A comparison of different scores for mortality prediction of acutely poisoned patients: a systematic review and meta-analysis

Acute poisoning typically accounts for 1%-3% of all emergency department (ED) visits, and comprise 4%-40% of admissions to intensive care units (ICU), with a mortality rate of 3%-6%. Accurate assessment of patient prognosis enables the early implementation of appropriate interventions and the effective allocation of limited resources, thereby preventing adverse outcomes. However, it remains unclear which tool offers superior predictive accuracy for the prognosis of poisoned patients. In this article, we review existing assessment tools used to predict mortality risk in poisoned patients and compare their performance. We conducted comprehensive searches in PubMed, EMBASE, Ovid, Scopus, Cochrane Library, CNKI, Wanfang Data, and SinoMed databases from their inception up to January 2025. Studies were included if they reported the performance of at least one scoring systems for predicting mortality in patients with acute poisoning. The PRISMA guidelines were followed (PROSPERO registration: CRD42024579941). Data of 60,403 patients across 65 studies were eligible for inclusion. The risk assessment tools reported in more than three studies included APACHE II (47), SOFA (19), SAPS II (11), PSS (16), MEWS (8), REMS (5), new-PMS (5). Significant heterogeneity was observed in the pooled analysis. In this study, PSS exhibited moderate sensitivity and specificity in predicting mortality among patients with acute poisoning, while MEWS demonstrated the highest sensitivity, and new-PMS showed the strongest specificity. The highest AUC values were observed for MEWS and APACHE II. Based on these findings, MEWS and new-PMS may represent the optimal tools for predicting in-hospital/28-day/30-day mortality in poisoned patients.