Premium
Refeeding Hypophosphatemia in Patients Receiving Parenteral Nutrition: Prevalence, Risk Factors, and Predicting Its Occurrence
Author(s) -
Wong Gabriel J. Y.,
Pang Janelle G. T.,
Li Yuan Y.,
Lew Charles Chin Han
Publication year - 2021
Publication title -
nutrition in clinical practice
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.725
H-Index - 71
eISSN - 1941-2452
pISSN - 0884-5336
DOI - 10.1002/ncp.10559
Subject(s) - medicine , nice , receiver operating characteristic , logistic regression , hypophosphatemia , confidence interval , retrospective cohort study , risk assessment , pediatrics , intensive care medicine , computer security , computer science , programming language
Background Patients receiving parenteral nutrition (PN) support may develop refeeding hypophosphatemia (RH), and its prevalence is highly variable in the literature. Identifying at‐risk patients is crucial to minimize clinical complications. The National Institute for Health and Clinical Excellence (NICE) guidelines are used widely to assess the risk of RH, but they lack validation. We aim to (1) identify the prevalence of RH by multiple diagnostic criteria; (2) assess the predictive ability of the NICE guidelines for RH; and (3) identify important risk factors for RH and evaluate their predictive abilities for RH in a new model. Methods This is a single‐center retrospective study on adult patients with PN ≥48 hours. Prevalence of RH was determined by 4 established diagnostic criteria. Prognostic accuracy of the NICE guidelines were assessed by the area under the receiver operating characteristic (ROC) curve. Multivariable logistic regressions were performed to develop a new risk‐assessment model. Results Of 149 enrolled patients, 23%–48% (35 to 72 of 149 patients) developed RH (depending on the diagnostic criteria used). The NICE guidelines demonstrated poor discrimination across all diagnostic criteria (ROC, 0.43–0.53). Critical illness, the use of diuretics, and hypomagnesemia prior to PN were independently associated with RH. These risk factors formed the new model for predicting RH and had good discrimination (ROC 0.74; 95% confidence interval, 0.66–0.82). Conclusion Prevalence of RH varied according to established diagnostic criteria. The current NICE guidelines poorly predict the occurrence of RH, and modification is likely beneficial. A new risk‐assessment model was developed; nevertheless, further validation is required.