A prospective key informant surveillance system to measure maternal mortality - findings from indigenous populations in Jharkhand and Orissa, India.

Abstract

In places with poor vital registration, measurement of maternal mortality and monitoring the impact of interventions on maternal mortality is difficult and seldom undertaken. Mortality ratios are often estimated and policy decisions made without robust evidence. This paper presents a prospective key informant system to measure maternal mortality and the initial findings from the system.
In a population of 228 186, key informants identified all births and deaths to women of reproductive age, prospectively, over a period of 110 weeks. After birth verification, interviewers visited households six to eight weeks after delivery to collect information on the ante-partum, intra-partum and post-partum periods, as well as birth outcomes. For all deaths to women of reproductive age they ascertained whether they could be classified as maternal, pregnancy related or late maternal and if so, verbal autopsies were conducted.
13 602 births were identified, with a crude birth rate of 28.2 per 1000 population (C.I. 27.7-28.6) and a maternal mortality ratio of 722 per 100 000 live births (C.I. 591-882) recorded. Maternal deaths comprised 29% of all deaths to women aged 15-49. Approximately a quarter of maternal deaths occurred ante-partum, a half intra-partum and a quarter post-partum. Haemorrhage was the commonest cause of all maternal deaths (25%), but causation varied between the ante-partum, intra-partum and post-partum periods. The cost of operating the surveillance system was US$386 a month, or US$0.02 per capita per year.
This low cost key informant surveillance system produced high, but plausible birth and death rates in this remote population in India. This method could be used to monitor trends in maternal mortality and to test the impact of interventions in large populations with poor vital registration and thus assist policy makers in making evidence-based decisions.

Full text

Page 1

BioMed Central
Page 1 of 8
(page number not for citation purposes)
BMC Pregnancy and Childbirth
Open Access
Research article
A prospective key informant surveillance system to measure
maternal mortality – findings from indigenous populations in
Jharkhand and Orissa, India
Sarah Barnett*1, Nirmala Nair2, Prasanta Tripathy2, Jo Borghi3, Suchitra Rath2
and Anthony Costello1
Address: 1UCL Centre for International Health and Development, Institute of Child Health, University College London, 30 Guilford Street,
London, WC1N 1EH, UK, 2Ekjut, 107, Laxman Singh Complex-I, Munirka, New Delhi, Pin-11007, India and 3Infectious Diseases Epidemiology
Unit, Department of Epidemiology and Population Health, London School of Hygiene & Tropical Medicine, Keppel Street, London, WC1E 7HT,
UK
Email: Sarah Barnett* - sarahalbarnett@googlemail.com; Nirmala Nair - nnair007@yahoo.co.in; Prasanta Tripathy - prasanta.ekjut@gmail.com;
Jo Borghi - Jo.Borghi@lshtm.ac.uk; Suchitra Rath - s_rath2006@yahoo.co.in; Anthony Costello - a.costello@ich.ucl.ac.uk
* Corresponding author
Abstract
Background: In places with poor vital registration, measurement of maternal mortality and
monitoring the impact of interventions on maternal mortality is difficult and seldom undertaken.
Mortality ratios are often estimated and policy decisions made without robust evidence. This paper
presents a prospective key informant system to measure maternal mortality and the initial findings
from the system.
Methods: In a population of 228 186, key informants identified all births and deaths to women of
reproductive age, prospectively, over a period of 110 weeks. After birth verification, interviewers
visited households six to eight weeks after delivery to collect information on the ante-partum,
intra-partum and post-partum periods, as well as birth outcomes. For all deaths to women of
reproductive age they ascertained whether they could be classified as maternal, pregnancy related
or late maternal and if so, verbal autopsies were conducted.
Results: 13 602 births were identified, with a crude birth rate of 28.2 per 1000 population (C.I.
27.7–28.6) and a maternal mortality ratio of 722 per 100 000 live births (C.I. 591–882) recorded.
Maternal deaths comprised 29% of all deaths to women aged 15–49. Approximately a quarter of
maternal deaths occurred ante-partum, a half intra-partum and a quarter post-partum.
Haemorrhage was the commonest cause of all maternal deaths (25%), but causation varied
between the ante-partum, intra-partum and post-partum periods. The cost of operating the
surveillance system was US$386 a month, or US$0.02 per capita per year.
Conclusion: This low cost key informant surveillance system produced high, but plausible birth
and death rates in this remote population in India. This method could be used to monitor trends
in maternal mortality and to test the impact of interventions in large populations with poor vital
registration and thus assist policy makers in making evidence-based decisions.
Published: 28 February 2008
BMC Pregnancy and Childbirth 2008, 8:6doi:10.1186/1471-2393-8-6
Received: 5 October 2007
Accepted: 28 February 2008
This article is available from: http://www.biomedcentral.com/1471-2393/8/6
© 2008 Barnett et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 2

BMC Pregnancy and Childbirth 2008, 8:6http://www.biomedcentral.com/1471-2393/8/6
Page 2 of 8
(page number not for citation purposes)
Background
Maternal mortality in the developing world remains high,
with little progress towards Millennium Development
Goal (MDG) 5 [1,2]. There remains little evidence of the
impact of interventions on maternal mortality at the pop-
ulation level in poor and remote areas, where ratios are
often highest. The main challenge in effectively monitor-
ing progress and the impact of interventions is the diffi-
culty in measuring maternal mortality.
Maternal death is a relatively rare event, so large sample
sizes are needed to monitor mortality ratios. Conven-
tional surveillance systems are expensive and logistically
challenging and consequently, there have only been three
published trials that have attempted to measure popula-
tion maternal mortality ratio [3-5]. Furthermore, under-
reporting is frequent with most deaths occurring outside
of the health system and in countries without efficient
vital registration systems. Cause of death assignment is
difficult: where death certificates are completed they do
not stipulate pregnancy status [6,7] and maternal deaths
may be intentionally misclassified [8]. Thus there is a
great need for alternative methods for measuring maternal
mortality [9].
India has the largest number of maternal deaths in the
world and accounts for 22% of all maternal deaths [10].
About two thirds of these occur in just nine states, which
include Jharkhand and Orissa [11]. There are an estimated
90 million indigenous people in India, referred to as
"scheduled tribes" or "adivasis". Research has shown that
an indigenous person in India is 1.2 times more likely to
experience excess mortality compared to a non-indige-
nous person with the same standard of living [12].
This paper presents a prospective key informant system of
birth and death identification that is designed to be
cheaper and simpler than conventional surveillance sys-
tems. It aims to: measure crude birth rates and maternal
mortality in a remote, predominantly indigenous popula-
tion in eastern India; identify pregnancy related and late
maternal deaths; ascertain the breakdown of maternal
deaths by cause and determine the timing and place of
death; and report the cost of the system.
Methods
A key informant surveillance system was established in
three contiguous districts in two states in India; West Sing-
hbhum and Saraikela-Kharswan in Jharkhand and Keon-
jhar in Orissa. Within each district, twelve population
clusters were selected; the average population per cluster
is 6338, covering a total population of 228,186. The clus-
ters are largely physically remote from health services and
comprise a high proportion (73%) of indigenous people.
Other inhabitants are largely from the scheduled (3%)
and other backward castes (24%).
The measurement of maternal mortality is just one com-
ponent of a larger key informant surveillance system
designed to identify all births, neonatal and maternal out-
comes for the duration of a randomised controlled trial.
The whole system is described below but this paper will
just present results and costs for the maternal mortality
component. The system comprises two stages; the identifi-
cation stage and the interviewing stage (Figure 1).
The identification stage utilised local key informants toi-
dentify all births and deaths to women of reproductive age
(15–49). Most key informants were traditional birth
attendants (TBAs). One key informant was recruited to
cover a geographic area of approximately 250 households.
They were paid an incentive of 30 rupees ($0.65) for every
accurate birth or death identification. The key informants
met with an interviewer two-three times a month to pass
on the relevant information and the interviewer in turn
visited the relevant households to verify the births and
deaths before paying the identifiers. The key informers
only missed one late maternal death where a woman had
migrated to work at a brick kiln. In this instance the local
auxillary nurse midwife informed the interviewer.
The interviewing stage utilised one full-time salaried inter-
viewer for each cluster. Their role was to verify the infor-
mation provided by key informants; to interview all
mothers six weeks after their delivery to collect detailed
information on the ante-partum, intra-partum, and post-
partum periods, as well as background information; and
to interview family members in the case of a death of a
women of reproductive age to ascertain whether she was
pregnant, or had recently given birth. The interviewer
ascertained whether a death was maternal, late maternal,
pregnancy related or none of these by interviewing rela-
tives of the deceased, without the use of a verbal autopsy
tool. Eighty two per cent of all interviews were conducted
between six to eight weeks after delivery.
One monitoring supervisor per district supervised the sur-
veillance system. In the case of a maternal, pregnancy
related or late maternal death they conducted verbal
autopsies with a close friend or relative of the mother, ide-
ally present at the time of death. 47 verbal autopsies were
with husbands; 55 with other relatives and one with a
friend. 94 of the 103 respondents were present at the time
of death. Supervisors also tried to conduct verbal autop-
sies with formal and informal care providers, present at
the time of death, but these proved more difficult and
only seventeen out of a possible 37 providers were inter-
viewed. 70% of maternal verbal autopsies were conducted
within eight weeks after delivery.

Page 3

BMC Pregnancy and Childbirth 2008, 8:6 http://www.biomedcentral.com/1471-2393/8/6
Page 3 of 8
(page number not for citation purposes)
Verbal autopsies were reviewed independently by two
local obstetricians from the study area. There was agree-
ment on primary causation in all but seven cases. In the
case of disagreement a third doctor was consulted.
Finally, the costs of the surveillance system incurred by
the project were explored. Finanacial data were obtained
from project accounts. Costs were classified in terms of:
identification of births and deaths; eliminating deaths to
women of reproductive age that were not maternal, late
maternal or pregnancy related by interviewers; interviews,
verbal autopsies; and data management (including data-
base design, data entry by data entry clerk and data analy-
sis by a number of project staff, and associated
equipment). The cost of verbal autopsies included the cost
of the supervisor and obstetricians time assigning cause of
death. Although, the obstetricians volunteered their time,
and were not paid by the project, their time was valued
using average salary rates as, in the case of scale-up to a
larger population, an incentive payment may be neces-
sary. Costs are presented as average monthly costs and as
unit costs (e.g. cost per identification). The costs of setting
up the system in terms of staff recruitment and training
were also estimated. Costs are presented in 2006 US dol-
lars ($1= IRS 44.1).
The surveillance system is on-going as it is being used to
monitor the impact of a randomized controlled trial, and
this paper refers to births occurring in the first 110 weeks,
from 21st November 2004 to 31st Dec 2006. Maternal and
pregnancy related deaths may occur up to 42 days after the
birth and are therefore included up until 11th February.
Late maternal deaths may occur up to one year after the
birth and have been followed up until 31st December
2007.
Results
Birth rates and maternal outcomes
In a population of 228 186, over a 110 week period, 13
602 births were identified, of which 13 160 were live
births (Table 1). The crude birth rate is 28 per 1000 pop-
ulation. 323 deaths to women of reproductive age were
identified, of these 29% (95) were classified as maternal
deaths, giving a maternal mortality ratio (MMR) of 722
per 100 000 live births (CI 591–882). Four additional
pregnancy related and four late maternal deaths were
identified.
Cause and timing of maternal deaths
A quarter of all maternal deaths (24%) occurred during
the ante-partum period, half during the intra-partum
Maternal Mortality Key Informant Surveillance System
Figure 1
Maternal Mortality Key Informant Surveillance System.
Key Informer
Identify all births, stillbirths, neonatal deaths
Key Informer
Identify all deaths to women of reproductive age
Compute MMR
Interviewer
Ascertain all ….
Maternal
Deaths
Pregnancy
Related
Deaths
Late
Maternal
Deaths
Interviewer
Interview family member regarding mother’s
background characteristics, ante-partum, intra-
partum and post-partum information
Supervisor
Conduct verbal autopsy
with family member
Supervisor
Conduct verbal autopsy
with provider
Obstetricians
Assign cause of death

Page 4

BMC Pregnancy and Childbirth 2008, 8:6 http://www.biomedcentral.com/1471-2393/8/6
Page 4 of 8
(page number not for citation purposes)
period and up to 48 hours afterwards (48%) and just over
a quarter up to 42 days post-partum (27%) (Table 2).
Haemorrhage was the primary cause of death for a quarter
of maternal deaths (25%); closely followed by malaria
(23%); and sepsis (17%). Malaria was the most common
cause of death during the ante-partum period (48%);
haemorrhage during the intra-partum period (39%); and
sepsis during the post-partum period (35%). Secondary
causes were also identified from the verbal autopsies.
Anaemia was a key underlying factor in 35% of deaths.
Malaria (22) and Sepsis (20) were also highlighted as key
underlying causes.
The additional pregnancy related deaths were due to sui-
cide, homicide, thyroid cancer and lightening and the late
maternal deaths were due to malaria, sepsis, embolism,
and tuberculosis.
220 additional deaths to women of reproductive age were
identified. The most common causes of death were
malaria (46), tuberculosis (28), accidental causes (29),
anaemia (22), gastro-intestinal conditions (20) and sui-
cide (15).
Place of delivery and death
85% of all of the deliveries took place at home, 15% took
place in a facility and 0.1% in transit. For those who died
the percentages were very similar with 87% of the deliver-
ies occurring at home and 13% in a facility.
60% of deaths occurred at home, 28% in a facility, 6% in
transit to a facility and 5% in transit from a facility (Table
3). There was very little variation in the place of death,
when broken down by timing with a similar pattern seen
during the ante-partum, intra-partum, and post-partum
periods.
Table 1: Birth rates and maternal outcomes
95% Confidence Interval
Total Population
Number of births
Crude birth rate per 1000 population
Number of live births
228,186
13602
28.2
13160
(27.7–28.6)
Number of deaths to women of reproductive age (15–49)
Number of maternal deaths
Number of pregnancy related deaths
Number of late maternal deaths*
323
95
99
4
Maternal mortality ratio per 100 000 live births (n) 722(591–882)
Table 2: Primary causes of maternal deaths by time of death
Primary cause of deathAnte-partum Intra-partum (< 48 hours)Post-partum (< 42 days) Total
% (n)%(n)% (n)% (n)
All Deaths 24.2(23)48.4(46) 27.4 (26)
Haemorrhage
Malaria
Sepsis
Hypertensive disorder
Obstructed Labour
Anaemia
Embolism
Abortion
Other direct*
Other indirect**
8.7
47.8
4.3
21.7
(2)
(11)
(1)
(5)
39.1
10.9
13.0
10.9
17.4
(18)
(5)
(6)
(5)
(8)
15.4
23.1
34.6
11.5
(4)
(6)
(9)
(3)
25.3
23.2
16.8
13.7
8.4
1.1
2.1
2.1
4.2
3.2
(24)
(22)
(16)
(13)
(8)
(1)
(2)
(2)
(4)
(3)
3.8
7.7
(1)
(2)
8.7(2)
6.5
2.2
(3)
(1)
3.8(1)
8.7(2)
n = 95
* Retained placenta and obstetric shock
** Tuberculosis and heart disease

Page 5

BMC Pregnancy and Childbirth 2008, 8:6 http://www.biomedcentral.com/1471-2393/8/6
Page 5 of 8
(page number not for citation purposes)
For those who delivered at home and died, 68% died at
home and 23% made it to a facility, while 3% died in tran-
sit to a facility and 3% in transit from a facility. For those
who delivered in a facility and died, most died in a facility
(89%) while 11% died in transit from the facility. For
those who died during pregnancy 62% died at home, 21%
at a facility, 12% in transit to a facility.
Cost of surveillance system
The average monthly cost of identification was US$ 386
(US$ 0.70 per identification) (Table 4). The average
monthly cost of the monitors eliminating deaths to
women of reproductive age that were not maternal, late
maternal or pregnancy related deaths was US$ 9 (US$ 1
per recording). The average monthly cost of interviews
was US$ 2,311 (US$ 0.17 per interview) and the average
cost of verbal autopsies was US$ 54 (US$ 0.44 per VA).
The average monthly cost of data management and anal-
ysis was US$ 343. The average monthly cost overall was
US$ 2,759 (US$ 0.20 per birth interviewed), or US$ 3,103
inclusive of data management (US$ 0.23 per birth inter-
viewed). Staff represented 84% of the cost; transport 9%
and supplies 5%. The one-off cost of recruiting and train-
ing all staff was US$ 7,355.
Discussion
This study has shown that even in remote, deprived pop-
ulations a key informant system can produce reliable and
plausible maternal mortality ratios at low cost. The key
informant method has several important advantages. The
system is designed to prospectively measure, rather than
estimate, levels of maternal mortality in a given popula-
tion. Most maternal mortality data for developing coun-
tries are estimates, derived retrospectively, from censuses,
facility records and indirect methods such as adding ques-
tions to household or Demographic and Health Surveys
Table 3: Place of death by time of death and place of delivery
Place of death
Home FacilityTransit to facilityTransit from facilityField
% (n)% (n)%(n)% (n)% (n)
All Deaths 60.2 6228.2295.864.9511
Time of Death
Ante-partum
Intra-partum (< 48 hours)
Postpartum (< 42 days)
Late maternal deaths (42 days – 1 year)
Pregnancy related
60.0
58.7
62.5
50.0
75.0
15
27
15
2
3
32.0
21.7
37.5
50.0
0
88.0
8.7
2
4 10
9
2
10.95
25.01
Place of delivery
Home
Facility
Died during pregnancy
68.341 23.3
88.9
20.6
14
8
7
3.32 3.3
11.1
5.9
2
1
2
1.71
61.821 11.84
n = 103
Table 4: Breakdown of the costs of the surveillance system by activity and resource
Cost componentIdentificationEliminating non-maternal deathsInterviewsVerbal Autopsies Total
Key informants
Interviewers
Supervisers
Obstertician
Supplies (Printing, stationary)
Supplies (Hats, badges, bags)
Communication
Refreshments
Transport
375
-
-
-
-
11
-
-
-
---
-
375
1 697
245
11
136
11
24
7
255
6.83
0.88
-
0.55
-
0.09
0.03
0.96
1 690
219
-
135
-
23
7
238
25
11
-
-
1
0
16
Total 386 9.342 31154 2 759