Survival Analysis of Under-five Mortality in West Sulawesi Indonesia

using Cox Regression

Mieke Nurmalasari

, Linta Ifada

and Setia Pramana

Department of Health Information Management, Faculty of Health and Science, Universitas Esa Unggul, Jl. Arjuna Utara

No. 9, Kebon Jeruk, Jakarta Barat 11510, Indonesia

BPS’ Statistics Indonesia, Jakarta, Indonesia

Department of Computational Statistics, Politeknik Statistika STIS, Jakarta, Indonesia

Keywords: Survival Analysis, Under-five Mortality, Cox Regression, Hazard Ratio, Mortality Rate.

Abstract: Reducing infant mortality is one of the targets of SDGs goal 3, promoting wellbeing for all at all ages and

ensuring healthy lives. Although the under-five mortality in Indonesia has declined, some provinces show a

high mortality rate such as West Sulawesi. Hence, efforts to decrease the rate is still needed by identifying

the main determinant of under-five mortality. The main objective of this study is to determine the main factors

that affect under-five mortality in West Sulawesi using Intercensal Population Survey 2015. In this study, the

impact of mother’s education, age of first delivery, previous birth interval, birth type, the gender of the under-

five child, and paramedics help during labor were investigated using the Cox Proportional Hazard Regression.

All variables impacted mostly to the survival rate of under-five children. Female under-five children had a

lower hazard (risk) of death compared to the males. Twin had a three times higher hazard of death as compared

to single born children. In addition, higher mother’s education tends to have a lower hazard.

1 INTRODUCTION

All health issues in the SDGs are integrated into goal

number 3, which is to ensure a healthy life and

wellbeing for all at all ages. Reducing the mortality

rate for infants and under-five children is one of the

target SDGs' goals (United Nations, 2017).

Countries around the world have been trying to

reduce child mortality in the last few decades. Child

mortality is an important indicator of children’s

health. The under-five mortality rate is the number of

under-five deaths (1 – 5 years) per 1000 live births

within one year. This mortality can describe the level

of health problems of children under five, the level of

primary health services and the success of primary

health (Danzhen You, Lucia Hug, Simon Ejdemyr,

Jan Beise and World, 2015).

In general, under-five mortality in Indonesia has

declined. However, some provinces have a high

mortality rate. The highest under-five mortality is

dominated by Provinces in Eastern Indonesia, one of

them is West Sulawesi Province. Based on the Health

Office report 2016, there has been a decrease in child

mortality rates over the past year in West Sulawesi

from 2010 to 2015. It was 16.4 per 1000 live births in

2010 and 14.26 per 1000 live birth in 2015 (Dinas

Kesehatan Provinsi Sulawesi Barat, 2016). However,

West Sulawesi became the province with the highest

number of infant deaths (IMR) in Indonesia which is

50 infant deaths that occurred in 1000 live births.

Infant mortality also increased in 2017 (Dinas

Kesehatan Provinsi Sulawesi Barat, 2017). Factors

influence this situation need to be investigated to

know the right policy in reducing the rate.

The main objective of this study is to determine

the main factors that affect or influence the survival

of under-five in West Sulawesi.

2 METHOD

This study used secondary data namely Intercensal

Population Survey 2015 in Indonesia performed by

BPS-Statistics Indonesia. The unit of this study is

under-five born children. There are 2549 cases in

West Sulawesi.

The response variable is the death risk of under-

fives which measured by under-fives survival time

from birth until death. The survival time is presented

in months which ranges between 0 to 60 months. The

Nurmalasari, M., Ifada, L. and Pramana, S.

Survival Analysis of Under-ﬁve Mortality in West Sulawesi Indonesia using Cox Regression.

DOI: 10.5220/0009566900920096

In Proceedings of the 1st International Conference on Health (ICOH 2019), pages 92-96

ISBN: 978-989-758-454-1

explanatory variables are the mother’s education,

mother’s first age of giving birth, previous birth

interval, birth type, the gender of the under-five child

and whether paramedics help labor. The impact of

these variables was investigated using the Cox

Proportional Hazard Regression.

The Cox proportional hazard model is a survival

model that measures the relationship between one or

more covariates with time. The risk of the event per

time unit changes over time at baseline levels of

covariates was denoted by h

(t). The hazard function

was represented by h(t). The hazard ratio was

expressed by h(t)/h

(t). The Cox proportional hazard

model was defined by the equation:

ℎ(𝑡)

ℎ

(𝑡)

= 𝑒

(𝛽

𝑥

+𝛽

𝑥

+⋯+𝛽

𝑝

𝑥

𝑝

)

(1)

Here h(t) hazard at time t, h

(t) is a baseline

hazard, x

is the p-th explanatory variable.

3 RESULT

3.1 Descriptive Analysis

From Table 1, we can observe that there are 254

(10%) infants could not survive before reaching five

years old. The percentage of infants who died before

5 years from the family whose mother has low

education (up to elementary school) is higher (12.3%)

than, other educational backgrounds.

The age of the mother at the first delivery seems

to be the risk of under-five mortality. Furthermore,

22.6% of the infant with the mother’s age at first

delivery above 35 years old was dead.

Moreover, infants with the birth type of multiple

(twin, triplets, etc) have almost triple risk to die as

compared to the single born infant. In addition, a male

infant has a higher risk to die before five as compared

to a female infant.

3.2 Survival Analysis

3.2.1 Kaplan Meier

The curve of Kaplan Meier in Figure 1 shows that at

the beginning of the first year the curve looked

decrease. It means that they are struggling to survive,

but afterward, the probability of the children to

survive is relatively stable until five years old or the

cumulative survival of the child was 60 months.

Figure 1: Overall Under-five child survival status in West

Sulawesi.

Table 1: Sample characteristics.

Variables

Died (n=254)

Alive (n=2295)

Total (n=2549)

n (%)

Mother’s education status

Low

158 (12.3%)

1126 (87.7%)

1284 (100%)

Middle

80 (8.2%)

901 (91.8%)

981 (100%)

High

16 (5.6%)

268 (94.4%)

284 (100%)

Mother’s age at the first delivery

20-35 years old

150 (9.6%)

1416 (90.4%)

1566 (100%)

<20 or >35 years old

97 (10.2%)

855 (89.8%)

952 (100%)

≥ 35 years old

7 (22.6%)

24 (77.4%)

31 (100%)

Previous birth interval

<24 months

117 (11.9%)

863 (88.1%)

980 (100%)

≥24 months

137 (8.7%)

1432 (91.3%)

1569 (100%)

Birth Type

Single

239 (9.6%)

2258 (90.4%)

2497 (100%)

Multiple

15 (28.8%)

37 (71.2%)

52 (100%)

Gender

Male

164 (12.2%)

1185 (87.8%)

1349 (100%)

Female

90 (7.5%)

1110 (92.5%)

1200 (100%)

Paramedics at Labor

Non-Medic

89 (11.2%)

704 (88.8%)

793 (100%)

Medic

165 (9.4%)

1591 (90.6)

1756 (100%)

Survival Analysis of Under-ﬁve Mortality in West Sulawesi Indonesia using Cox Regression

An overview of the under-five child in West

Sulawesi from the Kaplan Meier curve based on

explanatory variables can be seen in Figure 2 up to

Figure 6. The Kaplan Meier curve in Figure 2, it

appears that the survival curve of an under-five child

who has mothers with higher education is above the

survival curve of an under-five child who has mothers

with middle education and low. It could be concluded

that under-five child who has mothers with higher

education has a tendency to survive longer compared

to under-five who have mothers with middle and low

education.

Figure 2: Kaplan Meier curve for under-five survival in

West Sulawesi based on mother’s education.

According to Figure 3, it shows that mother which

age of first delivery above 35 years old tends to have

less survival compared to mother’s ideal first age of

giving birth (20-35).

Figure 3: Kaplan Meier curve for under-five survival in

West Sulawesi based on mother’s age of first delivery.

Figure 4 is the Kaplan-Meier curve which shows

that the survival curve for infants with previous birth

intervals more than 24 months is above the survival

curve for infants whose birth intervals are less than 24

months. This indicates that toddlers born to mothers

whose birth intervals are more than 24 months tend to

have the opportunity to survive longer.

Figure 4: Kaplan Meier curve for under-five survival in

West Sulawesi based on previous birth interval.

Kaplan Meier curve for under-five survival in

Figure 5 demonstrate that the curve of multiple birth

type is below the survival curve for a child with a

single birth type. Twin birth types are more likely

smaller for life longer than under-fives with a single

birth type.

Figure 5: Kaplan Meier curve for under-five survival in

West Sulawesi based on birth type.

Figure 6 shows that the cumulative survival for

under five for females seems to have lived longer than

the male because the survival curve of the female is

above the male.

Figure 6: Kaplan Meier curve for under-five survival in

West Sulawesi based on gender.

ICOH 2019 - 1st International Conference on Health

Table 2: Parameter estimation for Cox Regression Survival Model.

Variable

β (Se)

Wald

p-value

Hazard Ratio

Mother’s education status

Low*

Middle

High

-0.481 (0.141)

-0.897 (0.273)

11.643

10.792

0.001

0.618

0.408

Mother’s age of first delivery

20-35 years old*

< 20 years old

≥ 35 years old

-.0.027 (0.134)

0.986 (0.391)

0.041

6.356

0.840

0.012

0.973

2.681

Previous birth interval

< 24 months*

≥ 24 months

-.345 (0.131)

6.936

0.008

0.708

Birth type

Single*

Multiple

1.151 (0.271)

18.011

0,0001

3.161

Gender

Male*

Female

-0.510 (.131)

15.079

0,001

0.601

Paramedics labor

-0.080 (0.136)

0.347

0.556

.923

*Reference

3.2.2 Cox Regression Model

We use Cox regression to the model prediction of the

time (measured in weeks). We explore which factors

are associated with the risk of the under-five mortality

of children in West Sulawesi.

The result from the Cox Regression model in

Table 2 shows that almost all the explanatory

variables in the model were statistically significant

except variable paramedic labor.

The hazard ratio value indicates that mothers who

have higher education are less likely to have infant

deaths. This is in line with research conducted by

(Ettarh and Kimani, 2012), (Mwangi Muriithi, 2015)

and (Aheto, 2019). Mothers with secondary or higher

education have a higher desire to seek information or

knowledge about health care.

The hazard ratio based on the age of birth delivery

shows that mothers with age of first delivery > 35

years have a higher risk to not survive 2.681 times

compare to the ideal mother’s ages of first delivery

(20 – 35 years old).

Meanwhile, a mother who has a longer birth

interval with a previous birth, then the tendency for

children to die will decrease (Kayode, Adekanmbi

and Uthman, 2012). The close birth distance between

the first child to the next child can cause problems,

both mental and physical health. The World Health

Organization (WHO) and Badan Koordinasi

Keluarga Berencana Nasional (BKKBN) recommend

that the next birth interval should be two to three years

to minimize the risk of child and maternal mortality.

The type of births with twins has a greater chance

of dying 3.161 times compared to single births.

(Monden and Smits, 2017) also explain in their

research that twin births are more at risk of dying than

single childbirths.

The hazard ratio for gender indicates that the

female has a longer survival time than the male. The

results of this study are also the same as the research

conducted by (Ruggieri et al., 2016) and (Afeez et al.,

2018). Female toddlers are more able to survive than

male toddlers in terms of the immune system and

genetic factors. Female’s under-five children are

more immune to diseases.

4 CONCLUSIONS

All variables impacted mostly to the survival rate of

under-five children. Female under-five children had a

lower hazard (risk) of death compared to the males.

Twin under-five children had four times higher

hazard of death as compared to single born children.

Higher mother’s education tends to have a lower

hazard.

REFERENCES

Afeez, B. M. et al. (2018) ‘Reprospective Analysis of Some

Factors Responsible for Infant Mortality in Nigeria:

Evidence from Nigeria Demographic and Health

Survey (NDHS)’, American Journal of Mathematics

Survival Analysis of Under-ﬁve Mortality in West Sulawesi Indonesia using Cox Regression

and Statistics, 8(6), pp. 184–189. doi:

10.5923/j.ajms.20180806.04.

Aheto, J. M. K. (2019) ‘Predictive model and determinants

of under-five child mortality: Evidence from the 2014

Ghana demographic and health survey’, BMC Public

Health. BioMed Central Ltd. doi: 10.1186/s12889-019-

6390-4.

Danzhen You, Lucia Hug, Simon Ejdemyr, Jan Beise, P. I.

and World (2015) ‘Child Mortality 2015’, UNICEF

WHO World Bank Group United Nations, p. 74.

Available at: www.childmortality.org.

Dinas Kesehatan Provinsi Sulawesi Barat (2016): ‘Profil

Kesehatan Sulawesi Barat Tahun 2015’.

Dinas Kesehatan Provinsi Sulawesi Barat (2017) Profil

Kesehatan Provinsi Sulawesi Barat.

Ettarh, R. R. and Kimani, J. (2012) ‘Determinants of under-

five mortality in rural and urban Kenya’, Rural and

Remote Health.

Kayode, G. A., Adekanmbi, V. T. and Uthman, O. A.

(2012) ‘Risk factors and a predictive model for under-

five mortality in Nigeria: Evidence from Nigeria

demographic and health survey’, BMC Pregnancy and

Childbirth. doi: 10.1186/1471-2393-12-10.

Monden, C. W. S. and Smits, J. (2017) ‘Mortality among

twins and singletons in sub-Saharan Africa between

1995 and 2014: a pooled analysis of data from 90

Demographic and Health Surveys in 30 countries’, The

Lancet Global Health. doi: 10.1016/S2214-109X

(17)30197-3.

Mwangi Muriithi, D. (2015) ‘Determination of Infant and

Child Mortality in Kenya Using Cox-Proportional

Hazard Model’, American Journal of Theoretical and

Applied Statistics. Science Publishing Group, 4(5), p.

404. doi: 10.11648/j.ajtas.20150405.21.

Ruggieri, A. et al. (2016) ‘Monographic section the

influence of sex and gender on immunity, infection and

vaccination’, Ann Ist Super Sanità, 52(2), pp. 198–204.

doi: 10.4415/ANN_16_02_11.

United Nations (2017) The Sustainable Development Goals

Report 2017. New York: United Nations. Available at:

https://unstats.un.org/sdgs/files/report/2017/thesustain

abledevelopmentgoalsreport2017.pdf.

ICOH 2019 - 1st International Conference on Health