Development and Validation of a Nigerian-Specific Sepsis Mortality Prediction Model Incorporating Clinical, Microbiological, and Host Genomic Variables

Existing sepsis mortality prediction tools such as SOFA and APACHE II were developed and validated predominantly in high-income country intensive care settings and show suboptimal predictive accuracy in Nigerian resource-limited critical care environments characterised by different pathogen ecology, delayed presentation, and limited organ support. This study develops and validates a Nigerian-specific sepsis mortality prediction model incorporating clinical, microbiological, and host genomic variables in a multi-centre prospective cohort. Twelve intensive care units and high-dependency units across four geopolitical zones enrol 1,500 septic patients over 30 months. Candidate predictors include SOFA score, lactate, pre-hospital antibiotic use, time to first antibiotic dose, blood culture pathogen species and resistance profile, procalcitonin, host single-nucleotide polymorphisms in TNF-alpha and IL-6 promoter regions, and malaria co-infection status. The primary outcome is 28-day in-hospital mortality. Predictor selection uses LASSO regression, and the final model is developed using multivariable logistic regression with internal validation by bootstrap resampling and external validation in a hold-out cohort. Calibration is assessed using Hosmer-Lemeshow and calibration belt analysis. The Sepsis-3 definitions and Host-Pathogen Interaction Theory provide the study's theoretical underpinning. Original contributions include the first genomics-augmented sepsis prediction model calibrated to Nigerian critical care context. Findings will be implemented as a clinician decision support tool for sepsis triage. Keywords: sepsis, mortality prediction model, intensive care, Nigeria, host genomics.