The Norton Scale at Admission Improves APACHE II-Based Mortality Prediction in the Intensive Care Unit: A Retrospective Cohort Study of 5775 Patients

Background and Objectives: Malnutrition and frailty affect 30–55% of intensive care unit (ICU) patients, yet formal nutritional screening remains inconsistently implemented in routine ICU admission workflows. The APACHE II score, the standard measure of acute physiological severity, does not capture pre-existing nutritional status or functional reserve. The Norton scale, routinely recorded by nursing staff for pressure-ulcer screening, could serve as a pragmatic proxy for the nutritional-functional axis. We assessed its independent prognostic value at admission for in-hospital and post-ICU mortality. Materials and Methods: Retrospective cohort study of 5775 consecutive adult patients admitted to a Spanish tertiary polyvalent ICU between 2012 and 2019, with APACHE II and Norton scores recorded at admission. The Norton was analysed as continuous and categorised (minimal >14, medium 13–14, high 10–12, very high 5–9). Discrimination was assessed by AUC and DeLong’s test, predictive improvement by IDI and NRI, and internal validity by bootstrap resampling (B = 200). Results: Hospital mortality was 12.8% (n = 738), rising from 7.7% in patients with minimal-risk Norton to 34.5% in very high risk. After adjustment for APACHE II, each additional Norton point reduced the odds of death by 7.7% (adjusted OR = 0.923; 95% CI 0.903–0.943). Adding the Norton to APACHE II improved discrimination (AUC 0.865 → 0.872; DeLong p = 0.003; IDI = 0.011; continuous NRI = 0.30). In the highest APACHE II quartile, the absolute mortality difference between minimal and very high Norton categories reached 23.2 percentage points. The Norton’s prognostic effect was approximately twice as large for post-ICU mortality (ΔAUC +0.011) as for overall in-hospital mortality (ΔAUC +0.006). Conclusions: The Norton scale at admission improves the prognostic capacity of APACHE II in critically ill patients, particularly for post-ICU mortality. Its widespread availability without additional patient-level data collection positions it as a pragmatic candidate for routine prognostic assessment and for guiding targeted nutritional screening.

Description

Author Contributions: Conceptualization, M.T.-A., J.J.d.L.-B. and M.T.M.-L.; methodology, M.T.- A., J.J.d.L.-B. and M.P.-B.; software, J.P.-M. and I.V.; validation, M.P.-B., J.P.-M. and B.R.-G.; formal analysis, J.J.d.L.-B., M.P.-B. and I.V.; investigation, M.T.-A., S.G.-V., S.L.-A., A.A.T.-V. and B.R.-G.; resources, M.T.-A. and M.T.M.-L.; data curation, J.P.-M., S.L.-A. and A.A.T.-V.; writing—original draft preparation, M.T.-A., J.J.d.L.-B. and M.P.-B.; writing—review and editing, M.M.-P., A.M.P.-F., S.G.-V., B.R.-G., I.V. and M.T.M.-L.; visualization, M.P.-B. and J.P.-M.; supervision, M.T.-A., M.T.M.-L. and M.P.-B. All authors have read and agreed to the published version of the manuscript.