This is the authors' abstract. We don't add key points for this paper.
BACKGROUND: The prognostic performance of the Simplified Acute Physiology Score (SAPS) models has not been thoroughly investigated. We systematically reviewed and meta-analysed SAPS mortality prediction metrics across diverse intensive care unit (ICU) populations worldwide. METHODS: PubMed and Scopus were queried for external validations of SAPS in ICU settings. The "CHARMS" and "PROBAST" tools were applied for data extraction and risk-of-bias assessment, respectively. C-statistics and observed-to-expected ratios were summarised using random-effects models. Heterogeneity sources were explored using meta-regression. Evidence certainty was evaluated using the "GRADE" approach. PROSPERO registration CRD42023446721. RESULTS: Ninety studies (89 articles) were reviewed, totalling 565,994 ICU patients across 87 countries, and validating the SAPS II (n = 35), SAPS 3 global (n = 37), and/or SAPS 3 region-specific (n = 28) models. Endpoints were in-hospital (n = 70), in-ICU (n = 16), and/or 28/30-day mortality (n = 12). Nine studies performed recalibrations of SAPS II and 7 of SAPS 3. The meta-analyses provided high-certainty evidence of excellent discriminatory ability across all SAPS models in predicting in-hospital mortality, whereas data on unsatisfactory overall calibration were graded as low certainty. Successful recalibration was graded as high certainty for SAPS II and moderate certainty for SAPS 3. For ICU and 28/30-day mortality, the evidence was low to very low certainty for all SAPS models. Five decision curve analyses were conducted in three studies, all yielding positive net benefit across meaningful risk thresholds. CONCLUSIONS: SAPS models remain reliable as mortality risk-ranking tools in critical care. With adequate sample sizes, local recalibrations and dynamic updating of in-hospital mortality predictions are essential to ensure clinical utility. REGISTRATION: PROSPERO CRD42023446721.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Kokkinakis et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: