Comparing thermal imaging and non-contact infrared thermometers for monitoring skin temperature in a prospective cohort with lower limb cellulitis.

OBJECTIVE: Skin temperature assessment is essential for the diagnosis of cellulitis and monitoring treatment response, but is currently subjective and can contribute to overdiagnosis. We aimed to characterise skin temperature changes over time in cellulitis and compare two objective measurement approaches, a thermal imaging camera (TIC) and a non-contact infrared thermometer (NCIT). DESIGN: A device comparison study nested within a prospective cohort. We measured limb temperatures daily for 4 days using a TIC and two NCITs. SETTING: Two acute hospitals in the UK's National Health Service. PARTICIPANTS: 202 adults (age ≥18 years) diagnosed with lower limb cellulitis who attended hospital for antibiotic treatment. OUTCOME MEASURES: We used linear mixed-effects models to quantify changes in temperature over time and intraclass correlation coefficients (ICC) to assess reliability. We compared temperature measurements between devices using Lin's concordance coefficients and Bland-Altman plots with estimated 95% limits of agreement. RESULTS: 202 patients were included: 95% white ethnicity. Baseline limb temperature differences varied between 2.4°C and 3.4°C, depending on the device. All devices showed significant reductions in affected limb temperature per day, with the largest decrease recorded by the TIC (-0.34°C per day, 95% CI -0.48°C to -0.19°C, p<0.001). Only the TIC and NCIT-1 showed significant reductions in limb temperature difference per day. All devices had excellent reliability (ICCs ≥0.98). However, the magnitude of daily temperature change estimates was small relative to the repeatability coefficients of each device. The NCIT-2 consistently recorded the smallest differences in limb temperatures and demonstrated evidence of proportional bias. CONCLUSIONS: Daily temperature changes may be too small for reliable monitoring at the individual patient level, but cumulative changes from day 0 to day 3 may be sufficient for clinical interpretation, despite limitations in the precision of device measurements. NCITs' measurement capabilities differ widely, so these devices cannot be used interchangeably. Due to this and the potential benefits of advanced thermal image analysis, TICs should be prioritised for further study in cellulitis. Future research should confirm our findings in different skin tones and explore the clinical utility of thermal imaging in enabling earlier diagnosis or detecting signs of therapeutic failure.