Open Access Research

Maternal morbidity and mortality from

severe sepsis: a national cohort study
Colleen D Acosta,1 David A Harrison,2 Kathy Rowan,2 D Nuala Lucas,3
Jennifer J Kurinczuk,1 Marian Knight1

To cite: Acosta CD, ABSTRACT

Harrison DA, Rowan K, et al. Strengths and limitations of this study
Objectives: To describe the incidence, characteristics
Maternal morbidity and
and risk factors for critical care admission with severe ▪ The study used comprehensive and nationally
mortality from severe sepsis:
maternal sepsis in the UK. representative data with a large sample size and
a national cohort study. BMJ
Open 2016;6:e012323. Design: National cohort study. hence results have high external validity.
doi:10.1136/bmjopen-2016- Setting: 198 critical care units in the UK. ▪ This study had a robust case definition and
012323 Participants: 646 pregnant and recently pregnant results were validated in sensitivity analyses
women who had severe sepsis within the first using obstetric-specific systemic inflammatory
▸ Prepublication history for 24 hours of admission in 2008–2010. response syndrome criteria.
this paper is available online. Primary and secondary outcome measures: ▪ The data used were from 2008 to 2010, which
To view these files please Septic shock, mortality. may make results less generalisable to current
visit the journal online Results: Of all maternal critical care admissions, practice.
(http://dx.doi.org/10.1136/
14.4% (n=646) had severe sepsis; 10.6% (n=474) had ▪ Comparison of cohort data to aggregated
bmjopen-2016-012323). national data, and low statistical power due to
septic shock. The absolute risk of maternal critical care
admission with severe sepsis was 4.1/10 000 the small number of deaths, precluded multivari-
Received 18 April 2016
maternities. Pneumonia/respiratory infection able risk modelling for the outcomes of morbid-
Revised 13 July 2016
Accepted 14 July 2016 (irrespective of the H1N1 pandemic influenza strain) and ity and mortality.
genital tract infection were the most common sources ▪ Data on the characteristics of non-reporting hos-
of sepsis (40% and 24%, respectively). We identified a pitals were not available; thus, it was not pos-
significant gradient in the risk of severe maternal sepsis sible to conduct a sensitivity analysis to assess
associated with increasing deprivation (RR=6.5; 95% differences between reporting and non-reporting
CI 4.9 to 8.5 most deprived compared with most hospitals.
affluent women). The absolute risk of mortality was 1.8/
100 000 maternities. The most common source of
infection among women who died was pneumonia/ worldwide.1 2 Sepsis is a leading cause of
respiratory infection (41%). Known risk factors for direct maternal death in the UK,3 4 and is a
morbidity supported by this study were: younger age, major cause of direct maternal death in
multiple gestation birth and caesarean section. other high-resource and low-resource coun-
Significant risk factors for mortality in unadjusted
tries.5–8 While the absolute risk of death
analysis were: age ≥35 years (unadjusted OR (uOR)
=3.5; 95% CI 1.1 to 10.6), ≥3 organ system
from maternal sepsis is low in the UK (2.0/
dysfunctions (uOR=12.7; 95% CI 2.9 to 55.1), 100 000 maternities4), a study using the UK
respiratory dysfunction (uOR=6.5; 95% CI1.9 to 21.6), Obstetric Surveillance System (UKOSS),
renal dysfunction (uOR=5.6; 95% CI 2.3 to 13.4) and which captures data from obstetrician-led
haematological dysfunction (uOR=6.5; 95% CI 2.9 to maternity units, found that the magnitude of
14.6). severe morbidity is ∼50 times greater.9 The
1 Conclusions: This study suggests a need to improve rate of maternal critical care admissions with
National Perinatal
Epidemiology Unit (NPEU),
timely recognition of severe respiratory tract and genital severe sepsis in the UK, including women
Nuffield Department of tract infection in the obstetric population. The social admitted from maternity units as well as
Population Health, University gradient associated with the risk of severe sepsis other services among women outside the
of Oxford, Oxford, UK morbidity and mortality raises important questions immediate period of labour and delivery, is
2
Intensive Care National Audit regarding maternal health service provision and usage.
however unknown.
& Research Centre (ICNARC),
London, UK In countries that have relatively low mater-
3
Department of Anaesthesia, nal mortality rates, identification of risk
Northwick Park Hospital, factors for severe morbidity is critical to
Harrow, Middlesex, UK INTRODUCTION target points of intervention before progres-
Correspondence to
Even in an era of modern antibiotics and sion to more serious outcomes. The UKOSS
Professor Marian Knight; advanced medical care, the incidence of study investigated risk factors for severe mor-
marian.knight@npeu.ox.ac.uk sepsis has increased dramatically bidity; however, the study cohort was

Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323 1

Open Access

restricted to maternity units due to the nature of the critical care admissions19 and 7.3% of maternities.20
data collection system.9 Studies of maternal sepsis are Data comprising this study, therefore, were considered
also made more challenging due to the normal physio- representative of the UK maternity population, and
logical changes of pregnancy, which overlap with some results are generalised to that of the UK in discussion.
of the pathophysiological changes of sepsis. Currently,
there has been no national level study of the incidence Study design and case definition
or risk factors for admission to critical care with severe This was an anonymised cohort study of all pregnant
maternal sepsis in the UK. and ‘recently pregnant’ women who were reported in
The objectives of this study were to describe the inci- the CMP from 2008 to 2010, and who were either admit-
dence and characteristics of pregnant or recently preg- ted with or developed severe sepsis within the first
nant women who had severe sepsis within the first 24 hours of admission. Pregnant women were still preg-
24 hours following critical care admission in the UK, nant on admission to critical care. Recently pregnant
and to evaluate the risk factors for severe sepsis morbid- women were defined as having been pregnant within
ity on a national population level, in order to inform 42 days prior to admission to the critical care unit
strategies to improve prevention and outcomes. regardless of how the pregnancy ended (live birth, still-
birth, miscarriage or termination). Mortality was mea-
sured by death at ultimate discharge from acute
METHODS hospital, irrespective of direct or indirect causes. Severe
Data sources sepsis was defined according to a modified version of
This study was conducted using data on women admit- the protein C worldwide evaluation in severe sepsis
ted to critical care units from the Intensive Care (PROWESS) clinical trial definition:21 diagnosis of infec-
National Audit & Research Centre (ICNARC) Case Mix tion as primary or secondary reason for critical care unit
Programme (CMP) database. The CMP is the national admission and at least three systemic inflammatory
clinical audit for adult critical care units (including response syndrome (SIRS) criteria and evidence of at
intensive care and combined intensive care and high least one organ system dysfunction. Septic shock was
dependency units) in England, Wales and Northern defined as severe sepsis with cardiovascular organ system
Ireland, and is coordinated by ICNARC. The CMP data- dysfunction22 (systolic blood pressure <90 mm Hg or
base contains pooled case mix data, collected from the mean arterial pressure <70 mm Hg). Readmissions of
first 24 hours following admission to the critical care women to critical care during the same hospital stay
unit, and outcome data on consecutive admissions to were excluded to avoid double counting.
units participating in the CMP.10 The CMP database has
been independently assessed to be of high quality and Sample size and statistical analyses
regular assessment of data quality is ongoing.11 Data val- The sample size of this study is governed by the rate of
idation procedures and variable definitions have been critical care unit reporting coverage during the study
described in previous studies.10 12 All data used in this period. In 2008, 2009 and 2010, coverage of the database
study were validated and from units that had been was 65.0%, 75.2% and 80.2% of critical care units in
reporting to the CMP for at least 6 months. Support for England, Wales and Northern Ireland reported to the
the collection and use of patient-identifiable data CMP, respectively, with a total of 198 critical care units
without consent was obtained under section 251 of the contributing data during this time. The total number of
National Health Service (NHS) Act 2006 (approval maternal critical care admissions with severe sepsis was
number PIAG 2–10(f )/2005). estimated by extrapolating from the number of severe
National statistics were used for comparison with CMP maternal sepsis admissions observed in the CMP database
data in this study. Data on maternal age, deprivation each year, based on the CMP reporting coverage rate for
(measured using the Index of Multiple Deprivation each year and the total number of adult general critical
(IMD)), multiple births and stillbirths were obtained care units in England, Wales and Northern Ireland. The
from the Office for National Statistics (ONS) for England absolute risk or incidence of maternal critical care admis-
and Wales,13 and the Northern Ireland Statistics and sions with severe sepsis and maternal critical care deaths
Research Agency (NISRA).14 Data on ethnic group were from severe sepsis were estimated by dividing these extra-
obtained from ONS and extrapolated for Northern polated numbers by the total number of maternities
Ireland based on the reported ethnic population distribu- obtained from ONS13 and NISRA.14 Imputation methods
tion in the region.15 Data on mode of delivery were were not used to estimate the counts of severe maternal
obtained from Hospital Episode Statistics,16 StatsWales17 sepsis admissions and deaths from non-reporting hospi-
and extrapolated for Northern Ireland based on pub- tals because characteristics of these hospitals were
lished rates of mode of delivery in the region.18 unavailable and may have differed from reporting hospi-
Scottish critical care units do not participate in the tals (ie, data may not have been missing at random).
CMP and therefore data from Scotland are not included Using observed CMP data (not extrapolated data), fre-
in the study. Scotland accounts for ∼6% of all UK mater- quencies of demographic, clinical and delivery character-
nal critical care admissions, 2% of maternal sepsis istics were tabulated for all cases. Rates were compared

2 Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323

 Open Access

with nationally available data where possible. Using all critical care with severe sepsis if they were aged <20 or
maternities in England, Wales and Northern Ireland as the ≥40 years compared with women aged 25–29 years.
comparison group, the relative risk (RR) with 95% CI for Increased risk of severe sepsis was also significantly and
maternal critical care admission with severe sepsis was cal- progressively associated with lower socioeconomic status
culated for each variable. Characteristics of cohort survi- (figure 2). Of women who delivered by caesarean
vors were compared with those of non-survivors in a section (n=242), 33.1% (n=80) were admitted directly
univariable logistic regression model. from theatre for an emergency indication; 1.7% (n=4)
A sensitivity analysis was finally performed using the were admitted after elective surgery. The RR of admis-
obstetric-specific SIRS criteria defined by Waterstone sion to a critical care unit with severe sepsis for women
et al.23 This was carried out in order to assess whether who had a caesarean section compared with a vaginal
the more rigorous criteria were significantly associated delivery was 6.2 (95% CI 4.9 to 7.8).
with greater severity of sepsis or a different pattern of In assessing whether the rate of caesarean section may
maternal characteristics. Using the Waterstone criteria, have contributed to the social gradient associated with
elevated heart rate is defined as >100 bpm (compared to the risk of severe sepsis (since there was significant
>90) and an elevated white cell count is defined as overlap between the rate of deprivation and caesarean
>17 000 per µL (compared to >1200). section within the cohort), it was found that compared
Stata statistical software V.11 (StataCorp, College with affluent women (IMD quintiles 1–2), deprived
Station, Texas, USA) was used for all analyses. women (IMD quintiles 4–5) had a 1.9 (95% CI 1.2 to
3.1) times higher odds of caesarean section; of all
women in the cohort who had a caesarean section,
RESULTS 64.3% (n=151) were deprived compared with 19.2%
Incidence (n=45) who were affluent.
In the CMP database between 2008 and 2010, there
were 646 pregnant or recently pregnant women who Sources of infection and severity of illness
met the case definition for severe sepsis, which repre- Source of infection could be identified from the
sented 14.4% of maternal critical care unit admissions; primary reason for admission to the critical care unit for
10.6% (n=474) had septic shock. Extrapolating from 598 women (92.6%). Frequencies of the reported
these figures based on the database coverage, 873 source of infection and severity of illness among women
women in England, Wales and Northern Ireland were admitted with severe sepsis are shown in table 2. The
estimated to have had severe sepsis requiring critical most common source of infection was pneumonia/
care over the 3-year period, and 642 were estimated to respiratory infection (n=257; 39.8%). Of these, only 27
have had septic shock. The estimated absolute risk of were identified as laboratory confirmed cases of AH1N1
maternal critical care admission with severe sepsis was influenza; 2009 and 2010 were AH1N1 influenza pan-
4.1/10 000 maternities (95% CI 2.9 to 5.6). Sepsis admis- demic years. There was no significant difference in sever-
sion rates were the highest among women aged 16– ity of illness (Acute Physiology and Chronic Health
19 years (figure 1). One in three pregnant or recently Evaluation (APACHE) II score) between sources of
pregnant women in this age category admitted to critical infection; however, women with pneumonia or respira-
care had severe sepsis. tory infection had a significantly longer critical care unit
length of stay compared with other causes (median
Risk factors for admission to critical care with severe length of stay=4 days, IQR=2–9 days; p<0.001). There was
maternal sepsis a significantly different distribution of sources of infec-
Characteristics of the cohort are listed in table 1. tion between pregnant and recently pregnant women. A
Women had a significantly increased risk of admission to significantly greater proportion of pregnant women had

Figure 1 Rates and 95% CIs of

severe sepsis among maternal
critical care admissions and
maternities by 5-year age bands
from 2008 to 2010. ICU, intensive
care unit.

Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323 3

Open Access

Table 1 Characteristics and relative risks of severe sepsis among pregnant and recently pregnant women
Critical care admissions All maternities Relative risk
Total 646 (100.0) 2 186 818 (95% CI)
Recently pregnant 413 (63.9)
Maternal age, years (mean; SD) 28.3 (6.9) 29.0 (–)
<20 80 (12.4) 129 167 (5.9) 2.5 (1.9 to 3.3)
20–24 133 (20.6) 419 944 (19.2) 1.3 (1.0 to 1.6)
25–29 148 (23.0) 602 716 (27.6) 1
30–34 153 (23.7) 600 694 (27.5) 1.0 (0.8 to 1.3)
35–39 95 (14.7) 352 364 (16.1) 1.1 (0.9 to 1.4)
≥40 36 (5.6) 81 933 (3.7) 1.8 (1.2 to 2.6)
Non-white ethnicity 154 (23.8) 609 250 (27.9) 0.9 (0.8 to 1.0)
Index of multiple deprivation (quintiles)*
1 (least deprived) 63 (10.1) 589 784 (27.6) 1
2 75 (12.0) 475 573 (22.2) 1.5 (1.1 to 2.1)
3 109 (17.5) 395 137 (18.5) 2.6 (1.9 to 3.5)
4 148 (23.8) 346 002 (16.2) 4.0 (3.0 to 5.4)
5 (most deprived) 228 (36.6) 330 916 (15.5) 6.5 (4.9 to 8.5)
BMI, kg/m2 (median; IQR)* (N=312) 26 (22–30)
History of immunosuppression 15 (2.3)
Weeks gestation
Antenatal (median; IQR) 26 (20–31)
Postnatal (median; IQR) 38 (31–41)
Recently pregnant women only
Parity*
0 193 (48.4)
1 96 (24.1)
≥2 110 (27.6)
Assisted conception* 24 (8.1)
Mode of delivery†
Spontaneous vaginal 98 (25.9) 1 334 242 (61.0) 1
Assisted vaginal 28 (7.4) 273 340 (12.5) 1.4 (0.9 to 2.1)
Caesarean section 242 (64.0) 535 999 (24.5) 6.2 (4.9 to 7.8)
Unknown 10 (2.7)
All multiple births (live births and stillbirths) 28 (7.6) 34 663 (1.6) 4.4 (3.1 to 6.3)
Pregnancy outcomes
Live births 321 (77.7)
Stillbirths 47 (11.4) 11 697 (0.5%) 21.3 (16.3 to 27.9)
1st/2nd trimester loss 25 (6.1)
Ectopic pregnancy 10 (2.4)
Other‡ 2 (0.5)
Unknown 10 (2.4)
Hysterectomy* 20 (5.4)
Days since delivery (median; IQR) 3 (0–8)
Bold typeface indicates statistically significant results.
Figures are N (%) unless otherwise stated.
1=Reference Group.
*Of those reported.
†National rates are total deliveries.
‡Two women each had one live birth and one stillbirth from the most recent pregnancy.
BMI, body mass index.

pneumonia ( p<0.001), urinary tract infection/pyelo- maternities (95% CI 1.1 to 2.8). Pneumonia/respira-
nephritis ( p<0.001) or appendicitis ( p=0.04). In con- tory infection was the most common source of sepsis
trast, a significantly greater proportion of recently among women who died (N=12; 41.4%). Demographic
pregnant women had a genital tract infection ( p<0.001), and clinical characteristics of survivors and non-
an infection arising from surgical trauma ( p=0.02) or survivors are presented in table 3. Women aged
septicaemia ( p<0.05). ≥35 years, with ≥3 organ system dysfunctions, and
Of all severe maternal sepsis admissions, 4.6% those with respiratory, renal and/or haematological dys-
(N=29) died. The estimated absolute risk of acute hos- function had significantly higher unadjusted odds of
pital mortality of women admitted was 1.8/100 000 dying of severe sepsis.

4 Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323

 Open Access

admitted to intensive care (1 in 7 and 1 in 9 obstetric

intensive care unit admissions, respectively). The rate of
maternal death from ‘all-cause’ maternal sepsis is sub-
stantially higher than that from genital tract sepsis alone
(1.8/100 000 vs 0.5/100 000 maternities3 4), and similar
to the rate of maternal mortality from all infectious
causes in the UK in 2009–2012 (2.0/100 000 maternities;
95% CI 1.6 to 2.6).4 We further identified several find-
ings with clinical and healthcare policy implications:
pneumonia/respiratory infection is a leading source of
sepsis irrespective of epidemic influenza periods; and
there are major significant disparities in socioeconomic
Figure 2 Relative risks and 95% CIs of severe sepsis status and the risk of severe sepsis.
according to deprivation quintiles (reference group=1st IMD A strength of this study is that it used comprehensive
quintile). IMD, Index of Multiple Deprivation; RR, relative risk. and nationally representative data; this allowed for the
study of risk factors for severe maternal sepsis with data
of higher quality and greater clinical detail than could
Table 2 Characteristics of pregnant and postpartum be achieved using other approaches such as hospital dis-
women on admission to intensive care units in the UK from charge data. The study also used a robust case defin-
2008 to 2010 ition, and results were validated in sensitivity analyses
Total 646 (100.0) using the obstetric-specific SIRS criteria. Owing to its
Source of infection large sample size, the external validity of this study is
Pneumonia (chest infection) 257 (39.8) extremely high. However, there are several limitations in
Genital tract 157 (24.3) this study. We do not have data on organisms presumed
UTI/pyelonephritis 59 (9.1) or confirmed to be the causes of sepsis. Comparison of
Surgical trauma 24 (3.7) cohort data to aggregated national data, and low statis-
Septicaemia 20 (3.1) tical power due to the small number of deaths, pre-
Appendicitis 19 (2.9) cluded multivariable risk modelling for the outcomes of
Other infection 62 (9.6) morbidity and mortality. All cases of maternal sepsis
Unknown 48 (7.4) occurring in high dependency units within obstetric set-
Number of organ system dysfunctions
tings will not necessarily be captured within the
1 222 (34.4)
ICNARC data. It was also not possible to conduct a sensi-
2 234 (36.3)
≥3 189 (29.3) tivity analysis to assess the characteristics of non-
Organ system dysfunctions* reporting hospitals as these data were not available.
Cardiovascular 475 (73.5) Although we generalise the results in this study to the
Respiratory 381 (59.0) entire UK, we cannot rule out that rates and risk factors
Metabolic acidosis 345 (53.4) of severe sepsis may differ for Scotland. However, since
Renal 50 (7.7) Scotland accounts for a small proportion of total and
Haematological 65 (10.1) sepsis-specific critical care admissions, it is unlikely that
ICNARC physiology score (median; IQR) 14 (10–20) inclusion of data from Scotland would significantly affect
APACHE II score (median; IQR) 12 (10–16) the results.
Days in ICU (median; IQR) 2.8 (1.3–5.7)
Forty per cent of women with severe sepsis had pneu-
Deaths 29 (4.6)
monia/respiratory infection as the source of sepsis. This
Figures are N (%) unless otherwise stated.
*Organ system dysfunctions are not mutually exclusive. finding supports that of the recent UK and Ireland
ICNARC, Intensive Care National Audit & Research Centre; Confidential Enquiries into Maternal Deaths and
ICU, intensive care unit; UTI, urinary tract infection. Morbidity 2009–2012, which found that 54% of all
maternal sepsis deaths were caused by influenza (N=36;
Sensitivity analysis using obstetric-specific SIRS criteria 43%) or pneumococcal disease (N=9; 11%).4 In the
In a sensitivity analysis using the obstetric-specific SIRS present study, women with pneumonia/respiratory infec-
criteria defined by Waterstone et al,23 77 (11.9%) tion had a significantly longer length of critical care unit
women did not meet the more rigorous criteria; stay compared with all other causes, and the absolute
however, there were no significant changes in the distri- risk of maternal mortality was the largest compared with
bution of case characteristics or sources of infection. all other causes, comprising 41% of women who died
from severe sepsis. Despite the significant influenza epi-
demic, which occurred from 2009 to 2010, only 10.5%
DISCUSSION (27/257) of pneumonia/respiratory infection cases were
Severe sepsis and septic shock morbidity are common identified as being due to the pandemic AH1N1 strain,
among pregnant and recently pregnant women although reporting is likely to be incomplete. Incidence

Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323 5

Open Access

Table 3 Characteristics of survivors and non-survivors of severe obstetric sepsis following critical care admission
Severe sepsis survivors Severe sepsis deaths Unadjusted OR Adjusted OR
n=610 n=29 (95% CI) (95% CI)
Recently pregnant 387 (63.4) 22 (75.9) 1.8 (0.76 to 4.3) 1.1 (0.42 to 3.0)
Maternal age, years
<25 234 (38.4) 5 (17.2) 1* 1*
25–34 254 (41.7) 15 (51.7) 2.8 (0.99 to 7.7) 2.2 (0.71 to 7.0)
≥35 121 (19.9) 9 (31.0) 3.5 (1.1 to 10.6) 3.3 (0.94 to 11.2)
Non-white ethnicity 147 (24.1) 7 (24.1) 0.98 (0.39 to 2.5) 0.59 (0.21 to 1.6)
Deprivation (IMD quintiles 4 and 5) 354 (58.1) 17 (58.6) 1.02 (0.48 to 2.2) 2.6 (1.03 to 6.7)
BMI, kg/m2
Unknown 317 (52.1) 13 (44.8) 0.78 (0.17 to 3.7) 1.2 (0.15 to 9.1)
<25 126 (20.7) 3 (10.3) 1* 1*
≥25<30 90 (14.8) 7 (24.1) 3.3 (0.8 to 13.3) 5.2 (1.4 to 18.9)
≥30 76 (12.5) 6 (20.7) 3.5 (0.9 to 14.6) 6.3 (1.5 to 27.0)
History of immunosuppression 13 (2.1) 2 (6.9) 3.3 (0.73 to 15.7)
Weeks gestation
Antenatal
≥37 16 (7.0) 0 (0.0) –
25–36 104 (47.5) 3 (42.9) 1*
<25 99 (45.2) 4 (57.1) 1.4 (0.31 to 6.4)
Postnatal
≥37 202 (54.9) 14 (66.7) 1*
25–36 114 (31.0) 3 (14.3) 0.38 (0.11 to 1.3)
<25 52 (14.1) 4 (19.1) 1.1 (0.35 to 3.5)
Recently pregnant women only
Parity
0 183 (48.9) 9 (42.9) 1*
1 88 (23.5) 7 (33.3) 1.6 (0.58 to 4.5)
≥2 103 (27.5) 5 (23.8) 1.0 (0.32 to 3.0)
Assisted conception 24 (8.7) 0 (0.0) –
Mode of delivery
Spontaneous vaginal 89 (23.6) 8 (36.4) 1*
Assisted vaginal 28 (7.4) 0 (0.0) –
Caesarean section 227 (60.2) 12 (54.6) 0.59 (0.23 to 1.5)
Termination 23 (6.1) 2 (9.1) 0.97 (0.19 to 4.9)
Ectopic 10 (2.7) 0 (0.0) –
All multiple births 24 (7.0) 3 (15.0) 2.3 (0.64 to 8.6)
Stillbirth(s) 42 (21.9) 4 (33.3) 1.8 (0.5 to 6.2)
Hysterectomy 16 (4.6) 3 (15.0) 3.6 (0.96 to 13.7)
<24 hours since delivery 104 (27.6) 8 (36.4) 1.5 (0.61 to 3.7)
Source of infection†
Pneumonia (chest infection) 216 (35.4) 12 (41.4) 1.3 (0.60 to 2.7)
Intrauterine infection 69 (11.3) 2 (6.9) 0.58 (0.13 to 2.5)
Pelvic infection 47 (7.7) 0 (0.0) –
UTI/pyelonephritis 43 (7.1) 0 (0.0) –
Septicaemia‡ 17 (2.8) 2 (6.9) 2.6 (0.57-11.7)
Number of organ system dysfunctions
1 221 (36.2) 2 (6.9) 1*
2 224 (36.7) 8 (27.6) 3.9 (0.83 to 18.7)
≥3 165 (27.1) 19 (65.5) 12.7 (2.9 to 55.1)
Organ system dysfunction§
Cardiovascular 444 (72.8) 24 (82.8) 1.8 (0.68 to 4.8)
Respiratory 349 (57.2) 26 (89.7) 6.5 (1.9 to 21.6) 8.1 (1.8 to 36.0)
Metabolic acidosis 322 (52.8) 18 (62.1) 1.5 (0.68 to 3.1)
Renal 39 (6.4) 8 (27.6) 5.6 (2.3 to 13.4) 2.9 (0.94 to 9.3)
Haematological 52 (8.5) 11 (37.9) 6.5 (2.9 to 14.6) 5.7 (2.0 to 16.0)
Bold typeface indicates statistically significant results.
Figures are numbers (%) of women.‘–’ indicates OR estimate not possible due to zero incidence in either the case or control group.
*Reference group.
†Reasons for admission with a >5% frequency.
‡Admitted with or onset during critical care admission.
§Organ system dysfunctions not mutually exclusive.
BMI, body mass index; IMD, Index of Multiple Deprivation; UTI, urinary tract infection.

6 Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323

 Open Access

of primary pneumonia (bacterial and viral) is known to women in the population. In a cohort study of all
increase during seasonal and epidemic influenza women delivering in English hospitals, Alves et al27
periods,24 which is reflected in the increased proportion found that the most affluent women had a significantly
of severe sepsis from respiratory infection from 2009 to higher odds of caesarean section compared with women
2010. However, in the pre-epidemic year of 2008, respira- who were most deprived. Implications of our finding,
tory infection was also the largest source of severe mater- therefore, are that poorer women in general in the UK,
nal sepsis. and potentially poorer women who have a caesarean
Between 2011 and 2012, it was found that the largest section, are at increased risk of severe sepsis. It is import-
proportion of severe sepsis cases (31%) among women ant to note that the temporality of infection, whether
in UK maternity units was due to genital tract infection.9 the infection occurred after the caesarean section or if
The results of this study indicate that in addition to the woman had a caesarean section as a result of ante-
genital tract infection, respiratory infection is a major partum infectious morbidity, could not be determined
source of severe maternal sepsis irrespective of an influ- from these data. Each scenario has significant causal
enza epidemic. This should be considered in the implications. This is particularly relevant given that a
context that pregnant and peripartum women are at social gradient in the risk of severe maternal sepsis
greater risk of developing respiratory infection25 and an appears to be consistent with an overall social gradient
increased incidence of invasive streptococcal infections in health outcomes that exists in the UK and other
(caused by Streptococcus pneumoniae and Streptococcus pyo- advanced-economy countries.28 While the mechanism
genes (group A streptococcus)), occurring through behind this finding is unknown, women in more
community-acquired respiratory transmission, has also deprived geographic areas are known to have higher
been recently recorded in the UK.26 Immediate implica- rates of poor underlying health28 and decreased uptake
tions are that in addition to precautions for genital tract and continuity of maternity care.29 There is a need for
sepsis, there is clearly a precedent to improve prevention further work to investigate causative and potentially
and timely recognition of severe respiratory tract infec- modifiable factors relating to maternal sepsis among the
tion in pregnant and recently pregnant women.4 Most most deprived.
maternal deaths and critical illnesses from severe sepsis Known risk factors for maternal sepsis morbidity sup-
occur due to delay in recognition and diagnosis.3 ported by this study are: younger30 and older age,6 mul-
Obstetric and front-line clinicians should therefore tiple gestation birth6 and caesarean section.6 23 30
maintain a high index of suspicion, as well as alert preg- Additionally, we interpret the significantly higher rate of
nant and recently pregnant women to the possible sever- stillbirth in this cohort to be primarily an outcome of
ity of any infection, in particular the clinical symptoms severe maternal infection. Some of these effects, such as
of respiratory and genital tract infection and the the observed association with younger maternal age,
dangers of delay in seeking medical care. The recent may be mediated through other factors such as depriv-
report from the Mothers and Babies: Reducing Risk ation. However, although the younger women had a
through Audit and Confidential Enquiries in the UK higher rate of admission with sepsis, it is important to
(MBRRACE-UK) Confidential Enquiry into Maternal note that they had the lowest morality rate. Significant
Deaths4 highlighted the following as symptoms and signs unadjusted physiological risk factors for severe maternal
for women to be aware of: high temperature, chills and sepsis mortality are indicative of the continuum of sepsis
shivering, fast heartbeat, fast breathing, breathlessness, severity, eventually leading to multiple organ failure and
headache, severe abdominal pain and extreme sleepi- death.31 The risk factor of age ≥35 years should be eval-
ness. From an economic perspective, an improvement in uated in an adjusted model for mortality.
recognition before onset of severe infection would have
a substantial effect on intensive care resource usage.
The Confidential Enquiry also highlighted the import- CONCLUSIONS
ance of immunisation against influenza as a means of This study shows that the burden of severe sepsis on
preventing respiratory infection in pregnant women,4 pregnant and peripartum women admitted to critical
and this has been emphasised in public health care is significant. The finding that pneumonia/respira-
campaigns. tory infection caused the largest proportion of severe
In addition to the risk of severe maternal sepsis asso- sepsis cases and maternal sepsis deaths, irrespective of
ciated with respiratory infection, there was a clear gradi- epidemic influenza, indicates a critical need to improve
ent in the risk of severe sepsis associated with decreasing timely recognition of severe respiratory tract infection,
socioeconomic status. This represents a striking example in addition to genital tract infection, in the obstetric
of health inequities within a high-income country. population. Further studies should elucidate the respira-
Importantly, a significant proportion of women who had tory causative organisms to which this population is par-
a caesarean section (a well-established risk factor for ticularly susceptible. The social determinants that
severe sepsis) were also of low socioeconomic status potentially play a role in prepartum and postpartum
(64.3%). It is unlikely, however, that rates of caesarean maternal sepsis mortality must also be clarified using
section were simply higher among more deprived multivariable risk modelling in order to redress

Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323 7

 Open Access

inequalities in maternal health. The clear social gradient 8. Bauer ME, Bateman BT, Bauer ST, et al. Maternal sepsis mortality
and morbidity during hospitalization for delivery: temporal trends and
associated with risk of severe sepsis morbidity and mor- independent associations for severe sepsis. Anesth Analg
tality raises important questions regarding maternal 2013;117:944–50.
health service provision and usage, and may need to be 9. Acosta CD, Kurinczuk JJ, Lucas DN, et al. Severe maternal sepsis
in the UK, 2011–2012: a national case-control study. PLoS Med
addressed at the health policy level. 2014;11:e1001672.
10. Harrison DA, Brady AR, Rowan K. Case mix, outcome and length of
Acknowledgements The authors thank Susan M Sellers and Lisa Hinton for a stay for admissions to adult, general critical care units in England,
critical review of the manuscript. Wales and Northern Ireland: the Intensive Care National Audit &
Research Centre Case Mix Programme Database. Crit Care 2004;8:
Contributors CDA was involved with the study design, carried out the data R99–111.
interpretation and analysis and wrote the manuscript. DAH and KR 11. ICNARC. Quarterly Quality Reports, 2015. https://www.icnarc.org/
contributed to the data interpretation, analysis and writing of the manuscript. Our-Audit/Audits/Cmp/Reports/Quarterly-Quality-Report (accessed
DNL provided clinical input into the study and contributed to the writing of 16 Aug 2015).
12. Harrison DA, Welch CA, Eddleston JM. The epidemiology of severe
the manuscript. JJK and MK designed the study and supervised the analysis,
sepsis in England, Wales and Northern Ireland, 1996 to 2004:
writing and final review of the manuscript. secondary analysis of a high quality clinical database, the ICNARC
Funding This article presents independent research funded by the National Case Mix Programme Database. Crit Care 2006;10:R42.
13. Office for National Statistics. Maternities. Office for National
Institute for Health Research (NIHR) under the ‘Beyond maternal death: Statistics, 2012. http://www.ons.gov.uk/ons/taxonomy/index.html?
improving the quality of maternity care through national studies of ‘near-miss’ nscl=Maternities (accessed 1 Feb 2015).
maternal morbidity’ programme (programme grant RP-PG-0608-10038). Marian 14. Northern Ireland Statistics and Research Agency, editor. Births.
Knight is funded by an NIHR Research Professorship (NIHR-RP-011-032). 2012. http://www.nisra.gov.uk/demography/default.asp8.htm
(accessed 16 Aug 2015).
Disclaimer The views expressed in this publication are those of the author(s), 15. Northern Ireland Statistics and Research Agency. Census 2001
and not necessarily those of the NHS, the NIHR or the Department of Health. The results. NISRA, 2012. http://www.nisra.gov.uk/Census/agreement.
funder played no role in the study design; in the collection, analysis, and html (accessed 16 Aug 2015).
interpretation of data; in the writing of the report; and in the decision to submit 16. Hospital Episode Statistics. Hospital episode statistics: maternity
data. HES, 2012. http://www.hesonline.nhs.uk/ (accessed 16 Aug
the article for publication. All researchers are independent from the study funders. 2015).
Competing interests None declared. 17. StatsWales. Method of delivery. StatsWales, 2012. http://www.
statswales.wales.gov.uk/TableViewer/tableView.aspx?
Ethics approval Research ethics committee approval for use of anonymised ReportId=2021 (accessed 22 Jul 2015).
data was not required. Approval of the research protocol was obtained from 18. Black LA. Incidence of caesarean sections. Northern Ireland
Assembly, 2011. http://www.niassembly.gov.uk/Documents/RaISe/
the Intensive Care National Audit & Research Centre (ICNARC) prior to data
Publications/2011/Health/14711.pdf (accessed 16 Aug 2015).
extraction. 19. Scotland N. Scottish Confidential Audit of Severe Maternal Morbidity.
Provenance and peer review Not commissioned; externally peer reviewed. 6th Annual Report. Edinburgh: NHS QIS. 2010. http://www.
healthcareimprovementscotland.org/programmes/reproductive,
Data sharing statement ICNARC is committed to ensuring that data collected _maternal__child/programme_resources/scasmm.aspx (accessed 16
through the CMP are used as widely as possible by clinicians, Aug 2015).
20. Information Services Division. Maternity and births. ISD Scotland,
commissioners, providers, researchers and anyone with an interest in adult 2012. http://www.isdscotland.org/Health-Topics/Maternity-and-Births/
critical care. All requests for use of the data, including the data used for this Publications/data-tables.asp (accessed Aug 2015).
study, are considered by and subject to approval from the independent Data 21. Padkin A, Goldfrad C, Brady AR, et al. Epidemiology of severe
Access Advisory Group (DAAG). sepsis occurring in the first 24 hrs in intensive care units in England,
Wales, and Northern Ireland. Crit Care Med 2003;31:2332–8.
Open Access This is an Open Access article distributed in accordance with 22. Parker MM, Shelhamer JH, Bacharach SL, et al. Profound but
the terms of the Creative Commons Attribution (CC BY 4.0) license, which reversible myocardial depression in patients with septic shock.
permits others to distribute, remix, adapt and build upon this work, for Ann Intern Med 1984;100:483–90.
commercial use, provided the original work is properly cited. See: http:// 23. Waterstone M, Bewley S, Wolfe C. Incidence and predictors of
severe obstetric morbidity: case-control study. BMJ
creativecommons.org/licenses/by/4.0/ 2001;322:1089–93.
24. Morens DM, Taubenberger JK, Fauci AS. Predominant role of
REFERENCES bacterial pneumonia as a cause of death in pandemic influenza:
1. Senior K. In the dark about sepsis. Lancet Infect Dis 2012;12:751–2. implications for pandemic influenza preparedness. J Infect Dis
2. Martin GS. Sepsis, severe sepsis and septic shock: changes in 2008;198:962–70.
incidence, pathogens and outcomes. Expert Rev Anti Infect Ther 25. Poulakou G, Pérez M, Rello J. Severe acute respiratory infections in
2012;10:701–6. the postpandemic era of H1N1. Curr Opin Crit Care
3. Cantwell R, Clutton-Brock T, Cooper G, et al. Saving Mothers’ Lives: 2012;18:441–50.
reviewing maternal deaths to make motherhood safer: 2006–2008. 26. Zakikhany K, Degail MA, Lamagni T, et al. Increase in invasive
The Eighth Report of the Confidential Enquiries into Maternal Deaths Streptococcus pyogenes and Streptococcus pneumoniae infections
in the United Kingdom. BJOG 2011;118(Suppl 1):1–203. in England, December 2010 to January 2011. Euro Surveill 2011;16:
4. Knight M, Kenyon S, Brocklehurst P, et al. eds. Saving Lives, pii:19785.
Improving Mothers’ Care—Lessons learned to inform future 27. Alves B, Sheikh A. Investigating the relationship between affluence
maternity care from the UK and Ireland Confidential Enquiries into and elective caesarean sections. BJOG 2005;112:994–6.
Maternal Deaths and Morbidity 2009–12. Oxford: National Perinatal 28. Marmot M, Bell R. Fair society, healthy lives. Public Health 2012;126
Epidemiology Unit (NPEU), University of Oxford, 2014. (Suppl 1):S4–10.
5. Sullivan EA, Hall B, King JF. Maternal deaths in Australia 2003– 29. Hemingway H, Saunders D, Parsons L. Social class, spoken
2005. AIHW National Perinatal Statistics Unit: 1–51. 2008. http:// language and pattern of care as determinants of continuity of carer
www.preru.unsw.edu.au/PRERUWeb.nsf/resources/MD3/$file/md3a. in maternity services in east London. J Public Health Med
pdf (accessed 16 Aug 2015). 1997;19:156–61.
6. Kramer HMC, Schutte JM, Zwart JJ, et al. Maternal mortality and 30. Acosta CD, Bhattacharya S, Tuffnell D, et al. Maternal sepsis: a
severe morbidity from sepsis in the Netherlands. Acta Obstet Scottish population-based case-control study. BJOG
Gynecol Scand 2009;88:647–53. 2012;119:474–83.
7. Saucedo M, Deneux-Tharaux C. Épidémiologie des morts 31. Geller SE, Rosenberg D, Cox SM, et al. The continuum of maternal
maternelles en France, 2001–2006. Bull Epidémiol Hebd 2010;no. morbidity and mortality: factors associated with severity.
2–3:10–14. Am J Obstet Gynecol 2004;191:939–44.

8 Acosta CD, et al. BMJ Open 2016;6:e012323. doi:10.1136/bmjopen-2016-012323