Repeatability of RRate measurements in children during triage in two Ugandan hospitals

Abstract

Pneumonia is the leading cause of death in children globally. In low- and middle-income countries (LMICs) pneumonia diagnosis relies on accurate assessment of respiratory rate, which can be unreliable when completed by nurses with less-advanced training. To inform more accurate measurements, we investigate the repeatability of the RRate app used by nurses in Ugandan district hospitals. This secondary analysis included 3,679 children aged 0-5 years. The dataset had two sequential measurements of respiratory rate collected by 14 nurses using the RRate app. We measured agreement between respiratory rate observations while indicating observations' clustering around WHO fast-breathing thresholds. WHO thresholds are 60 breaths per minute (bpm) for under two months (Age-1), 50 bpm for two to 12 months (Age-2), and 40 bpm for 12.1 to 60 months (Age-3). We assessed the repeatability of the paired measurements per user through the Intraclass Correlation Coefficient (ICC) and calculated an overall ICC value. The respiratory rate measurement took less than 15 seconds for 7,277 (98.9%) of the measurements. Despite respiratory rates clustering around WHO thresholds, breathing classification based on the thresholds (Fast vs normal) was altered between sequential measurements in only 12.6% of children. The mean (SD) respiratory rate by age group was 60 (13.1) bpm for Age-1, 49 (11.9) bpm for Age-2, and 38 (10.1) for Age-3, and the bias (Limits of Agreements) were 0.3 (-10.8-11.3) bpm, 0.4 (-8.5-9.3) bpm, and 0.1 (-6.8, 7.0) bpm for Age-1, Age-2, and Age-3 respectively. The repeatability of the paired respiratory rate measurements was high, with an ICC ≥ 90% for 12 of 14 users and an overall ICC value (95% CI) of 0.95 (0.94-0.95). The RRate measurements were efficient and repeatable. The simplicity, repeatability, and efficiency support its usage in LMICs healthcare facilities, and endorses a more widespread clinical adoption.

Fig 1. Consort diagram for children included in the study.

Fig 2. Bland Altman plot for RR-1 and RR-2 across all children.

Limits of Agreement (LoAs) and bias are under repeatability conditions, considering one set of samples as the test and the other as the reference. U95 and L95 represent the upper and lower bounds of the 95% CI of the limits of agreements respectively.

Fig 3. Bland Altman plots for Age-1, Age-2 and Age-3 (from left to right).

Threshold for fast breathing is 60, 50 and 40 bpm respectively. Red dots represent children with two fast breathing measurements, green dots – two normal breathing, and blue dots – one fast and one normal. U95 and L95 represent the upper and lower bounds of the 95% CI of the limits of agreements respectively.