An Application for Spatial Frailty Models: An Exploration with Data on Fungal Sepsis in Neonates

Area of Science:

• Neonatal intensive care
• Epidemiology
• Biostatistics

Background:

• Neonatal fungal sepsis (NFS) is a significant global cause of neonatal mortality, especially in neonatal intensive care units (NICUs).
• Limited research exists on applying spatial frailty models with a Bayesian approach to identify NFS hotspots and mortality risk factors.

Purpose of the Study:

• To explore the utility of Bayesian spatial frailty models in identifying risk factors and geographical hotspots for neonatal mortality due to fungal sepsis.
• To analyze data from a cohort of neonates diagnosed with fungal sepsis in Tamil Nadu, India.

Main Methods:

• Utilized Bayesian spatial frailty models (Log-logistic, Log-normal, Weibull proportional hazard models) for analyzing a cohort of 80 neonates with fungal sepsis (2018-2020).
• Employed R version 4.1.3 and QGIS version 3.26 for risk factor identification and hotspot mapping.
• Assessed the effectiveness of parametric distributions within spatial frailty models.

Main Results:

• Abnormal activated thromboplastin levels significantly increased mortality risk across all models (HRs ranging from 18.49 to 22.12).
• Hemorrhage was also identified as a risk factor in Log-normal and Weibull models (HRs 1.65 and 1.75, respectively).
• Villivakkam, Tiruvallur, and Poonamallee blocks were identified as high-risk areas for NFS mortality.

Conclusions:

• Bayesian spatial frailty models are effective for identifying NFS mortality risk factors and quantifying their impact.
• Early detection and management of identified risk factors are crucial for improving survival rates in neonates with fungal sepsis.
• The study highlights the need for targeted interventions in identified high-risk geographical areas.