Citation: Folorunso Serifat A., et al. “Application of Bayesian Nonparametric Estimation on Cervical Cancer: A Case Study of University
College Hospital, Nigeria”. EC Nursing and Healthcare 3.3 (2021): 161-167.
162
Application of Bayesian Nonparametric Estimation on Cervical Cancer: A Case Study of University College Hospital, Nigeria
to be a threat to women’s health in developing countries. Marc., et al
mortality from it occurred in 2018. In Malawi, 132 (44%) patients in stage 1 cervical cancer and 168 (56%) patients in stage 1 - 4 were
reported for the study periods of 310 patients (Pandaora., et al. 2017). Cervical cancer grows in the cervix (the entrance from the vagina
an extremely contagious virus that is transmitted through sexual contact. Once diagnosed, as long as it is treated early and controlled
According to Human Papillomavirus and Related Cancers, Fact Sheet 2018, it was reported that there are 50.33 million women aged
cervical cancer every year and 10403 dies from the illness. It is estimated that approximately 3.5% of women in the general population in
Nigeria are projected at a given time to have harbor cervical HPV-16/18 infection and 66.9% of invasive cervical cancers are due to HPVs
-
ease and screening methods.
Materials and Methods
University College Hospital (UCH) of the University of Ibadan, Nigeria is a leading tertiary Centre for cancer care in Nigeria and as
extracted from medical records of cervical cancer in this hospital. This study examined 310 patients with cervical cancer from reproduc-
tive-age women and other data with considered variables such as age and survival time which is the period used in admission and this is
measured in a month. The months are categorized as follows: stage I = 12 months, stage II = 24 months, stage III = 36 months, stage IV =
48 months, and stage V = 60 months and above. The ages of the patients are also grouped into 4 as shown in table 1. The mortality by the
disease is a binary event indicator (response) measured alive = 1 and dead = 0 with continuous explanatory variables that are captured
nonparametrically.
The Bayesian nonparametric regression model was applied to estimate cervical cancer women with their age and the survival period
with cervical cancer measured in several months. Age of the woman and this period are modeled by basic functions (
) ap-
is the seasonal effect
and
j
-
proximated by a polynomial spline of degree
min 0 1 1 max
...
mm
xx
ζζ ζ ζ
−
= < << < =
. To ensure
-
that assumption of a strictly linear effect on the continuous variables (woman‘s age and length of stay) is not appropriate, non-parametric
modeling of cervical cancer is then developed and compactly written as
( ) ( )
period ' age (1)
ij
mortality f time f woman s= +
The estimated nonlinear functions
depend considerably on the hyperprior assumptions with the choice of hyperparam-
eters
the patient status regresses nonparametrically on woman’s age and survival period of the disease in Bayesian approach.