^{1}

^{2}

^{2}

^{1}

^{2}

Edited by: Eric Giannoni, Centre Hospitalier Universitaire Vaudois (CHUV), Switzerland

Reviewed by: Martin Chalumeau, Université Paris Descartes, France; Antonio Rodriguez-Nunez, Universidade de Santiago de Compostela, Spain

Specialty section: This article was submitted to Pediatric Critical Care, a section of the journal Frontiers in Pediatrics

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

We hypothesized that current vital sign thresholds used in pediatric emergency department (ED) screening tools do not reflect observed vital signs in this population. We analyzed a large multi-centered database to develop heart rate (HR) and respiratory rate centile rankings and

Initial HR and respiratory rate data entered into the Cerner™ electronic health record at 169 participating hospitals’ ED over 5 years (2009 through 2013) as part of routine care were analyzed. Analysis was restricted to non-admitted children (0 to <18 years). Centile curves and

HR and RR centiles and

Heart and respiratory rate centiles derived from a large real-world non-hospitalized ED pediatric population can inform the modification of electronic and paper-based screening tools to stratify children by the degree of deviation from normal for age rather than dichotomizing children into groups having “normal” versus “abnormal” vital signs. Furthermore, these centiles also may be useful in paper-based screening tools and bedside alarm limits for children in areas other than the ED and may establish improved alarm limits for bedside monitors.

Vital sign thresholds are incorporated into various screening tools to help identify those children at higher risk of serious medical or surgical illness (

Most screening tools use vital sign thresholds that fail to consider the physiologic stress response of a child seen in the ED. Thus, the upper “normal” vital sign thresholds observed in ED patients were higher than observed in children who were hospitalized on the ward or who were ambulatory (

In sepsis, early identification of children at risk is a key recommendation for optimal management (

Recent studies empirically derived centile ranks and, in some cases,

We hypothesized that current pediatric HR and RR vital sign thresholds used in Pediatric Advanced Life Support (PALS) or derived from low-acuity ED patients do not accurately reflect empirically derived HR and RR centiles. To develop empirically derived thresholds that could be incorporated into ED screening tools and may inform monitor alarm limits, we analyzed a very large multi-institutional database to derive HR and RR centile ranks and

Data in Cerner Health Facts^{®} (^{®}.

The HR and RR distributions by age were modeled using the generalized additive models for location, scale, and shape (GAMLSS) methodology and software (

Sample raw distributions of individual heart rate (HR) and respiratory rate (RR) parameters for a single age group.

For modeling the distributions of RR, a natural logarithm transformation was required for model convergence. It was also necessary to introduce Gaussian statistical noise of up to ±2 breaths/min (with a mean value of 0) to overcome the reduced variation due to digit bias in the raw RR measurements (

For each vital sign, our modeling process entailed a stepwise fitting procedure to calculate optimal age-specific fits for mean, SD, skew, and kurtosis, with an additive term that used “penalized B-splines” to create smooth centile curves (

Because our original intent was to examine both raw and TMP-corrected initial vital signs, our data capture was restricted to initial encounters having HR, RR, and TMP values taken within 15 min of one another. Encounters having two or more distinct measurements recorded for the same vital sign at the same date and time were not uncommon (comprising about 7% of the total). In such cases, we selected the average value, provided that the range of simultaneous values did not exceed 10% of the largest value for HR or RR (i.e., approximately 10–20 bpm for HR or 2–5 breaths/min for RR), or 3% of the largest value for TMP (i.e., approximately 1°C), arbitrarily chosen to exclude likely erroneous outliers; if exceeded, the encounter was excluded. Records in one or more of the following categories also were similarly excluded as likely outliers: (1) extreme (likely spurious) values of HR (<30 or >300 bpm), RR (0 or ≥120 breaths/min), or TMP (<30 or >46°C); (2) encounters classified as “Trauma Center” cases; or (3) encounters where the patient had a diagnosis of a chronic heart or respiratory condition present on admission. The latter two exclusion types (collectively ~0.4% of cases) were excluded since we did not want to include children who were more likely to have very abnormal vital signs. A summary of the selection and exclusion criteria used to determine the final data set for our study is given in Data Sheet S2 in Supplementary Material with details on specific diagnostic exclusions in Data Sheet S3 in Supplementary Material. A sensitivity analysis of the effects of these exclusions on the final modeled centile results was conducted, as described below.

To test model reproducibility, we performed a stratified split-sample validation whereby the full data set used for each vital sign was divided into a “training”subset consisting of two-thirds of randomly selected records from each age group, and a “test” data subset consisting of the remaining one-third of records. The training subset was modeled by GAMLSS to generate centile cutoffs, and the percentage of records with vital sign values above and below these modeled cutoffs for the 95th, 99th, 5th, and 1st centiles, respectively, were compared between the training and test data subsets using exact chi-square tests. Holm (step-down Bonferroni) (

To determine the effect of excluding encounters classified as “Trauma Center,” or those where the patient had a diagnosis of a chronic heart or respiratory condition, the modeling of HR and RR distributions was repeated including these encounters. The resultant modeled centile values, rounded to the nearest whole number for HR and RR, were then compared between datasets of children with and without exclusions.

We graphically plotted the empirically derived 5th, 50th, and 95th centiles by age compared with the same centiles calculated by O’Leary et al. (

The Health Facts^{®} data used comprises initial vital sign values and relative measurement times for HR and RR collected from ED encounters involving children (ages 0–17 years) at 169 U.S. hospitals (five were children’s hospitals) for calendar years 2009 through 2013. For all encounters, only patient types classified as “Emergency” were selected, which excludes children who were hospitalized from the ED since the Health Facts database does not specifically identify those admitted patients who entered the hospital

Our preliminary analysis of patient TMPs taken in the ED found that TMP distributions varied according to patient age and route of measurement, making generalized TMP corrections of HR and RR problematic. Therefore, we analyzed HR and RR without TMP correction.

A total of 1,203,042 encounters were used to study the distribution of initial HR by age, with slightly fewer (1,202,984) available for an analysis of RR. Patient information and encounters included in our study for each of the vital signs examined are presented in Tables

Number of contributing encounters by patient age group.

Age description | HR, RR^{a} |
Age description | HR, RR^{a} |
---|---|---|---|

<1 month | 12,860 (1.1%) | 2 to <3 years | 93,406 (7.8) |

1 to <2 months | 13,016 (1.1%) | 3 to <4 years | 82,813 (6.9) |

2 to <3 months | 13,204 (1.1%) | 4 to <5 years | 74,328 (6.2) |

3 to <4 months | 11,255 (0.9%) | 5 to <6 years | 68,517 (5.7) |

4 to <5 months | 11,846 (1.0%) | 6 to <7 years | 59,316 (4.9) |

5 to <6 months | 12,317 (1.0%) | 7 to <8 years | 50,664 (4.2) |

6 to <7 months | 12,895 (1.1%) | 8 to <9 years | 46,448 (3.9) |

7 to <8 months | 13,354 (1.1%) | 9 to <10 years | 44,818 (3.7) |

8 to <9 months | 13,580 (1.1%) | 10 to <11 years | 44,093 (3.7) |

9 to <10 months | 13,349 (1.1%) | 11 to <12 years | 43,225 (3.6) |

10 to <11 months | 13,234 (1.1%) | 12 to <13 years | 42,813 (3.6) |

11 to <12 months | 13,002 (1.1%) | 13 to <14 years | 45,766 (3.8) |

12 to <13 months | 13,016 (1.1%) | 14 to <15 years | 50,007 (4.2) |

13 to <14 months | 12,078 (1.0%) | 15 to <16 years | 53,940 (4.5) |

14 to <15 months | 11,548 (1.0%) | 16 to <17 years | 59,621 (5.0) |

15 to <16 months | 11,115 (0.9%) | 17 to <18 years | 65,264 (5.4) |

16 to <17 months | 10,805 (0.9%) | ||

17 to <18 months | 10,351 (0.9%) | ||

18 to <19 months | 10,012 (0.8%) | ||

19 to <20 months | 9,476 (0.8%) | ||

20 to <21 months | 9,085 (0.8%) | ||

21 to <22 months | 9,011 (0.7%) | ||

22 to <23 months | 8,878 (0.7%) | ||

23 to <2 years | 8,716 (0.7%) |

^{a}Overall numbers for respiratory rate (RR) are slightly less than for heart rate (HR) (1,202,984 versus 1,203,042), but percentages of total for each age group as rounded are identical for HR and RR

Contributing encounters: patient demographics and length of stay.

Encounter characteristics | HR, RR^{a} |
---|---|

Mean/Median | |

Length of Stay (hours) | 5.7/2.6 |

Gender | N (% of total) |

Male | 627,095 (52.1%) |

Female | 575,834 (47.9%) |

Missing/Unknown/Other | 113 (0.0%) |

Race/ethnicity | |

Caucasian | 508,180 (42.2%) |

African American | 423,607 (35.2%) |

Missing/Unknown/Other | 140,270 (11.7%) |

Hispanic | 100,929 (8.4%) |

Asian | 17,524 (1.5%) |

Native American | 12,532 (1.0%) |

Total | 1,203,042^{a} |

^{a}Overall numbers for respiratory rate (RR) are slightly less than for heart rate (HR) (1,202,984 versus 1,203,042), but mean/median length of stay and percentages of total for each demographic characteristic as rounded are identical for HR and RR

Representative raw distributions (histograms) for the HR and RR (Figure

For HR and RR (Figures

Modeled centile curves for pediatric vital sign metrics.

The results of the split-sample validation of model result reproducibility (Table

Validation of vital sign centile modeling results.

Vital sign metric | Modeled^{a} |
Training set, ^{b} |
Test set, ^{b} |
Raw ^{c} |
Holm-adjusted ^{d} |
---|---|---|---|---|---|

HR | >99th | 6,653 (0.83) | 3,325 (0.83) | 0.99 | 1.0 |

HR | >95th | 42,540 (5.30) | 20,980 (5.23) | 0.09 | 0.37 |

HR | <5th | 36,890 (4.60) | 18,477 (4.61) | 0.85 | 1.0 |

HR | <1st | 7,771 (0.97) | 3,776 (0.94) | 0.15 | 0.44 |

RR | >99th | 7,862 (0.98) | 3,961 (0.99) | 0.70 | 0.86 |

RR | >95th | 39,921 (4.98) | 19,828 (4.94) | 0.43 | 0.86 |

RR | <5th | 31,091 (3.88) | 15,834 (3.95) | 0.06 | 0.22 |

RR | <1st | 5,246 (0.65) | 2,692 (0.67) | 0.28 | 0.83 |

^{a}Using cutoffs obtained by modeling data in training set

^{b}Number of encounters with vital sign metric greater than (for 95th and 99th) or less than (for 5th and 1st) modeled cutoff

^{c}Comparison of cases above or below cutoff between training and test sets using exact chi-square (df = 1)

^{d}Holm (step-down Bonferroni) corrections, performed separately for heart rate (HR) and respiratory rate (RR) were employed to adjust for multiple testing (99th, 95th, 5th, and 1st centiles) of each metric

Table

The modeled values of HR, and RR from the 1st to 99th centiles for each age group are provided as Tables

Heart rate (HR, bpm) centiles^{a}

Age (years) midpoint | Age/units midpoint | Age description | C1 | C2.5 | C5 | C10 | C25 | C50 | C75 | C90 | C95 | C97.5 | C99 |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

0.042 | 0.5 months | <1 month | 106 | 114 | 121 | 128 | 140 | 153 | 165 | 177 | 184 | 191 | 198 |

0.125 | 1.5 months | 1 to <2 months | 112 | 119 | 126 | 133 | 144 | 156 | 168 | 180 | 187 | 194 | 202 |

0.208 | 2.5 months | 2 to < 3 months | 108 | 115 | 121 | 128 | 139 | 151 | 163 | 176 | 184 | 192 | 201 |

0.292 | 3.5 months | 3 to <4 months | 105 | 111 | 117 | 124 | 135 | 147 | 160 | 173 | 182 | 190 | 200 |

0.375 | 4.5 months | 4 to <5 months | 102 | 109 | 115 | 121 | 133 | 145 | 159 | 172 | 181 | 190 | 200 |

0.458 | 5.5 months | 5 to <6 months | 101 | 107 | 113 | 120 | 131 | 144 | 158 | 173 | 182 | 190 | 201 |

0.542 | 6.5 months | 6 to <7 months | 100 | 106 | 111 | 118 | 130 | 143 | 158 | 173 | 182 | 191 | 201 |

0.625 | 7.5 months | 7 to <8 months | 98 | 104 | 110 | 117 | 128 | 142 | 157 | 172 | 182 | 190 | 201 |

0.708 | 8.5 months | 8 to <9 months | 97 | 103 | 109 | 115 | 127 | 141 | 157 | 172 | 182 | 190 | 201 |

0.792 | 9.5 months | 9 to <10 months | 96 | 102 | 107 | 114 | 126 | 141 | 157 | 172 | 182 | 190 | 201 |

0.875 | 10.5 months | 10 to <11 months | 95 | 101 | 106 | 113 | 125 | 140 | 157 | 172 | 182 | 191 | 201 |

0.958 | 11.5 months | 11 to <12 months | 94 | 100 | 106 | 112 | 125 | 140 | 157 | 173 | 183 | 191 | 201 |

1.042 | 12.5 months | 12 to <13 months | 93 | 100 | 105 | 112 | 124 | 140 | 158 | 174 | 183 | 192 | 202 |

1.125 | 13.5 months | 13 to <14 months | 93 | 99 | 105 | 112 | 124 | 140 | 158 | 174 | 184 | 193 | 203 |

1.208 | 14.5 months | 14 to <15 months | 92 | 98 | 104 | 111 | 124 | 140 | 158 | 174 | 184 | 193 | 204 |

1.292 | 15.5 months | 15 to <16 months | 92 | 98 | 103 | 110 | 123 | 140 | 158 | 174 | 184 | 193 | 204 |

1.375 | 16.5 months | 16 to <17 months | 91 | 97 | 103 | 110 | 123 | 139 | 157 | 174 | 184 | 193 | 203 |

1.458 | 17.5 months | 17 to <18 months | 90 | 96 | 102 | 109 | 122 | 138 | 157 | 173 | 183 | 192 | 203 |

1.542 | 18.5 months | 18 to <19 months | 89 | 95 | 101 | 108 | 121 | 137 | 156 | 172 | 182 | 191 | 202 |

1.625 | 19.5 months | 19 to <20 months | 89 | 95 | 100 | 107 | 120 | 136 | 154 | 171 | 181 | 190 | 201 |

1.708 | 20.5 months | 20 to <21 months | 88 | 94 | 99 | 106 | 119 | 135 | 153 | 170 | 180 | 189 | 200 |

1.792 | 21.5 months | 21 to <22 months | 87 | 93 | 98 | 105 | 118 | 134 | 152 | 169 | 179 | 188 | 198 |

1.875 | 22.5 months | 22 to <23 months | 86 | 92 | 98 | 104 | 117 | 133 | 151 | 167 | 178 | 187 | 197 |

1.958 | 23.5 months | 23 months to <2 years | 86 | 92 | 97 | 104 | 116 | 132 | 150 | 166 | 176 | 185 | 196 |

2.5 | 2.5 years | 2 to <3 years | 82 | 88 | 93 | 99 | 111 | 126 | 143 | 159 | 169 | 178 | 189 |

3.5 | 3.5 years | 3 to <4 years | 77 | 82 | 87 | 93 | 103 | 118 | 133 | 149 | 158 | 167 | 177 |

4.5 | 4.5 years | 4 to <5 years | 73 | 78 | 82 | 88 | 98 | 112 | 127 | 142 | 152 | 160 | 170 |

5.5 | 5.5 years | 5 to <6 years | 70 | 75 | 79 | 84 | 94 | 107 | 122 | 137 | 146 | 154 | 165 |

6.5 | 6.5 years | 6 to <7 years | 67 | 71 | 76 | 81 | 90 | 103 | 118 | 132 | 141 | 149 | 160 |

7.5 | 7.5 years | 7 to <8 years | 65 | 69 | 73 | 78 | 87 | 99 | 113 | 128 | 137 | 145 | 155 |

8.5 | 8.5 years | 8 to <9 years | 63 | 67 | 71 | 76 | 85 | 97 | 110 | 124 | 133 | 141 | 152 |

9.5 | 9.5 years | 9 to <10 years | 62 | 66 | 70 | 74 | 83 | 94 | 107 | 121 | 130 | 138 | 149 |

10.5 | 10.5 years | 10 to <11 years | 60 | 64 | 68 | 73 | 81 | 92 | 105 | 118 | 127 | 136 | 146 |

11.5 | 11.5 years | 11 to <12 years | 59 | 63 | 67 | 71 | 79 | 90 | 103 | 116 | 125 | 133 | 143 |

12.5 | 12.5 years | 12 to <13 years | 57 | 61 | 65 | 69 | 77 | 88 | 100 | 113 | 122 | 130 | 140 |

13.5 | 13.5 years | 13 to <14 years | 56 | 59 | 63 | 67 | 75 | 86 | 98 | 111 | 119 | 127 | 138 |

14.5 | 14.5 years | 14 to <15 years | 54 | 58 | 62 | 66 | 74 | 84 | 96 | 109 | 117 | 125 | 135 |

15.5 | 15.5 years | 15 to <16 years | 53 | 57 | 61 | 65 | 73 | 84 | 96 | 108 | 116 | 124 | 134 |

16.5 | 16.5 years | 16 to <17 years | 53 | 57 | 60 | 65 | 73 | 83 | 96 | 108 | 116 | 124 | 133 |

17.5 | 17.5 years | 17 to <18 years | 53 | 57 | 60 | 65 | 73 | 84 | 96 | 109 | 117 | 124 | 134 |

^{a}Centiles abbreviated as C1 (first centile) to C99 (99th centile)

Each sensitivity analysis compared modeled centiles between data sets with and without predefined exclusions for 11 centile levels and 40 age categories (as presented in Tables

Respiratory rate (RR, breaths/minute) centiles^{a}

Age (years) midpoint | Age/units midpoint | Age description | C1 | C2.5 | C5 | C10 | C25 | C50 | C75 | C90 | C95 | C97.5 | C99 |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|

0.042 | 0.5 month | <1 month | 19 | 22 | 24 | 27 | 33 | 39 | 47 | 54 | 59 | 64 | 69 |

0.125 | 1.5 months | 1 to <2 month | 20 | 22 | 24 | 27 | 32 | 38 | 45 | 53 | 58 | 63 | 69 |

0.208 | 2.5 months | 2 to <3 month | 19 | 21 | 23 | 26 | 30 | 37 | 44 | 52 | 58 | 63 | 70 |

0.292 | 3.5 months | 3 to <4 month | 19 | 21 | 23 | 25 | 29 | 35 | 42 | 51 | 57 | 63 | 71 |

0.375 | 4.5 months | 4 to <5 month | 18 | 20 | 22 | 24 | 28 | 34 | 41 | 49 | 56 | 62 | 71 |

0.458 | 5.5 months | 5 to <6 month | 18 | 20 | 21 | 23 | 27 | 33 | 40 | 48 | 54 | 61 | 71 |

0.542 | 6.5 months | 6 to <7 month | 18 | 20 | 21 | 23 | 27 | 32 | 38 | 46 | 53 | 60 | 70 |

0.625 | 7.5 months | 7 to <8 month | 18 | 19 | 21 | 23 | 26 | 31 | 37 | 45 | 51 | 58 | 69 |

0.708 | 8.5 months | 8 to <9 month | 18 | 19 | 20 | 22 | 26 | 30 | 36 | 44 | 50 | 57 | 67 |

0.792 | 9.5 months | 9 to <10 month | 17 | 19 | 20 | 22 | 25 | 30 | 35 | 43 | 49 | 56 | 66 |

0.875 | 10.5 months | 10 to <11 month | 17 | 19 | 20 | 22 | 25 | 29 | 35 | 42 | 48 | 55 | 65 |

0.958 | 11.5 months | 11 to <12 month | 17 | 19 | 20 | 21 | 24 | 29 | 34 | 41 | 47 | 54 | 64 |

1.042 | 12.5 months | 12 to <13 month | 17 | 18 | 20 | 21 | 24 | 28 | 34 | 40 | 46 | 53 | 64 |

1.125 | 13.5 months | 13 to <14 month | 17 | 18 | 20 | 21 | 24 | 28 | 33 | 40 | 46 | 52 | 63 |

1.208 | 14.5 months | 14 to <15 month | 17 | 18 | 20 | 21 | 24 | 28 | 33 | 39 | 45 | 51 | 62 |

1.292 | 15.5 months | 15 to <16 month | 17 | 18 | 19 | 21 | 24 | 27 | 32 | 39 | 44 | 51 | 62 |

1.375 | 16.5 months | 16 to <17 month | 17 | 18 | 19 | 21 | 23 | 27 | 32 | 38 | 44 | 50 | 61 |

1.458 | 17.5 months | 17 to <18 month | 17 | 18 | 19 | 21 | 23 | 27 | 32 | 38 | 43 | 50 | 61 |

1.542 | 18.5 months | 18 to <19 month | 17 | 18 | 19 | 20 | 23 | 27 | 31 | 37 | 43 | 49 | 60 |

1.625 | 19.5 months | 19 to <20 month | 17 | 18 | 19 | 20 | 23 | 26 | 31 | 37 | 42 | 48 | 59 |

1.708 | 20.5 months | 20 to <21 month | 16 | 18 | 19 | 20 | 23 | 26 | 30 | 36 | 41 | 48 | 59 |

1.792 | 21.5 months | 21 to <22 month | 16 | 18 | 19 | 20 | 23 | 26 | 30 | 36 | 41 | 47 | 58 |

1.875 | 22.5 months | 22 to <23 month | 16 | 18 | 19 | 20 | 22 | 26 | 30 | 35 | 40 | 47 | 58 |

1.958 | 23.5 months | 23 month to <2 years | 16 | 17 | 19 | 20 | 22 | 25 | 29 | 35 | 40 | 46 | 57 |

2.5 | 2.5 years | 2 to <3 years | 16 | 17 | 18 | 19 | 21 | 24 | 28 | 33 | 37 | 43 | 53 |

3.5 | 3.5 years | 3 to <4 years | 15 | 16 | 17 | 18 | 20 | 23 | 26 | 30 | 34 | 38 | 47 |

4.5 | 4.5 years | 4 to <5 years | 15 | 16 | 17 | 18 | 20 | 22 | 25 | 28 | 31 | 35 | 43 |

5.5 | 5.5 years | 5 to <6 years | 14 | 16 | 17 | 18 | 19 | 21 | 24 | 27 | 30 | 33 | 39 |

6.5 | 6.5 years | 6 to <7 years | 14 | 15 | 16 | 17 | 19 | 21 | 23 | 26 | 28 | 31 | 37 |

7.5 | 7.5 years | 7 to <8 years | 14 | 15 | 16 | 17 | 18 | 20 | 22 | 25 | 27 | 30 | 35 |

8.5 | 8.5 years | 8 to <9 years | 14 | 15 | 16 | 17 | 18 | 20 | 22 | 25 | 27 | 29 | 33 |

9.5 | 9.5 years | 9 to <10 years | 13 | 15 | 15 | 16 | 18 | 20 | 22 | 24 | 26 | 28 | 32 |

10.5 | 10.5 years | 10 to <11 years | 13 | 14 | 15 | 16 | 18 | 19 | 21 | 23 | 25 | 28 | 31 |

11.5 | 11.5 years | 11 to <12 years | 13 | 14 | 15 | 16 | 17 | 19 | 21 | 23 | 25 | 27 | 30 |

12.5 | 12.5 years | 12 to <13 years | 13 | 14 | 15 | 16 | 17 | 19 | 20 | 22 | 24 | 26 | 29 |

13.5 | 13.5 years | 13 to <14 years | 13 | 14 | 15 | 15 | 17 | 18 | 20 | 22 | 24 | 25 | 28 |

14.5 | 14.5 years | 14 to <15 years | 13 | 14 | 14 | 15 | 17 | 18 | 20 | 22 | 23 | 25 | 28 |

15.5 | 15.5 years | 15 to <16 years | 13 | 14 | 14 | 15 | 17 | 18 | 20 | 22 | 23 | 25 | 27 |

16.5 | 16.5 years | 16 to <17 years | 12 | 14 | 14 | 15 | 17 | 18 | 20 | 21 | 23 | 24 | 27 |

17.5 | 17.5 years | 17 to <18 years | 12 | 14 | 14 | 15 | 16 | 18 | 19 | 21 | 23 | 24 | 27 |

^{a}Centiles abbreviated as C1 (first centile) to C99 (99th centile)

In infants, Figure

In children (≥1 year), Figure

We hypothesized that empirically derived HR and RR distributions would deviate from PALS recommended distributions, and values derived from a low-acuity population of children seen in the ED. Ideally, an effective screening tool should balance high sensitivity to detect children at risk of deterioration while limiting too many false-positive patients leading to alert fatigue (

As seen in Figures

Similarly, the empirically derived 95th centile RR (Figure

Thus, it is not surprising that in a ward inpatient sample of over 116,000 observations, 54% of the HR and 40% of the RR vital signs observed in hospitalized children were outside textbook “normal” vital sign distributions and 38% of HR and 30% of RR observations would have resulted in increased early warning scores (

Several groups have used data from EHR’s to redefine the distribution of vital signs observed in children in the ED (

We believe our empirically derived centiles based on a very large, multi-centered database of ED visits by children provide evidence-based vital sign parameters to use in either EHR or paper-based early warning scores, to set monitor alarm limits and to risk-stratify children for clinical trials or epidemiologic studies. Rather than using dichotomous threshold values, which are often set by consensus, to define “normal” from “abnormal,” we believe that these data can lead to the development of more useful objective risk stratification tools.

Although using vital sign parameter

The potential value of using risk-stratified assessment of vital signs was seen in an ED-based study of 1,750 febrile children evaluated using seven different vital sign modeling strategies to identify children with serious bacterial infections (

Although we generated smooth curves for the individual vital signs, the raw vital sign distributions for RR shown in Figure

We performed a stratified split-sample validation that showed (Table

Most screening tools do not TMP adjust the HR or RR. It is well known that increased TMP increases metabolic demand, thus increasing HR and RR (

We planned to analyze the effect of TMP on HR and RR, but we were not confident in adjusting oral TMP to core TMP and did not know how to correct for axillary or temporal artery TMPs. A recent systematic review of the accuracy of peripheral versus core TMPs in adults and children (

Since we did not have precise ages, we used 1-month ranges (if <2 years of age) or 1-year age ranges for older children. It seems likely that the large sample size and smoothing of the data will limit the impact of including children within an entire year, especially since within any year range there is sizable normal variation based on child size and height as seen with blood pressure (

We chose to use extreme exclusion criteria for HR and RR. This may have resulted in the inclusion of data from children with erroneous or very abnormal vital sign values, but again the very large data set likely limits the impact of these outliers.

Further studies are needed to determine if using empirically derived vital sign thresholds would result in a lower sensitivity to identify high-risk children. Our previous analysis using empirically derived higher vital sign thresholds based on data from a single ED noted improved positive predictive value for severe sepsis/septic shock identification without affecting sensitivity (

In summary, the derived HR and RR centiles and

The datasets used and analyzed during the current study are available from the corresponding author, but restrictions apply to the availability of these data since they were obtained through a written agreement with Cerner Corporation and so are not publicly available. Data may be available from the author on reasonable request and with permission of Cerner Corporation.

Since we analyzed de-identified clinical data, this is not human subjects research.

RS, SG, and AZ each made substantial contributions to the conception and design of the study. RS conducted the analyses and derived the centile curves and

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The authors acknowledge the support of Cerner Corporation.

The Supplementary Material for this article can be found online at

ED, Emergency Department; EHR, electronic health record; HR, heart rate; GAMLSS, generalized additive models for location, scale, and shape; RR, respiratory rate; TMP, temperature; BCPE, Box–Cox Power Exponential; BCT, Box–Cox “t”; PALS, Pediatric Advanced Life Support; PEARS, Pediatric Emergency Assessment, Recognition and Stabilization; CCS, Clinical Classifications Software; ICU, intensive care unit; C, Centile (percentile).