Annisa Auliarahmi, Purhadi, Shofi Andari, Muhammad Fikri, Pitriani

Identifikasi Faktor Risiko Kematian Ibu dan Neonatal di Kalimantan Menggunakan Model BPGIGR dengan Algoritma BHHH

Introduction

Identifikasi faktor risiko kematian ibu dan neonatal di kalimantan menggunakan model bpgigr dengan algoritma bhhh. Identifikasi faktor risiko kematian ibu & neonatal di Kalimantan menggunakan model BPGIGR dengan algoritma BHHH. Menemukan faktor signifikan seperti posyandu aktif, BBLR, ASI eksklusif, kunjungan K4, & tablet zat besi.

Abstract

Poisson regression is widely applied for modeling count data and requires the strict assumption of equidispersion, meaning that the mean and variance of the data must be equal. In practice, this condition is rarely satisfied. To address this issue, the Bivariate Poisson Generalized Inverse Gaussian Regression (BPGIGR) model was developed by combining the Poisson distribution with the Generalized Inverse Gaussian (GIG) distribution to overcome overdisperion in two correlated response variables. This study aims to obtain parameter estimates and corresponding test statistics for the BPGIGR model by incorporating two exposure variables to account to account for differences in population size across analytical units. Parameter estimation is performed using the Maximum Likelihood Estimation (MLE) method with the Berndt-Hall-Hall-Hausman (BHHH) algorithm. The BPGIGR model is implemented on maternal and neonatal deaths in Kalimantan in 2024 to identify the significant contributing factors. The results indicate that the model is influenced by the percentages of active posyandu, low birth weight, complete neonatal visits, exclusive breasfeeding, K4 visits, and pregnant women receiving iron tablets with an AICc of 9.719,092.

Review

This study presents a timely and methodologically advanced approach to understanding the complex factors contributing to maternal and neonatal mortality in Kalimantan. Acknowledging the common challenge of overdispersion in count data, which often violates the strict assumptions of standard Poisson regression, the authors employ the Bivariate Poisson Generalized Inverse Gaussian Regression (BPGIGR) model. This sophisticated statistical framework is particularly apt for simultaneously modeling two correlated response variables—maternal and neonatal deaths—while effectively addressing issues of overdispersion. The application of such a robust model to a critical public health issue like maternal and neonatal mortality underscores the study's significant practical relevance. The methodological core of this research lies in the BPGIGR model, which ingeniously combines the Poisson distribution with the Generalized Inverse Gaussian (GIG) distribution to accommodate overdispersed, correlated count data. A key strength is the incorporation of two exposure variables, crucial for accounting for population size variations across analytical units and thus ensuring more accurate rate estimation. Parameter estimation is rigorously performed using the Maximum Likelihood Estimation (MLE) method, facilitated by the Berndt-Hall-Hall-Hausman (BHHH) algorithm. This choice of algorithm suggests a computationally intensive, yet robust, approach to deriving model parameters, which is appropriate for complex multivariate models dealing with non-standard distributions. The findings of this study provide valuable insights into the determinants of maternal and neonatal deaths in Kalimantan, specifically identifying several key modifiable factors. The model indicates that the percentages of active posyandu, low birth weight, complete neonatal visits, exclusive breastfeeding, K4 visits (antenatal care), and pregnant women receiving iron tablets are significant contributors. These results are highly actionable, pointing towards critical areas for public health intervention and resource allocation to improve maternal and child health outcomes. The reported AICc of 9,719.092 provides a metric for model fit, suggesting a reasonably well-fitting model within its context. Overall, this paper offers a significant contribution by applying an advanced statistical model to a vital public health problem, yielding evidence-based recommendations for policy and practice.

Full Text

Comments

You need to be logged in to post a comment.