DOI: 10.1111/1471-0528.13058 www.bjog.org
The impact of socioeconomic position on severe maternal morbidity outcomes among women in Australia: a national case–control study A Lindquist,a,b,* N Noor,a,* E Sullivan,c M Knighta a
National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK b Department of Obstetrics and Gynaecology, Monash Health, Melbourne, Vic., Australia c Faculty of Health, University of Technology Sydney, Sydney, NSW, Australia Correspondence: Prof M Knight, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Old Rd campus, Oxford OX3 7LF, UK. Email [email protected] Accepted 13 July 2014. Published Online 17 September 2014.
Objective Studies in other developed countries have suggested that
socioeconomic position may be a risk factor for poorer pregnancy outcomes. This analysis aimed to explore the independent impact of socioeconomic position on selected severe maternal morbidities among women in Australia. Design A case–control study using data on severe maternal
morbidities associated with direct maternal death collected through the Australasian Maternity Outcomes Surveillance System. Setting Australia. Population 623 cases, 820 controls. Methods Logistic regression analysis to investigate differences in
outcomes among different socioeconomic groups, classified by Socio-Economic Indexes for Areas (SEIFA) quintile. Main outcome measures Severe maternal morbidity (amniotic
fluid embolism, placenta accreta, peripartum hysterectomy, eclampsia or pulmonary embolism).
Results SEIFA quintile was statistically significantly associated with maternal morbidity, with cases being twice as likely as controls to reside in the most disadvantaged areas (adjusted OR 2.00, 95%CI 1.29–3.10). Maternal age [adjusted odds ratio (aOR) 2.20 for women aged 35 or over compared with women aged 25–29, 95% CI 1.64–3.15] and previous pregnancy complications (aOR 1.30, 95%CI 1.21–1.87) were significantly associated with morbidity. A parity of 1 or 2 was protective (aOR 0.58, 95%CI 0.43–0.79), whereas previous caesarean delivery was associated with maternal morbidity (aOR 2.20 for women with one caesarean delivery, 95% CI 1.44–2.85, compared with women with no caesareans). Conclusion The risk of severe maternal morbidity among women
in Australia is significantly increased by social disadvantage. This study suggests that future efforts in improving maternity care provision and maternal outcomes in Australia should include socioeconomic position as an independent risk factor for adverse outcome. Keywords Australia, maternal morbidity, socioeconomic position.
Please cite this paper as: Lindquist A, Noor N, Sullivan E, Knight M. The impact of socioeconomic position on severe maternal morbidity outcomes among women in Australia: a national case–control study. BJOG 2014; DOI: 10.1111/1471-0528.13058.
Introduction Australians generally enjoy high standards of living and have one of the longest life expectancies at birth in the world, at just over 80 years.1 Despite this, important health disparities exist between different groups of Australians, with, for example, the average life expectancy of Indigenous Australians 17 years less than that of non-Indigenous Australians.2 Research conducted in Australia in 2000 looking at health and socioeconomic status suggested that despite substantial improvements in the *These authors contributed equally to this manuscript.
ª 2014 Royal College of Obstetricians and Gynaecologists
health of the Australian population in previous years, inequalities in socioeconomically related mortality had increased for some conditions.3 Although studies in Australia have investigated risk factors for severe maternal morbidity,4 there is no national research into the association between socioeconomic position and severe maternal morbidity. Studies in other developed countries have suggested that socioeconomic position may be a risk factor for poorer outcomes. The most recent Confidential Enquiry into Maternal Deaths in the UK showed that unemployed women and those in routine/ manual occupations were six times more likely to die during pregnancy and labour than women from the highest
1
Lindquist et al.
socioeconomic group.5 A recent study investigated the social gradient of severe maternal morbidity in the UK and the results suggested that independent of ethnicity, BMI and age, low socioeconomic position was a risk factor for severe maternal morbidity.6 Identifying high-risk women is critical for the prevention of adverse outcomes, as it allows targeted interventions and intensive clinical management of specific groups of women. The Australasian Maternal Outcomes Surveillance System (AMOSS) was established for the purpose of investigating risk factors, management and outcomes of different maternal morbidities. Based on the methodology of the UK Obstetric Surveillance System (UKOSS), AMOSS enables nationwide research and surveillance into a range of pregnancy complications across both Australia and New Zealand to be carried out. The aim of this analysis was to explore the independent impact of socioeconomic position on selected severe maternal morbidities among women in Australia.
Methods AMOSS is a national system to study selected severe maternal morbidities in Australia that was established in 2009.7 Negative surveillance of conditions, whereby nil reports are also collected, is conducted monthly across eligible maternity units with >50 births per year, covering about 96% of eligible sites nationally. Where a case is identified, a data collection form is completed, detailing demographic, risk and pregnancy factors, management and outcomes of the condition and pregnancy. Information on control women is also collected for specific studies. We performed a case–control study using data on specific severe maternal morbidities associated with direct maternal death extracted from the AMOSS studies database. We collectively included women who had amniotic fluid embolism, placenta accreta, peripartum hysterectomy, eclampsia or pulmonary embolism as ‘severe maternal morbidity’ cases (n = 623). Women with more than one of the specific morbidities were included only once in the analysis. Definitions of each condition are shown in Box 1. Controls were identified as the two women who had delivered immediately prior to the identified cases of placenta accreta and peripartum hysterectomy in the same hospitals (n = 820). We collected identical data for cases and controls, apart from the details of the management and outcomes of the severe morbidity defining the case. The main exposure variable of socioeconomic status was the Socio-Economic Indexes for Areas (SEIFA) quintile, derived from the maternal postcode of residence. It is created from four socioeconomic indexes and is a measure of socioeconomic
2
disadvantage, with the lowest quintile (1) indicating the most disadvantaged areas8,9 All analyses were performed using R version 2.15.1. Box 1 Definitions of severe maternal morbidities Amniotic fluid embolism: Case defined by either a clinical diagnosis of AFE (acute hypotension or cardiac arrest, acute hypoxia or coagulopathy in the absence of any other potential explanation for the symptoms and signs observed) or a postmortem diagnosis (presence of fetal squames or other debris in the pulmonary circulation). Peripartum hysterectomy: Any woman whose pregnancy terminates and who has a hysterectomy in the same clinical episode or within 6 weeks postpartum when the indication for hysterectomy is related to the pregnancy, e.g. secondary postpartum haemorrhage, or any woman giving birth and undergoing a hysterectomy in the same clinical episode or within 6 weeks postpartum when the indication for hysterectomy is related to the birth, e.g. secondary postpartum haemorrhage Placenta accreta: Any women identified as having placenta accreta (or increta or percreta) diagnosed by antenatal imaging, or diagnosed at operation or diagnosed by pathology specimen Eclampsia: Any woman having convulsions during pregnancy or in the first 10 days postpartum, together with at least two of the following features within 24 hours of the convulsion(s): 1 Hypertension (a booking diastolic pressure of 42 iu/l) Antenatal pulmonary embolism: Cases were defined as any woman in whom 1 Pulmonary embolism was confirmed using suitable imaging (angiography, computed tomography, echocardiography, magnetic resonance imaging or ventilation-perfusion scan showing a high probability of pulmonary embolism). 2 Or pulmonary embolism was confirmed at surgery or post-mortem. 3 Or a clinician made a diagnosis of pulmonary embolism with signs and symptoms consistent with PE present, and the patient received a course of anticoagulation therapy (>1-week duration).
We performed a descriptive analysis to examine the distribution of maternal characteristics and SEIFA quintiles in the case and control groups using chi-squared tests. We performed unconditional logistic regression analysis to investigate the association between severe maternal morbidity and SEIFA quintiles, adjusted for potential confounding factors. Results are presented as odds ratios (OR) with 95% confidence intervals (CI). We included variables that were likely to confound the association between maternal morbidity and socioeconomic
ª 2014 Royal College of Obstetricians and Gynaecologists
Severe maternal morbidity in Australia
status in a multivariable model. Variables included were: maternal age, country of birth, maternal body mass index (BMI), Aboriginal or Torres Strait Islander status, remoteness of residence classified using the Australian Standard Geographical Classification – Remoteness Area,10 parity, smoking, previous medical and pregnancy complications, multiple pregnancy, history of previous caesarean delivery, number of previous caesarean deliveries and insurance status. We used a univariable analysis to determine variables that were to be added in the multivariable model. Maternal age and BMI were continuous variables that were tested for the linearity assumption. BMI exhibited a non-linear association with maternal morbidity. Owing to its non-linearity, we categorised this variable into four groups. Maternal age showed a linear relationship with the outcome variable. We therefore analysed this variable in both continuous and categorical terms, but for ease of interpretation it has been presented as a categorical variable in the tables. Variables that had a high proportion of missing data (>1%, Table 1), such as BMI, smoking and previous pregnancy complications, were considered unlikely to be missing at random; we therefore created a separate category of ‘missing’ for these variables. We built a multivariable model using unconditional logistic regression. We retained variables that had been
Table 1. Distribution of missing data for maternal morbidity cases and controls Variable
SEIFA quintile Country of birth Maternal age Aboriginal or Torres Strait Islander status Remoteness of residence Smoking BMI Parity Multiple pregnancy Previous pregnancy complications Previous medical/obstetric complications History of previous caesarean delivery Number of previous caesarean deliveries Health insurance
Number of cases with missing information (% of total cases)
Number of controls with missing information (% of total controls)
4 (0.6) 6 (1.0) – 32 (5.1)
3 (0.4) 8 (0.1) – 40 (4.9)
2 60 78 – – 39 8
2 70 96 1 2 33 7
(0.3) (9.6) (12.5)
(6.2) (1.2)
(0.2) (8.5) (11.7) (0.12) (0.2) (4.0) (0.9)
8 (1.3) 9 (1.4)
7 (0.8) 7 (0.8)
3 (0.5)
4 (0.5)
ª 2014 Royal College of Obstetricians and Gynaecologists
established as confounders through previous literature (BMI, smoking and remoteness) in the final model. Other potential confounders were added to the simple model containing maternal morbidity status and SEIFA quintiles, and the effect of each variable on the model was assessed using Akaike information criterion (AIC) values and likelihood ratio tests. Variables significantly affecting the fit of the model (P < 0.05) were retained in the final model. We examined all possible interaction terms but none were found to be statistically significant. We assessed the fit of the final multivariable mode using the Hosmer–Lemeshow goodness-of-fit test. Given the 623 cases and 820 controls, and a prevalence of 8.9% for the lowest socioeconomic quintile, our analysis had an 80% power to detect as statistically significant (P < 0.05) odds of 1.36 or greater.
Results Table 2 shows the characteristics of cases and controls. In comparison with controls, cases were more likely to be aged 35 or over, belong to a lower SEIFA quintile, to be an Aboriginal or Torres Strait Islander, have a parity of 3 or more, previous pregnancy complications and a history of previous caesarean delivery. Table 3 shows the morbidities each woman had, including those women who had more than one morbidity. The multivariable model examined the relationship between maternal morbidity and socioeconomic status, while controlling for potential confounders and other risk factors (Table 4). Maternal age was significantly associated with maternal morbidity, with women aged 30–34 being 1.44 (95%CI 1.04–1.99) times more likely and women aged 35 or over being 2.2 (95%CI 1.64–3.15) times more likely to be cases, compared with controls after adjustment. Socioeconomic status was statistically significantly associated with maternal morbidity, with cases being twice (95% CI 1.29–3.10) as likely to belong to the lowest socioeconomic group (SEIFA quintile 1), 1.56 (95%CI 1.02–2.38) times more likely to fall in SEIFA quintile 2 and 1.79 (95% CI 1.26–2.54) times more likely to be in quintile 3, compared with controls. Having a parity of one or two was found to be protective [adjusted odds ratio (aOR) 0.58, 95%CI 0.43–0.79] against maternal morbidity, once adjusted for age and previous caesarean section and pregnancy complications, whereas women who had reported previous pregnancy complications were 1.3 (95%CI 1.21– 1.87) times more likely to be cases compared with controls. The number of previous caesarean deliveries was also statistically significantly associated with maternal morbidity, with women with one caesarean delivery having double the odds of morbidity (95%CI 1.44–2.85), two caesarean deliveries having 4.10 times the odds (95%CI 2.60–6.59) and three or more previous caesarean deliveries having 9.35
3
Lindquist et al.
Table 2. Maternal characteristics associated with maternal morbidity Characteristic
Age
The impact of socioeconomic position on severe maternal morbidity outcomes among women in Australia: a national case–control study A Lindquist,a,b,* N Noor,a,* E Sullivan,c M Knighta a
National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK b Department of Obstetrics and Gynaecology, Monash Health, Melbourne, Vic., Australia c Faculty of Health, University of Technology Sydney, Sydney, NSW, Australia Correspondence: Prof M Knight, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Old Rd campus, Oxford OX3 7LF, UK. Email [email protected] Accepted 13 July 2014. Published Online 17 September 2014.
Objective Studies in other developed countries have suggested that
socioeconomic position may be a risk factor for poorer pregnancy outcomes. This analysis aimed to explore the independent impact of socioeconomic position on selected severe maternal morbidities among women in Australia. Design A case–control study using data on severe maternal
morbidities associated with direct maternal death collected through the Australasian Maternity Outcomes Surveillance System. Setting Australia. Population 623 cases, 820 controls. Methods Logistic regression analysis to investigate differences in
outcomes among different socioeconomic groups, classified by Socio-Economic Indexes for Areas (SEIFA) quintile. Main outcome measures Severe maternal morbidity (amniotic
fluid embolism, placenta accreta, peripartum hysterectomy, eclampsia or pulmonary embolism).
Results SEIFA quintile was statistically significantly associated with maternal morbidity, with cases being twice as likely as controls to reside in the most disadvantaged areas (adjusted OR 2.00, 95%CI 1.29–3.10). Maternal age [adjusted odds ratio (aOR) 2.20 for women aged 35 or over compared with women aged 25–29, 95% CI 1.64–3.15] and previous pregnancy complications (aOR 1.30, 95%CI 1.21–1.87) were significantly associated with morbidity. A parity of 1 or 2 was protective (aOR 0.58, 95%CI 0.43–0.79), whereas previous caesarean delivery was associated with maternal morbidity (aOR 2.20 for women with one caesarean delivery, 95% CI 1.44–2.85, compared with women with no caesareans). Conclusion The risk of severe maternal morbidity among women
in Australia is significantly increased by social disadvantage. This study suggests that future efforts in improving maternity care provision and maternal outcomes in Australia should include socioeconomic position as an independent risk factor for adverse outcome. Keywords Australia, maternal morbidity, socioeconomic position.
Please cite this paper as: Lindquist A, Noor N, Sullivan E, Knight M. The impact of socioeconomic position on severe maternal morbidity outcomes among women in Australia: a national case–control study. BJOG 2014; DOI: 10.1111/1471-0528.13058.
Introduction Australians generally enjoy high standards of living and have one of the longest life expectancies at birth in the world, at just over 80 years.1 Despite this, important health disparities exist between different groups of Australians, with, for example, the average life expectancy of Indigenous Australians 17 years less than that of non-Indigenous Australians.2 Research conducted in Australia in 2000 looking at health and socioeconomic status suggested that despite substantial improvements in the *These authors contributed equally to this manuscript.
ª 2014 Royal College of Obstetricians and Gynaecologists
health of the Australian population in previous years, inequalities in socioeconomically related mortality had increased for some conditions.3 Although studies in Australia have investigated risk factors for severe maternal morbidity,4 there is no national research into the association between socioeconomic position and severe maternal morbidity. Studies in other developed countries have suggested that socioeconomic position may be a risk factor for poorer outcomes. The most recent Confidential Enquiry into Maternal Deaths in the UK showed that unemployed women and those in routine/ manual occupations were six times more likely to die during pregnancy and labour than women from the highest
1
Lindquist et al.
socioeconomic group.5 A recent study investigated the social gradient of severe maternal morbidity in the UK and the results suggested that independent of ethnicity, BMI and age, low socioeconomic position was a risk factor for severe maternal morbidity.6 Identifying high-risk women is critical for the prevention of adverse outcomes, as it allows targeted interventions and intensive clinical management of specific groups of women. The Australasian Maternal Outcomes Surveillance System (AMOSS) was established for the purpose of investigating risk factors, management and outcomes of different maternal morbidities. Based on the methodology of the UK Obstetric Surveillance System (UKOSS), AMOSS enables nationwide research and surveillance into a range of pregnancy complications across both Australia and New Zealand to be carried out. The aim of this analysis was to explore the independent impact of socioeconomic position on selected severe maternal morbidities among women in Australia.
Methods AMOSS is a national system to study selected severe maternal morbidities in Australia that was established in 2009.7 Negative surveillance of conditions, whereby nil reports are also collected, is conducted monthly across eligible maternity units with >50 births per year, covering about 96% of eligible sites nationally. Where a case is identified, a data collection form is completed, detailing demographic, risk and pregnancy factors, management and outcomes of the condition and pregnancy. Information on control women is also collected for specific studies. We performed a case–control study using data on specific severe maternal morbidities associated with direct maternal death extracted from the AMOSS studies database. We collectively included women who had amniotic fluid embolism, placenta accreta, peripartum hysterectomy, eclampsia or pulmonary embolism as ‘severe maternal morbidity’ cases (n = 623). Women with more than one of the specific morbidities were included only once in the analysis. Definitions of each condition are shown in Box 1. Controls were identified as the two women who had delivered immediately prior to the identified cases of placenta accreta and peripartum hysterectomy in the same hospitals (n = 820). We collected identical data for cases and controls, apart from the details of the management and outcomes of the severe morbidity defining the case. The main exposure variable of socioeconomic status was the Socio-Economic Indexes for Areas (SEIFA) quintile, derived from the maternal postcode of residence. It is created from four socioeconomic indexes and is a measure of socioeconomic
2
disadvantage, with the lowest quintile (1) indicating the most disadvantaged areas8,9 All analyses were performed using R version 2.15.1. Box 1 Definitions of severe maternal morbidities Amniotic fluid embolism: Case defined by either a clinical diagnosis of AFE (acute hypotension or cardiac arrest, acute hypoxia or coagulopathy in the absence of any other potential explanation for the symptoms and signs observed) or a postmortem diagnosis (presence of fetal squames or other debris in the pulmonary circulation). Peripartum hysterectomy: Any woman whose pregnancy terminates and who has a hysterectomy in the same clinical episode or within 6 weeks postpartum when the indication for hysterectomy is related to the pregnancy, e.g. secondary postpartum haemorrhage, or any woman giving birth and undergoing a hysterectomy in the same clinical episode or within 6 weeks postpartum when the indication for hysterectomy is related to the birth, e.g. secondary postpartum haemorrhage Placenta accreta: Any women identified as having placenta accreta (or increta or percreta) diagnosed by antenatal imaging, or diagnosed at operation or diagnosed by pathology specimen Eclampsia: Any woman having convulsions during pregnancy or in the first 10 days postpartum, together with at least two of the following features within 24 hours of the convulsion(s): 1 Hypertension (a booking diastolic pressure of 42 iu/l) Antenatal pulmonary embolism: Cases were defined as any woman in whom 1 Pulmonary embolism was confirmed using suitable imaging (angiography, computed tomography, echocardiography, magnetic resonance imaging or ventilation-perfusion scan showing a high probability of pulmonary embolism). 2 Or pulmonary embolism was confirmed at surgery or post-mortem. 3 Or a clinician made a diagnosis of pulmonary embolism with signs and symptoms consistent with PE present, and the patient received a course of anticoagulation therapy (>1-week duration).
We performed a descriptive analysis to examine the distribution of maternal characteristics and SEIFA quintiles in the case and control groups using chi-squared tests. We performed unconditional logistic regression analysis to investigate the association between severe maternal morbidity and SEIFA quintiles, adjusted for potential confounding factors. Results are presented as odds ratios (OR) with 95% confidence intervals (CI). We included variables that were likely to confound the association between maternal morbidity and socioeconomic
ª 2014 Royal College of Obstetricians and Gynaecologists
Severe maternal morbidity in Australia
status in a multivariable model. Variables included were: maternal age, country of birth, maternal body mass index (BMI), Aboriginal or Torres Strait Islander status, remoteness of residence classified using the Australian Standard Geographical Classification – Remoteness Area,10 parity, smoking, previous medical and pregnancy complications, multiple pregnancy, history of previous caesarean delivery, number of previous caesarean deliveries and insurance status. We used a univariable analysis to determine variables that were to be added in the multivariable model. Maternal age and BMI were continuous variables that were tested for the linearity assumption. BMI exhibited a non-linear association with maternal morbidity. Owing to its non-linearity, we categorised this variable into four groups. Maternal age showed a linear relationship with the outcome variable. We therefore analysed this variable in both continuous and categorical terms, but for ease of interpretation it has been presented as a categorical variable in the tables. Variables that had a high proportion of missing data (>1%, Table 1), such as BMI, smoking and previous pregnancy complications, were considered unlikely to be missing at random; we therefore created a separate category of ‘missing’ for these variables. We built a multivariable model using unconditional logistic regression. We retained variables that had been
Table 1. Distribution of missing data for maternal morbidity cases and controls Variable
SEIFA quintile Country of birth Maternal age Aboriginal or Torres Strait Islander status Remoteness of residence Smoking BMI Parity Multiple pregnancy Previous pregnancy complications Previous medical/obstetric complications History of previous caesarean delivery Number of previous caesarean deliveries Health insurance
Number of cases with missing information (% of total cases)
Number of controls with missing information (% of total controls)
4 (0.6) 6 (1.0) – 32 (5.1)
3 (0.4) 8 (0.1) – 40 (4.9)
2 60 78 – – 39 8
2 70 96 1 2 33 7
(0.3) (9.6) (12.5)
(6.2) (1.2)
(0.2) (8.5) (11.7) (0.12) (0.2) (4.0) (0.9)
8 (1.3) 9 (1.4)
7 (0.8) 7 (0.8)
3 (0.5)
4 (0.5)
ª 2014 Royal College of Obstetricians and Gynaecologists
established as confounders through previous literature (BMI, smoking and remoteness) in the final model. Other potential confounders were added to the simple model containing maternal morbidity status and SEIFA quintiles, and the effect of each variable on the model was assessed using Akaike information criterion (AIC) values and likelihood ratio tests. Variables significantly affecting the fit of the model (P < 0.05) were retained in the final model. We examined all possible interaction terms but none were found to be statistically significant. We assessed the fit of the final multivariable mode using the Hosmer–Lemeshow goodness-of-fit test. Given the 623 cases and 820 controls, and a prevalence of 8.9% for the lowest socioeconomic quintile, our analysis had an 80% power to detect as statistically significant (P < 0.05) odds of 1.36 or greater.
Results Table 2 shows the characteristics of cases and controls. In comparison with controls, cases were more likely to be aged 35 or over, belong to a lower SEIFA quintile, to be an Aboriginal or Torres Strait Islander, have a parity of 3 or more, previous pregnancy complications and a history of previous caesarean delivery. Table 3 shows the morbidities each woman had, including those women who had more than one morbidity. The multivariable model examined the relationship between maternal morbidity and socioeconomic status, while controlling for potential confounders and other risk factors (Table 4). Maternal age was significantly associated with maternal morbidity, with women aged 30–34 being 1.44 (95%CI 1.04–1.99) times more likely and women aged 35 or over being 2.2 (95%CI 1.64–3.15) times more likely to be cases, compared with controls after adjustment. Socioeconomic status was statistically significantly associated with maternal morbidity, with cases being twice (95% CI 1.29–3.10) as likely to belong to the lowest socioeconomic group (SEIFA quintile 1), 1.56 (95%CI 1.02–2.38) times more likely to fall in SEIFA quintile 2 and 1.79 (95% CI 1.26–2.54) times more likely to be in quintile 3, compared with controls. Having a parity of one or two was found to be protective [adjusted odds ratio (aOR) 0.58, 95%CI 0.43–0.79] against maternal morbidity, once adjusted for age and previous caesarean section and pregnancy complications, whereas women who had reported previous pregnancy complications were 1.3 (95%CI 1.21– 1.87) times more likely to be cases compared with controls. The number of previous caesarean deliveries was also statistically significantly associated with maternal morbidity, with women with one caesarean delivery having double the odds of morbidity (95%CI 1.44–2.85), two caesarean deliveries having 4.10 times the odds (95%CI 2.60–6.59) and three or more previous caesarean deliveries having 9.35
3
Lindquist et al.
Table 2. Maternal characteristics associated with maternal morbidity Characteristic
Age
