678 - Introducing New Intrauterine Growth Charts Based on the U.S. Data from Pediatrix Medical Group for Use in NICU and Nursery

Publication Number: 678.5169

Fu-Sheng Chou, Kaiser Permanente Bernard J. Tyson School of Medicine, Riverside, CA, United States; Hung-Wen Yeh, Children's Mercy Hospitals and Clinics, Kansas City, MO, United States; Crystal Hsueh, University of California, Los Angeles, Los Angeles, CA, United States; Veeral Tolia, Pediatrix, Dallas, TX, United States; Irene Olsen, Drexel University, Philadelphia, PA, United States; Reese H. Clark, Duke University School of Medicine, Marietta, SC, United States

Background: Existing growth charts for assessing birth size in the NICU and nursery have limitations: 1) The 2010 Olsen charts use limited selection criteria to exclude newborns presumed to have abnormal growth and rely on completed weeks, 2) The 2013 Fenton charts are based on a meta-analysis of 6 studies, each with its selection criteria, and extrapolate curves beyond 36 weeks to connect with WHO charts, 3) The WHO charts lack GA-specific percentile curves at birth.

Objective: We aimed to develop new growth charts based on newborns requiring NCU care and those admitted to nurseries for comparison.

Design/Methods: Deidentified data from newborns admitted to Pediatrix NICUs in 2010-2022 were used to develop NICU growth charts. For nursery charts, data were collected from uncomplicated newborns in 2021-2024 admitted to Pediatrix nurseries with a discharge age <= 2 (vaginal) or 3 (Cesarean) days. The data were split into training and validation sets. The training set was used to create growth models based on GA in days using P-splines. The validation set assessed bias by analyzing the mean z-score distribution. The percentages of newborns classified as LGA (>90th percentile) or SGA ( < 10th percentile) in the validation set calculated using new charts, 2010 Olsen, and 2013 Fenton charts were compared. Bootstrapping was used to calculate confidence intervals.

Results: Data from 1,152,720 newborns between 22w0d and 41w6d were used for NICU charts, and from 129,150 newborns between 35w0d and 41w6d for nursery charts. Maternal/neonatal characteristics are in Table 1. The training and validation data were well-matched and showed no bias (Fig 1A,2A). Unlike nursery charts (Fig 2B), we observed bulging between the 3rd and 97th percentile curves in NICU charts between 35-38 weeks, suggesting increased variance (Fig 1B,2B). This coincided with a higher proportion of infants of diabetic mothers and those with growth restriction (Fig 1C). In the validation sets, we found that both new charts, unlike 2010 Olsen or 2013 Fenton charts, correctly identified the expected 10% of newborns as SGA or LGA (Fig 1D,2C). In a separate group of newborns initially admitted to nursery but developed hypoglycemia or required transfer for higher levels of care, we identified a higher proportion of newborns being SGA or LGA when using the nursery charts (Fig 2D).

Conclusion(s): Intrauterine growth charts developed using data from newborns requiring NICU care are significantly different from those developed using newborns admitted to nurseries. The new NICU and nursery charts better represent the newborns admitted to each unit for size categorization.

Table 1. Maternal and Neonatal Characteristics.


Figure 1. New intrauterine growth charts for newborns admitted to the neonatal intensive care units.
(A) The mean z-scores along with the 99.75% confidence interval (based on the adjusted α of 0.0025 per group for 20 completed gestational week groups) for males (left panel) and females (right panel) were calculated from the training dataset (grey) and the validation dataset (black) using the birth weight growth curves, showing confidence intervals crossing 0 and overlapping intervals between the training and validation datasets, suggesting non-biased weight curves. Similar findings were observed for length and head circumference (not shown here). (B) The representative percentile curves of the 3rd, 10th, 25th, 50th, 75th, 90th, and 97th percentiles from bottom to top for female weight (left upper), female length and head circumference (right upper), male weight (left lower), and male length and head circumference (right lower). (C) Maternal and fetal characteristics stratified by completed gestational week for the subset of infants in the 35-38- and 39-41-week groups. Newborns exposed to pregnancy complications leading to increased likelihood of delivering heavier or lighter newborns were more likely to be delivered between 35-38 weeks. (D) Left panel: The percentages of infants from the validation dataset below the 10th-percentile curves of the 2010 Olsen charts (2010 Olsen), the 2013 Fenton charts (2013 Fenton), and the new birth weight charts developed in this study (2024 Chou-Validation) were plotted against completed gestational week for comparison. The percentages of infants from the training dataset below the 10th-percentile curves of the new charts (2024 Chou-Training) were also calculated and plotted for comparison. Right panel: The percentages of infants from the validation dataset above the 90th-percentile curves of the 2010 Olsen charts (2010 Olsen), the 2013 Fenton charts (2013 Fenton), and the new birth weight charts developed in this study (2024 Chou-Validation) were plotted against completed gestational week for comparison. The percentages of infants from the training dataset above the 90th-percentile curves of the new charts (2024 Chou-Training) were also calculated and plotted for comparison. The error bars represent 99.75% confidence intervals (based on the adjusted α of 0.0025 per group for 20 completed gestational week groups) calculated from bootstrapping of samples 100 times with replacement. The results indicate that the new growth charts, but not the 2010 Olsen or the 2013 Fenton charts, were able to correctly identify the expected 10% for infants that are <10th or >90th percentiles.

Figure 2. New intrauterine growth charts for uncomplicated newborns admitted to the nursery.
Figure 2. New intrauterine growth charts for uncomplicated newborns admitted to the nursery. (A) The mean z-scores along with the 99.3% confidence interval (based on the adjusted α of 0.007 per group for the 7 completed gestational week groups between 35 and 41 weeks) for males (left) and females (right) are calculated from the training dataset (light color) and the validation dataset (dark color) using the weight curves of the intrauterine growth charts developed in this study. Similar findings were observed for length and head circumference (not shown here). (B) The intrauterine growth charts for uncomplicated male (left, blue curves) and female (right, red curves) newborns on weight (top), length (middle), and head circumference (bottom) were plotted on the newly developed intrauterine growth charts for newborns admitted to the NICU (black curves) for comparison. (C) Top panel: the percentages of infants from the validation dataset below the 10th-percentile curves of the 2010 Olsen charts (2010 Olsen–Validation), the 2013 Fenton charts (2013 Fenton–Validation), the 2024 Chou charts for NICU infants (2024 Chou-NICU–Validation), and the new weight charts developed in this study (2024 Chou-nursery–Validation) were plotted against gestational week for comparison. The percentages of infants from the training dataset below the 10th-percentile curves of the new charts (2024 Chou-nursery–Training) were also calculated and plotted for comparison. Bottom panel: the percentages of infants from the validation dataset above the 90th-percentile curves of the 2010 Olsen charts (2010 Olsen–Validation), the 2013 Fenton charts (2013 Fenton–Validation), the 2024 Chou charts for NICU infants (2024 Chou-NICU–Validation), and the new weight charts developed in this study (2024 Chou-nursery–Validation) were plotted against gestational week for comparison. The percentages of infants from the training dataset above the 90th-percentile curves of the new charts (2024 Chou-nursery–Training) were also calculated and plotted for comparison. The error bars represent 99.3% confidence intervals (based on the adjusted α of 0.007 per group for 7 gestational week groups between 35 and 41 weeks) calculated by bootstrapping of samples with replacement 500 times. (D) The number (percentage) of infants with hypoglycemia, requiring transfer, or both who were also small or large for gestational age based on the indicated growth charts.


Table 1. Maternal and Neonatal Characteristics.


Figure 1. New intrauterine growth charts for newborns admitted to the neonatal intensive care units.
(A) The mean z-scores along with the 99.75% confidence interval (based on the adjusted α of 0.0025 per group for 20 completed gestational week groups) for males (left panel) and females (right panel) were calculated from the training dataset (grey) and the validation dataset (black) using the birth weight growth curves, showing confidence intervals crossing 0 and overlapping intervals between the training and validation datasets, suggesting non-biased weight curves. Similar findings were observed for length and head circumference (not shown here). (B) The representative percentile curves of the 3rd, 10th, 25th, 50th, 75th, 90th, and 97th percentiles from bottom to top for female weight (left upper), female length and head circumference (right upper), male weight (left lower), and male length and head circumference (right lower). (C) Maternal and fetal characteristics stratified by completed gestational week for the subset of infants in the 35-38- and 39-41-week groups. Newborns exposed to pregnancy complications leading to increased likelihood of delivering heavier or lighter newborns were more likely to be delivered between 35-38 weeks. (D) Left panel: The percentages of infants from the validation dataset below the 10th-percentile curves of the 2010 Olsen charts (2010 Olsen), the 2013 Fenton charts (2013 Fenton), and the new birth weight charts developed in this study (2024 Chou-Validation) were plotted against completed gestational week for comparison. The percentages of infants from the training dataset below the 10th-percentile curves of the new charts (2024 Chou-Training) were also calculated and plotted for comparison. Right panel: The percentages of infants from the validation dataset above the 90th-percentile curves of the 2010 Olsen charts (2010 Olsen), the 2013 Fenton charts (2013 Fenton), and the new birth weight charts developed in this study (2024 Chou-Validation) were plotted against completed gestational week for comparison. The percentages of infants from the training dataset above the 90th-percentile curves of the new charts (2024 Chou-Training) were also calculated and plotted for comparison. The error bars represent 99.75% confidence intervals (based on the adjusted α of 0.0025 per group for 20 completed gestational week groups) calculated from bootstrapping of samples 100 times with replacement. The results indicate that the new growth charts, but not the 2010 Olsen or the 2013 Fenton charts, were able to correctly identify the expected 10% for infants that are <10th or >90th percentiles.

Figure 2. New intrauterine growth charts for uncomplicated newborns admitted to the nursery.
Figure 2. New intrauterine growth charts for uncomplicated newborns admitted to the nursery. (A) The mean z-scores along with the 99.3% confidence interval (based on the adjusted α of 0.007 per group for the 7 completed gestational week groups between 35 and 41 weeks) for males (left) and females (right) are calculated from the training dataset (light color) and the validation dataset (dark color) using the weight curves of the intrauterine growth charts developed in this study. Similar findings were observed for length and head circumference (not shown here). (B) The intrauterine growth charts for uncomplicated male (left, blue curves) and female (right, red curves) newborns on weight (top), length (middle), and head circumference (bottom) were plotted on the newly developed intrauterine growth charts for newborns admitted to the NICU (black curves) for comparison. (C) Top panel: the percentages of infants from the validation dataset below the 10th-percentile curves of the 2010 Olsen charts (2010 Olsen–Validation), the 2013 Fenton charts (2013 Fenton–Validation), the 2024 Chou charts for NICU infants (2024 Chou-NICU–Validation), and the new weight charts developed in this study (2024 Chou-nursery–Validation) were plotted against gestational week for comparison. The percentages of infants from the training dataset below the 10th-percentile curves of the new charts (2024 Chou-nursery–Training) were also calculated and plotted for comparison. Bottom panel: the percentages of infants from the validation dataset above the 90th-percentile curves of the 2010 Olsen charts (2010 Olsen–Validation), the 2013 Fenton charts (2013 Fenton–Validation), the 2024 Chou charts for NICU infants (2024 Chou-NICU–Validation), and the new weight charts developed in this study (2024 Chou-nursery–Validation) were plotted against gestational week for comparison. The percentages of infants from the training dataset above the 90th-percentile curves of the new charts (2024 Chou-nursery–Training) were also calculated and plotted for comparison. The error bars represent 99.3% confidence intervals (based on the adjusted α of 0.007 per group for 7 gestational week groups between 35 and 41 weeks) calculated by bootstrapping of samples with replacement 500 times. (D) The number (percentage) of infants with hypoglycemia, requiring transfer, or both who were also small or large for gestational age based on the indicated growth charts.