Anaesthesia 2012, 67, 1218 doi:10.1111/j.1365-2044.2011.06896.

Original Article
A validation study of the CEMACH recommended modified early obstetric
warning system (MEOWS)*
S. Singh,1 A. McGlennan,2 A. England2 and R. Simons2

1 Consultant Anaesthetist, Barnet Hospital, Herts, UK. 2 Consultant Anaesthetist, Royal Free Hospital, London, UK

Summary
The 20032005 Condential Enquiry into Maternal and Child Health report recommended the introduction of the modied
early obstetric warning system (MEOWS) in all obstetric inpatients to track maternal physiological parameters, and to aid early
recognition and treatment of the acutely unwell parturient. We prospectively reviewed 676 consecutive obstetric admissions,
looking at their completed MEOWS charts for triggers and their notes for evidence of morbidity. Two hundred patients (30%)
triggered and 86 patients (13%) had morbidity according to our criteria, including haemorrhage (43%), hypertensive disease of
pregnancy (31%) and suspected infection (20%). The MEOWS was 89% sensitive (95% CI 8195%), 79% specic (95% CI 76
82%), with a positive predictive value 39% (95% CI 3246%) and a negative predictive value of 98% (95% CI 9699%). There
were no admissions to the intensive care unit, cardio respiratory arrests or deaths during the study period. This study suggests
that MEOWS is a useful bedside tool for predicting morbidity. Adjustment of the trigger parameters may improve positive
predictive value.
. .........................................................................................................................................................................
Correspondence to: A. McGlennan
Email: alan.mcglennan@royalfree.nhs.uk
*Presented in part at the Obstetric Anaesthetists Association Annual Meeting, May 2010, Newcastle.
Accepted: 28 May 2011

The development of early warning systems from simple
bedside observation charts arose from the knowledge that
physiological abnormalities precede critical illness [1, 2]. A
variety of early warning systems, developed in the non-
obstetric adult patient population, assign weighted values to
a number of physiological parameters according to their
degree of deviation from the normal. When the measure-
ment reaches a dened threshold, a mandatory action is
initiated to expedite further assessment of the patient by a
suitably qualied clinician [3, 4]. It is thought that early
intervention will result in improved patient outcome.
A number of non-obstetric adult studies have attempted
to validate early warning systems with variable success.
Whilst sensitivity has been shown to be generally acceptable,
positive predictive values are consistently low when death or
ICU admission is applied as the primary outcome measure
[4]. As ICU admissions, cardiorespiratory arrests and death
are rare in the obstetric population in developed countries
[5, 6], obstetric morbidity has become increasingly impor-
tant as an outcome measure.
The 20032005 triennial Condential Enquiry into
Maternal and Child Health (CEMACH) report recommended
the routine use of the modied early obstetric warning system
(MEOWS), which is an early warning system adapted for the
obstetric population [5]. Although no previous studies have
validated such charts, a number of hospitals in the UK already
use them in their obstetric units [7].
The aim of our study was to evaluate the MEOWS as a
tool for predicting maternal morbidity, by measuring its
sensitivity, specicity and predictive value.
Methods
Ethical approval was obtained from the local ethics
committee. Monitoring of parturients using MEOWS chart
is an established aspect of patient care at our hospital. This
chart was adapted from the seventh CEMACH report
(20032005) [5]. Midwives and clinicians were made aware
of the importance of accurately charting patients parame-
ters, the need for mandatory callout and review by a clinican
following a trigger. Both midwives and clinicians were

12 Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland

Singh et al. | A validation study of MEOWS Anaesthesia 2012, 67, 1218

informed that compliance would be audited. However, to blood sampling where applicable), conscious level (AVPU:
minimise observer bias, they were not briefed about the Alert, responds to Voice or Pain, and Unresponsive) and
proposed study. Before the collection of data, compliance of pain scores (0 = no pain, 1 = slight pain on movement,
the MEOWS chart was audited daily until a target of 100% 2 = intermittent pain at rest moderate pain on movement)
was achieved. A weekly audit was subsequently carried out were documented at least every 12 h (Appendix 1, Table 1).
to ensure that standards were maintained. A trigger was dened as a single markedly abnormal
Over a period of 2 months, all women between observation (red trigger), or the combination of two
20 weeks gestation and 6 weeks postpartum, who were simultaneous mildly abnormal observations (two yellow
admitted as an inpatient to the maternity unit, were triggers). A trigger prompted urgent medical assessment
included in the study. (Appendix 2). Outcome at 30 days (morbidity, death,
Measurement of temperature (oral), blood pressure and intensive care unit (ICU) admission, discharged alive) was
heart rate (DINAMAP, Critikon, Tampa, FL, USA), respi- retrieved from the hospital record system and notes. The
ratory rate, oxygen saturation (pulse oximeter and arterial denitions of maternal morbidity were agreed jointly by the
authors at the beginning of the study (Table 2).
The data were entered onto a password-protected,
Table 1. Limits of trigger thresholds for MEOWS parame- encrypted spreadsheet. Age was analysed with ANOVA and
ters (see Appendix 1). length of stay was analysed using the MannWhitney test.
The chi-squared test was used to compare the development
Red trigger Yellow trigger
of morbidity and the presence of emergency intervention in
Temperature; C < 35 or > 38 3536
Systolic BP; mmHg < 90 or > 160 150160 women who triggered, compared with those who did not. In
or 90100 all statistical tests, we regarded a value of p < 0.05 as
Diastolic BP; mmHg > 100 90100
Heart rate; < 40 or > 120 100120 statistically signicant. We also tested for association
beats.min)1 or 4050 between an abnormal parameter and presence of morbidity
Respiratory rate; < 10 or > 30 2130
by calculating relative risk for individual parameters. All
breaths.min)1
Oxygen saturation; < 95 statistical analysis was carried out using GraphPad Prism
% version 5.00 for Windows (GraphPad Software, San Diego,
Pain score 23
Neurological Unresponsive, Voice CA, USA, http://www.graphpad.com), and was veried by
response pain an independent statistician. Sensitivity, specicity, positive
predictive values and negative predictive values were
BP, blood pressure
calculated for the sample.

Table 2. Diagnostic denitions of obstetric morbidity.

Obstetric morbidity Diagnostic criteria

Obstetric haemorrhage [9]
Pre-eclampsia [10]
Suspected infection [11]
Pulmonary embolus [12]
Cerebral venous sinus
thrombosis
Intracranial bleed
Acute asthma [13]
Status epilepticus [13]
Diabetic ketoacidosis [13]
Myocardial infarction [13]
Pulmonary oedema [13]
Anaesthetic complications [8]
Other diagnosis
Documented estimated blood loss > 1500 ml, drop in haemoglobin concentration 3 g. dl)1,
or need for blood transfusion
BP systolic 160 mmHg, or diastolic 110 mmHg plus proteinuria 0.3 g.day)1 (+ 2 dipstick) or
hypertension ( 140 90 mmHg) and proteinuria with at least one of the following: headache;
visual disturbance; epigastric pain; clonus; platelet count 100 109.l)1;
AST > 50 iu.l)1; Cr > 100 lmol.l)1; or CrCl < 80.8 ml.min)1
Clinically suspected focus of infection positive laboratory culture, treated with antibiotics
(excluding commensals and antibiotic prophylaxis)
CTPA (pulmonary angiography), or V Q scan with high probability
CT MRI confirmed
CT MRI confirmed
History of asthma and audible expiratory wheeze, with reduced PEFR
History of epilepsy, prolonged multiple seizures
Hyperglycaemia, metabolic acidosis, ketone in urine
Symptoms with increased serum troponin and new ECG changes
Breathlessness, crepitations, requiring diuretics
High spinal epidural, aspiration after difficult or failed intubation

BP, blood pressure; AST, aspartate aminotransferase; Cr, creatinine; CrCL, creatinine clearance; CTPA, computed tomography pulmonary
angiogram; V/Q, ventilation/perfusion; MRI, magnetic resonance imaging; PEFR, peak expiratory ow rate; ECG, electrocardiography.

Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland 13

Anaesthesia 2012, 67, 1218 Singh et al. | A validation study of MEOWS

43% Table 3. Risk of developing morbidity in the presence of an

abnormal parameter. Values are relative risk (95% CI).
31%
Relative risk
20% Parameter of morbidity p value
Heart rate > 100 7.0 (4.910.1) 0.0001
beats.min)1
4% 2% Diastolic BP > 90 mmHg 6.6 (4.79.4) 0.0001
Haemorrhage PET Infection Anaesthetic Acute asthma
Systolic BP > 150 mmHg 5.4 (3.87.8) 0.0001
Respiratory rate > 22 4.8 (2.98.0) 0.0001
Figure 1 Distribution of maternal morbidity according breaths.min)1
Temperature > 38 C 3.4 (2.05.6) 0.0003
to dened diagnostic criteria. Values are proportion.
Systolic BP < 90 mmHg 2.4 (1.53.7) 0.0013
PET, pre-eclampsia. Oxygen saturation < 95% 1.3 (0.27.9) 0.56
Pain score 23 2.7 (0.88.4) 0.17
Responds to voice, pain 0.0 1.0
Results or unresponsive
The MEOWS charts were completed for 673 of 676 patients.
The charts of the remaining three patients were missing
from the notes at the time of analysis. These were
reconstructed from observations documented in the notes presence denitively. For a screening tool to be of value, it
and were included in the nal analysis. must be cost effective, safe and validated. The validity of an
Two hundred (30%) women triggered and the call-out early warning system is assessed by its sensitivity, specicity
algorithm was initiated. Eighty-six (13%) women tted our and overall accuracy. This is the rst study attempting to
criteria for morbidity. The mean (SD) age of women in the validate an obstetric early warning chart.
trigger and non-trigger groups was 30 (6) and 31 (5) years, Sensitivity addresses the question what proportion of
respectively. Overall, women who triggered were more likely patients with dened morbidity had triggered the MEOWS
either to have or to develop morbidity than those who did chart? The MEOWS demonstrated a much higher
not (39% vs < 1%, respectively, p < 0.0001). They were also sensitivity than non-obstetric early warning systems that
more likely to have undergone an emergency obstetric are currently used in the adult population. We attribute this
intervention (caesarean section, ventouse or forceps deliv- to the use of morbidity as our primary end point, rather
ery): 46% vs 16%, respectively, p < 0.0001. The median (IQR than death or ICU admissions; both are rare occurrences in
[range]) hospital stay in women who triggered was 3 (14 obstetric patients. In addition, MEOWS applies triggering
[127]) days vs 1 (1-2 [121]) days in the women who did criteria that are recognised as early signs of morbidity.
not trigger (p < 0.0001). Previously, both have been cited as reasons for the failure of
The most common morbidity was haemorrhage (43%), an early warning system to demonstrate sensitivity in the
followed by hypertensive disease of pregnancy (31%) and adult population. Lowering the threshold at which morbi-
suspected infection (20%, Fig. 1). There were no ICU dity is dened, or lowering the threshold at which patients
admissions, cardiorespiratory arrests or deaths. trigger, may increase sensitivity further. This would reduce
The most frequent trigger was high blood pressure the numbers of false negatives. However, it would also have
(42%), followed by tachycardia (28%) and low blood the effect of increasing the number of false positives, thereby
pressure (18%). Temperature, respiratory rate and oxygen decreasing specicity.
saturation were the least frequent triggers (6%, 4% and 2%, Good specicity is desirable in an early warning system
respectively). The relative risk of a patients developing to minimise unnecessary cost and workload, as well as
morbidity for individual parameters triggered is shown in minimising the emotional burden placed on the patient by
Table 3. preventing unnecessary investigations. The specicity of
The overall sensitivity of MEOWS in predicting mor- MEOWS is reasonable and comparable to other adult early
bidity was 89% (95% CI 8195%), specicity 79% (95% CI warning systems, though there is scope for further rene-
7682%), positive predictive value 39% (95% CI 3246%) ment. Our study showed that some triggered parameters
and negative predictive value 98% (95% CI 9699%). performed better than others. High blood pressure, tachy-
cardia, tachypnoea and pyrexia were better indicators of
Discussion maternal morbidity, whilst hypotension was not as sensitive
The MEOWS is a simple bedside screening tool for maternal an indicator as we expected for our morbidity criteria. We
morbidity. Screening identies individuals likely to have hypothesise that the trigger threshold for low blood pressure
morbidity, whilst a diagnostic test seeks to conrm its is incorrectly set on the MEOWS chart for the obstetric

14 Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland

Singh et al. | A validation study of MEOWS
population, and increasing its threshold will result in
signicantly lower rates of false positives and improved
specicity. The optimal trigger threshold for low blood
pressure will be the subject of a future study.
The ideal MEOWS chart would have a sensitivity and
specicity close to 100%, such that most, if not all, triggered
patients are correctly identied as having morbidity, and
very few have misleading triggers. In practice, this is rarely
the case, and striking a suitable balance between sensitivity
and specicity is necessary. In a maternal early warning
system, whilst a false positive may place a burden on
resources and create unnecessary anxiety, a false negative
could have catastrophic consequences for the patient.
Therefore, a system that favours sensitivity over specicity
would be more prudent.
Positive predictive value addresses the question, how
likely is it that the patient has morbidity given that there is
a trigger? Positive and negative predictive values indicate
the accuracy of the MEOWS chart. The usefulness of
accuracy is limited by the prevalence of morbidity in a
population. The higher the prevalence of morbidity, the
greater the accuracy. The prevalence of morbidity can
change with time and population type, and therefore the use
of sensitivity and specicity is preferable to determine
validity. Our predictive values reect a low prevalence of
maternal morbidity. A number of false positives were due to
single triggers that were not reproduced, and we have re-
appraised midwives of the need to ensure that parameters
are measured accurately, and a single unexplained trigger
repeated before callout.
The promising results of this validation study have
encouraged the midwives and obstetricians in our institution
to continue implementing the MEOWS chart in every
obstetric patient. To minimise workload and unnecessary
duplication of documentation, the chart has been incorpo-
rated into the maternal notes. Although some centres
advocate the use of early warning charts in high-risk patients
only, in our hospital every obstetric patient has a MEOWS
chart started at the rst antenatal visit and vital signs
documented until discharge. A number of preventable
deaths cited by CEMACH [5] occurred due to failure to
recognise women who were at high risk of maternal
complications. The charts stay with the mother until
postnatal discharge. This provides a visual trend of individ-
ual physiology, and allows assessment and treatment based
on what is abnormal for the patient, not just the population
as a whole.
There are some drawbacks to our study. The triggers
that we used, as laid out in the CEMACH report, are set
close to the values that dene morbidity. Thus, a positive
trigger, e.g. high blood pressure, which is associated with
Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland
Anaesthesia 2012, 67, 1218
morbidity, often becomes a self-fullling prophecy. This is a
single centre study, based in a tertiary referral centre and
may suffer from population bias. We may have lost a
proportion of women to follow-up, who may have presented
elsewhere with morbidity. The 30-day limit to patient review
may also mean that we have missed some morbidity. Finally,
our denitions of morbidity have incorporated nationally
accepted diagnostic criteria as far as possible, but there is no
universal denition of obstetric morbidity; some of these
denitions are a slightly arbitrary, which will inuence
whether a woman enters the morbidity group.
Despite these limitations, our results strongly support
the use of the MEOWS chart for all obstetric patients, as
recommended by the CEMACH report [5]. In its current
state, the chart has high sensitivity, reasonable specicity
and an accuracy that reects a low prevalence of morbidity
in our maternal population. Further evaluation will focus on
rening the cut-off values for low blood pressure, and this
may substantially reduce the false positive rate and the
burden associated with it.
Acknowledgements
We are grateful to Paul Bassett (Statsconsultancy Ltd) for his
help in statistical analysis of the manuscript. No external
funding or competing interests declared.
References
1. Goldhill D R, White S A, Sumner A. Physiological values and
procedures in the 24 h before ICU admission from the ward.
Anaesthesia 1999; 54: 52934.
2. Cullinane M, Findlay G, Hargraves C, Lucas S. An Acute Problem?.
London: National Confidential Enquiry into Patient Outcome and
Death, 2005.
3. Goldhill D R, McNarry A F, Mandersloot G, McGinley A. A physiolog-
ically-based early warning score for ward patients: the association
between score and outcome. Anaesthesia 2005; 60: 54753.
4. Gao H, McDonnell A, Harrison DA, et al. Harvey Systematic review
and evaluation of physiological track and trigger warning systems for
identifying at risk patients on the ward. Intensive Care Medicine
2007; 33: 66779.
5. Lewis G (ed.) Saving Mothers Lives: Reviewing maternal Deaths to
make Motherhood Safer 20032005. The Seventh Confidential
Enquiry into Maternal Deaths in the United Kingdom. London:
CEMACH, 2007.
6. Harrison DA, Penny JA, Yentis SM, et al. Case mix, outcome and
activity for obstetric admissions to adult, general Critical Care units: a
secondary analysis of the ICNARC Case Mix Programme Database.
Critical Care 2005; 9 (Suppl 3): S25S37.
7. Swanton RDJ, Al-Rawi S, Wee MYK. A national survey of obstetric
early warning systems in the United Kingdom. International Journal
of Obstetric Anesthesia 2009; 18: 2537.
8. Brace V, Penney G, Hall M. Quantifying severe maternal morbidity: a
Scottish population study. British Journal of Obstetrics and Gynae-
cology 2004; 111: 4814.
9. Grady K, Howell C, Cox C. Managing Obstetric Emergencies &
Trauma- The MOET Course Manual, 2nd edn. London: RCOG Press,
2007.
10. Tuffnell DJ, Shennan AH, Waugh JJS, Walker JJ. The Management of
Severe Pre-Eclampsia and Eclampsia, Guideline 10(A). London: RCOG
Press, 2006.
15
Anaesthesia 2012, 67, 1218
11. Horan TC, Andrus M, Dudeck MA. CDC NHSN Surveillance definition
of health care associated infection and criteria for specific types of
infection in the acute care setting. American Journal of Infection
Control 2008; 36: 30932.
12. British Thoracic Society Standards of Care Committee Pulmonary
Embolism Guideline Development Group. British thoracic society
16
Singh et al. | A validation study of MEOWS
guidelines for the management of suspected acute pulmonary
embolism. Thorax 2003; 58: 47084.
13. Hope RA, Longmore JM, Hodgetts TJ, Ramrakha PS. Oxford
Handbook of Clinical Medicine, 6th edn. Oxford: Oxford University
Press, 2004.
Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland
 Name: Ward: Consultant: Date:

Time:

Hospital Number: Date of Bir th: Height:

URINE passed (Y/N) passed (Y/N)

protein ++ protein ++
Date: Proteinuria

Time: protein >++ protein >++

Urinalysis
39 39
38 38 Individual
Parameters Oedema (Y/N)
37 37 o

36 36 >37

0
35 35 Clear/Pink Clear/Pink

Temperature C
Amniotic
fluid
Green Green
200 200
190 190
180 180 Alert Alert
170 170
160 160 Voice Voice
150 150 NEURO
RESPONSE
140 140 ( )
Pain Pain
130 130
Singh et al. | A validation study of MEOWS

120 120
110 110 Unresponsive Unresponsive
100 100

Systolic blood pressure

90 90
01 01
80 80 Individual
Pain Score
70 70 Parameters (no.)

60 60 SBP> 23 23
50 50
Normal Normal

MAP mmHg Lochia

Heavy/Fresh/ Heavy/Fresh/
Offensive Heavy/Fresh/Offensive
Offensive
130 130
120 120 NO ( ) NO ( )

Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland

110 110 Looks
unwell
100 100
YES ( ) YES ( )
90 90
80 80
70 70 Individual NO ( ) NO ( )
60 60 Parameters Trigger
50 50 DBP:

Diastolic blood pressure

YES ( ) YES ( )
40 40

170 170
160 160 Wound site check
150 150
140 140 Blood glucose
130 130
120 120
Nausea score
110 110
100 100

Heart rate
90 90 Bowel action
80 80
70 70 Individual
Daily weight
60 60 Parameters
50 50 Pulse Rate:
40 40
PAIN SCORE (assess pain on movement, deep breathing or coughing)
Neuro Responses
No pain at rest or on movement 0
>30 >30 No pain at rest, slight pain on movement 1
RESP Alert Patient is alert and concious
(write rate 2130 2130 Intermittent pain at rest, moderate pain on movement 2
in corresp. 1120 1120 Intermittent pain at rest, moderate pain on movement 3
Individual Verbal Patient responds to verbal stimulus
box)
010 010 NAUSEA SCORE
Parameters
95100% 1120 Pain Patient responds to painfull stimulus
Saturations Resp. Rate: None 0
<95% 95100% Nausea 1
Unresponsive Patient is unresponsive to any stimulus Vomiting 2
Administered 02 (L/min.) %

Appendix 1 The MEOWS observation chart.

Anaesthesia 2012, 67, 1218

17
Anaesthesia 2012, 67, 1218 Singh et al. | A validation study of MEOWS

Maternity Observation Chart

If Unresponsive Call 2222

Patient Triggers?
Action:
(1 Red Score or 2 Yellow Scores)
Attend within 10 min or send deputy
Confirm observations
Take history & examination
Decide on differential diagnosis
YES

Options:
* Reset trigger levels
* Make intervention (fluid, oxygen, etc)
* Decide on relocation (CLOMA, theatre,ITU
* Consider involvement of PART team on bleep 2525
Midwife action:
* Make referral and consider appropriate escalation
Start 1/2 hourly observation Call Anaesthetic SpR
Give oxygen at 10 l.min1 on bleep 1901
Must:
Review observation chart AND Call Obstetric SpR
* Decide when to reveiw
Review drug prescription chart on bleep 2345
* Write clear plan in notes
If pregnant, then tilt

Appendix 2 Callout algorithm for MEOWS triggers.

18 Anaesthesia 2011 The Association of Anaesthetists of Great Britain and Ireland