Predicting hospital length of stay at admission using global and country-specific competing risk analysis of structural, patient, and nutrition-related data from utritionDay 2007-2015

Journal article

Hospital length of stay (LOS) is an important clinical and economic outcome and knowing its predictors could lead to better planning of resources needed during hospitalization. This analysis sought to identify structure, patient, and nutrition-related predictors of LOS available at the time of admission in the global nutritionDay dataset and to analyze variations by country for countries with n > 750. Data from 2006-2015 (n = 155,524) was utilized for descriptive and multivariable cause-specific Cox proportional hazards competing-risks analyses of total LOS from admission. Time to event analysis on 90,480 complete cases included: discharged (n = 65,509), transferred (n = 11,553), or in-hospital death (n = 3199). The median LOS was 6 days (25th and 75th percentile: 4-12). There is robust evidence that LOS is predicted by patient characteristics such as age, affected organs, and comorbidities in all three outcomes. Having lost weight in the last three months led to a longer time to discharge (Hazard Ratio (HR) 0.89; 99.9% Confidence Interval (CI) 0.85-0.93), shorter time to transfer (HR 1.40; 99.9% CI 1.24-1.57) or death (HR 2.34; 99.9% CI 1.86-2.94). The impact of having a dietician and screening patients at admission varied by country. Despite country variability in outcomes and LOS, the factors that predict LOS at admission are consistent globally.

Authors: Kiss, N; Hiesmayr, M; Sulz, I; Bauer, P; Heinze, G

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: MDPI
• Journal: Nutrients
• Volume: 13
• Issue: 11
• Article number: 4111
• Place of publication: Switzerland
• Publication date: 2021-11-16
• Acceptance date: 2021-11-09
• DOI: 10.3390/nu13114111
• EISSN: 2072-6643
• ISSN: 2072-6643
• Pmid: 34836366
• Language: English
• Keywords: Aged; 80 and over; length of stay; FFR; nutritional status; diagnostic tests; predictive value of tests; proportional hazards models; nutrition assessment; routine; hospital mortality; patient admission; time factors; risk assessment