Increased heart rate is a normal physiological response associated with fever. Clinicians observe that children experience tachycardia when their temperature is raised, even during a self-limiting febrile illness. However, tachycardia can also reflect a compensatory cardiovascular response to maintain blood pressure and tissue perfusion in children with serious infection.1^–3 So while heart rate is rightly considered an essential part of the examination of seriously ill children, some clinicians seldom measure it, seeing it only as a non-discriminatory sign of fever.4

Distinguishing children with serious infections from the majority who present with minor or self-limiting infections is an everyday task in primary and secondary care.5 6 The diagnostic task is difficult as serious infections are relatively uncommon, but the consequences of failing to recognise them can be fatal. For example, half of children with a subsequent diagnosis of meningococcal disease were sent home after initial consultation in primary care.7^–9 Concerns about the late diagnosis of children with serious infections led to the recently published National Institute for Health and Clinical Excellence (NICE) guidelines on the assessment of children with feverish illness in primary and secondary care.10 The guidelines recommend that the measurement of heart rate should be part of the examination in all febrile children but acknowledge the difficulties with the current reference values.

The published reference values for normal heart rates in children are age specific, but they do not take into account the effect of temperature.11^–14 There is evidence from historic studies, as well as one recent paediatric study, that heart rate increases with temperature15^–17; measuring heart rate without taking account of temperature therefore adds little useful diagnostic information. We report a cross–sectional primary care study with the purpose of deriving centile charts for heart rates in febrile children so that clinicians in all settings can assess whether a child’s heart rate is higher than would be expected for a given level of fever.

METHODS

Study sample

We asked general practitioners, nurse practitioners and research nurses to recruit children aged 3 months–12 years presenting with fever to 10 surgeries in Oxfordshire, Buckinghamshire and Somerset, and two out-of-hours centres in Oxfordshire. We obtained informed consent from all parents. We recruited from December 2003 to March 2006. In the first year we recruited all-comers but in the final 18 months we asked staff to prioritise recruitment of children aged <2 years of age and/or with temperatures over 38.0°C to try to achieve a recruitment target of 30–60 children in each age–temperature category (four age bands: 3–12 months, 1–2 years, 2–5 years, 5–12 years; four temperature bands: <37.0°C, 37.0°C–37.9°C, 38.0°C–38.9°C, ⩾39.0°C), the sample size suggested by Virtanen for constructing regression-based reference limits.18 We recorded a total of 2123 episodes of care on 2011 children during the study. We reviewed the details of the 112 children who had more than one episode of care and selected one episode for the analysis (the visit with the highest temperature, or if the temperatures were identical, the first visit). We excluded a further 78 children due to absence of a recorded heart rate (30), temperature (35) or a date of birth within the eligible age range (13). The final study sample comprised 1933 children, of whom 1050 (54.3%) were recruited from in-hours general practice surgeries and 883 (45.7%) from out-of-hours centres. There were slightly more boys (1027, 53.1%) than girls.

Measurement of temperature and heart rate

We measured temperature using WelchAllyn SureTemp Plus axillary electronic thermometers (WelchAllyn, San Diego, CA) which were calibrated by the manufacturer at the start of the study and every 6 months during the study. We asked recruiters to place the thermometer high in the axilla, adduct and hold the arm close to the chest wall for the 10 s reading time. We measured heart rate using Nonin 8500 pulse oximeters (Nonin Medical, Plymouth, MN), placing the pulse oximeter probe on the finger or toe depending on age.

Recording of clinical outcome

We asked the recruiters to record the clinical diagnosis made at presentation. We subsequently reviewed the practice records of each child seen during general practice surgeries to identify any hospital attendances. For children seen at the two out-of-hours clinics in Oxfordshire, we reviewed the patient information system at the local hospital to which children would have been referred (and to which parents would almost certainly have taken their child if not referred) to identify visits to the emergency department or hospital admissions in the 7 days following the date of recruitment.

Statistical analysis

We excluded three groups of children before creating the centile charts: 119 children who were referred to the hospital by the general practitioner or who attended hospital within 7 days of recruitment, 123 children whose final diagnosis was not an infection (of whom eight had been referred or attended hospital) and 14 children whose heart rates were more than 3 standard deviations (SD) from the mean rate for their age group and temperature band.19 We subsequently also excluded the 96 children who were older than 10 years 11 months of age as there were only 11 children over this age who had temperatures of 38.0°C or higher, and we considered that the original 5–12-year age group was too broad. The final sample used to create the centile charts (n = 1589) had the following age distribution: 254, 3–12 months; 254, 1–2 years; 538, 2–5 years; and 543, 5–10 years. A total of 622 children had a temperatures under 37.0°C, 609 had temperatures between 37.0°C and 37.9°C, 221 had temperatures between 38.0°C and 38.9°C, and 137 had temperatures of 39.0°C or higher. We calculated correlation coefficients using Spearman’s r (rank correlation) for non-parametric variables and linear regression to estimate regression coefficients for the relationship between heart rate and temperature in each age group. We calculated the median and upper centiles (75th, 90th and 97th) of heart rate at a given temperature for children in each of the four age groups using LmsChartMaker Pro (Medical Research Council, UK) based on the method of Cole and Green.20 The final models used for each age group were checked using Z-score graphs, detrended Q-Q plots and Q-statistic curves for the parameters in the model (L, M, S).

RESULTS

The most frequent diagnoses for the children included in the centile chart analysis were upper respiratory tract infection (859, 54.1%), non-specific viral illness (215, 13.5%), lower respiratory tract infections (125, 7.9%), or diarrhoea and/or vomiting (82, 5.2%). Heart rate was significantly negatively correlated with age (r = −0.62) and positively correlated with temperature (r = 0.49). The correlation between heart rate and temperature was significant for all four age groups but was smaller in children aged <1 year (r = 0.41) and 1–2 years (r = 0.42) than in those in the 2–5-year (r = 0.65) or 5–10-year (r = 0.59) age group.

In the combined group of 1589 children, heart rate increased by 13.7 bpm (95% CI 12.5 to 14.9) with each 1°C increment in temperature. The incremental increases in heart rate in each age group were: 3–12 months, 12.1 bpm (95% CI 9.2 to 15.0) per 1°C increase in temperature; 1–2 years, 9.9 bpm (95% CI 7.3 to 12.5); 2–5 years, 14.1 bpm (95% CI 12.7 to 15.5); and 5–10 years,14.1 bpm (95% CI 12.6 to 15.6). The 50th, 75th, 90th and 97th centiles of heart rate plotted against temperature for children in the four age groups are displayed in fig 1. In all four age groups, heart rate increased with temperature most rapidly between 37.0°C and 38.0°C before levelling off slightly. This phase of rapid rise in heart rate was most evident in children under 1 year of age. In all age groups, there was a fairly consistent relationship between the 50th and 97th centiles, ranging from 37–41 bpm for children aged <1 year, 38–42 bpm for children aged 1–2 years, 29–32 bpm for those aged 2–5 years, and 27–34 bpm for those aged 5–12 years. The values of heart rate at the 50th (ie, median), 75th, 90th and 97th centiles are displayed in table 1. The rate selected in each temperature band is the heart rate at the upper end of that temperature band, rounded to the nearest whole number. The heart rates that are within the Advanced paediatric life support (APLS) upper limits of normal for that age group are indicated by italics in table 1.11

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Median and upper centiles of heart rate expected at different temperatures for children with acute self-limiting infections in primary care.

DISCUSSION

Principal findings

This study confirms that clinicians need to consider the effects of temperature as well as age when interpreting heart rate in children presenting with acute infections. Measurement of heart rate without adjusting for age and temperature will mislead and provides little useful diagnostic information for a clinician. The charts present a simple visual method to allow clinicians to identify children who have a higher heart rate than would be expected for their age and temperature. In the context of a health service in which a child may be assessed over a short period of time by different professionals, the charts will make comparisons over time easier irrespective of variation in temperature during the course of the illness (eg, a seriously ill child will remain on a high centile even if the temperature and heart rate are reduced by use of an antipyretic).

Strengths and weaknesses

A potential criticism of the study is that the children recruited were not an entirely representative sample from primary care. Recruitment was not systematic, the proportion of children consulting out-of-hours care was high, and we set minimum recruitment targets for each age–temperature combination. However, the range of diagnoses is typical of first-contact care settings and we have no reason to believe that a child’s physiological response is likely to differ according to the measurement setting. We consider that the sampling is unlikely to have introduced any systematic bias.

We consider that presenting the data as centile charts is appropriate for clinical use, but it has the weakness of not indicating the potential statistical error in their derivation. Our recruitment target was based on statistical simulation studies which show that, in constructing regression-based reference limits, estimates of variance become stable with a sample size of about 30, while increasing the sample size above 60 achieves negligible improvement in precision.18 We met the higher target of 60 children in 10 out of 16 age-temperature bands and fell below 30 in only two bands (temperature >39.0°C and ages below 2 years). The centile estimates are therefore likely to be statistically reliable across most of their range but least precise in the youngest children and in those with high fever.

The potential weakness of axillary measurement is that it may be affected by changes in skin perfusion. However, we were swayed by the inconsistency of aural measurement and the logistic infeasibility of rectal measurement in a primary care setting. It is a strength that the charts are based on a measurement method that is commonly employed in primary and secondary care and recommended by NICE. Our aim was to derive centile charts which apply to clinicians providing first contact care to children. Using pulse oximeters is likely to have given similar values to measurement by palpation or auscultation, with greater precision than simple clinical counting. We did not collect details of medications that might have affected either temperature or heart rate (eg, antipyretics, β₂ agonist inhalers) as, again, this would have restricted the applicability of the centile charts to a minority of children who do not receive such medications.

Comparison with previous research

Despite the widely quoted normal values of heart rate in paediatric texts and resuscitation guidelines, there has been no systematic attempt to define the relationship between heart rate and temperature in children. The widely used APLS cut-off values for heart rate in children11 are exceeded by the majority of feverish children in our study, and thus have extremely low specificity. This may explain why many primary care doctors do not measure heart rate and do not agree on threshold values for tachycardia.4 Aside from historic studies of fever induced artificially in adults,16 17 21 22 there is only a single study describing the relationship between heart rate and temperature in children.15 Hanna et al found that heart rate increased linearly by 9.6 bpm per 1°C rise in temperature in infants under 12 months of age, which is consistent with our results. Our centile charts demonstrate that the relationship between heart rate and temperature is not linear, so that adding a set value to correct for temperature may be insufficient.

Clinical and research implications

While we cannot determine the positive predictive value for serious illness in a child of a temperature measurement which is on the higher centiles (nor the negative predictive value of a measurement on a lower centile), we believe that these children require a clinician to perform a comprehensive assessment of all vital signs and to remain especially alert to the possibility of serious illness. Even if further examination reveals no other cause for concern, they may be at higher risk for serious illness and merit re-assessment within a few hours. However, clinicians also need to be aware that heart rate can be quite labile in children, therefore single highly abnormal results (for example in an anxious child) should be repeated. It is possible that a series of measurements of heart rate and temperature over a time period may yield more valuable information than isolated measurements, similar to the interpretation of growth charts. In febrile children a rise across the heart rate centiles adjusted for temperature may indicate deterioration in the child’s cardiovascular state, while a rise in heart rate with an associated rise in temperature may find the measurement on a similar centile and provide reassurance.

Further research is needed to determine the predictive values of all vital signs including respiratory rate, peripheral perfusion and oxygen saturation as well as temperature and heart rate in a range of acute paediatric settings. Until such research has been completed, we recommend the use of our simple temperature–heart rate centile charts to all those involved in the assessment of febrile children and encourage further reports of the diagnostic value of the charts in primary and secondary paediatric care.