Esophageal thermometry outperforms bladder and tympanic monitoring during targeted temperature management and shock in the ICU.

BACKGROUND

Accurate core temperature management is fundamental to intensive care unit (ICU) practice, yet the dynamic reliability of less-invasive surrogate metrics during critical illness remains under-investigated. This study evaluated the accuracy and potential confounders of esophageal, bladder, and tympanic thermometry against intravascular temperature.

METHODS

A prospective longitudinal study was conducted in 22 ICU patients receiving concurrently multi-site core temperature monitoring. High-resolution temperature and hemodynamic data were extracted over up to 72 h. Absolute agreement and dynamic trending were assessed using Bland-Altman and four-quadrant analyses, with confounders evaluated via Spearman correlation matrices.

RESULTS

Among 833 paired measurements, esophageal thermometry demonstrated near-perfect agreement with intravascular temperature (bias 0.11 °C; 95% limits of agreement [LoA] -0.19 to 0.41 °C; Lin's concordance correlation coefficient [CCC] = 0.957). In contrast, bladder and tympanic modalities showed wider 95% LoAs (-0.32 to 0.82 °C and -0.62 to 0.55 °C, respectively) and context-dependent reliability. During targeted temperature management (TTM), bladder temperature exhibited marked thermal lag (trending concordance rate, 50.2%), driven by cardiac output, where decreased cardiac output increased delays between core and bladder temperature during cooling and re-warming (r = -0.42, P < 0.05). Tympanic thermometry errors were independent of hemodynamic parameters, showing no correlation with cardiac index (r = 0.08, P = 0.218) or norepinephrine dosage (r = 0.13, P = 0.052).

CONCLUSIONS

Esophageal thermometry is a highly reliable non-invasive surrogate for intravascular temperature. Conversely, bladder thermometry is susceptible to flow-dependent thermal inertia during shock and TTM. Meanwhile, tympanic measurements exhibit wide, hemodynamically-independent variability, rendering them unreliable for continuous monitoring.