Page 45 - ISMCON souvenir 2021

ISMSCON - 2021

OS1: WEIGHTED PROPENSITY SCORE MATCHING FOR LOST-
FOLLOWUP INFORMATION OF TIME TO EVENT OUTCOME

Abhipsa Tripathy Bhrigu Kumar Rajbongshi Atanu Bhattacharjee Gajendra K. Vishwakarma a
a
a
b c
a Department of Mathematics & Computing, Indian Institute of Technology Dhanbad, Dhanbad-826004, India, Mail
id: abhips.trips@gmail.com
b Section of Biostatistics, Centre for Cancer Epidemiology, Tata Memorial Centre, Navi Mumbai-410210, India,
c Homi Bhabha National Institute, Mumbai, India.

The propensity score, which is the probability that an individual is assigned one of the treatments as a
function of observed covariates, is commonly employed to adjust for confounding in large non randomized
studies. Typically, the propensity scores are employed either by stratification or matching. In presence of
censoring in survival data to reduce the effect of bias we take the help of the propensity score matching
(PSM) method to update the data set that allows comparison between treated and control subjects.

Aim of the work is to update the censored observation in a given dataset and apply several methods of a
multistate model to evaluate hazard rates and survival functions.

In this work, we are proposing the application of PSM in updating individual patients’ information about
censoring for different types of events and applying various models for the multi-state to estimate the
hazard rates in presence of covariates. A threshold probability is assumed to compare with the distribution
of the propensity scores and depending on the score function the censored information is updated.
Methods have been applied on a simulated dataset as well as a package dataset.

Mean square error of the regression coefficients reduced after using propensity score matching. Mean
estimated regression coefficients for covariate x1 obtained before and after updation of dataset using
propensity score matching are -0.4067 and -0.4504 respectively. The estimated value of the regression
coefficients for covariate “match” increases from -0.0171to -0.0055, hence the hazard ratio decreases
slightly.

Underestimating the disease complexity might give highly biased results. Hence it is important to classify
the doubtful censored cases as confirmed death to have better estimates of hazards at different time
points. PSM method has often resulted in decreasing the bias compared to crude analysis of the dataset.
Keywords: Propensity score matching, censoring, multistate model, stratified Cox PH, simulation

OS2: TITLE: FACTORS ASSOCIATED WITH ANAEMIA IN

WOMEN OF REPRODUCTIVE AGE GROUP (15-49 YEARS) OF
SOUTH-WEST BIHAR DURING COVID-19 PANDEMIC: A CROSS

SECTIONAL STUDY

Dr. Abhishek Kumar1, Dr. Swati Shikha2, Dr. D K Yadava3
1&2 Assistant Professor, Department of Community Medicine, NMCH, Sasaram, 2 Retd. Professor & Head,
Department of Community Medicine, NMCH, Sasaram
Email of presenting author: ak07mail@gmail.com

Abstract

Background: The dietary intake of women in India are often found to be much less than their nutritional
requirement. About 22.9% of women in reproductive age (15 – 49 years) in India and about 30.4% of
women in Bihar are undernourished, with a body mass index (BMI) of less than 18.5 kg/m2. Objectives:
To find out the association of various factors with anaemia among the women of reproductive age group

CONFERENCE SOUVENIR 43