The association between nutritional adequacy and 28-day mortality in the critically ill is not modified by their baseline nutritional status and disease severity

Background During the initial phase of critical illness, the association between the dose of nutrition support and mortality risk may vary among patients in the intensive care unit (ICU) because the prevalence of malnutrition varies widely (28 to 78%), and not all ICU patients are severely ill. Therefore, we hypothesized that a prognostic model that integrates nutritional status and disease severity could accurately predict mortality risk and classify critically ill patients into low- and high-risk groups. Additionally, in critically ill patients placed on exclusive nutritional support (ENS), we hypothesized that their risk categories could modify the association between dose of nutrition support and mortality risk. Methods A prognostic model that predicts 28-day mortality was built from a prospective cohort study of 440 patients. The association between dose of nutrition support and mortality risk was evaluated in a subgroup of 252 mechanically ventilated patients via logistic regressions, stratified by low- and high-risk groups, and days of exclusive nutritional support (ENS) [short-term (≤ 6 days) vs. longer-term (≥ 7 days)]. Only the first 6 days of ENS was evaluated for a fair comparison. Results The prognostic model demonstrated good discrimination [AUC 0.78 (95% CI 0.73–0.82), and a bias-corrected calibration curve suggested fair accuracy. In high-risk patients with short-term ENS (≤ 6 days), each 10% increase in goal energy and protein intake was associated with an increased adjusted odds (95% CI) of 28-day mortality [1.60 (1.19–2.15) and 1.47 (1.12–1.86), respectively]. In contrast, each 10% increase in goal protein intake during the first 6 days of ENS in high-risk patients with longer-term ENS (≥ 7 days) was associated with a lower adjusted odds of 28-day mortality [0.75 (0.57–0.99)]. Despite the opposing associations, the mean predicted mortality risks and prevalence of malnutrition between short- and longer-term ENS patients were similar. Conclusions Combining baseline nutritional status and disease severity in a prognostic model could accurately predict 28-day mortality. However, the association between the dose of nutrition support during the first 6 days of ENS and 28-day mortality was independent of baseline disease severity and nutritional status. Electronic supplementary material The online version of this article (10.1186/s13054-019-2500-z) contains supplementary material, which is available to authorized users.