Man versus machine: comparison of naked-eye estimation and quantified capillary refill

BACKGROUND: Capillary refill (CR) time is traditionally assessed by ‘naked-eye’ inspection of the return to original colour of a tissue after blanching pressure. Few studies have addressed intra-observer reliability or used objective quantification techniques to assess time to original colour. This study compares naked-eye assessment with quantified CR (qCR) time using polarisation spectroscopy and examines intra-observer and interobserver agreements in using the naked eye. METHOD: A film of 18 CR tests (shown in a random fixed order) performed in healthy adults was assessed by a convenience sample of 14 doctors, 15 nurses and 19 secretaries (Department of Emergency Medicine, Linköping University, September to November 2017), who were asked to estimate the time to return to colour and characterise it as ‘fast’, ‘normal’ or ‘slow’. The qCR times and corresponding naked-eye time assessments were compared using the Kruskal-Wallis test. Three videos were shown twice without observers’ knowledge to measure intra-observer repeatability. Intra-observer categorical assessments were compared using Cohen’s Kappa analysis. Interobserver repeatability was measured and depicted with multiple-observer Bland-Altman plotting. Differences in naked-eye estimation between professions were analysed using ANOVA. RESULTS: Naked-eye assessed CR time and qCR time differ substantially, and agreement for the categorical assessments (naked-eye assessment vs qCR classification) was poor (Cohen’s kappa 0.27). Bland-Altman intra-observer repeatability ranged from 6% to 60%. Interobserver agreement was low as shown by the Bland-Altman plotting with a 95% limit of agreement with the mean of ±1.98 s for doctors, ±1.6 s for nurses and ±1.75 s for secretaries. The difference in CR time estimation (in seconds) between professions was not significant. CONCLUSIONS: Our study suggests that naked-eye-assessed CR time shows poor reproducibility, even by the same observers, and differs from an objective measure of CR time.