Baseline Clinical Predictors and Toxicological Drivers of In‐Hospital Poisoning Mortality in Resource‐Constrained Emergency Care: A Firth Penalized Regression Analysis

Background: Evidence regarding clinical characteristics and determinants of mortality among patients with acute poisoning in agricultural regions of Ethiopia remains limited. This study aimed to evaluate baseline clinical predictors associated with in-hospital mortality among acutely poisoned patients in a low-resource tertiary emergency department. Methods: A retrospective observational study was conducted among consecutive acute poisoning cases presenting to Asella Referral and Teaching Hospital (October 2022-September 2024). Of 372 screened cases, 7 patients who left against medical advice were excluded, yielding a final analytical cohort of 365 patients. Sociodemographic, toxicological, and clinical data were extracted from medical records. The primary outcome was in-hospital mortality. Multivariable Firth penalized logistic regression was performed to minimize small-sample bias given sparse outcome events (15 deaths). A sensitivity model additionally adjusted for sex and rural residence. Results: The median age of patients was 22.0 years (IQR: 19.0-27.0); 230 (61.8%) were aged 15-24 years, 199 (53.5%) were female, and 245 (65.9%) were rural residents. Insecticides were the predominant poisoning category (157; 42.2%). Aluminum Phosphide (AlP) exposure occurred in 43 (11.6%) patients, resulting in 12 deaths (case-fatality rate: 27.9%; 95% CI 15.3%-43.7%). Overall in-hospital mortality was 4.1% (15/365; 95% CI 2.3%-6.6%), with AlP accounting for 80.0% of all cohort deaths. In the primary multivariable Firth model, AlP exposure (aOR 30.48, 95% CI 7.88-117.95) and admission Glasgow Coma Scale < 15 (aOR 19.90, 95% CI 5.09-77.88) were strongly and independently associated with in-hospital mortality. These associations remained robust in sensitivity analyses. Conclusions: Impaired consciousness at admission (GCS < 15) and Aluminum Phosphide exposure serve as critical baseline indicators for rapid, low-cost risk stratification. While these exploratory parameters highlight high-risk groups, prospective multi-center validation across diverse hospital tiers is required before formal integration into clinical triage algorithms.