their condition is so severe that they are unlikely to survive without hospital care. The second concept is that of met need for life-saving surgery, referring to conditions that are thought to have a high probability of dying if the woman fails to obtain a major surgical intervention.5,16 In our study, all women who survived and who experienced a near miss or received life-saving obstetric surgery for an absolute maternal indication before being a near miss were categorised as having a life-threatening complication. A near miss was dened adapting the criteria proposed by Mantel et al.19 and Penney and colleagues20 This denition was based on criteria for organ dysfunction, including cardiac, pulmonary, vascular, immunological, respiratory, renal, liver, coagulation and cerebral dysfunction (Table 1). The organ dysfunction can occur at any time during early or late pregnancy, during labour or delivery or in the immediate postpartum. Detailed criteria have been reported elsewhere.15 We aimed at dening extreme conditions based on clinical criteria, such as the presence of shock or coma.15 Because this was not always possible, we also relied on management criteria, including blood transfusion, emergency hysterectomy, intensive care admission and need for resuscitation. In this study, women who had received four or more units of blood were considered to be a near miss. All cases of eclampsia, uterine rupture and ectopic pregnancy were also considered to be a near miss. Near miss do not capture all cases of absolute life-threatening maternal morbidity since some women may receive a life-saving intervention for an absolute indication before they become acutely ill. For example, a woman with a complete placenta praevia will have a high probability of dying in the absence of a caesarean section, but she may not become a near miss if the caesarean section is performed in time. We therefore expanded the denition of life-threatening morbidity to include women who received life-saving surgery for absolute maternal indications using the denitions proposed by the Unmet Obstetric Need Network.5,16 Life-saving surgery includes caesarean section, laparotomy, hysterectomy,

Methods
Study population
The study was conducted in Serang and Pandeglang districts in West Java, Indonesia. Serang district, with a population of 1.8 million, is situated 100 km west of Jakarta, the national capital. Pandeglang district borders Serang to the South and

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

83

Ronsmans et al.

Table 1. Criteria for inclusion of near miss cases (modied from Mantel et al.)19 Organ dysfunction Cardiac dysfunction Pulmonary embolism Vascular dysfunction

nancy and postpartum, regardless of the underlying cause. All cases of life-threatening morbidity and death at the population level were combined into an indicator called pregnancy-related illness and mortality (PRIAM).

Immunological dysfunction Respiratory dysfunction

Pulmonary oedema, cardiac arrest, cardiac failure Hypovolaemia requiring four or more units of blood, blood loss with hypovolaemic shock (systolic blood pressure ,90 mmHg or undetectable pulse), infusion and/or transfusion of 1 l in 2 hours and free-ow infusion* Septic shock

Data collection
We used four sources of data: hospital case notes for the identication of life-threatening morbidity, communitybased data for the ascertainment of maternal deaths,21 National Statistical Ofce data for population size in the two districts and a population-based survey reporting births between April 2004 and March 2006.14 Data on all pregnancy-related admissions between December 2005 and May 2006 were collected prospectively from four hospitals: one large tertiary hospital, two smaller private hospitals (Kencana and Budi Asih hospitals) in Serang and the district hospital in Pandeglang. These four hospitals cover almost all hospital admissions related to pregnancy in the two districts. Eight medical doctors screened all the registers from all hospital wards on a daily basis to ascertain possible cases of life-threatening morbidity. They then reviewed the case notes for anyone reported to have a complication on the register and for all major surgical interventions. Special efforts were made to identify all maternal deaths in the hospitals.22 All maternal deaths in the two districts occurring between January 2004 and December 2005 were identied using an informant-based approach. Detailed methods have been reported elsewhere;21 but in brief, in all 708 villages of the two districts, village informantshealth post volunteers and unpaid officialscollected information about deaths to women of reproductive age during the past 2 years in their community. They also ascertained whether the deaths occurred from onset of pregnancy to 42 days postpregnancy or delivery. The total number of deaths was estimated using a capturerecapture technique based on linked data from two different informant networks.

Intubation or ventilation for reasons other than general anaesthesia and oxygen saturation on pulse oximetry ,90% leading to ventilation Renal dysfunction Oliguria ,30 ml per hour or ,400 ml per 24 hours, shock not responsive to intravascular rehydration or diuresis and haemodialysis Liver dysfunction Jaundice in pre-eclampsia and abnormal liver function tests Coagulation Acute thrombocytopenia, prolonged dysfunction bleeding time, abnormal activated partial thromboplastine time (APTT) or prothrombine time (PT) and coagulopathy Cerebral dysfunction Coma, cerebral oedema, seizures other than eclampsia Management-based criteria Intensive care admission, emergency hysterectomy, needs resuscitation and anaesthetic accident Clinical diagnosis Eclampsia, uterine rupture and ectopic pregnancy *Free-ow infusion refers to a massive infusion of uids in case of shock.

symphysiotomy, craniotomy and internal version (even though internal versions are not surgical interventions, they are generally retained in the numerator). Maternal indications are selected only if they are absolute, that is for conditions that are thought to have a high probability of dying if the woman fails to obtain a major surgical intervention. Absolute maternal indications included in this study are severe antepartum haemorrhage due to placenta praevia or placental abruption, major cephalopelvic disproportion, transverse lie and brow presentation. Deaths in women of reproductive age were dened as maternal when they occurred from onset of pregnancy to 42 days postpregnancy or delivery. This time-of-death denition is consistent with our denition of life-threatening morbidity, which includes all life-threatening conditions during preg-

Data analysis
The target population was dened by area of residence, separating villages into Serang urban, Serang rural, Pandeglang urban, Pandeglang rural and Pandeglang rural remote (dened as >33 km from the nearest hospital for villages in Pandeglang only). All pregnancy-related hospital admissions were matched to their village of residence in the two districts, and women living outside Serang and Pandeglang or with unknown district of residence were excluded. Hospital admissions with known district of residence but missing data on village were assigned an area of residence through multiple imputations using STATA 10. The expected number of births in each village was calculated by multiplying the population size with the crude birth rate estimated from the population survey.14

84

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

Pregnancy-related illness and mortality (PRIAM)

We rst examined the completeness of the hospital admission data by examining the proportion of births in the two districts that took place in hospital comparing two data sources: those obtained from hospital admission data (with Poisson exact condence intervals) and those obtained from the population-based survey.14 Condence intervals for the population-based survey were estimated with allowance for village clustering and stratication. We then described the medical causes of pregnancy-related admissions to hospital, pooling data from all four hospitals. Admissions are described by medical diagnosis, separating near miss, those who received life-saving surgery for absolute maternal indications (but were not near miss) and maternal deaths in hospital. Women with more than one complication were assigned a mutually exclusive diagnosis based on an algorithm identifying the underlying diagnosis. Finally, to examine the incidence of life-threatening complications in the two districts and its relationship with maternal mortality, we report the incidence of life-threatening complications, maternal mortality and PRIAM by area of residence. Morbidity, mortality and PRIAM are expressed per 100 000 births, and condence intervals were calculated assuming a Poisson distribution. Because not all cases of eclampsia or ectopic pregnancy may be absolute (i.e. some women may survive without receiving hospital care), this analysis was repeated excluding cases of eclampsia and ectopic pregnancy for which we had no explicit signs of organ failure from the denition of near miss.

Results
There were 3006 pregnancy-related admissions in the four hospitals in Serang and Pandeglang between December 2005 and May 2006 (1282 in Serang Hospital, 734 in Pandeglang Hospital, 604 in Kencana Hospital and 386 in Budi Asih Hospital). Of these, 11 women were admitted to one hospital and then referred to another, so their rst admission was dropped from further analysis, and 167 were excluded

because the women did not live in Serang or Pandeglang (n = 100) or the district of residence was unknown (n = 67). Of the remaining 2828 admissions, the village of residence was known for 2239 (79.2%) and 589 (20.8%) women were assigned an area of residence because they were known to be from Serang and Pandeglang, but their village of residence was unknown. The majority of those with missing village data were non-life-threatening hospital admissions (91%). Nearly, two-thirds (59.6%) of women admitted to hospital gave birth in the hospital. Using hospital data for the numerator and expected number of births in each village of residence, the proportion of births taking place in hospital was 4.6 (95% CI 4.44.8) compared with 4.2% (95% CI 3.05.8) using the population-based survey. These proportions differed significantly by area of residence but not by data source (Table 2). Table 3 shows the distribution of life-threatening complications and institutional maternal deaths by type of complication. There were 518 cases of life-threatening morbidity: 331 (63.9%) were near miss and another 187 (36.1%) women had received life-saving surgery for an absolute maternal indication before reaching the acute state of near miss. The most common indication for near miss was retained placenta (16.3%), followed by ectopic pregnancy (13.9%), eclampsia (13.3%) and abortion (11.5%). Near miss due to dystocia (7.3%) or obstetric infection (2.1%) were uncommon. The most common indication for life-saving surgery in women without near miss was cephalopelvic disproportion (51.9%) and placenta praevia (29.9%). Overall, dystocia was the most common indication for life-threatening morbidity (31.5%), followed by antepartum (18.3%) and postpartum haemorrhage (15.8%) (Figure 1). Thirty-ve (1.2%) women died in hospital, 3 (8.6%) of which followed major surgery. Nonobstetric medical conditions were the most common cause of maternal death in hospital (42.8%; Table 3 and Figure 1). The proportion of hospital admissions that were life threatening or died by area of residence of the women is shown in Figure 2. In urban Serang, where 13.2% of deliveries occurred

Table 2. Proportion of women who deliver in a hospital by data source and area of residence Area of residence Population survey of women with a recent birth Number of births in survey between April 2004 and March 2006 Serang urban Pandeglang urban Serang rural Pandeglang rural Pandeglang ruralremote Total 334 95 296 108 401 1234 % of women who deliver in hospital (95% CI) 12.3 (8.517.4) 4.2 (1.312.4) 1.7 (0.64.6) 3.7 (1.68.5) 2.0 (0.94.5) 4.2 (3.05.8) Prospective hospital study Expected number of births between December 2005 and May 2006 5953 2412 17 163 5627 5504 36 658 % of women who deliver in hospital (95% CI) 13.2 (12.314.2) 6.5 (5.57.6) 2.7 (2.52.9) 3.0 (2.53.5) 2.0 (1.72.4) 4.6 (4.44.8)

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

85

Ronsmans et al.

Table 3. Causes of life-threatening complications and maternal deaths in four hospitals in Serang and Pandeglang district (December 2005 to May 2006) Complications Life-threatening complications Near miss (A) (%) Life-saving surgery but no near miss (B) (%) All (A1B) (%) Institutional maternal deaths (%)

Early pregnancy loss Abortion Ectopic pregnancy Antepartum haemorrhage Placenta praevia Placental abruption Unspecied Postpartum haemorrhage Uterine atony Retained placenta Tear Unspecied Hypertensive diseases Pre-eclampsia and hypertension Eclampsia Dystocia Uterine rupture Bandls ring Cephalopelvic disproportion Malpresentation Breech/footling Prolonged labour Obstetric infection Other Asthma Heart failure Diabetes Trauma Cerebrovascular accident and stroke Other infectious disease Other Unknown Total

38 (11.5) 46 (13.9) 23 (7.0) 9 (2.7) 3 (0.9) 10 (3.0) 54 (16.3) 7 (2.1) 11 (3.3) 14 (4.2) 44 (13.3) 17 (5.1) 2 (0.6) 5 (1.5) 2 (0.6) 3 (0.9) 7 (2.1) 7 (2.1) 1 (0.3) 8 (2.4) 1 (0.3) 5 (1.5) 14 (4.2) 331

56 (29.9) 4 (2.1) 2 (1.1) 97 (51.9) 28 (15.0) 187

38 (7.3) 46 (8.9) 79 (15.3) 13 (2.5) 3 (0.6) 10 (1.9) 54 (10.4) 7 (1.4) 11 (2.1) 14 (2.7) 44 (8.5) 17 (3.3) 4 (0.8) 102 (19.7) 30 (5.8) 3 (0.6) 7 (1.4) 7 (1.4) 1 (0.2) 8 (1.5) 1 (0.2) 5 (1.5) 14 (2.7) 518

1 (2.9) 1 (2.9) 1 (2.9) 1 (2.9) 1 (2.9) 3 (8.6) 4 (11.4) 5 (14.3) 2 (5.7) 1 (2.9) 2 (5.7) 1 (2.9) 8 (22.9)* 3 (8.6) 1 (2.9) 35

*Including meningitis, tuberculosis and hepatitis.

in a hospital (Table 2), 12.9% of pregnancy-related hospital admissions were associated with life-threatening complications or death. In the other areas, where the proportion of births taking place in hospital were much lower, these proportions were higher (22.3, 24.3, 29.3 and 33.3% for urban Pandeglang, rural Serang, rural Pandeglang and remote Pandeglang, respectively, Figure 2). The incidence of life-threatening complications at the population level was 1416 per 100 000 births. The incidence was much higher in urban (2654 per 100 000) than in rural areas (1050 per 100 000, P < 0.001; Table 4), and the lowest ratio

was seen in rural Serang (880 per 100 000). We identied 474 maternal deaths between January 2004 and December 2005 in the two districts, resulting in a maternal mortality ratio of 421 per 100 000 live births (95% CI 376483).21 The mortality patterns were the reverse of those seen for life-threatening complications; urban ratios (226 per 100 000) being much lower than rural ratios (479 per 100 000, P < 0.001), but mortality did not compensate for the variation in the incidence of lifethreatening complications (Figure 3). The overall incidence of PRIAM was much lower in rural than in urban areas (1529 and 2880 per 100 000, respectively, P < 0.001), and it was lowest in

86

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

Pregnancy-related illness and mortality (PRIAM)

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Near miss (A) Life-saving All lifeMaternal deaths surgery without threatening in hospital near miss (B) complic ations (A+B) Obstetric infection Hypertensive diseases Antepartum haemorrhage Abortion

% of pregnacy-related hospital admissions

100%

80% Other pregnancyrelated hospital admissions Death 40% Life-threatening complications

60%

20%

0%

ba

ba

ra

ra

ur

ur

ru

ru

an

an

an

an

Se r

Se r

gl

gl

de

de

Pa n

Pa n

Medical conditions Dystocia Postpartum haemorhage Ectopic pregnancy

Figure 2. Pregnancy-related hospital admissions in four hospitals in Serang and Pandeglang district (December 2005 to May 2006) by place of residence and severity of complication.

Figure 1. Causes of life-threatening complications in women with near miss, life-saving surgery without near miss, all life-threatening complications and maternal deaths in four hospitals in Serang and Pandeglang districts (December 2005 to May 2006).

rural Serang (1304 per 100 000). Excluding eclampsia and ectopic pregnancy for which we had no explicit signs of organ failure from all cases of near miss did not alter this pattern (data not shown).

Discussion
This is the rst study to report the incidence of life-threatening obstetric morbidity in a poor population with very low access to professional delivery care. In the two districts studied in this study, of 100 000 women who gave birth, 1416 had a life-threatening complication and 421 died as a consequence of pregnancy, resulting in an overall estimate of PRIAM of 1837 per 100 000 births. This ratio ranged from 1304 to 2983 per 100 000 depending on area of residence. Our overall estimate of PRIAM comes very close to the natural levels of maternal mortality observed in populations with no access to obstetric interventions. In Somerset, England, in the 16th century, maternal mortality was 2350 per 100 000 births at a time when very few women had access to obstetric care.23 In Europe between 1500 and 1850, maternal mortality was estimated to be 2000 per 100 000 live births, although estimates for Sweden between 1750 and 1800 are much lower at around 900 per 100 000.24,25 In Afghanistan, maternal mortality was 2182 per 100 000 in a nomadic community with very little access to care and 6507 per 100 000 in a population living in extremely remote mountainous areas.26

In this study, we aimed at dening complications that are so severe that their probability of dying is close to 100% in the absence of any major intervention. We call those complications absolutely life threatening such that women who experience these problems are unlikely to survive if they do not receive care in a hospital. This concept is important since a less stringent denition would have resulted in some women surviving in the community, and a count in hospital would have misrepresented the actual incidence. Complications such as dystocia, placenta praevia, unsafe abortion or pregnancyinduced hypertension, for example, are not considered absolutely life threatening, unless they are associated with a severe organ dysfunctionand are thus unlikely to survive in the community. We also include life-saving surgery for absolute maternal indications because such cases have a high probability of dying in the absence of surgery. Prophylactic measures in hospital that can prevent the occurrence of near miss, such as active management of the third stage in women at high risk of postpartum haemorrhage, antibiotics in women with early signs of potentially lethal chorioamnionitis, magnesium sulphate or delivery in a woman with impending eclampsia are not included because many women would have survived even in the absence of these interventions. Our study has a number of limitations. First, one critical assumption is that the conditions selected are so severe that they require hospital care for the women to survive. This is not likely to be the case for uncomplicated eclampsia and ectopic pregnancy since many will survive in the absence of professional care.27,28 However, excluding these cases did not alter the overall patterns. Haemorrhage was by far the

Pa

nd

eg

la

ng

ru

ra

re

m ot

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

87

Ronsmans et al.

Table 4. Population-based incidence per 100 000 births of life-threatening complications and maternal deaths in Serang and Pandeglang districts, Indonesia Life-threatening complications (95% CI) All urban Serang Pandeglang All rural Serang Pandeglang Pandeglang (remote) All areas 2654 (23173027) 2772 (23653228) 2364 (17903062) 1050 (9341176) 880 (7451032) 1493 (11911848) 1127 (8641444) 1416 (12971543) Maternal deaths (95% CI) 226 (173287) 212 (153282) 261 (160391) 479 (411556) 424 (348513) 501 (369661) 628 (489791) 421 (367483) Women with life-threatening complications or death (95% CI) 2880 (25293269) 2983 (25673463) 2624 (20073342) 1529 (13901682) 1304 (11401488) 1993 (16392395) 1754 (14292150) 1837 (17001980)

dominant cause of near miss, and our denition may not have been sufciently stringent: 59.0% of the cases of haemorrhage were in hypovolaemic shock, 13.7% had acute blood loss with haemoglobin levels less than 5 g/dl, 5.9% were associated with other organ dysfunctions, but in 21.4% of haemorrhage cases, we relied on blood transfusion only (data not shown). Use of blood transfusion has been a consistent criterion of near miss in other sites.1820,29 Transfusing four or more units is only performed in extreme cases in the hospitals in the study area, and it is unlikely that such women would have survived outside hospital. The condition that is perhaps the most difcult to ascertain reliably is cephalopelvic disproportion, the primary indication for life-saving surgery in our study. The equivocal nature of the denition of cephalopelvic disproportion is well known,30 and some cases may have been wrongly classied as life threatening and requiring life-saving surgery.5 Second, we relied on case notes for the ascertainment of life-threatening morbidity, and some of the information reportedthere may have been unreliable. For example, three cases of near miss were reported to have occurred after a prolonged labour, which seems doubtful. We classied these cases as near miss because of the organ dysfunction associated with them (all three women were reported to have been in shock), even though the underlying obstetric diagnosis is likely to have been misreported. An investment in near-miss audits may ultimately improve the quality and reliability of information on obstetric complications. Third, the large variation between urban, rural and remote areas in the incidence of PRIAM is not surprising, although we would have expected to nd a lower incidence in urban areas, which was not the case. We have no reason to believe that the populations in these areas are biologically different, but better access to care in urban areas should have led to more cases being prevented, and a lower incidence of PRIAM. The variation appears to follow the pattern of uptake of hospital care, with a higher incidence where more women give birth in hospital. This suggests that more cases may have been missedand survived in the communitywhere fewer

women make it into hospital or that severe morbidity may be overdiagnosed as more women give birth in hospital. Alternatively, it could suggest the presence of iatrogenic disease in urban populations. Even in areas with similarly low uptake of hospital care, however, such as in rural Serang and rural Pandeglang, the variation in the incidence of PRIAM was substantial (1304 and 1993 per 100 000, respectively). Some women in rural Serang may have sought care elsewhere, although our estimates of hospital births are consistent with those from the population-based survey. Finally, there may have been a bias in our estimates of maternal mortality, although we see no reason why this would be different between areas of residence. Unfortunately, the causes of maternal death occurring outside health facilities were not available in this study, and we were not able to compare the causes of maternal death with those of life-threatening morbidity. Our near-miss rates are much lower than the rates of severe maternal morbidity observed in the population-based study from West Africa where the incidence was 6170 per 100 000 live births.10 Denitions in the latter study were less stringent, however, since any complication leading to hospitalisation
3500 3000
Maternal mortality ratio Life-threatening complications

Per 100,000 births

2500 2000 1500 1000 500 0

Figure 3. Incidence of life-threatening complications and maternal deaths per 100 000 births in Serang and Pandeglang districts. Vertical bars are 95% CI for all life-threatening complications and death.

88

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

Pregnancy-related illness and mortality (PRIAM)

was included. Our rates are much higher, however, than those reported in studies from low-mortality countries using similar organ-based denitions, such as in Scotland (380 per 100 000 births).20 Clearly, variations in case denitions preclude meaningful comparisons, and there is a clear need to set standard criteria for cases of life-threatening morbidity.1

no responsibility for the information provided or views expressed in this paper. The views expressed herein are solely those of the authors.

Acknowledgements
We thank Drs R Nababan, M Bangun, K Marbun, R Patiassina, W Handayani, S Tessy and Sherlyana for doing the medical record review and data extraction; Drs MJN Mamahit, M Baharuddin, D Danukusumo, A Sastrowardoyo and E Harianto for their expertise on near-miss criteria and Veronique Filippi, Cynthia Stanton and Vincent De Brouwere for their very valuable comments on an earlier draft of this paper. This work was undertaken as part of Immpact, funded by the Bill and Melinda Gates Foundation, the UK Department for International Development, the European Commission and USAID. Immpact is an international research programme, which also provides technical assistance through its afliate organisation, Ipact. The funders have no responsibility for the information provided or views expressed in this paper. The views expressed herein are solely those of the authors. j

Conclusions
The measurement of life-threatening maternal morbidity in populations with low access to obstetric care remains challenging. The approach tested in this studyrelying on conditions that are absolutely life-threatening such that their count in hospitals can be used to represent the incidence in the general populationis promising but needs further testing in populations with varied disease epidemiology and access to care. While the reliance on facility-based data improves the robustness of case definitions, the continuing need to partially depend on management criteria19,20 and the inevitable human element in the diagnosis of some obstetric complications30 makes standardisation of case denitions difcult. There is no doubt, however, that the investigation of life-threatening morbidity is extremely useful for the purpose of auditing the quality of care in hospitals, both in developed20 and in developing country hospitals.18,19,29 Interviewing near-miss survivors will also help identify the barriers and delays in the health system that may contribute to the development of near miss. An investment in near-miss audits may ultimately improve the quality and reliability of information on obstetric complications and facilitate the development of rigorous and standard criteria for the denition of lifethreatening morbidity.

References
1 Say L, Pattinson RC, Gulmezoglu AM. WHO systematic review of maternal morbidity and mortality: the prevalence of severe acute maternal morbidity (near miss). Reprod Health 2004;17:3 2 Ronsmans C, Filippi V. Reviewing severe maternal morbidity: learning from women who survive life threatening complications. In: Beyond the Numbers. Reviewing Maternal Deaths and Complications to Make Pregnancy Safer. Geneva, Switzerland: World Health Organisation; 2004, pp103-124. 3 Minkauskien M, Nadisauskiene R, Padaiga Z, Makari S. Systematic review on the incidence and prevalence of severe maternal morbidity. Medicina 2004;40:299309. 4 Souza JP, Cecatti JG, Parpinelli MA, de Sousa MH, Serruya SJ. Systematic review of near miss maternal morbidity. Cad Saude Publica 2006; 22:25564. 5 Unmet Obstetric Need Network. Guide 1: Tackling Unmet Obstetric Needs Part 1: Concepts, General Principles and International Network (87 KB). Antwerp: Institute for Tropical Medicine, 1999 [http: //wwwitgbe/uonn/eng/home2html]. Accessed 16 November 2008. 6 Filippi V, Ronsmans C, Gandaho T, Graham W, Allihonou E, Santos P. Womens reports of severe (near-miss) obstetric complications in Benin. Stud Fam Plann 2000;31:30924. 7 Ronsmans C, Achadi E, Sutratikto G, Zazri A, McDermott J. Womens recall of obstetric complications in south Kalimantan, Indonesia. Stud Fam Plann 1997;28:20314. 8 Sloan N, Amoaful E, Arthur P, Winikoff B, Adjei S. Validity of womens self-reported obstetric complications in rural Ghana. J Health Popul Nutr 2001;19:4551. 9 Stewart M, Stanton C, Festin M, Jacobson N. Issues in measuring maternal morbidity: lessons from the Philippines Safe Motherhood Survey Project. Stud Fam Plann 1996;27:2935. 10 Prual A, Bouvier-Colle M-H, De Bernis L, Breart G. Severe maternal morbidity from direct obstetric causes in West-Africa: incidence and case fatality rates. Bull World Health Organ 2000;78:593602. 11 Bang RA, Bang AT, Reddy MH, Deshmukh MD, Baitule SB, Filippi V. Maternal morbidity during labour and the puerperium in rural homes and the need for medical attention: a prospective observational study in Gadchiroli, India. BJOG 2004;111:2318. 12 AbouZahr C. Global burden of maternal death and disability. Br Med Bull 2003;67:111.

Disclosure of interest
The authors have no conict of interest

Contribution to authorship
The study was designed by C.R. The eldwork and data collection were supervised by A.A., P.D. and F.N. The analysis was performed by S.S., A.A. and P.D. The rst draft of the paper was written by C.R. and all authors reviewed various versions.

Details of ethics approval

Funding
This work was undertaken as part of Immpact, funded by the Bill and Melinda Gates Foundation, the UK Department for International Development, the European Commission and USAID. Immpact is an international research programme, which also provides technical assistance through its afliate organisation, Ipact. The funders have

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology

89

Ronsmans et al.

13 Immpact, 2005. [www.immpact-international.org]. Accessed 16 November 2008. 14 Achadi E, Scott S, Pambudi ES, Makowiecka K, Marshall T, Adisasmita A, et al. Midwifery provision and uptake of maternity care in Indonesia. Trop Med Int Health 2007;12:14907. 15 Adisasmita A, Deviany PE, Nandiaty F, Stanton C, Ronsmans C. Obstetric near miss and deaths in public and private hospitals in Indonesia. BMC Pregnancy Childbirth 2008;8:10. 16 Ronsmans C, De Brouwere V, Dubourg D, Dieltiens G. Measuring the need for life-saving obstetric surgery in developing countries. BJOG 2004;111:102730. 17 Makowiecka K, Achadi E, Izati Y, Ronsmans C. Midwifery provision in two districts in Indonesia: how well are the rural villages served? Health Policy Plan 2008;23:6775. 18 Filippi V, Ronsmans C, Gohou V, Goufodji S, Lardi M, Sahel A, et al. Maternity wards or emergency obstetric rooms? Incidence of near-miss events in African hospitals. Acta Obstet Gynecol Scand 2005;84:1116. 19 Mantel GD, Buchmann E, Rees H, Pattinson RC. Severe acute maternal morbidity: a pilot study of a denition of a near miss. Br J Obstet Gynaecol 1998;105:98590. 20 Brace V, Penney G, Hall M. Quantifying severe maternal morbidity: a Scottish population study. BJOG 2004;111:4814. 21 Ronsmans C, Scott S, Qomariyah SN, Achadi E, Braunholz D, Marshall T, Pambudi E, Witten K, Graham W. Midwife-led community care and maternal mortality in Indonesia. Bull World Health Organ; in press.

22 Rapid Ascertainment Process for Institutional Deaths (RAPID). Immpact Toolkit: Module 4 Version 2, 2007. [www.immpact-international.org/ toolkit/module4/rapid/]. Accessed 16 November 2008. 23 Willmott Dobbie BM. An attempt to estimate the true rate of maternal mortality, sixteenth to eighteenth centuries. Med Hist 1982;26: 7990. 24 Peller S. Births and Deaths Among Europes Ruling Families Since 1500. Chicago, IL: Aldine Publishing Co, 1965. 25 Hogberg U, Wall S. Secular trends in maternal mortality in Sweden from 1750 to 1980. Bull World Health Organ 1986;64:7984. 26 Bartlett L, Mawji S, Whitehead S, Crouse C, Dalil S, Ionete D, et al. Where giving birth is a forecast of death: maternal mortality in four districts of Afghanistan, 1999-2002. Lancet 2005;365:86470. 27 Loudon I. Death in Childbirth. An International Study of Maternal Care and Maternal Mortality 1800-1950. Oxford, UK: Clarendon press, 1992. 28 Shalev E, Peleg D, Tsabari A, Romani S. Spontaneous resolution of ectopic tubal pregnancy: natural history. Fertil Steril 1995;63:1519. 29 Okong P, Byamugisha J, Mirembe F, Byaruhanga R, Bergstrom S. Audit of severe maternal morbidity in Ugandaimplications for quality of care. Acta Obstet Gynecol Scand 2006;85:797804. 30 Crowther C, Enkin M, Keirse MJNC, Brown I. Monitoring the progress of labour. In: Chalmers I, Enkin M, Keirse MJNC, editors. Effective Care in Pregnancy and Childbirth, Volume 2, Childbirth. Oxford, UK: Oxford University Press; 1991. pp. 83345.

90

2008 The Authors Journal compilation RCOG 2008 BJOG An International Journal of Obstetrics and Gynaecology