International Journal of Epidemiology
© International Epidemiological Association 1997
Vol. 26, No. 2
Printed in Great Britain
High Maternal Mortality Levels
and Additional Risk from Poor
Accessibility in Two Districts
of Northern Province, Zambia
F LE BACQ AND A RIETSEMA
Le Bacq F (Kasama District Health Services, Box 410056, Kasama, Zambia) and Rietsema A. High maternal mortality
levels and additional risk from poor accessibility in two districts of Northern Province, Zambia. International Journal of
Epidemiology 1997; 26: 357–363.
Background. Maternal mortality ratios in Kasama and Kaputa Districts, two remote rural areas of Northern Province,
Zambia, were suspected to be very high. In order to evaluate the impact of a referral system baseline maternal mortality
levels and additional maternal mortality risk arising from poor accessibility were estimated.
Methods. The sisterhood method was applied to a random population sample of 3123 respondents in Kasama District
and to 2953 in Kaputa District during May and June 1995. For Kasama also hospital-based maternal mortality was calculated from record analysis from 1 January 1991 up to 31 December 1995. Population attributable risk and population
etiological fraction were calculated for Kasama District.
Results. Maternal mortality ratio for Kasama District was 764 per 100 000 live births and 1549 for Kaputa District. Kasama hospital-based maternal mortality was 543 per 100 000 live births. In Kasama District population attributable risk of
maternal mortality from poor accessibility was 220 maternal deaths per 100 000 live births, and the population etiological
fraction was 29%. In Kaputa District population attributable risk was 1006 maternal deaths per 100 000 live births, and
the population etiological fraction was 65%.
Conclusions. This study suggests that solving the accessibility problem would decrease the mortality burden from
maternal causes with at least 29% in Kasama District and 65% in Kaputa District.
Keywords: maternal mortality, accessibility, Zambia
Consequently, the second objective of our study was to
estimate the additional risk of maternal mortality arising from this physical accessibility problem.
Kasama is the provincial district in Northern Province with a population of about 200 000 people. Kaputa
is a district in the far northwest of that province with
about 60 000 people (Figures 1 and 2). All health centres
in both districts are accessible only by dirt roads and, at
the time of study there was neither telephone nor radio
communication. Public transport is almost non-existent.
Kaputa has no theatre facility anywhere in the district.
In Kasama District theatre facilities are only available
at the general hospital itself.
Two studies from Western Province in Zambia reveal
very high estimates of maternal mortality: 889 maternal
deaths per 100 000 live births in the provincial Mongu
district (unpublished), and 1193 per 100 000 in Kalabo
district in the far west.1 The first objective of our study
was therefore to confirm our suspicion of an equally
high maternal mortality ratio in the remote Northern
Province. Zambian health reforms aim to bring health
care as close to the family as possible.2 In line with this
policy maternal health care could be improved by decentralizing both maternity and family planning services.3 Although access to care is more than a problem
of distance,4 and a decision to seek health care takes
place in a complex web of relationships, large distances
affect profoundly women’s willingness and ability to
seek care, particularly when appropriate transportation
is very limited, communication difficult, and terrain
inhospitable as in the Northern Province of Zambia.
METHODS
During May and June 1995 two retrospective
community-based surveys were carried out using the
sisterhood method5 in Kasama and Kaputa Districts. In
addition Kasama General Hospital based maternal
mortality was calculated retrospectively by analysis of
Kasama District Health Services, Box 410056, Kasama, Zambia.
357
358
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
FIGURE 1 Geographical location of Zambia
FIGURE 2 Geographical location of Kasama and Kaputa in Northern Province, Zambia
MATERNAL MORTALITY AND ACCESSIBILITY
all records from 1 January 1991 up to 31 December
1995. Also, the catchment area in terms of institutional
deliveries was determined for Kasama General Hospital by studying the residential address of the last
1000 deliveries.
Sampling
In Kasama District, a list of 1086 villages, with an
average population of 100 children and 100 adults, constituted the sampling frame. From these 30 villages
were originally selected to ensure at least 3000 respondents. A random number table was used to select the
first village between number 1 and 36 on our list, with
a sampling interval of 36 to identify the other 29 villages. In addition a random strategy was worked out to
add neighbouring villages to the selected clusters in
case the number of adults eligible for interview fell
short of the expected number of 100 per village. Eventually 44 villages were included in the sample. In
Kaputa District interviews were carried out in all
villages that also have a primary school. The eligible
44 villages were evenly scattered across Kaputa District.
Training and Data Collection Tool
In Kasama 15 nursing students were trained for two
half-days, which included field testing in the villages
surrounding the nursing school area, in administering
the following questionnaire in the local language:
a. Date of birth of respondent
b. Sex of respondent
c. How many sisters have you ever had who were born
to the same mother?
d. How many of these sisters reached age 15, including
those who are now dead?
e. How many of these sisters reaching age 15 are alive
now?
f. How many of these sisters reaching age 15 are dead
now?
g. How many of these dead sisters, who reached age
15, died while pregnant, during delivery, or during
the first 6 weeks after delivery?
All interviewers had to indicate on the form whether
the sum of e + f = d, and d was øc. In that way
discrepancies were sorted out in the field. During the
training it was also stressed that a female respondent
should not include herself in reporting the number of
sisters born to her mother.
In Kaputa District 80 primary school teachers were
trained in a similar way, including a small field test,
in 11 groups. Afterwards they administered the questionnaire in the village in which the primary school was
situated.
359
Analysis and Definitions
All data were analysed on a computer with Epi-info,
Version 5.0. Confidence intervals (CI) and comparisons
were made using the Normal approximation to the
binomial distribution as described by Osborn. 6 A
maternal death was defined according to the ICD-10 as
the death of a woman while pregnant or within 42 days
of termination of pregnancy, irrespective of the duration and the site of the pregnancy, from any cause
related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes.7
Maternal mortality ratio (MMR) was defined as the
number of maternal deaths per 100 000 live births.
Maternal mortality rate was calculated as total fertility
rate multiplied by maternal mortality ratio.8 The total
fertility rate of 7.1 for rural Zambia was used; as estimated by the Zambian demographic and health survey
of 1992.9 Population attributable risk (PAR) was defined as risk in the total population minus risk in the
unexposed population. The risk in the total population
was estimated separately for Kasama and Kaputa Districts using the community-based sisterhood method.
The exposure factor was defined as ‘living more than
2 hours walking to the nearest hospital’. The unexposed
population consequently were the women living within
2 hours walking distance to the nearest hospital, which
in Kasama applies to almost all the women delivering
in the hospital as will be seen from the Results. Population etiological fraction (PEF) was defined as PAR/
Risk in total population. Stratified analysis of Kasama
data was done using two areas: one with poor, and
another with very poor access to health services. This
was based on the plotting of a measles epidemic which
reached its peak during the sisterhood survey. Starting
in Kasama town, the spreading epidemic was limited by
two rivers only. Measles immunization coverage is as
high beyond those rivers as within them. This coincidental proof of effectiveness of these natural barriers also
indicated that it was more difficult for mothers living
beyond those rivers to reach Kasama Hospital for an
emergency.
Maternal death figures from Kasama General Hospital were compiled mainly retrospectively from the
records from 1 January 1991 up to 31 December 1995.
The MMR for the hospital was calculated as the number
of maternal deaths divided by the number of live births
which took place at the hospital during the same period
of 5 years.
RESULTS
In Kasama and Kaputa Districts interviews were completed for 3123 and 2953 respondents respectively.
360
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 1 Community-based maternal mortality, Kasama District, Zambia
Age group
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–54
55–59
60+
Total
a
No. of
respondents
b
Sisters
reaching
age 15
c
Maternal
deaths
d
Adjustment
factor
e
Sister units
of risk
exposure
401
551
499
417
366
164
191
114
159
261
3123
1258
1729
1558
1357
1195
538
604
369
454
738
9801
13
40
48
35
48
12
21
10
22
23
272
0.107
0.206
0.343
0.503
0.664
0.802
0.900
0.958
0.986
1.000
135
356
534
683
793
431
544
354
448
738
5016
Note: total life time risk of maternal mortality (c/e) is 0.054226. When this risk is divided by the total fertility rate (7.1) and multiplied by 100 000,
then the number of maternal deaths per 100 000 live births is 764.
TABLE 2 Community-based maternal mortality, Kaputa District, Zambia
Age group
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–54
55–59
60+
Total
a
No. of
respondents
b
Sisters
reaching
age 15
c
Maternal
deaths
d
Adjustment
factor
e
Sister units
of risk
exposure
179
397
450
434
440
317
250
181
129
176
2953
588
1304
1401
1443
1485
1092
874
601
386
523
9696
10
56
100
93
106
76
57
54
30
41
623
0.107
0.206
0.343
0.503
0.664
0.802
0.900
0.958
0.986
1.000
63
269
481
726
986
876
787
576
381
523
5666
Note: total life time risk of maternal mortality (c/e) is 0.10995. When this risk is divided by the total fertility rate (7.1) and multiplied by 100 000,
then the number of maternal deaths per 100 000 live births is 1549.
Total maternal deaths divided by total sister units of
risk exposure yielded a lifetime risk of dying from
maternal causes of 5.4% for Kasama and 11.0% for
Kaputa. Division of this risk by the total fertility rate of
7.1 and multiplication by 100 000 leads to 764 maternal
deaths per 100 000 live births for Kasama District and
1549 for Kaputa District (Tables 1 and 2). The average
age of Kasama respondents was 35.2 ± 0.3 years, and
33.5 ± 0.3 years for Kaputa respondents. The MMR for
Kasama General Hospital was 543 per 100 000 live
births, based on 60 deaths and 11 041 live births during a 5-year period. In Kasama the respondents’ sex
was recorded for 2947 (94.4%) individuals: 1166 male
and 1781 female. In Kaputa the sex was recorded for
2953 (89.3%) individuals: 1195 male and 1442 female.
No significant difference for MMR was found between
male and female respondents in both Kasama (P =
0.25428) and Kaputa (P = 0.25428). In Kasama 1902 respondents came from the poor accessibility zone, and
1221 from the very poor accessibility zone. Maternal
361
MATERNAL MORTALITY AND ACCESSIBILITY
TABLE 3 Maternal mortality ratios and 95% confidence limits for different data sets
Data source
a
Maternal
deaths
b
Sister units
of risk
exposure
c
Maternal
mortality
ratio
d
Lower 95%
confidence
limit
272
108
143
154
118
623
268
282
60
5016
1875
2858
3121
1894
5666
2335
2693
–
764
811
705
695
877
1549
1617
1475
543
673
658
589
585
719
1427
1423
1303
406
Kasama total
Kasama male
Kasama female
Kasama – PAa
Kasama – VPAb
Kaputa total
Kaputa male
Kaputa female
KGHc
e
Upper 95%
confidence
limit
855
964
820
805
1036
1670
1810
1647
681
a
Poor accessibility.
Very poor accessibility.
c
Kasama General Hospital.
The maternal mortality ratio is calculated as follows: [a/(b × 7.1)] × 100 000, where 7.1 is the total fertility ratio for rural Zambia. The ratio for KGH
is calculated directly from the 11 041 live births which took place in the hospital between 1991 and 1995 as follows: (a/11 041) × 100 000.
b
TABLE 4 Relative risks (RR) of maternal mortality with 95% confidence limits and significance testing of comparisons between several
areas
Comparisons
Kasama – Kaputa
Kasama: male–female
Kaputa: male–female
Kasama Access: PAa – VPAb
KGH–Kasama PA
KGH–Kasama VPA
KGH–Kaputa
P value
, 0.00001
0.254280
0.254280
0.049996
, 0.00001
, 0.00001
, 0.00001
RR
Lower limit
Upper limit
2.03
1.67
2.48
1.26
1.28
1.61
2.85
0.89
0.86
1.06
2.10
1.77
1.98
2.55
4.11
a
Poor accessibility.
Very poor accessibility.
c
Kasama General Hospital.
b
mortality ratios according to location, sex, and accessibility are summarized in Table 3. Significantly different MMR was found between the two accessibility
zones, between Kasama General Hospital and each of
the accessibility areas in Kasama District, and between
Kasama General Hospital and Kaputa District (Table 4).
The same Table shows relative risks (RR) with 95%
confidence limits for all significant differences, e.g. a
woman faces an RR of 2.03 (1.67–2.48) of dying from
any maternal cause in Kaputa as compared with a
mother in Kasama. The number of sisters who died at
age ù15 from any cause was 1421 in Kasama and
2094 in Kaputa. The number of sisters dying from a
maternal cause was 272 in Kasama and 623 in Kaputa.
Consequently, the proportional maternal mortality rate
is 19.1% in Kasama and 29.8% in Kaputa.
Retrospective study of the distribution of women
delivering in Kasama General Hospital by residential
address showed that 94% came from within 2 hours
walking distance. Retrospective study of mothers dying
in the same hospital showed that 64% came from within
the same area.
The risk of dying from a maternal cause is 764 per
100 000 live births in Kasama District. The same risk
for women delivering in Kasama General Hospital is
543. Therefore the population attributable risk (PAR) of
dying from a maternal cause in Kasama District, related
to remoteness is 220 maternal deaths per 100 000 live
362
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
births. The population etiological fraction (PEF) is 29%.
In Kaputa District, which has no hospital, the risk in the
total population is 1549 per 100 000 live births. The
PAR there is 1006 maternal deaths, and the PEF is 65%.
DISCUSSION
Comparison with other studies shows that rural Zambia
has one of the highest maternal mortality ratios in
Africa: 570 in Ethiopia,10 287 in Zimbabwe,11 426 in
Malawi,12 297 in Tanzania,13 740 in Djibouti.14 This
study estimated the contribution of poor accessibility in
terms of remoteness, to this problem.
Previous studies in other parts of Africa have used
the sisterhood method and found it to be reliable and
robust. The good quality of our community-based data
is supported by the absence of a significant difference
between male- and female-based MMR in both Kasama
and Kaputa Districts. The catchment area for all maternal deaths in Kasama General Hospital is clearly bigger than for all mothers delivering live babies in the
same institution. The hospital-based maternal mortality
is therefore likely to be overestimated because of its
limited referral function for complicated maternity cases.
Using the procedure as described by Graham et al.5
for calculating the time-location, the community-based
estimates of maternal mortality for Kasama and Kaputa
refer to a period about 10.8 and 10.1 years prior to data
collection respectively. The hospital data refer to a
period 5–10 years later. It is unlikely that the population of Northern Province, Zambia, has experienced a
significant change in the overall level of maternal mortality in the last 10 years. Therefore the time-location
dof the community-based study and the hospital study
are reasonably comparable.
Hospital delivery in Kasama seems to be possible
only by and large for mothers living within walking
distance of that institution. This is supported by the
hospital distribution of mothers’ place of residence.
The difference in mortality risk for mothers living just
beyond the walking distance area, or far across the two
rivers which effectively divide the district is only just
significant. The measles epidemic, which led to stratification of the data in the first place, was transmitted
from person to person. Maternal problems like obstructed labour or post partum haemorrhage have to be
brought to the hospital for a better chance of survival.
If there are few means of transport then it makes only a
small difference on which side of the river one lives.
The data from Kaputa District illustrate that maternal
mortality can double when there is no hospital in the
district in addition to poor transport. Comparison of
data from Kasama General Hospital, the poor access
stratum of Kasama District, the very poor access
stratum, and Kaputa District, show a dose-response
relationship between distance and maternal mortality.
Also, of all women dying, proportionally more die of
maternal causes in the more remote areas.
The PAR indicates the additional risk of dying from
maternal causes for women living in remote areas. The
PEF suggests that maternal mortality could be reduced
by 29% if hospital services, as they currently exist in
Kasama General Hospital could be brought within
reach of all mothers in Kasama District. The real benefit
could be bigger because the hospital-based maternal
mortality rate is likely to be overestimated because of
the different catchment areas of numerator and denominator. However, this may be a maximum figure because
it implies that all mothers would make use of the
offered maternity service to the same extent as they do
now. At present the institutional delivery rate is approximately 25% with all the accessibility problems. It can
therefore be anticipated that the benefits in terms of
maternal mortality reduction, from an effective referral
system will be about 29%. The relatively high maternal
mortality levels already found amongst women with
good access to Kasama General Hospital also suggests
that reduction beyond 29% in Kasama District requires
more than solving physical accessibility problems, e.g.
by offering better quality of maternity services. In
Kaputa District, which at present has no hospital, providing more accessible services would have an even more
profound impact as can be seen from the calculated
etiological fraction of 65%. The outcome of this study
has encouraged both districts to set up a better referral
system, e.g. through installation of short wave radios in
outlying health centres, start up of strategic blood banks
in the district, and an ambulance service. In the future
we want to repeat this maternal mortality study and evaluate the impact versus the cost of our referral system.
ACKNOWLEDGEMENTS
A special word of thanks goes to the nursing school
students of Kasama and the teachers of Kaputa District
who collected the raw data in their respective districts.
We also want to thank Dr Chelemu, Provincial Medical
Officer for Northern Province, Zambia, and Dr Peepercorn, Senior Medical Officer for Northern Province,
Zambia, for their encouraging comments and support.
REFERENCES
1
Vork F, Kyanamina S. Maternal Mortality in Kalabo District,
Zambia: The Sisterhood Method Survey. Lusaka, Zambia:
Ministry of Health, 1995.
MATERNAL MORTALITY AND ACCESSIBILITY
2
Musowe V. Strategic Health Plan . Lusaka, Zambia: Ministry of
Health, 1995, p. 3.
3
Fortney J A. The importance of family planning in reducing
maternal mortality. Stud Fam Plann 1987; 18: 109–14.
4
Timyan J, Griffey Brechin S J, Measham D M, Ogunleye B. The
Health of Women: A Global Perspective. Oxford: Westview
Press, 1995, pp. 217–33.
5
Graham W, Brass W, Snow R W. Estimating maternal mortality:
the sisterhood method. Stud Fam Plann 1989; 20: 125–35.
6
Osborn J F. Manual of Medical Statistics. Vol. 1. London: London
School of Hygiene and Tropical Medicine, 1986, p. 45.
7
Fortney J A. Implications of the ICD-10 definitions related to
death in pregnancy, childbirth or the puerperium. World
Health Stat Q 1990; 43: 246–48.
8
Graham W, Airey P. Measuring maternal mortality: sense and
sensitivity. Health Policy Plann 1987; 2: 323–33.
9
Gaisie K, Cross A R, Nsemukila G. Zambia Demographic and Health
Survey 1992. Columbia: Macro International, 1992, p. 27.
10
363
Shiferaw T, Tessema F. Maternal mortality in rural communities
of Illubabor, southwestern Ethiopia: as estimated by the
sisterhood method. Ethiop Med J 1993; 31: 239–49.
11
Oosterhuis J W A. Estimating maternal mortality by sisterhood method in rural Zimbabwe. Trop Doc 1993; 23:
67–68.
12
Chiphangwi J D, Zamaere T P, Graham W J, Duncan B, Kenyon T,
Chinyama R. Maternal mortality in the Thyolo district of
southern Malawi. E A Med J 1992; 69: 675–79.
13
Walraven G E L, Mkanje R J B, Van Roosmalen J, Van Dongen
P W J, Dolmans W M V. Assessment of maternal mortality
in Tanzania. Brit J Obstet Gynaecol 1994; 101: 414–17.
14
David P, Kawar S, Graham W. Estimating maternal mortality in
Djibouti: an application of the sisterhood method. Int J
Epidemiol 1991; 20: 551–57.
(Revised version received May 1996)