- Top of page
Objectives: To analyse risk factors of perinatal death, with an emphasis on potentially avoidable risk factors, and differences in the frequency of suboptimal care factors between maternity units with different levels of care.
Methods: Six hundred and eight pregnancies (2001–2005) in South Australia resulting in perinatal death were described and compared to 86 623 live birth pregnancies.
Results: Two hundred and seventy cases (44.4%) were found to have one or more avoidable maternal risk factors, 31 cases (5.1%) had a risk factor relating access to care, while 68 cases (11.2%) were associated with deficiencies in professional care. One hundred and four women (17.1% of cases) presented too late for timely medical care: 85% of these did have a sufficient number of antenatal visits.
The following independent maternal risk factors for perinatal death were found: assisted reproductive technology (adjusted odds ratio (AOR) 3.16), preterm labour (AOR 22.05), antepartum haemorrhage (APH) abruption (AOR 6.40), APH other/unknown cause (AOR 2.19), intrauterine growth restriction (AOR 3.94), cervical incompetence (AOR 8.89), threatened miscarriage (AOR 1.89), pre-existing hypertension (AOR 1.72), psychiatric disorder (AOR 1.85) and minimal antenatal care (AOR 2.89). The most commonly found professional care deficiency in cases was the failure to act on or recognise high-risk pregnancies/complications, found in 49 cases (8.1%).
Conclusion: Further improvements in perinatal mortality may be achieved by greater emphasis on the importance of antenatal care and educating women to recognise signs and symptoms that require professional assessment. Education of maternity care providers may benefit from a further focus on how to recognise and/or manage high-risk pregnancies.
- Top of page
Perinatal mortality is a useful indicator of standards of perinatal care. Perinatal audits can be useful in identifying suboptimal, and therefore potentially avoidable, care factors that contributed to a perinatal death.1
The perinatal mortality in Australia is low compared to other western countries in the world. The perinatal mortality rate (these rates are expressed as deaths per 1000 births for births of at least 400 g birthweight (or if birthweight unavailable, 20 weeks gestation) and neonatal deaths within the first 28 days of life) in South Australia (9.4 deaths per 1000 in 2004) is generally lower than the Australian rate (10.5 deaths per 1000 in 2004).2
To decrease the perinatal death rate in South Australia even further, information is needed about potentially avoidable versus unavoidable risk factors. Avoidable risk factors could, at least in theory, open pathways to primary or secondary prevention.
Several systems of classifying perinatal deaths exist for epidemiological surveillance and perinatal audits.3 The Perinatal Society of Australia and New Zealand guideline for institutional perinatal mortality audit divides the factors relating to care into three major groups: factors relating to the woman/her family/her social situation, factors relating to access to care and factors relating to professional care.4 The Confidential Enquiry into Stillbirths and Deaths in Infancy (CESDI) 8th annual report provides a number of suboptimal care factors most frequently listed by assessment panels, for example, failure to recognise or act on high risk pregnancies, poor communication between care providers and poor attendance for antenatal checks by mothers.5
The aim of this audit was to analyse: (i) patterns of risk factors present in cases of perinatal death, with an emphasis on potentially avoidable risk factors, and (ii) differences in the frequency of suboptimal care factors between maternity units with different levels of care.
- Top of page
We reviewed all 608 pregnancies that resulted in perinatal deaths occurring between 1-1-01 and 31-12-05 in South Australia. In South Australia, perinatal death includes stillbirths of at least 20 weeks gestation or 400 g birthweight and neonatal deaths within the first 28 days of life. Deaths attributed to congenital abnormalities and terminations of pregnancies were excluded from our study. A twin or triplet pregnancy was considered as one pregnancy even if all babies died. These 608 pregnancies were compared to 86 623 live birth pregnancies in 2001–2005 in South Australia (after excluding all neonatal deaths and postneonatal deaths attributed to a congenital abnormality).
The Pregnancy Outcome Unit from the South Australian Department of Health provided 608 summaries based on records obtained by the Unit regarding each perinatal death. The South Australian Pregnancy Outcome Unit receives its birth data via the standardised birth summary sheets completed by midwives in all public and private South Australian Maternity Units.
The deaths summaries are reviewed on a regular basis by the Perinatal Subcommittee of the Maternal, Perinatal and Infant Mortality Committee (MP&IMC) to classify the cause of death and to identify avoidable factors. The perinatal subcommittee consists of senior obstetricians, perinatal pathologists, neonatologists, epidemiologists and midwives.
These summaries were used to obtain information about the mother, her social background, the pregnancy, time-related events intrapartum and the post-partum stage.
All data were extracted from the perinatal death summaries by two trained research officers following standardised criteria.
To analyse patterns of risk factors present in cases of perinatal death, the first aim of this audit, only categorical data available in both controls and cases could be used for initial χ2 testing followed by multivariate analysis logistic regression.
In order to analyse differences in the frequency of suboptimal care factors between maternity units with different levels of care (second aim), only data from the perinatal death summaries were used by default.
The specifications of some characteristics studied as potential risk factors for perinatal death are presented in Table 1. Data on smoking (yes/no) and illicit drug use (yes/no) are included as ‘mandatory’ points in the standard perinatal summaries, data on domestic violence were only available in the perinatal death summaries, and could only be positively identified if specifically mentioned as relevant factor by the care providers involved.
Table 1. Specifications of some of the characteristics studied as potential risk factors for perinatal death
|• Preterm birth||Gestation < 37|
|• Grande multiparity||Parity > 4|
|• Assisted reproductive technology||Intracytoplasmic sperm injection or in vitro fertilisation|
|• Minimal antenatal care||In deaths: if clearly mentioned in the summaries.|
|In live births: < 3 antenatal visits.|
|• Late presentation for medical care in a timely manner||Loss of fetal movements > 24 h for patients in the third trimester.|
|Per vagina fluid loss, blood loss or severe abdominal pain > 4 hours|
|Access to care (only for cases)|
|• Distance or remoteness|| |
|• Lack of transport|| |
|• Lack of local resources||Such as inadequate equipment or insufficient staff|
|Professional care factors (only for cases)|
|• Poor communication between care providers (peripartum)||Either written or oral communication|
|• Failure to recognize or act on high risk pregnancies or complications (peripartum)||Such as major CTG misinterpretation, doctors delay and/or inappropriate procedures|
|• Inappropriate hospital level or care provider (peripartum)||Providing peripartum care outside existing protocols for level 1–3 hospitals|
|• Deficiencies in antenatal care||Such as failure to recognize major IUGR, insufficient management of maternal problems, not referring or recognizing high risk pregnancies.|
In this audit, suboptimal care factors were only positively identified if there were no reasons for any doubt, that is, data were only positively identified on the basis on non-ambiguous statements or quantitative statements in the perinatal death summaries. For both live births and perinatal deaths, data were also linked to the hospital level where the infant was born: levels 1, 2 and 3 maternity units, private and country hospitals.
Ethics approval was not sought, as this audit conforms to the standards established by the National Health and Medical Research Council for ethical quality.6 The MP&IMC gave permission for the use of the perinatal death summaries.
Statistical analyses were performed using stata Statistical Software 8 (StataCorp, College Station, TX, USA). Comparisons between the live births and perinatal deaths were made using the χ2 test. A P-value < 0.05 was considered statistically significant. Odds ratios (OR) with 95% confidence intervals (CI) were used to describe the association between perinatal death and different variables. Multivariate logistic regression was based on backward elimination of non-significant variables and yielded adjusted OR and 95% CI. The Hosmer–Lemeshow goodness-of-fit test was performed for this model.
For the comparison of the deficiencies in peripartum professional care per hospital type, individual cells were tested using Haberman's adjusted standardised residual. An allowance for multiple comparisons was made using a Bonferroni adjustment, resulting in a critical value of 2.576, indicating a significant difference between the expected and observed outcomes. An adjusted residual higher than 2.576 will indicate a significantly high amount of care deficiencies, and an adjusted residual lower than –2.576 will indicate a significantly low amount of care deficiencies.
- Top of page
The 608 pregnancies reviewed included 614 perinatal deaths, of which 441 were stillbirths and 173 were neonatal deaths. Potentially avoidable risk factors in the maternal subgroup were found in 270 cases (44.4%). Thirty-one cases (5.1%) had one or more risk factors relating to access to care; 24 cases suffered a lack of local resources, 12 cases had problems related to distance or remoteness and four cases had transport problems.
Fifty-five cases (9.0%) experienced one or more deficiencies in peripartum care, and 16 cases (2.6%) experienced a deficiency in antenatal care. Considering peripartum and antenatal care as professional care in general, 68 cases (11.2%) suffered a deficiency in professional care, possibly contributing to perinatal death. Among the 170 term or post-term perinatal deaths in our study, 27.1% of cases were associated with professional care deficiencies.
The distribution of potentially avoidable maternal risk factors in 600 cases (8 home/interstate pregnancies were excluded) over the different types of hospitals of birth is presented in Fig. 1. The mothers giving birth in country hospitals had the greatest percentage of smokers (40.6%), while private hospitals had the smallest percentage (3.2%). The illicit drug use in the case group was the highest in level 1 hospitals (13.9%) while none of the women with a perinatal death in any private hospital reported illicit drug use. The proportion of cases in which the mothers had minimal antenatal care prior to the perinatal death was the greatest in level 1 hospitals (19.4%) closely followed by country hospitals (16.7%), while private hospitals reported the lowest proportion of cases with minimal antenatal care (1.6%). The proportion of women presenting too late for medical care in the presence of major concerning signs/symptoms was greatest in level 3 (18.8%) and smallest in level 2 hospitals (10.8%).
From women who had perinatal death in a country hospital, six women (6.3%) suffered from domestic violence, while women in level 1 and private hospitals reported no domestic violence.
The associations between perinatal death and different variables are shown in Table 2.
Table 2. Risk factors for perinatal death, 608 cases and 86 623 controls, South Australia, 2001–2005†
|Characteristics||Number of cases (%)||Number of controls (%)||Crude OR (95% CI)||Adjusted OR (95% CI)|
|Previous perinatal death||35 (5.8%)||1180 (1.4%)||4.42 (3.04–6.26)|| |
|Teenage pregnancy||47 (7.7%)||4618 (5.3%)||1.49 (1.08–2.01)||1.13 (0.82–1.57)|
|Indigenous status||52 (8.6%)||2220 (2.6%)||3.55 (2.61–4.75)||1.57 (1.12–2.21)|
|Primiparity||275 (45.2%)||36299 (41.9%)||1.14 (0.97–1.35)|| |
|Grande Multiparity||12 (2.0%)||1143 (1.3%)||1.51 (0.77–2.66)|| |
|ART||33 (5.4%)||485 (0.6%)||10.19 (6.87–14.67)||3.16 (2.09–4.76)|
|Multiple pregnancy||63 (10.4%)||1455 (1.7%)||6.77 (5.10–8.84)|| |
|Preterm labour||438 (72.2%)||6079 (7.0%)||34.34 (28.59–41.26)||22.05 (18.16–26.76)|
|APH abruption||77 (12.7%)||591 (0.7%)||21.11 (16.19–27.25)||6.40 (4.80–8.55)|
|APH praevia||10 (1.6%)||523 (0.6%)||2.75 (1.31–5.14)|| |
|APH other cause||70 (11.5%)||2106 (2.4%)||5.22 (3.99–6.74)||2.19 (1.66–2.89)|
|PI hypertensive disorder||52 (8.6%)||6782 (7.8%)||1.10 (0.81–1.47)||0.50 (0.37–0.69)|
|Gestational diabetes||16 (2.6%)||3295 (3.8%)||0.68 (0.39–1.12)||0.59 (0.35–1.00)|
|IUGR||112 (18.4%)||2467 (2.9%)||7.70 (6.19–9.51)||3.94 (3.12–4.99)|
|Cervical incompetence||28 (4.6%)||144 (0.2%)||28.99 (18.46–44.09)||8.89 (5.46–14.51)|
|Threatened miscarriage||29 (4.8%)||1792 (2.1%)||2.37 (1.57–3.45)||1.89 (1.25–2.84)|
|Anaemia||32 (5.3%)||6254 (7.2%)||0.71 (0.48–1.02)||0.41 (0.28–0.60)|
|Pre-existing hypertension||24 (4.0%)||1074 (1.2%)||3.27 (2.07–4.95)||1.72 (1.07–2.77)|
|Pre-existing diabetes||12 (2.0%)||438 (0.5%)||3.96 (2.02–7.04)|| |
|Asthma||30 (4.9%)||5965 (6.9%)||0.70 (0.47–1.01)||0.53 (0.36–0.78)|
|Psychiatric disorder||40 (6.6%)||2089 (2.4%)||2.85 (2.01–3.94)||1.85 (1.29–2.65)|
|Smoking||172 (28.3%)||15499 (17.9%)||1.81 (1.51–2.17)|| |
|Illicit drug use||38 (6.3%)||853 (1.0%)||6.70 (4.66–9.39)|| |
|Minimal antenatal care||64 (10.5%)||1074 (1.2%)||9.37 (7.07–12.51)||2.89 (2.10–4.00)|
In the multivariate analysis, significant risk factors for perinatal death were: indigenous status, assisted reproductive technology (ART), preterm labour, antepartum haemorrhage (APH) as a result of abruption, APH from other or unknown cause, intrauterine growth restriction (IUGR), cervical incompetence, threatened miscarriage, pre-existing hypertension, psychiatric disorders and minimal antenatal care.
In the univariate model there was no significant relationship found between perinatal death and pregnancy-induced hypertensive disorders, gestational diabetes, anaemia and asthma. However, after adjustment, these factors were significantly protective factors for the occurrence of perinatal death (borderline for gestational diabetes).
The Hosmer–Lemeshow goodness-of-fit test showed adequate model calibration.
Figure 2 shows deficiencies in peripartum care, which might have contributed to a perinatal death, by hospital type of birth. The percentages in the figure are given as percentages of the total number of births in that hospital. Level 2 and 3 hospitals were found to have significantly lower amounts of deficiencies in peripartum care where country hospitals were found to have significantly higher amounts of peripartum deficiencies.
Figure 2. Perinatal death percentages and peripartum care by hospital type of birth. () Appropriate peripartum care and () deficiencies in peripartum care.
Download figure to PowerPoint
The ‘failure to act on or recognise complications or high risk pregnancies’ factor was the most common deficiency (Table 3), found in 49 of the total 53 hospital cases that were associated with a peripartum care deficiency.
Table 3. Deficiencies in peripartum care in 600 hospital births by hospital type of birth in South Australia, 2001–2005†
|Hospital level of birth||Total number of births||Number of PNDs||Perinatal mortality rate per 1000 births||PNDs with peripartum care deficiencies||% peripartum care deficiencies of PNDs||Failure to recognize or act on complications (%)‡||Poor communicatio n (%)||Inappropriate level of care or care provider (%)|
|Level 1||5 069||36||7.1|| 6||16.7%|| 6 (100%)||1 (16.7%)|| 2 (33.3%)|
|Level 2|| 8 405|| 65||7.7|| 0||0|| 0||0||0|
|Level 3||30 391||341|| 11.2|| 21||6.2%||21 (100%)||0|| 1 (4.8%)|
|Private||23 569|| 62||2.6|| 9||14.5%|| 9 (100%)||0|| 4 (44.4%)|
|Country||19 514|| 96||4.9|| 17||17.7%||13 (76.5%)||1 (5.9%)||10 (58.8%)|
- Top of page
Several perinatal audits have been done in which the possible preventability of perinatal deaths was described.7–9 A Dutch study7 distinguished between ‘definitely avoidable’, ‘probably avoidable’, ‘possibly avoidable’, ‘definitely not avoidable’ and ‘undetermined’ perinatal deaths. Myers et al.8 determined a perinatal death to be potentially avoidable if any factor was determined that possibly could have influenced the outcome.
Instead of investigating the potential avoidability in a perinatal death, we chose to look exclusively at the presence of avoidable risk factors in each case. By identifying solely the presence of suboptimal care instead of estimating the influence of a contributing factor, either from the mother or from the care provider, we tried to avoid the difficult discussion of whether the absence of suboptimal care would have altered the adverse outcome. A comparable approach was used in an Icelandic study where Georgdottir et al.10 mentioned that perinatal deaths may not have been avoidable even if the care had been optimal. This approach could partially explain our relatively high proportion of avoidable maternal risk factors. In contrast, the summaries used for this audit can lack information about suboptimal care and existing maternal conditions possibly contributing to a perinatal death, and therefore may lead to underreporting.
The overall perinatal mortality rate in the various hospitals (Fig. 2, Table 3) shows the expected high number of deaths in the level 3 units having the neonatal intensive care units for extremely preterm neonates.
The avoidable maternal risk factors depicted in Figure 1 may have been biased by the fact that we only considered the hospital of birth, not what type of patient the cases usually were (in other words, a private patient could have been giving birth in a level 3 hospital, and was then considered a level 3 patient). Domestic violence,11 illicit drug use,12 alcohol abuse and smoking are likely to be under reported, regardless of the hospital of birth.
The overall pattern of risk factors (Table 2) is mostly in line with expectations. Other studies have shown comparable results about teenage pregnancies (as univariate risk factor) being a risk factor for perinatal mortality.13,14 However, after adjustment in multivariate analysis, being a teenage mother was not an independent risk factor. An interesting finding is the elimination of smoking in the multivariate analysis, probably because of its association with IUGR and placental abruption. In the multivariate analysis, both pregnancy-induced hypertensive disorders and gestational diabetes were significantly ‘protective’ for perinatal death. A reduced risk for perinatal death (although not significant) was reported previously for pregnancy-induced hypertensive disorders in South Australia,15 and may be because of the intensive antenatal care and mostly appropriate referral of these high-risk pregnancies, and/or the fact that the intrinsic effects of maternal hypertension actually convey some fetal protection.15
We had difficulty comparing our results to other audits on this topic, because of the use of different populations, different variables and sometimes the degree of avoidability in perinatal mortality used in other audits. Delke et al.16 investigated 133 term or post-term perinatal deaths and found preventable maternal risk factors based on non-compliance, minimal antenatal care and late presentation for care in 39.1% of the cases. Wolleswinkel et al.17 included smoking as suboptimal care and found 40% of maternal suboptimal care. Our study found a slightly higher percentage of suboptimal care (44.4%), partly explained by our broader variable list of maternal avoidable risk factors and stricter criteria.
Potentially avoidable risk factors according to access to care were present in 5% of cases in Wolleswinkel et al.17; this did not differ from our results (5.1%), while we expected to find a higher amount because of South Australia's vast distances and remote areas. Apparently existing retrieval services are effective in overcoming difficulties because of remoteness.
Delke et al.16 found deficiencies in professional care in 38.3%, while we found 11.2%. This difference is partially the result of our inclusion of preterm births, where professional care often could not have altered the adverse outcome. Among the 170 term or post-term perinatal deaths in our study, 27.1% of cases were associated with professional care deficiencies. Miranda et al.,18 in a comparable audit in Granada, found suboptimal professional care to be present in 12.8% of perinatal deaths, also Myers et al.8 found 15%, both studies in line with the 11.2% in South Australia. Antenatal care deficiencies appear to be rare (2.6%) in South Australia's perinatal deaths compared to other studies,10,17 which may reflect good referral practices among antenatal care providers.
Any national or statewide audit on this topic will have some unavoidable weaknesses. Our choice of the late presentation for care criteria may be arbitrary, although we feel that the cut-offs we proposed are realistic as many studies have shown that a major decrease or loss of fetal movements is associated with adverse pregnancy outcome.19
It should also be noted that the identification of minimal antenatal care in cases had to be based on the perinatal death summaries, while for live birth the actual number of antenatal visits was known (minimal antenatal care defined as < 3 antenatal visits).
Difference in ascertainment is another unavoidable weakness in this type of study, that is, the fact that live birth data were obtained via the standardised perinatal sheets completed by midwives in all South Australian maternity units, while the summaries on perinatal deaths were based on often quite complete access to medical records, detailed reports by the obstetricians involved etc.
The results of this five-year statewide audit show that 17.1% of the cases presented too late for timely medical care; 85% of these had sufficient antenatal visits. These data indicate that further improvements in perinatal outcome might be achieved with expanding patient education focusing on concerning signs and symptoms indicating fetal and maternal compromise.
The presence of professional care deficiencies in 11.2% of cases indicates that further improvements may be achieved with further professional education for all maternity care providers, with perhaps a particular emphasis on country hospitals. The newly introduced statewide web-based cardiotocograph accreditation is an important step in this direction.