04 Dec White Paper: iThermonitor Core Temperature Accuracy in NICU Patients
Neonatal Intensive Care Unit, Chengdu Maternity and Child Health Care Hospital (Raiing Medical)-
BACKGROUND: Core temperature is normally measured at the rectum in NICUs, but patients’ rectal mucosa is vulnerable when using this measuring method. We therefore determined the precision and accuracy of a novel wireless axillary thermometer, the iThermonitor, to determine its suitability for use in NICU patients.
METHODS: We enrolled 40 patients admitted to the NICU. Core temperature was measured every 4 hours by the Philips Esophageal/Rectal Temperature Probe reusable, infant, continuous monitoring, (21076A), and estimated at the axilla with a wireless iThermonitor WT701 (Raiing Medical, Boston MA) at 1-hour intervals. The rectal temperature was recovered to 1-hour intervals using linear interpolation during statistical analysis. Pairs of axillary and reference rectal temperatures were compared and summarized using linear regression and repeated-measured Bland and Altman methods. We a priori determined that the iThermonitor would have clinically acceptable accuracy if most estimates were within ±0.5°C of the rectal reference, and suitable precision if the limits-of-agreement were within ±0.5°C.
RESULTS: There were 489 sets of paired temperatures. Axillary and rectal temperatures were similar, with a mean difference (rectal minus axillary) of only 0.01°C ± 0.24°C (standard deviation). The Bland-Altman 95% limits-of-agreement were reasonably narrow, with the estimated upper limit at 0.45°C and the lower limit at ‑0.48°C, thus ±0.46°C, indicating good agreement across the range of mean temperatures from 35.2°C to 37.7°C.The absolute difference was within 0.5°C in 96% of the measurements.
CONCLUSIONS: Axillary temperature, as recorded by the iThermonitor WT701, well represents core temperature in NICU and thus appears suitable for clinical use.
There were a total of 489 sets of iThermonitor and reference measurements from 40 patients. As shown in Figure 1, the iThermonitor temperature and the reference rectal temperature agreed well overall. The iThermonitor temperatures ranged from 35.2 to 37.7°C, while the rectal temperatures ranged from 35.2 to 37.9°C. The bias was 0.01°C with an SD of 0.24 (Table 1).
Figure 1 Scatter plot of iThermonitor axillary temperature against reference rectal core temperature
The 95% limits of agreement for the iThermonitor were relatively narrow, with the estimated upper limit at 0.45°C (95% CI: 0.42, 0.47) and the lower limit at -0.48 °C (95% CI: -0.52, -0.43), indicating good (±0.46°C) agreement across the range of mean temperatures from 35.2 to 37.9°C. No obvious change of bias or variability was observed with increasing temperature in Bland-Altman plot (Figure 2). The proportion of measurements difference within 0.5°C was 96% (95% CI: 93, 98) comparing iThermonitor measurements to rectal temperatures.
Figure 2 Bland-Altman plot for iThermonitor axillary temperature versus reference rectal core temperature.
The X axis shows the average of the iThermonitor and reference rectal temperatures; the Y axis is the difference between the two measurements. The black dashed line shows mean bias; the red dashed lines show 95% limits of agreement; shaded area shows the 95% for limits of agreement.
Table 1. Agreement between iThermonitor axillary thermometer and rectal reference core temperature1
1489 records of iThermonitor and 489 records of rectal measurements from 40 patients
2Bias was calculated as the average difference between the iThermonitor axillary thermometer and the reference rectal temperature across all measurements for each patient. The mean of the bias and the standard deviation (SD) for single differences on different patients were presented and used to calculate the Bland-Altman limits of agreement.
3Bland-Altman limits of agreement for repeated measures, adjusting for within-patient correlation. The 95% confidence intervals (CIs) for limits were estimated using the method of variance estimates recovery (MOVER).