Penerapan Regresi Conway Maxwell Poisson untuk Mengatasi Overdispersi pada Jumlah Kematian Bayi di Provinsi Jawa Barat

Infant Mortality Rate (IMR) is a key public health indicator reflecting the social, economic, environmental, and healthcare service quality conditions of a population. In 2023, West Java recorded the highest number of infant deaths in Indonesia. These data are count-type in nature and are commonly analyzed using Poisson regression. However, due to the frequent occurrence of overdispersion, the Poisson method becomes less appropriate. As an alternative, the Conway-Maxwell Poisson (CMP) regression is employed, offering greater flexibility in handling violations of the equidispersion assumption. This study aims to apply CMP regression to address overdispersion in the number of infant deaths in West Java Province using the Maximum Likelihood (ML) estimation method. The data used in this study comprise the total number of infant deaths in 2023 across 27 districts and cities in West Java Province. The ML parameter estimation analysis shows that the dispersion parameter values obtained from the CMP and Poisson models are 10.92 and 126.49, respectively. In terms of model evaluation criteria, the CMP model yields an AIC of 402.455 and BIC of 415.41, whereas the Poisson model shows an AIC of 4183.46 and BIC of 4195.12. These results indicate that the CMP model outperforms the Poisson model in handling infant mortality data. Furthermore, four variables are found to be statistically significant in explaining the number of infant deaths in West Java Province, namely the percentage of antenatal care coverage (K4), the number of health facilities by district/city, the percentage of households with clean and healthy living behavior (PHBS), and the percentage of neonatal asphyxia complications, with a significance level of alpha = 5%.