Inter-Observer Agreement in Measuring Respiratory Rate

Respiratory rate (RR) is an important vital sign which is strongly correlated with in-hospital mortality. At the same time, RR is the most likely vital sign to be omitted when assessing a patient. We believe that one reason for this could be the difficulty in measure the RR, since it is not read off a monitor, but counted manually. Also there is the possibility of assessment bias and the inter-observer reliability becomes important. We therefore set out to investigate how the nursing staff counting the actual number of respirations per minute would agree with the nursing staff using a predefined ordinal scale.

For this prospective study, we recorded five videos of a young healthy man breathing approximately 5, 10, 15, 30 and 60 times per minute. The videos were shown in a random order to a suitable sample of the nursing staff. The participants were randomized into two groups; one to count the exact number of breaths per minute, and one to use a predefined ordinal scale.

Comparing the exact number of breaths per minute, the Intra Class Coefficient (ICC) was 0.99 (95% CI: 0.97-1.00). Comparing the RR using the predefined scale, the overall Kappa Fleiss Coefficient was 0.75.

The inter-observer agreement was high when comparing the use of the actual number of breaths per minute and substantial when comparing the use of the predefined scale. This is the largest inter-observer study on RR to date. However, further studies on the use of scaled comparisons of RR are needed.