Paper Type

Doctoral Dissertation


Brooks College of Health

Degree Name

Doctor of Clinical Nutrition (DCN)


Nutrition & Dietetics

NACO controlled Corporate Body

University of North Florida. Department of Nutrition & Dietetics

First Advisor

Dr. Lauri Wright

Second Advisor

Dr. Philip Foulis

Third Advisor

Dr. Susan Goldsmith

Fourth Advisor

Dr. James Epps

Department Chair

Dr. Lauri Wright

College Dean

Dr. Curt Lox


Background: Muscle mass is highly correlated with patient outcomes. Techniques to identify patients with low muscularity include computed tomography (CT) and bioelectrical impendence analysis (BIA) however disadvantages of cost, exposure to radiation and access make these measurements unavailable to the average dietitian. Urinary creatinine excretion (UCE) and estimation of creatinine height index (CHI) are strongly associated with muscularity and outcomes, however, require a 24-hour urine collection. The postulation that UCE may be predicted from patient variables, through mathematical modeling, would avoid the need for a 24-hour urine collection and may be clinically useful.

Methods: Input variables of age, height, weight, gender, plasma creatinine, urea nitrogen, glucose, sodium, potassium, chloride and carbon dioxide from a deidentified data set of 967 patients who had UCE measured were used to develop models to predict UCE. The model identified with the best predictive ability was validated using four-fold cross validation and a separate data set not used to construct the model. Model predicted UCE and CHI were compared to measures of muscularity. The model was then retrospectively applied to a convenience sample of 120 critically ill veterans to examine prevalence of low muscle mass and if UCE and CHI were associated with outcomes.

Results: A model to estimate UCE was identified utilizing the input variables of plasma creatinine, plasma BUN, age and weight was found to be highly correlated, moderately predictive of UCE and statistically significant. Model predicted UCE was found to be highly correlated with accepted measures of muscularity. Applying the model to a cohort of subjects identified that 44.2% of subjects had severe sarcopenia. Subjects with model estimated CHI 60% were found to have significantly lower body weight, BMI, plasma creatinine, albumin and prealbumin levels. Subjects with CHI 60% were found to be 8.0 times more likely to be diagnosed with malnutrition and 2.6 times more likely to be readmitted in 6 months. Subjects with low CHI trended towards longer hospital and ICU LOS, however it did not meet statistical significance.

Conclusion: The development of a model which predicts UCE and correlates with muscle mass offers a novel method for the RDN to identify patients with sarcopenia on hospital admission. This method could allow the RDN to quickly screen new admissions without the use of CT or DEXA scans and without the inconvenience of a 24-hour urine collection by using readily available patient variables.