Comparing predictive energy-expenditure equations with indirect calorimetry in older adults living in long-term care facilities

Accurate estimation of resting energy expenditure (REE) in long-term care facilities (LTCFs) is challenging, and indirect calorimetry (IC) is seldom available. This cross-sectional study assessed the agreement between predictive equations and IC in older adults residing in LTCFs. REE was measured by IC and compared with estimates from 22 predictive equations. Agreement was evaluated using linear regression (R 2 ), root mean square error (RMSE), Bland–Altman analysis, individual-level accuracy (90–110% of measured REE), mean percentage bias (%Bias), Lin's concordance correlation coefficient (CCC), and extreme prediction errors. A total of 182 residents were enrolled (72% women; mean age 81.24 ± 9.21 years; BMI 23.05 ± 5.11 kg/m 2 ). Across equations, R 2 ranged from 0.24 to 0.41, RMSE from 230 to 317 kcal/d, and individual accuracy from 27% to 45%. Mean percentage bias ranged from −6% to +18%, indicating systematic under- and overestimation across models. The most consistent overall performance was observed for Lührmann, Müller, and Liu equations. Nonetheless, Bland–Altman analysis revealed wide limits of agreement, and CCC values indicating poor-to-moderate concordance (0.465–0.624). Several equations also produced extreme errors exceeding ±1100 kcal/d, demonstrating limited interchangeability with IC at the individual level. In people living in LTCFs, predictive equations showed modest explanatory power and low individual accuracy, with wide limits of agreement indicating that none can reliably replace IC when precise estimation of REE is required. Clinically, pREE should be regarded as an estimate rather than a dose, requiring cautious adjustment and verification against short-term monitoring of body weight, intake, and clinical status. When precise energy dosing is critical, IC should be used whenever feasible. • Indirect calorimetry is rarely available in long-term care facilities. • Predictive equations explained ≤41% of REE variability. • The best-performing formulas were Lührmann, Müller, and Liu. • Individual accuracy was low, with ≤45% within ±10% of measured REE. • Bland–Altman showed wide limits, limiting clinical interchangeability.