Birth-size reference charts for newborns admitted to the neonatal intensive care units

Abstract

Background

Our original birth-size reference charts (the Olsen charts), published in 2010 (weight, length, head circumference) and 2015 (body mass index), had limitations, including manually defined exclusion criteria and the use of completed gestational weeks. This project aimed to update the charts and address those limitations.

Methods

We analyzed de-identified data from infants admitted to Pediatrix NICUs between 2010 and 2022. In contrast to prior work, newborns with mortality before discharge, multiple births, and congenital anomalies were included; only statistical outliers (defined as two times the interquartile range below the first quartile or above the third quartile) were excluded. Gestational age was recorded in days, and penalized b-splines were used as curve smoothers. Birth weight, length, head circumference, and BMI models were developed using a randomly selected subset and validated by the remaining data. Proportions of SGA and LGA infants were compared to the 2010 Olsen and 2013 Fenton charts.

Results

One million one hundred fifty-two thousand seven hundred twenty infants born between 22 and 41 weeks of gestation were included. No overfitting bias was observed in the new charts. These charts more accurately identified the expected 10% of newborns as SGA/LGA compared to existing charts.

Conclusions

The use of refined methods in new charts improved birth size referencing for NICU infants.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12887-026-06889-1.

Background

Birth-size reference charts (also known as intrauterine growth charts in the literature) are used to assess size at birth and monitor the growth of preterm infants [1–4]. In 2010, we reported birth-size reference charts based on anthropometric measures performed in neonatal intensive care units (NICUs) across the United States from 1998 to 2006 [1]. This was our prior attempt to develop anthropometric reference charts using a large, racially diverse dataset of cross-sectional birth data. We plotted the growth charts for weight, length, and head circumference by completed week of gestation and sex to estimate fetal growth [1]. In 2015, we also reported birth-size charts for body mass index (BMI) [5]. Our goal in this original work was to provide a reference tool to assess size at birth and monitor postnatal growth.

Over the last decade, limitations regarding our original work have been identified. These limitations include the definition of the sub-populations considered to exhibit “normal” vs. “abnormal” growth and the bias introduced by the exclusion criteria [2, 6]. Additionally, using completed rather than average gestational weeks may lead to misclassification of birth size [7]. In this study, we aimed to develop a new set of birth-size reference charts to address the limitations of the old charts and to provide more current charts.

Methods

Data source and dataset curation

This study utilized a de-identified, cross-sectional dataset of birth anthropometric measures from the Pediatrix Medical Group Clinical Data Warehouse (CDW). The dataset included infants born between 2010 and 2022 with a gestational age (GA) ranging from 22 0/7 to 41 6/7 weeks and admitted to US NICUs. Available data from the CDW comprised birth weight (measured on an electronic scale to the nearest 10 g), length (taken with measuring tape or a length board to the nearest millimeter), and head circumference (measured using measuring tape to the nearest millimeter). The attending neonatologist assigned the estimated GA based on the best assessments, including obstetric history, obstetric examination, prenatal ultrasound, and postnatal physical examination. Both completed weeks and days were recorded (for instance, 36 weeks and 4 days) to facilitate modeling by gestational age in days.

The dataset was randomly split into a training set and a validation set. The training set consisted of data from 50% of randomly selected infants for each gestational day and was used to develop the initial growth models for bias analysis. If there were fewer than 200 infants for a specific gestational day, we used data from 80% of the randomly selected infants to optimize the number of data points in the training set. The remaining infants were included in the validation set. The final growth charts were developed using all available data points from both training and validation sets, except for outliers.

New growth chart development

Each infant contributed a single value for birth weight, length, or head circumference. BMI was calculated using the formula: . Outliers were identified as values located two times the interquartile range below the first quartile or above the third quartile for each GA in days and were subsequently excluded. The remaining measurement data and the corresponding GA in days were used in generalized additive models for location, scale, and shape (GAMLSS). Separate growth charts were developed for males and females. The L (lambda, representing the skewness), M (mu, representing the mean), and S (sigma, representing the variance) values for each gestational day were then extracted from the model to construct the LMS table. Penalized b-splines were utilized as the default curve smoother unless stated otherwise. All models were developed using the gamlss package in R.

Model validation

We assessed model bias using three validation metrics: mean square error (MSE), mean absolute error (MAE), and mean z-scores. First, we calculated the MSE and MAE for each gestational day based on the residuals of the training set. Subsequently, we applied the validation set to the models to calculate MSE and MAE for validation. When plotted against GA in days, the MSEs and MAEs computed using the training and validation sets are expected to overlap for the model to be considered non-biased.

Additionally, we calculated z-scores of measurements in the training and validation datasets, and then grouped the z-scores according to completed GA weeks (from 22 to 41 weeks) to obtain the mean and the corresponding 99.75% confidence interval (CI) for each completed gestational week group by using an adjusted α of 0.0025 for 20 comparisons (22 to 41 weeks), as previously established [1].

Comparing the performances of the new growth charts and the existing growth charts

Using the validation dataset, we compared the percentages of infants below the 10^th percentile and above the 90^th percentile curves of the original growth charts developed in 2010 (2010 Olsen charts) [1], the 2013 growth charts developed by Fenton et al. (2013 Fenton charts) [2], and the new charts using the training dataset in this study. We first calculated the z-score/percentile for each measurement in both the training and validation datasets using the new growth charts, and the z-score/percentile for each measurement in the validation dataset using the 2010 Olsen and 2013 Fenton charts based on the actual gestation at birth in weeks and days. We then calculated the percentages of infants below or above the indicated percentile curves for each completed gestational week group. Since the LMS values of the 2010 Olsen charts are only available by completed gestational weeks, the z-score/percentile calculation was based on the completed gestational week when the measurements were taken (rather than the GA in days). We performed bootstrapping, involving 100 iterations of repeated sampling with replacement, to obtain the 99.75% CI as we did for the z-score described above.

Results

We analyzed 1,152,720 infants born with a GA between 22 0/7 and 41 6/7 weeks from 2010 to 2022. Maternal and neonatal characteristics are provided in Table 1. All 1,152,720 infants had weight data, with 9,240 (0.8%) categorized as outliers; 1,072,545 (93.0%) had length data, with 9,548 (0.9%) identified as outliers; and 1,078,644 (93.6%) had head circumference data, with 9,059 (0.8%) considered outliers.

Table 1.

Maternal and neonatal characteristics

Number	1,152,720
Maternal
Maternal age (years), median (IQR)	29 [24, 33]
Antenatal steroids, n (%)	371,198 (32.2)
Chorioamnionitis, n (%)	52,818 (4.6)
Pre-eclampsia, n (%)	122,117 (10.6)
Smoking, n (%)	66,853 (5.8)
Diabetes mellitus, n (%)	162,533 (14.1)
Obesity, n (%)	73,330 (6.4)
Fetal growth restriction, n (%)	78,015 (6.8)
Multiple gestations, n (%)	155,432 (13.5)
Neonatal
Female sex, n (%)	510,710 (44.3)
Race/ethnicity, n (%)
White	534,353 (46.3)
Black	202,376 (17.6)
Hispanic	238,576 (20.7)
Asian	37,104 (3.2)
Other	140,311 (12.2)
Major congenital anomalies, n (%)	129,190 (11.2)
In-hospital mortality, n (%)	18,133 (1.6)

Model development

We used all data points in the training dataset, excluding outliers, to develop growth models for weight, length, head circumference, and BMI (Fig. 1). In the modeled curves, we observed a rapid widening of the distance between percentile curves between 35 and 38 weeks, indicating greater variability in measurements, followed by a narrowing at later GAs, suggesting reduced variability. This observation was most pronounced in weight curves. It was associated with increases in the proportions of mothers affected by gestational diabetes mellitus, pre-eclampsia, intrauterine growth restriction (based on obstetric assessments), multiple gestations, and fetuses affected by major anomalies between 35–38 completed weeks compared to 39–41 completed weeks (Table 2). This observation was evident only with the penalized B-spline smoothing technique, which outperformed cubic splines in fitting the unsmoothed percentiles calculated from the raw data (Supplemental Fig. 1).

Fig. 1.

Modeled birth-size reference charts. The modeled birth-size reference charts for weight (A for males, B for females), length (C for males, D for females), head circumference (E for males, F for females), and body mass index (G for males, H for females) are presented. The representative curves displayed include the 3^rd, 10^th, 50^th, 90^th, and 97^th percentiles, arranged from bottom to top

Table 2.

Maternal and fetal characteristics stratified by completed gestational week

Completed gestational week group	< 35 weeks (n = 467,561)	35–38 weeks (n = 453,556)	39–41 weeks (n = 231,603)
Maternal age, median [IQR]	29 [24, 33]	29 [24, 33]	28 [23, 32]
Chorioamnionitis, n (%)	13,192 (2.8)	14,898 (3.3)	24,728 (10.7)
Pre-eclampsia, n (%)	74,807 (16.0)	43,512 (9.6)	3,798 (1.6)
Smoking, n (%)	28,487 (6.1)	26,626 (5.9)	11,740 (5.1)
Diabetes mellitus, n (%)	63,989 (13.7)	80,286 (17.7)	18,258 (7.9)
Obesity, n (%)	31,687 (6.8)	31,030 (6.8)	10,613 (4.6)
Fetal growth restriction, n (%)	36,924 (7.9)	37,583 (8.3)	3,508 (1.5)
Multiple gestations, n (%)	108,652 (23.2)	46,514 (10.3)	266 (0.1)
Major anomalies, n (%)	48,858 (10.4)	56,038 (12.4)	24,294 (10.5)

IQR Interquartile range

Model validation

We calculated MSEs and MAEs from the training and validation datasets. When plotted against GA in days, we found that the MSEs and MAEs from the training set closely overlapped with those from the validation set (Supplemental Fig. 2–5).

We also plotted the mean z-scores and CIs for each completed gestational week group, calculated from the training and validation datasets, for comparison. The mean z-scores are expected to be 0. The CIs all crossed zero and showed significant overlap between the training and validation datasets (Fig. 2 for weight; Supplemental Fig. 6 for length, head circumference, and BMI).

Fig. 2.

Mean birth weight z-scores by completed gestational week. The mean z-score, along with the 99.75% confidence interval (based on an adjusted α of 0.0025 per group for 20 completed gestational-week groups from 22 to 41 weeks), for males (A) and females (B) is calculated from the training dataset (grey) and the validation dataset (black) using the LMS values of the newly created nursery charts

Comparison of the new charts with existing birth-size reference charts

Using the training and validation datasets, we first compared the percentages of infants below the 10^th percentile and above the 90^th percentile on the new charts. We found the percentages comparable between the two datasets at around 10% for most completed gestational week groups [compare between 2025 Chou-NICU (Training set) and 2025 Chou-NICU (Validation set) in Fig. 3].

Fig. 3.

The percentages of newborns whose birth weights fall below the 10^th or above the 90^th percentile curves of the indicated charts. The percentages of newborns from the validation dataset (A) that fall below the 10^th percentile and (B) that exceed the 90^th percentile curves of the 2010 Olsen charts (white bars), the 2013 Fenton charts (light grey bars), and the 2025 NICU charts (black bars) are plotted against completed gestational weeks for comparison. Newborns from the training dataset (dark grey bars) are also calculated for comparison. The error bars represent 99.75% confidence intervals (based on an adjusted α of 0.0025 per group for 20 completed gestational-week groups between 22 and 41 weeks), calculated from bootstrapping of samples 100 times with replacement

Using the validation dataset, the percentages of infants below the 10^th and above the 90^th percentile curves for weight from the indicated reference charts were compared. While the percentages across completed gestational weeks remained close to 10% with the new charts [bars for the 2025 Chou-NICU (Validation set) in Fig. 3], the percentages varied across completed gestational weeks when using the 2010 Olsen charts and the 2013 Fenton charts (bars for the 2010 Olsen and 2013 Fenton in Fig. 3). Specifically, the 2010 Olsen charts overestimated the number of male and female newborns below the 10^th percentile among those born at < 28 weeks and between 35–37 weeks. In contrast, the 2013 Fenton charts underestimated male and female newborns below the 10^th percentile between 28 and 33 weeks and overestimated male and female newborns below the 10^th percentile between 35 and 37 weeks. The Fenton charts also overestimated the number of male newborns below the 10^th percentile born at 40–41 weeks. The 2013 Fenton charts underestimated male and female newborns above the 90^th percentile at 23–35 and 40–41 weeks, while overestimating those above the 90^th percentile at 37–39 weeks. Fluctuations were also observed in length and head circumference measurements when using the 2010 Olsen and 2013 Fenton charts (Supplemental Figs. 7 and 8).

Distribution of weight percentile for newborns affected by various conditions

We calculated the birth size percentile for each newborn affected by mortality, major congenital anomalies, multiple gestation, maternal diabetes mellitus, and maternal pre-eclampsia to assess the distribution (Fig. 4 for weight, Supplemental Fig. 9 for length, and Supplemental Fig. 10 for head circumference). We found that newborns affected by each examined condition were born across the full spectrum of percentiles, from 0 to 100. For those newborns who died, they were slightly more likely to be born below the 20^th percentile, with a prevalence above 5% among those born below the 2^nd percentile. A higher percentage of newborns affected by major congenital anomalies were born below the 2^nd or above the 98^th percentile. Birth weights of newborns born as multiple gestations spanned the entire percentile range. Among newborns below the 50^th percentile, 15–20% were born as a result of multiple gestations; conversely, among newborns above the 95^th percentile, fewer than 5% were born as a result of multiple gestations. Among newborns affected by maternal diabetes mellitus, 23.4% were born above the 90^th percentile (instead of the expected 10%). Conversely, 17.3% of newborns exposed to pre-eclampsia were born below the 10^th percentile (instead of the expected 10%).

Fig. 4.

Distribution of weight percentiles among newborns affected by mortality before discharge, major anomalies, multiple births, maternal diabetes mellitus, and maternal pre-eclampsia. The histograms show the sex-specific percentages for each condition (y-axis) across percentiles (x-axis) from 0 to 100

2025 birth-size reference charts for NICU newborns

The weight, length, head circumference, and BMI curves for the 3^rd, 10^th, 25^th, 50^th, 75^th, 90^th, and 97^th percentiles are presented in Supplemental Fig. 11 (females) and Supplemental Fig. 12 (males) in a more user-friendly format suitable for clinical settings. The LMS tables are included in Supplemental Tables 1–8.

Discussion

In this study, we developed new birth-size reference charts based on GA in days, using an updated group of US NICU infants born between 22 0/7 and 41 6/7 weeks of gestation from 2010 to 2020. Similar to our original work, only infants admitted to the NICU were included. Unlike the original work, we used (1) GA in days (vs. completed gestational weeks), (2) the penalized b-splines smoothing technique, and (3) all available data points except for outliers (i.e., no exclusions based on maternal or fetal conditions). With these improvements, our goal was to achieve a better fit and provide more accurate birth percentile assignment across the GA range for infants admitted to the NICUs. The resultant curves fit the input raw data better and are non-biased. This method makes the percentile curves appear less smooth, but they are more consistent with real-world data. Because no overfitting bias was observed at the two GA extremes with this modeling method, the new charts expanded the low GA limit to 22 0/7 weeks (compared with the 2010 Olsen charts, which started at 24 weeks of gestation) to improve birth-size referencing for the increasing numbers of infants born at 22 and 23 weeks [8].

NICUs today use charts based on weekly data to assess the birth size of preterm and term infants and the postnatal growth of preterm infants [1, 2, 5, 9, 10]. These curves were derived from data with varying patient-selection criteria and were smoothed using cubic splines [1, 2, 5, 9, 10] or polynomials [9], methods that are prone to information loss. The 2013 and 2025 Fenton charts provided the most significant contrast in patient selection [2, 10], with the 2025 version excluding a limited list of conditions thought to affect fetal growth—congenital anomalies, Caesarean birth without a trial of labor, maternal smoking, maternal hypertension before or during pregnancy, and maternal diabetes. In our study, we included all newborns other than outliers, and employed penalized b-splines for curve smoothing, which partly explains the improved accuracy in identifying infants < 10th percentile (widely used to define small for gestational age, SGA) and > 90th percentile (widely used to define large for gestational age, LGA) when compared to either the 2010 Olsen charts or the 2013 Fenton charts. We further examined the distribution of birth weight for several conditions that may be associated with fetal growth, with the goal of emphasizing that infants affected by these conditions have birth weights spanning the full range of percentiles, indicating varying degrees of effect on fetal growth, and that these conditions are not the sole drivers of fetal growth. For example, newborns affected by maternal diabetes may be born LGA, but may also be born appropriate for gestational age, likely with altered body composition [11]. If a condition is more common in heavier infants, excluding those infants would inevitably shift the percentile curves downward. With this downward shift, the percentile ranking of the birth size of infants with the condition in question would become even higher. In other words, excluding any of these conditions may shift the final reference curves in the opposite direction, exaggerating the impact of the condition on fetal growth. Moreover, the list of conditions that may affect fetal growth is non-exhaustive, raising further concerns about selection bias. Several studies have also demonstrated the differences between the SGA and LGA assignments based on existing growth chart references, which employed various patient exclusion criteria [4, 12]. The method used to develop growth curves can influence how infants are classified based on their birth measurements.

Our new approach to growth chart development has identified the expected 10% of infants born SGA and LGA, including the range between 35 and 38 weeks, where older charts overestimated both SGA and LGA infants. This improvement is primarily attributed to capturing a greater degree of variability in birth size between 35 and 38 weeks in the new charts with an inclusive approach to patient selection and with the use of penalized b-splines. Further examination of maternal pregnancy complications has led to an important observation that the impact of maternal risk factors on fetal growth rate is GA-dependent in cross-sectional studies. For example, most infants born to mothers with diabetes were delivered between 35 and 38 completed weeks. Mothers with diabetes are more likely to deliver LGA infants if glucose control during gestation is not well controlled [13]. Infants of diabetic mothers may be at greater risk of experiencing respiratory distress, hypoglycemia, and feeding problems requiring NICU admissions [13]. When compared to infants born at 39–41 completed weeks, infants born between 35 and 38 completed weeks were also more likely to be born to mothers carrying multiples, those with pre-eclampsia, and those with fetal growth restriction. These infants are more likely to be born SGA. Together, these findings explain the biphasic change in variability (difference between the 10th and 90th percentile) in our new charts between 35 and 41 completed weeks, as well as the improved accuracy in SGA/LGA assignment in this GA range.

Because the effect of maternal risk factors on birth size depends on GA, the selection bias introduced by excluding infants born to mothers with these risk factors, as seen in the older charts, would also be GA-dependent. For instance, the overestimation of SGA infants before 28 weeks’ gestation by the 2010 Olsen charts may be due to the use of exclusion criteria (multiple gestation, NICU mortality, and congenital anomalies), as infants born to mothers carrying multiples tend to be born earlier than those born as singletons. Mortality is more prevalent among infants delivered at an earlier GA. Similarly, the underestimation of both SGA infants born between 28 and 34 weeks’ gestation and LGA infants born before 34 weeks’ gestation by the 2013 Fenton charts may be due to the inclusion of six studies, each with varying selection criteria, resulting in lower estimates in the low-percentile curves. Although not directly assessed in this study, the underestimation of SGA infants born between 28 and 34 weeks’ gestation is likely not seen when the 2025 Fenton charts are used for reference; instead, an overestimation of SGA infants is expected because the low-percentile curves are much higher in the 2025 Fenton charts (compared to the 2013 charts) due to the exclusion of newborns exposed to maternal hypertensive disorders of pregnancy and maternal smoking. Moreover, the overestimation and underestimation of SGA and LGA infants delivered after 34 weeks’ gestation by the 2013 Fenton charts may be attributed to several factors, including the meta-analysis of six studies, the application of cubic splines, the inclusion of uncomplicated newborns, and the smoothing of two datasets from 36 to 50 weeks. The new charts developed in this study help mitigate these biases and serve as a new and bona fide reference tool that enables more accurate SGA/LGA assignment across the entire GA range for NICU newborns. Since SGA and LGA are categorical variables derived from continuous measures and depend on the population included in the sample, we recommend using percentiles or z-scores for predictive modeling of outcomes. Whether z-scores or SGA/LGA are used to describe birth size, it is crucial to define the goal of size designation and select the charts best suited to meeting that goal. Our new charts provide a reference tool for comparing newborn size to all newborns in our database, whereas the 2025 Fenton charts provide a reference for comparing to a selected subgroup of newborns based on predetermined criteria.

We published the first reference charts for BMI in 2015 [5]. Using the modeling approach established in this study, we updated the BMI reference charts in this report. With the increasing availability of body composition measurement tools in the NICU [13], predicting body composition with the combined use of birth weight, length, and BMI charts may provide additional utility for these reference tools in estimating body composition in newborns without direct measurement.

In addition to their use in birth-size assessment, birth-size reference charts are currently used for postnatal growth assessment of preterm infants in the NICU. This approach may be problematic for several reasons. First, the curves of the cross-sectional birth-size reference charts derived from NICU infants born beyond 34 weeks’ gestation are biased because most infants born after 34 weeks’ gestation are not admitted to the NICU. Second, we found that pregnancy complication profiles differ among GA groups, making the population heterogeneous across GA groups. In addition to these two issues, each gestational day consists of infants with distinct demographic compositions due to the cross-sectional nature of the dataset. Black infants are overrepresented among extremely low gestational age infants [14, 15]. Black infants also have lower birth weight compared to White and Asian infants [16, 17]. As a result, the curves may better estimate the growth of Black fetuses in the early GA range than in the later GA range. The composition of the birth cohort changes as GA progresses, affecting the birth size distribution at each GA [18]. It is crucial to understand that the calculated z-scores for any measurement using birth-size reference charts only adjust for sex and GA relative to the birth cohort of that specific GA. The heterogeneity of the cohort across GA groups was not taken into account when calculating the z-scores. Using a meta-analysis to combine curves with different selection criteria, as done in the Fenton 2013 charts [17], compounds these problems.

Growth assessment starts at birth and continues regularly throughout the NICU stay. Birth-size reference charts are created from birth data, making them the most appropriate tool for assessing newborn size at birth. Developing a consensus on defining optimal postnatal growth for preterm infants and the best monitoring tools remains elusive. Most agree that the fetal growth rate remains an appropriate tool for assessing postnatal growth of preterm infants, as recommended by the American Academy of Pediatrics [19]. However, fetal growth rate is not a constant. The intrauterine environment influences fetal growth rate throughout gestation [20]. Although it is a common practice, postnatal use of birth-size reference charts for longitudinal growth assessment of preterm infants has limitations. One of the most important reasons is that, while most newborns below 35 weeks were admitted to the NICU, most infants born at 35 weeks or above were not. Therefore, when applying our birth-size reference charts for postnatal growth assessment, it may be more appropriate to limit their use to infants younger than 35 weeks postmenstrual age and to combine them with growth velocity assessment [20]. Additionally, growth assessment in the NICU may be more comprehensive when birth-size reference charts are combined with our recently published 2023 Postnatal Growth Charts for Preterm Infants and the WHO charts [15, 21, 22].

Despite significant improvements, our new birth-size reference charts have limitations. First, the input data were derived from infants admitted to the NICU. As with our original work, this means the growth charts may not be suitable for uncomplicated newborns admitted to the nursery. Additionally, infants who were born alive but died before NICU admission were excluded, which may introduce further bias, particularly in the extremely low GA groups. Similar to any administrative dataset, underreporting and inaccurate reporting are possible despite internal data consistency checks, low rates of missing data, and prospective data entry. This report did not compare the new charts with Intergrowth-21st [10] and the 2025 Fenton charts [9] because these charts were designed to be growth standards—how infants should grow under optimal conditions. Our charts were intended to be a size reference for this population.

Conclusion

This report provides updated birth-size reference charts based on a recent NICU population in the United States. This new set of sex-specific birth-size reference charts is all-inclusive and intended as a reference (not a standard) tool for birth-size referencing and comparison. The updated charts and the corresponding LMS values are accurate to gestational day. These updated charts offer a new and enhanced reference tool for assessing birth size in the NICU.

Abbreviations

NICU

Neonatal intensive care units

BMI

Body mass index

CDW

Clinical data warehouse

GA

Gestational age

GAMLSS

Generalized additive models for location, scale, and shape

LMS

Lambda, mu, sigma

MSE

Mean squared error

MAE

Mean absolute error

CI

Confidence interval

Authors’ contributions

Dr. Fu-Sheng Chou designed the study, performed the analysis, interpreted data, and drafted and critically reviewed the initial and revised manuscript. Dr. Hung-Wen Yeh supervised data analysis and critically reviewed the initial manuscript. Ms. Crystal Hsueh contributed to data analysis and critically reviewed the initial manuscript. Drs. Irene E. Olsen and Reese H. Clark contributed to the study design, interpreted data, and drafted and critically reviewed the initial and revised manuscript. Dr. Veeral N. Tolia interpreted data and critically reviewed the initial and revised manuscript.

Funding

No relevant financial support was granted for this project.

Data availability

The data that support the findings of this study are available from Pediatrix Medical Group but restrictions apply to the availability of these data, which were used under contract for the current study, and so are not publicly available. Data are however available for request from Pediatrix Medical Group at https://mednax2ndhipaaaccount-gajdh.formstack.com/forms/research\_application\_complete.

Declarations

Ethics approval and consent to participate

This data-only study with a deidentified dataset was deemed non-human research by the Kaiser Permanente Southern California Institutional Review Board (https://kpscnursingresearch.org/kpsc-institutional-review-board-irb/). As such, informed consent to participate was not required, including from the parents or legal guardians of any participant under the age of 16.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.