AN INVESTIGATION OF ARTIFICIAL INTELLIGENCE-ASSISTED PREDICTION MODELS FOR NEONATAL SEPSIS IN INTENSIVE CARE UNITS

Abstract

Neonatal sepsis remains a major clinical challenge in intensive care units because delayed recognition and treatment can rapidly lead to severe complications and mortality. Conventional approaches to identifying neonatal sepsis rely on clinical assessment, laboratory investigations, and continuous monitoring, yet these approaches may be constrained by nonspecific early symptoms and the complex clinical condition of neonates. Artificial intelligence (AI)-assisted prediction models offer an emerging approach for identifying combinations of clinical indicators that may signal elevated sepsis risk before overt deterioration becomes apparent. This study investigates healthcare professionals' perceptions of AI-assisted neonatal sepsis prediction models and examines the factors influencing their clinical adoption in intensive care settings. A quantitative cross-sectional survey design was adopted, targeting hospital-based healthcare professionals working in Lahore, Pakistan, including pediatricians, neonatologists, nurses, medical officers, and other clinical staff involved in neonatal care. For the illustrative analysis, data from 200 respondents were considered. A structured questionnaire measured perceived predictive usefulness, perceived accuracy, interpretability, trust, perceived ease of use, organizational readiness, and intention to adopt AI-assisted prediction models. Descriptive statistics, Pearson correlation, multiple regression, and subgroup analysis were used to examine relationships among the study variables. The illustrative findings indicate that perceived predictive usefulness, model accuracy, interpretability, professional trust, and organizational readiness were positively associated with clinical adoption intention. Perceived usefulness emerged as the strongest predictor, followed by trust and organizational readiness. Respondents generally supported AI as a clinical decision-support tool but emphasized that AI-generated predictions should complement rather than replace professional judgment. The study concludes that successful adoption of AI-assisted neonatal sepsis prediction models depends not only on predictive performance but also on transparency, explainability, clinician trust, workflow compatibility, staff training, and institutional readiness. The findings provide a Lahore-based perspective on the organizational and professional requirements for responsible implementation of AI-enabled neonatal clinical decision support.