Introduction: Systemic disparities in healthcare delivery and adoption of best practices affect mortality rates in critically ill patients. Higher mortality in low- and middle-income countries (LMICs) compared to high-income countries (HICs) can be largely explained due to this. Clinical scores like the SOFA and APACHE II, developed in resource-rich environments, may not perform equally across diverse clinical settings. To address variability in care, checklist-based QI initiatives such as the Checklist for Early Recognition and Treatment of Acute Illness and iNjury (CERTAIN) aim to standardize evidence-based practices. We assessed the predictive validity of these scores across mixed-income settings and evaluated whether the implementation of CERTAIN enhances their prognostic performance.
Methods: Secondary analysis of the multinational CERTAIN study database, which included data from 34 ICUs across 18 countries (11 LMICs and 5 HICs) between 2013 and 2017. SOFA and APACHE II scores were documented at ICU admission, and corresponding ICU and hospital mortality outcomes were collected. The discriminative performance of the scores was assessed using AUROC, with comparisons between HICs and LMICs, as well as pre- and post-CERTAIN implementation, analyzed using DeLong's test. Calibration was evaluated using the Hosmer-Lemeshow test.
Results: Among the 4,204 patients, the median age was 62 years (47.1 - 74.7); 59% were female, and 77% were from LMICs. The SOFA score showed fair discrimination for ICU mortality (AUROC 0.68; 95% CI, 0.66-0.70), performing better in HICs than LMIC (0.79 vs. 0.65; p < 0.0001). Its AUROC slightly increased post-CERTAIN (0.66 to 0.70; p = 0.08) with no significant difference by income group. APACHE II showed modest discrimination for hospital mortality (AUROC 0.66; 95% CI, 0.64-0.68), with similar performance across settings (0.65 vs. 0.66; p = 0.80). Post-CERTAIN, APACHE II AUROC improved overall (0.62 to 0.69; p < 0.001), mainly in LMICs (0.61 to 0.69; p < 0.001), with no change in HICs (0.65 to 0.67; p = 0.77).
Conclusions: Implementation and adoption of CERTAIN can enhance predictive validity of APACHE II in critically ill patients, particularly in LMICs. These results advocate for scalable, evidence-based interventions to promote equitable critical care outcomes globally.
