Prediction of early- and long-term mortality in adult patients acutely admitted to internal medicine: NRS-2002 and beyond

BACKGROUND & AIMS: In hospitalized patients malnutrition is a risk factor for adverse clinical outcomes. The Nutritional Risk Screening 2002 (NRS-2002) represents a quick and simple tool to identify malnutrition risk in this population. No study tested the predictive power of NRS-2002 on mortality adjusting for confounders related to patient's complexity, thus considering conditions such as functional status, illness-related severity and inflammation. The aim of this study was to explore the independent prognostic power and the relative weight of NRS-2002 screening tool to predict inhospital and post-discharge (up to 1 year) mortality, adjusting for variables representing the non-disease specific multidimensional complexity of patients admitted to Internal Medicine wards. METHODS: Retrospective observational study including 5698 consecutive patients acutely admitted to an Internal Medicine Department. Logistic regression models were run to test the predictive power of the NRS-2002 on patient mortality at different time intervals, adjusted for age, sex, Charlson comorbidity index, Glasgow Prognostic Score (GPS), BUN/creatinine ratio, Modified Early Warning Score (MEWS), and Norton index. The performance of the logistic models in predicting mortality was measured through the c-statistic. The different time of death between patients scored upon admission as NRS-2002 < 3 or ≥3 was evaluated through crude Kaplan-Meier curves and multivariate Cox proportional hazard analysis. RESULTS: Patients classified at high malnutrition risk (NRS-2002 ≥ 3) showed a higher and earlier mortality (Log-rank test: p < 0.001) compared to subjects in the NRS-2002 "low-risk" group. NRS-2002 ≥ 3 was an independent significant (p < 0.01) predictor of mortality in logistic regression at every time interval. Among the considered covariates, Charlson index, GPS and Norton scale showed a steadily higher OR than NRS-2002 in predicting both early and late mortality. The multivariate models demonstrated a very good discrimination for hospital and mid-term (up to 90 days) mortality. Being classified at risk for malnutrition (NRS-2002 ≥ 3) on admission independently increased the risk of one-year death (HR = 1.431; 95% CI: 1.277-1.603; p < 0.001) compared to the patients who were scored at low malnutrition risk. CONCLUSIONS: Malnutrition risk identified upon hospital admission by NRS-2002 independently contributes to early and late mortality in a population including a majority of elderly. However, risk of malnutrition has to be considered according to other factors related to comorbidities, functional status, illness severity and inflammation which reciprocally interact, concurring at worsening patient's outcome.