The reference ranges for heart rate and respiratory rate cited in international paediatric guidelines, such as those presented in Web Box 1
, are widely used as the basis for clinical decisions when interpreting vital signs in children. For example, the widely used PAWS and Brighton PEWS2,3
assessment tools refer to APLS reference ranges. We have shown that: 1) there is considerable disagreement between these reference ranges; 2) they appear to be inconsistent with existing evidence on heart and respiratory rates in healthy children.
For clinicians involved in the assessment of children, our findings suggest that current consensus-based reference ranges for heart rate and respiratory rate should be updated with new thresholds based on our proposed centile charts, particularly for age groups where our findings show that many children are likely to be misclassified. Normal ranges, such as those published in textbooks and clinical handbooks, should also be updated in the light of our results. To assist the development of cut-offs for use in clinical settings, we present values corresponding to the median and six different centiles for both heart rate and respiratory rate for 13 age groups between birth and 18 years of age.
By providing several different centiles for children of all ages, we have given clinicians and those responsible for developing clinical guidelines and early warning scores sufficient information to select cut-offs that are most appropriate to the type of clinical setting in which they are likely to be used. Selection of an appropriate cut-off should take into account the likely derangement in vital signs associated with the level of illness that is to be detected, and the penalty associated with misclassifying both healthy and unwell children. Further research may be required in some areas to ascertain this. Where multiple measurements are made over time, the centile charts may also be used to assess the magnitude of changes in heart rate or respiratory rate.
Clinicians who wish to carry out accurate measurements of heart rate in children should be aware that manual measurement of heart rates, which is common practice in many settings, may underestimate the true rates. In these children, measurement of heart rate by automated methods provides more accurate results. Professional bodies responsible for publishing guidelines and scoring systems should consider revising current thresholds, by selecting heart rate and respiratory rate values that represent an upper centile for each age group. To facilitate this, the authors propose to make the data used to create and freely available upon request.
A key strength of our approach is that the centile charts were created using kernel regression, a non-parametric modelling technique which avoids imposing any particular form onto the shape of the centile charts. This is particularly important for this type of data, as there is no reason to expect that it will follow an analytical function such as a straight line or exponential. However, a number of limitations to the method are worth noting. Our systematic review involved an extensive search of the available literature using three large databases, with no restriction on language or country of publication. However, it is possible that our search strategy and inclusion criteria may have missed relevant studies, particularly studies published before 1960. We excluded 13 studies as we were not able to extract the required data or could not obtain full copies, and we did not attempt to contact original authors to obtain individual patient data, as this would not have been feasible given the number of included studies, some of which were published over 25 years ago. We observed marked heterogeneity in the settings in which the children were measured, their state of wakefulness, and the method of measurement, all of which may have an effect on the measured variables. As reported, subgroup analysis showed that the setting, method of measurement, and economic development had a significant effect on heart rate in children, but not on respiratory rate (for more detail see Web Annex 2
). We excluded children with illnesses that might affect the heart rate or respiratory rate, and measurements known to be made during exertion, but many studies did not report whether children were quiet or agitated during measurement, which may have introduced additional heterogeneity that could not be assessed. However, by using the subgroup analysis on wakefulness as a proxy for agitation, it is unlikely that this would have a significant effect on the results. The heterogeneity of the data can also be interpreted as a strength, making the centiles more relevant to a wide spectrum of clinical settings.
Our centile charts have been developed using data from normal children. As with all clinical measurements, they should be used as part of an overall assessment of a child’s condition, and interpretation of measured values should also take into account any factors which might be expected to affect the measured value. For example, measurements of heart rate may be increased in the presence of fever,27
anxiety, or if measured with automated methods or in developing countries. These should therefore inform the selection of appropriate centiles for use as cut-offs in such situations.
Further research should consider assessing the benefit of integrating our centiles into early warning scores. Improvement in sensitivity and specificity will be age-dependent and will depend on the accuracy of the previous reference ranges. For the existing APLS reference ranges, which were observed to have the greatest agreement with our centiles, and suggest that a large number of children are currently misclassified. For example, at the age of ten years, the APLS cut-off for heart rate classifies approximately 40% of normal children as abnormal, and the APLS cut-off for respiratory rate misclassifies approximately 63% of normal children. Furthermore, based on the age distribution of children presenting to primary care in a previously reported study,27
we estimate that the specificity of APLS could be improved by as much as 20% for heart rate, and 51% for respiratory rate if revised centile charts are used. Further research, in both healthy children, and those presenting with a spectrum of conditions, should test the validity of our centiles and any cut-offs derived from them.
In conclusion, we have shown that existing reference ranges for heart rate and respiratory rate in children are inconsistent, and do not agree with centile charts derived from a systematic review of observational studies. This has potentially wide-ranging implications for clinicians involved in the assessment of children, and for the design of resuscitation guidelines, triage scores, and early warning systems.