Table 3.
Accuracy of the methods used to assess EE, expressed as bias, precision, and accuracy and inaccuracy rates
| Bias | Precision | Accuracy quantified | ||||||
|---|---|---|---|---|---|---|---|---|
| Mean difference in kcal/day, 95% CI | Mean difference (% of gold standard EE) | SD of bias (Levene’s F-testa) | Bland–Altman limits of agreement | Accuracy rates | Inaccuracy rates | |||
| <10 % | <15 % | >25 % | >30 % | |||||
| Method | ||||||||
| EE:VCO2 b | +141, +107 to +174 | 7.7 % | 153 | −166 to +447 | 61 % | 79 % | 2 % | 0 % |
| EE:HBc | +246 | 13.5 % | 263 | −722 to +280 | 31 % | 52 % | 13 % | 5 % |
| −303 to −189 | p < 0.001 | (F 14.1) | p = 0.001 | p < 0.001 | p < 0.002 | p < 0.01 | ||
| p < 0.001 | ||||||||
| EE:Esp25d | +156, | 8.6 % | 344 | −531 to +843 | 40 % | 56 % | 25 % | 14 % |
| +81 to +230 | p = 0.709 | (F 31.1) | p = 0.009 | p = 0.002 | p < 0.001 | p < 0.001 | ||
| p < 0.001 | ||||||||
| EE:Faisye | +176, | 9.7 % | 218 | −260 to + 612 | 45 % | 61 % | 17 % | 12 % |
| +129 to +233 | p = 0.226 | (F 9.0) | p = 0.01 | p = 0.003 | p < 0.001 | p < 0.001 | ||
| p = 0.003 | ||||||||
| EE:PSUf | −22, | 1.2 % | 254 | −529 to +458 | 54 % | 75 % | 10 % | 6 % |
| −77 to +33 | p < 0.001 | (F 11.9) | p = 0.341 | p = 0.582 | p = 0.05 | p = 0.023 | ||
| p < 0.001 | ||||||||
| Post hoc calculation | ||||||||
| EE:HB15g | −10 | 0.5 % | 257 | −524 to +504 | 55 % | 71 % | 10 % | 6 % |
| −66 to +46 | p < 0.001 | (F 12.3) | p = 0.435 | p = 0.285 | p = 0.05 | p = 0.023 | ||
| p < 0.001 | ||||||||
EE energy expenditure, EE:Esp25 energy expenditure calculated with the European Society for Clinical Nutrition and Metabolism guideline equation of 25 kcal/kg/day, EE:Faisy energy expenditure calculated with the Faisy equation, EE:HB15 energy expenditure calculated with the Harris–Benedict equation with 15 % added, EE:PSU energy expenditure calculated with the Penn State University 2003b equation, EE:VCO 2 energy expenditure from ventilator-derived volume of carbon dioxide and nutritional respiratory quotient, HB Harris–Benedict equation
Less than 10 % and less than 15 % accuracy rates represent the proportion of patients for which EE:VCO2 (or equation-based EE) predicted EE within 10 % and within 15 %, respectively, of gold standard EE:Calorimetry. Greater than 25 % and greater than 30 % inaccuracy rates represent the proportion of patients for whom EE:VCO2 (or equation-based EE) differed by more than 25 % and more than 30 %, respectively, from gold standard EE:Calorimetry
All p values are relative to EE:VCO2.
F-test and p value reflect the comparison of the variance of the mean difference of EE:VCO2 and EE from equations. The higher the F-value, the higher the difference of the variances. p < 0.05 indicates that the variance of the mean difference is significantly different from EE:VCO2.
aLevene’s F-test on similar variances
Bias: b vs. c p < 0.001; b vs. d p = 0.709; b vs. f p = <0.001; b vs. e p < 0.001; b vs. g p = 0.226; c vs. d p < 0.001; c vs. f p < 0.001; c vs. e p < 0.001; c vs. g p < 0.001; d vs. f p = 0.001; d vs. e p < 0.001; d vs. g p = 0.652; f vs. e p = 0.762; e vs. g p < 0.001