Criterion validity of the visual estimation method for determining patients' meal intake in a community hospital

The accuracy of the visual estimation method is unknown, even though it is commonly used in hospitals to measure the dietary intake of patients. We aimed to compare the difference in the validity of visual estimation according to the raters' job categories and tray divisions, and to demonstrate associations between meal characteristics and validity of visual estimation in a usual clinical setting in a community hospital.

We collected patients' dietary intake data in usual clinical settings for each tray in 3 ways: visual estimation by nursing assistants, visual estimation by dietitians, and weighing by researchers (reference method). Dietitians estimated the dietary intake using 2 divisions, namely, whole tray and food items. Then we compared the weights and visual estimation data to evaluate the validity of the visual estimation method.

Mean nutrient consumption of target trays was significantly different when using the visual estimation of target trays than when using the weighed method (visual estimation by nursing assistants [589 ± 168 kcal, 24.3 ± 7.0 g/tray, p < 0.01], dietitians' whole trays [561 ± 171 kcal, 23.0 ± 6.9 g/tray, p < 0.05], food items [562 ± 171 kcal/tray, p < 0.05], and dietitians' food items [23.4 ± 7.3 g/tray, p = 0.63]). Spearman's correlations for both methods were very high for energy (ρ = 0.91-0.98, p < 0.01) and protein intakes (ρ = 0.88-0.96, p < 0.01), respectively. The limits of agreement in the Bland-Altman plot for both dietary intake categories were -121 kcal to 147 kcal/tray and -6.4 g to 7.0 g/tray (nursing assistants, whole division), -122 kcal-106 kcal/tray and -6.7 g to 5.5 g/tray (dietitians, whole divisions), and -82 kcal to 66 kcal/tray and -4.3 g to 3.9 g/tray (dietitians, food items divisions). High intake rate of grains was significantly associated with decreased odds of a difference between two methods based on the nursing assistant's whole tray evaluation (odds ratio [OR]: 0.85; 95% confidence interval [CI]: 0.76-0.94) and the dietitians' whole tray (OR: 0.80; 95% CI: 0.72-0.89) and food items evaluations (OR: 0.64; 95% CI: 0.56-0.73), respectively. In addition, minced meals were also associated with a difference between two methods, for the nursing assistants' whole tray (OR: 3.53; 95% CI: 1.66-7.51) and dietitians' food items (OR: 2.92; 95% CI: 1.37-6.22).

Visual estimation by nursing assistants and dietitians correlated highly with the weighing method although the limits of agreement were wide. Nursing assistants and dietitians should pay attention to low consumption and modified texture meals when evaluating dietary intake using the visual estimation method.