Abstract
Objectives:
We evaluated the prevalence of high weight status in children ages 0–24 months (m) using data from electronic health records (EHR) and NHANES. We also examined relationships between weight status during infancy and obesity at 24m of age.
Methods:
EHR data from 4 institutions in North and South Carolina included patients born 1/1/2013–10/10/2017 (N=147,290). NHANES data included study waves from 1999–2018 (unweighted N=5,121). We calculated weight-for-length (WFL), weight-for-age (WFA), and body mass index (BMI), excluding implausible values, and categorized weight status (<85th, 85th to <95th, or ≥95th percentile), assessing prevalence at birth, 6m, 12m, 18m, and 24m. Utilizing individual, longitudinal EHR data, we used separate regression models to assess obesity risk at 24m based on anthropometrics at birth, 6m, 12m, and 18m, adjusting for sex, race/ethnicity, insurance, and health system.
Results:
Prevalence of BMI≥95th percentile in EHR data at 6m, 12m, 18m, and 24m were 9.7%, 15.7%, 19.6%, and 20.5%, respectively. With NHANES the prevalence was 11.6%, 15.0%, 16.0%, and 8.4%. For both, the prevalence of high weight status was higher in Hispanic children. In EHR data, high weight status at 6, 12, and 18 months was associated with obesity at 24m, with stronger associations as BMI category increased and as age increased.
Conclusions:
High weight status is common in infants and young children, although lower at 24m in NHANES than EHR data. In EHR data, high BMI at 6m, 12m, and 18m was associated with increased risk of obesity at 24m.
Keywords: infancy, obesity, children, EHR data
INTRODUCTION
Obesity prevention during infancy and early childhood is a public health priority.1 2 However, pediatric providers have no guidelines for classifying infants or children less than 2 years of age as having obesity3 and no consistent understanding of which infants or young children at highest risk of developing obesity and obesity-related comorbidities. Moreover, nationally-representative data sources that measure children’s weight and length directly, such as the National Health and Nutrition Examination Survey (NHANES), have typically pooled all children under 2 years of age into the same category when reporting results and do not examine differences by age between 6 and 24 months.4–6
There remains controversy in the literature regarding which growth measures during infancy and early childhood are most appropriate for predicting obesity in later childhood and adulthood. Gold standard measures of body composition are impractical for clinical use.2,7 Despite weight-for-length being recommended for clinical use in the first two years of life,8 recent studies suggest body mass index (BMI) is a better correlate of body composition and may be a more appropriate predictor of subsequent obesity risk.8,9 Furthermore, there is a paucity of both clinical and research data that includes infant anthropometrics and provides appropriate racial, ethnic, and socioeconomic representation of children in the US in the 21st century. To address these gaps, we 1) evaluated prevalence of “high” (85th to <95th percentile and ≥95th percentile) weight and adiposity measures (BMI, weight-for-length [WFL], and weight-for-age [WFA]) in infants and young children from electronic health record (EHR) data associated with four academic medical centers in North and South Carolina and from NHANES, a nationally representative sample of children in the United States; and 2) examined the relationship of weight status during infancy with obesity at 2 years of age among young children in the Carolinas.
METHODS
Data
The Carolinas Collaborative is a dual state regional clinical data research network built on a learning health system model. This federated data network includes electronic health record (EHR) data from Health Sciences South Carolina, Duke Health, University of North Carolina at Chapel Hill, and Wake Forest Baptist Health. Health Sciences South Carolina consists of 7 large health systems and 3 research-intensive universities in South Carolina. The North Carolina institutions represent 3 of the 4 large academic medical centers in North Carolina and are health systems with large pediatric primary care populations. Data are maintained in the domains of demographics, encounters, diagnoses, procedures, laboratory tests, medications, and vital signs and are based on the PCORnet common data model.10 From patients born between 1/1/2013 and 10/10/2017 we retrospectively extracted weight, height/length, age in days at encounter, ethnicity, race, sex, and payor/insurance status from the Carolinas Collaborative Data Model EHR data. This time frame was chosen as it represents the time that EHR data was consistently available from those institutions at the time of the data request in October 2017. The data request was fulfilled between April 2018 and January 2020, depending on site. This study was approved by the Institutional Review Boards at Wake Forest School of Medicine and Duke University School of Medicine.
We also evaluated data from repeated cross-sections of NHANES, a study of the non-institutionalized U.S. population that consists of a questionnaire, physical examination, and laboratory analysis. Participants under 2 years in the NHANES study had recumbent length and weight measured using a standard protocol of a digital scale and an infantometer.11 We included children ages 0–26 months who were measured in NHANES and had a recorded length and weight between 1999 and 2018 (unweighted N=5,121), the most recent available data.
For both data sources we examined data from children at 6, 12, 18, and 24 months; data from EHR age was available in days and NHANES records age in whole months. We considered visits to be at these time points if they fell within the following periods: 6 months (150–240 days [EHR data] or 5–8 months [NHANES data]), 12 months (305–425 days or 10–14 months), 18 months (480–600 days or 16–20 months), and 24 months (670–790 days or 22–26 months) of age. In the case of multiple visits during each of the 4 time periods, the visit closest to the target date was used. For example, if an infant had weight and length recorded at 160 days and 175 days, only the weight and length at 175 days was included. Additionally, for the EHR data, we examined data from the newborn visit (0–7 days). Newborns were not included in the NHANES data because the number of infants who were enrolled and completed anthropometrics by 7 days of life was very small (unweighted N=109). Race and ethnicity were categorized as non-Hispanic white, non-Hispanic Black, Hispanic, non-Hispanic Other, non-Hispanic Unknown race, or Unknown race and ethnicity. These categories were chosen due to the data available in NHANES and in the EHR. Insurance status was categorized as public/government insurance (referred to as Medicaid throughout manuscript) or other (including private/commercial, other, and unknown).
Statistical Analyses
We used extracted data to calculate BMI (kg/m2) for both data sets, and used the World Health Organization growth charts for age and sex to generate BMI for age and sex, WFA, and WFL percentiles.12 For NHANES data, length, weight, BMI, WFL, and WFA were considered to be biologically implausible values if they were >5 standard deviations (SD) from the means at each age. At each time point, ≤ 5 values were excluded for biologically implausible values. For EHR data, newborns were excluded if the weight was <1.5 kg. The growthcleanr module was then utilized, which excluded 2,381 participants (549,363 total observations) for length and weight values that were deemed to be duplicate, carried forward, a minimum length change, or extreme. This cleaning method was developed to evaluate childhood growth and has been used previously to improve the quality and inferences from electronic health record data.13,14 Subsequently, observations were excluded if there was not a length and weight available or if observations had a calculated z-score < −5 or > 5. The total sample size for EHR data included 147,290 unique patients and 285,771 total observations.
We used cross-tabulations to assess the prevalence of participants with a weight status of 85th to <95th percentile and ≥95th percentile using BMI, WFL, and WFA at each time point, and for NHANES, we utilized a second order Rao and Scott correction to account for the complex survey design. Additionally, for the EHR data, we used separate modified Poisson regression models to assess the risk of obesity at 2 years based on weight status (85th to <95th percentile and ≥95th percentile by BMI, WFA, and WFL) at birth, 6, 12, and 18 months, adjusting for sex, race/ethnicity, insurance status, and site. Poisson regression allowed us to calculate risk ratios (RR) with corresponding 95% confidence intervals (CI), which we present for each model. Analyses to compare longitudinal across time were not possible for NHANES due to its repeated cross-sectional design. All analyses were performed in R 4.0.0 (R Core Team, 2020).
RESULTS
Sample Characteristics
Patients with data extracted from the EHR were 53.5% male, 45.8% non-Hispanic white, 23.6% non-Hispanic Black, and 13.5% Hispanic. Less than half of children (34.9%) were insured by Medicaid (Table 1).
Table 1:
Demographics of Subjects Included in Analyses
| Carolinas Collaborative N=147,290 N (%) | NHANES Unweighted N=5,121 N (Weighted %) | |
|---|---|---|
| Male Sex | 78,772 (53.5%) | 2,610 (51.6%) |
| Race/Ethnicity | ||
| Non-Hispanic white | 67,470 (45.8%) | 1,621 (53.5%) |
| Non-Hispanic Black | 34,769 (23.6%) | 1,057 (12.7%) |
| Hispanic | 19,825 (13.5%) | 1,998 (25.2%) |
| Non-Hispanic Other | 6,748 (4.6%) | 445 (8.5%) |
| Non-Hispanic Unknown race | 11,102 (7.5%) | - |
| Unknown (both race and ethnicity) | 7,376 (5.0%) | - |
| Medicaid | 51,419 (34.9%) | 2,703 (43.8%) |
Demographics of NHANES participants appropriately reflect the population of the United States with subjects being 51.6% male, 53.5% non-Hispanic white, 12.7% non-Hispanic Black, and 25.2% Hispanic. A larger proportion of children included in NHANES received Medicaid insurance than children in the EHR sample (43.8%) (Table 1, weighted percentages).
Proportion of participants with high weight status
Proportions of WFL, WFA, and BMI in the ranges of 85th to <95th percentile and ≥95th percentile are presented in Table 2. In general, proportions of children with weight status in the 85th to <95th percentile and 95th percentile increased with increasing age of the child, with the exception of 24 months in NHANES data. Proportions of children with a BMI ≥95th percentile in the EHR data were 2.4% at birth and at 6, 12, 18, and 24 months were 9.7%, 15.7%, 19.6%, and 20.5%, respectively. In NHANES, proportions of BMI≥95th percentile were 11.6%, 15.0%, 16.0%, and 8.4% at 6, 12, 18, and 24 months, respectively (Table 2). There were small differences in proportions of high weight status as defined differently by WFL, WFA, and BMI; for example, fewer children were in the ≥95th percentile when defined by WFA than when defined by WFL or BMI (Table 2).
Table 2:
Prevalence of participants with weight status of 85th to <95th percentile and ≥95th percentile at various time points
| 85th to <95th percentile, Carolinas Collaborative (%) | 85th to <95th percentile, NHANES (%) | ≥95th percentile, Carolinas Collaborative (%) | ≥95th percentile, NHANES (%) | |
|---|---|---|---|---|
| Weight-For-Length (WFL) | ||||
| Newborn | 5.2 | - | 3.4 | - |
| 6 Months | 12.6 | 14.8 | 11.5 | 11.3 |
| 12 Months | 16.1 | 16.1 | 14.7 | 10.2 |
| 18 Months | 17.2 | 14.0 | 17.3 | 19.1 |
| 24 Months | 17.4 | 12.6 | 17.9 | 8.3 |
| Weight-For-Age (WFA) | ||||
| Newborn | 6.0 | - | 2.7 | - |
| 6 Months | 10.2 | 15.9 | 6.2 | 9.9 |
| 12 Months | 14.0 | 10.0 | 9.4 | 5.0 |
| 18 Months | 15.4 | 8.9 | 11.5 | 4.9 |
| 24 Months | 15.1 | 11.5 | 11.5 | 7.4 |
| Body Mass Index (BMI) | ||||
| Newborn | 4.8 | - | 2.4 | - |
| 6 Months | 11.2 | 13.1 | 9.7 | 11.6 |
| 12 Months | 16.1 | 17.7 | 15.7 | 15.0 |
| 18 Months | 17.5 | 20.3 | 19.6 | 16.0 |
| 24 Months | 18.2 | 15.0 | 20.5 | 8.4 |
Proportions of children with weight status in the 85th to <95th percentile are presented in Table 3 by race/ethnicity. Proportions of children with BMI ≥95th percentile tended to be higher among Hispanic children compared to non-Hispanic white and non-Hispanic Black infants (Table 4). Among 24-month-olds in the EHR data, 27.3% of Hispanic children had a BMI ≥95th percentile, compared to 20.3% of non-Hispanic Black children and 19.5% of non-Hispanic white children. Among 24-month-olds in NHANES, 12.9% of Hispanic children had a BMI ≥95th percentile, compared to 9.5% of non-Hispanic Black children and 6.0% of non-Hispanic white children.
Table 3:
Prevalence of participants with weight status of 85th to <95th percentile by race/ethnicity
| Hispanic Carolinas Collaborative (%) | Hispanic NHANES (%) | Non-Hispanic White Carolinas Collaborative (%) | Non-Hispanic White NHANES (%) | Non-Hispanic Black Carolinas Collaborative (%) | Non-Hispanic Black NHANES (%) | |
|---|---|---|---|---|---|---|
| Weight-For-Length (WFL) | ||||||
| Newborn | 7.5 | - | 4.3 | - | 5.9 | - |
| 6 Months | 15.5 | 19.5 | 11.9 | 12.0 | 12.5 | 18.8 |
| 12 Months | 18.3 | 14.9 | 15.9 | 17.4 | 16.0 | 18.7 |
| 18 Months | 19.0 | 13.3 | 17.5 | 13.4 | 16.6 | 16.1 |
| 24 Months | 17.9 | 15.1 | 17.5 | 12.2 | 16.9 | 10.7 |
| Weight-For-Age (WFA) | ||||||
| Newborn | 6.5 | - | 7.1 | - | 4.3 | - |
| 6 Months | 12.6 | 17.5 | 9.7 | 15.9 | 10.0 | 15.7 |
| 12 Months | 15.3 | 9.7 | 14.3 | 9.3 | 13.5 | 12.2 |
| 18 Months | 16.2 | 11.1 | 15.5 | 9.4 | 15.1 | 5.8 |
| 24 Months | 14.9 | 11.9 | 15.5 | 10.7 | 15.0 | 13.8 |
| Body Mass Index (BMI) | ||||||
| Newborn | 7.0 | - | 4.4 | - | 4.4 | - |
| 6 Months | 14.3 | 16.9 | 10.4 | 10.9 | 11.1 | 16.0 |
| 12 Months | 18.3 | 16.6 | 16.0 | 18.2 | 15.7 | 21.7 |
| 18 Months | 20.0 | 19.8 | 17.6 | 20.6 | 16.7 | 20.9 |
| 24 Months | 19.8 | 15.4 | 18.3 | 16.2 | 17.2 | 13.4 |
Table 4:
Prevalence of participants with weight status of ≥95th percentile by race/ethnicity
| Hispanic Carolinas Collaborative (%) | Hispanic NHANES (%) | Non-Hispanic White Carolinas Collaborative (%) | Non-Hispanic White NHANES (%) | Non-Hispanic Black Carolinas Collaborative (%) | Non-Hispanic Black NHANES (%) | |
|---|---|---|---|---|---|---|
| Weight-For-Length (WFL) | ||||||
| Newborn | 5.1 | - | 2.6 | - | 3.9 | - |
| 6 Months | 15.0 | 12.1 | 9.9 | 10.0 | 12.9 | 16.3 |
| 12 Months | 18.2 | 13.8 | 13.8 | 19.8 | 15.4 | 9.2 |
| 18 Months | 22.4 | 11.9 | 16.1 | 18.5 | 17.6 | 7.0 |
| 24 Months | 24.3 | 11.7 | 16.9 | 6.7 | 17.5 | 8.8 |
| Weight-For-Age (WFA) | ||||||
| Newborn | 3.0 | - | 3.2 | - | 1.7 | - |
| 6 Months | 7.9 | 9.9 | 5.4 | 9.3 | 6.3 | 13.2 |
| 12 Months | 10.8 | 6.1 | 8.7 | 4.8 | 10.2 | 6.3 |
| 18 Months | 13.7 | 15.6 | 10.9 | 3.7 | 11.9 | 6.2 |
| 24 Months | 15.1 | 9.0 | 10.5 | 6.4 | 11.7 | 9.9 |
| Body Mass Index (BMI) | ||||||
| Newborn | 3.7 | - | 2.0 | - | 2.5 | - |
| 6 Months | 12.8 | 12.9 | 8.4 | 10.5 | 10.5 | 15.7 |
| 12 Months | 19.5 | 18.3 | 14.6 | 15.3 | 16.6 | 13.9 |
| 18 Months | 25.0 | 19.4 | 18.4 | 14.5 | 19.7 | 14.4 |
| 24 Months | 27.3 | 12.9 | 19.5 | 6.0 | 20.3 | 9.5 |
Risk of obesity at 2 years of age with EHR data
The risk of obesity at 2 years of age was higher with increasing age for infants with both 85th to <95th percentile and ≥95th percentile compared to other (<85th percentile) using WFL, WFA, and BMI (Table 5). Having a BMI between the 85th and <95th percentiles at birth, 6 months, 12 months, and 18 months was associated with significant increased risk of obesity at 2 years of age compared to those with BMI < 85th percentile (RR 1.60, 2.64, 3.45, and 4.86, respectively). Having a BMI ≥95th percentile at birth, 6 months, 12 months, and 18 months was associated with significant increased risk of obesity at 2 years of age compared to those with BMI < 85th percentile (RR 1.88, 3.82, 6.26, and 11.27, respectively). At almost all time points the risk of obesity at 2 years of age was higher when using WFL or BMI cut points of high weight status (85th to <95th percentile and ≥95th percentile) compared to WFA (Table 5).
Table 5:
Risk of obesity (BMI ≥95th percentile) at 2 years of age based on presence of weight status of 85th to <95th percentile and ≥95th percentile compared to other (<85th percentile) at each time point controlling for sex, race/ethnicity, insurance status, and site
| 85th to <95th percentile, Carolinas Collaborative RR (95% CI) | ≥95th percentile, Carolinas Collaborative RR (95% CI) | |
|---|---|---|
| Weight-For-Length (WFL) | ||
| Newborn | 1.39 (1.24, 1.55) | 1.68 (1.50, 1.90) |
| 6 Months | 2.62 (2.44, 2.81) | 3.74 (3.51, 3.98) |
| 12 Months | 3.54 (3.30, 3.79) | 6.45 (6.08, 6.85) |
| 18 Months | 4.91 (4.53, 5.33) | 11.25 (10.50, 12.06) |
| Weight-For-Age (WFA) | ||
| Newborn | 1.50 (1.34, 1.68) | 1.80 (1.57, 2.07) |
| 6 Months | 2.59 (2.43, 2.78) | 3.47 (3.24, 3.71) |
| 12 Months | 3.07 (2.89, 3.27) | 4.86 (4.59, 5.14) |
| 18 Months | 3.54 (3.32, 3.78) | 6.45 (6.10, 6.81) |
| Body Mass Index (BMI) | ||
| Newborn | 1.60 (1.44, 1.78) | 1.88 (1.65, 2.13) |
| 6 Months | 2.64 (2.46, 2.83) | 3.82 (3.59, 4.07) |
| 12 Months | 3.45 (3.21, 3.70) | 6.26 (5.89, 6.64) |
| 18 Months | 4.86 (4.46, 5.29) | 11.27 (10.48, 12.13) |
DISCUSSION
High weight status is common in children <2 years of age in NHANES and in EHR data of children in the Carolinas, although the prevalence was lower at 24m in NHANES. Proportions of children with weight and adiposity measures of 85th to <95th percentile and ≥95th percentile generally increased with age and tended to be higher in Hispanic children in both datasets. There were small differences in proportions of high weight status as defined differently by WFL, WFA, and BMI; for example, fewer children had weight status ≥95th percentile by WFA than by WFL or BMI. For children in the Carolinas, the risk of weight status of 85th to <95th percentile and ≥95th percentile were higher with increasing age using WFL, WFA, and BMI. At 6, 12, and 18 months, the risk of developing obesity at 2 years of age was higher when using a cutpoint of ≥95th percentile compared to of 85th to <95th percentile and when using WFL or BMI compared to WFA.
The similarities we observed in some areas of comparison of weight status of 85th to <95th percentile and ≥95th percentile between EHR data of children in the Carolinas and the nationally representative sample of NHANES may suggest that pooled EHR data across institutions can be used to make population-level assessments of infants and children’s growth. EHR epidemiology has been proposed as an important component of population health research, as the benefits of large sample sizes and generalizable patient populations are unique to EHR research,15 and is increasingly being used for obesity surveillance.13,16 Importantly, EHR data has the ability to include populations traditionally underrepresented in longitudinal research studies but at highest risk for obesity.17 Additionally, studies utilizing EHR are able to leverage clinical data and are less expensive and faster to perform than other study types, such as randomized controlled trials or prospective cohort studies that have previously been used to obtain similar data.
There were some important differences noted between NHANES and the EHR data of children in the Carolinas. The prevalence of a weight status ≥95th percentile generally increased between 6 months and 24 months, with the highest prevalence noted at 24 months using WFL, WFA, and BMI in the EHR data. In NHANES, the prevalence at 18 and 24 months was lower than in the EHR data. While comparing national level data to state level data has limitations, to our knowledge, there is no North Carolina-wide assessment of childhood weight/length for children under the age of 2. Many surveys that allow for state-level estimates, such as the National Survey of Children’s Health, rely on caregiver-reported height and weight. For children enrolled in WIC in 2016, 13.9% of US WIC participants aged 2 to 4 years had obesity, compared to a similar proportion in NC at 14.2%.18 In older children ages 10–17 participating in The National Survey of Children’s Health in 2018, children in North Carolina have a similar obesity prevalence (16.1%) to the US as a whole (15.5%).19 The differences in prevalence of high weight status in children 18–24 months may be due to true differences in prevalence between the Carolinas and the U.S. overall, differences in the measurement of clinical compared to survey data, non-random sampling of children represented in the Carolinas EHR data, or a combination of causes.
The proportion of participants with weight status ≥95th percentile was higher among children with a Hispanic ethnicity in EHR data from the Carolinas compared to NHANES, and at 24 months the prevalence of weight status ≥95th percentile was higher among children of all ethnicities in EHR data from the Carolinas. While the populations were generally similar between the two data sets, there was a higher proportion of Medicaid recipients and children of Hispanic ethnicity in NHANES, whereas the EHR data had a higher proportion of non-Hispanic Black children. In the EHR data, 35% of children had Medicaid insurance, which is similar to statewide proportions of 40%.20 In the NHANES data, 44% of children had Medicaid compared to national proportions of 39.1%.21 While we would have expected NHANES to have a lower percentage of Medicaid recipients as coverage has generally increased over time since 1999,21 NHANES does oversample Hispanic and non-Hispanic Black participants,11 who generally have higher proportions of Medicaid coverage.21 While children in the EHR data have higher proportions of a weight status ≥95th percentile, this is most pronounced for non-Hispanic white children, and is reflected in the similar overall rates of obesity at age 2 among non-Hispanic Black and non-Hispanic white children.
In recent years, advances in medical informatics have greatly improved our ability to identify implausible values in growth data from EHRs and therefore improve the quality of the data being analyzed for research purposes. The algorithm (“growthcleanr”) used in our study, developed by Daymont and colleagues, identifies implausible values by comparing length/weight values to previous and future measurements and excludes values that were recorded repeatedly without re-measurement.13 A similar methodology recently utilized by the ADVANCE Clinical Data Research Network found that using longitudinal (vs population-based) outlier methods can reduce the underestimation of anthropometric changes in children’s growth data.16 While these methodologies significantly improve the accuracy of using clinical data for research purposes, they cannot address all data quality issues. For example, if measurements are taken in a consistently biased way, algorithms will not detect this and the results will be biased. It remains important for clinical data to come from pediatric providers with training in obtaining accurate measurements in children of all ages. Future research should utilize EHR data and longitudinal outlier methods to evaluate changes in high weight status prevalence in infancy over time, growth trajectory differences by geography or patient demographics, and changes in child growth after local policy changes (e.g., changes to SNAP/WIC nutritional packages22) or significant events (e.g., schools closing for the COVID-19 pandemic).
We found that the risk of obesity at 2 years of age was very similar when using WFL or BMI cut points. A previous prospective study of two birth cohorts in the US found that defining overweight using WFL and BMI in the first two years of life had similar associations with cardiometabolic outcomes during early adolescence.23 Another study of a pre-birth cohort in Colorado found that WFA, WFL, and BMI were all poor surrogates for fat mass at the newborn period. By 5 months of age, however, WFL and BMI were both strongly associated with adiposity, however, change in BMI z-score was the best predictor of fat accrual throughout the first 5 months of life.24 A separate study of healthy term-born infants from two cohorts found that BMI was a better predictor of adiposity at 1 month of age compared to WFL.8 EHR data of infants from a large pediatric network found that a high BMI at 2 months of age (both using cut-points of the 85th percentile and the 97.7th percentile) had a higher positive predictive value for obesity at 2 years compared to WFL.25
Numerous other studies have shown that obesity and rapid infant weight gain during infancy are highly predictive of obesity later in childhood.26,27 Therefore, it has been suggested that WHO BMI-for-age percentile curves should be used to monitor weight gain in infants and toddlers.9 The NHANES data in our study demonstrated that the prevalence of BMI and WFL >85th and >95th percentile are lower at 24 months than they are at 12 and 18 months. Other NHANES data shows that a higher proportion of children ages 2–5 have a BMI>95th percentile than infants aged 0–23 mo.4 Using cut-points of the 97.7th percentile and 85th percentile of BMI for obesity and at risk for developing overweight or obesity, respectively, have been suggested as possible cut points.9 However, given the high risk of obesity at 2 years of age using cut-points of the 85th percentile throughout infancy in our study, and the use of the 85th and 95th percentile of BMI to define overweight and obesity at 2 years of age and thereafter, we propose using the cut-points of the 85th percentile and 95th percentile of BMI for overweight and obesity during infancy. Further research is needed for clinicians to understand how best to intervene on obesity during infancy to promote optimal growth in childhood, balancing the potentially competing risks of persistent obesity during childhood/adulthood and restricting intake during critical periods of growth and development in young children. We recommend all pediatric providers review the WHO BMI growth charts throughout infancy and discuss the child’s growth with their family.
Our study was strengthened by the use of both a nationally representative sample and a large and racially diverse EHR sample including 147,290 unique patients and 285,771 total observations and also by our use of longitudinal outlier methods to clean the EHR data. However, our study does have limitations. Consistent follow-up information was only available on participants through the first 2 years of life in this EHR dataset, due to the relatively recent adoption of EHR implementation at these institutions and technical complexities of merging legacy, often homegrown, electronic health record infrastructure. Future studies should continue to follow these children over time to assess long-term obesity risk associated with overweight and obesity during infancy. Additionally, we were limited by data that was available in this federated common data model, which was unlikely to be designed with childhood growth measures and outcomes as a priority. Unlike prospective research studies, EHR data does not yet consistently include other factors known to influence infant and early childhood growth, such as household income, social determinants of health, rurality, maternal data, parental BMI, feeding characteristics, and body composition data, all of which should be assessed in future prospective and retrospective studies, especially as more EHRs are beginning to link maternal and child charts and more information about feeding characteristics during infancy. Future research should also explore additional analytic techniques to compare prevalence estimates between population-based surveys and electronic health records.28 Additionally, EHR relies on lengths and weights obtained at clinical visits, which unlike the high-quality prospective measurements of NHANES, may be less accurate and reliable.11 Finally, NHANES uses only recumbent length for children under age 2, while we are unable to assess specific measurement procedures in EHR data. Although this may have introduced error at the 24-month time frame, this is likely non-differential misclassification.
CONCLUSION
In conclusion, high weight status is common in children <2 years of age in NHANES and in EHR data of children in the Carolinas, although the prevalence was lower at 24m in NHANES; in both, prevalence of high weight status (85th to <95th percentile and ≥95th percentile) increased with age and were higher in Hispanic children. For children in the Carolinas, the risk of high weight status was higher with increasing age, while at almost all time points the prevalence of obesity at 2 years of age was higher when using WFL or BMI cut points compared to WFA. EHR data represent a practical and effective way of making population level estimates of obesity during infancy and early childhood, especially when combined with longitudinal outlier methods.
What’s New.
High weight status is common in children <2 in NHANES and in EHR data from the Carolinas, although lower at 24m in NHANES. In EHR data, high weight status at 6, 12, and 18m was associated with increased risk of obesity at 24m.
Acknowledgements
This study was supported in part by the Carolinas Collaborative, a regional clinical data research network funded by the Duke Endowment and managed by Health Sciences South Carolina. Dr. Brown was supported by a grant from the National Institute for Child Health (grant 1K23HD099249).
The authors gratefully acknowledge the data extraction services of the Wake Forest Clinical and Translational Science Institute, which is supported by the National Center for Advancing Translational Sciences (NCATS), National Institutes of Health, through Grant Award Number UL1TR001420. The Duke Biostatistics, Epidemiology, and Research Design Methods Core’s support of this project was made possible in part by CTSA Grant (UL1TR002553) from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH), and the NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of NCATS or NIH.
Funding:
This study was supported in part by the Carolinas Collaborative, a regional clinical data research network funded by the Duke Endowment and managed by Health Sciences South Carolina. Dr. Brown was supported by a grant from the National Institute of Child Health and Human Development (grant 1K23HD099249). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of Interest: The authors have no conflicts of interest to report.
References:
- 1.Daniels SR, Hassink SG, Abrams SA, et al. The role of the pediatrician in primary prevention of obesity. Pediatrics. 2015;136(1):e275–e292. [DOI] [PubMed] [Google Scholar]
- 2.Lumeng JC, Taveras EM, Birch L, Yanovski SZ. Prevention of obesity in infancy and early childhood: a National Institutes of Health workshop. JAMA pediatrics. 2015;169(5):484–490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Krebs NF, Himes JH, Jacobson D, Nicklas TA, Guilday P, Styne D. Assessment of child and adolescent overweight and obesity. Pediatrics. 2007;120(Supplement 4):S193–S228. [DOI] [PubMed] [Google Scholar]
- 4.Ogden CL, Fryar CD, Martin CB, et al. Trends in obesity prevalence by race and hispanic origin—1999–2000 to 2017–2018. Jama. 2020;324(12):1208–1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Fryar C, Carrol M, Ogden C. Prevalence of high weight-for-recumbent length among infants and toddlers from birth to 24 months of age: United States, 1971–1974 through 2015‐2016. Health E-Stats Hyattsville, MD: National Center for Health Statistics. 2018. [Google Scholar]
- 6.Akinbami LJ, Kit BK, Carroll MD, Fakhouri THI, Ogden CL. Trends in Anthropometric Measures Among US Children 6 to 23 Months, 1976–2014. Pediatrics. 2017;139(3). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Demerath EW, Fields DA. Body composition assessment in the infant. American Journal of Human Biology. 2014;26(3):291–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Roy SM, Fields DA, Mitchell JA, et al. Body mass index is a better indicator of body composition than weight-for-length at age 1 month. The Journal of pediatrics. 2019;204:77–83. e71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Woo JG, Daniels SR. Assessment of Body Mass Index in Infancy: It Is Time to Revise Our Guidelines. J Pediatr. 2019;204:10–11. [DOI] [PubMed] [Google Scholar]
- 10.Corley DA, Feigelson HS, Lieu TA, McGlynn EA. Building data infrastructure to evaluate and improve quality: PCORnet. Journal of oncology practice. 2015;11(3):204–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey (NHANES) Anthropometry Procedures Manual. US Department of Health and Human Services, Centers for Disease Control and Prevention. 2013. [Google Scholar]
- 12.De Onis M, World Health Organization. WHO child growth standards: length/height-for age, weight-for-age, weight-for-length, weight-for-height and body mass index-for-age: methods and development. 2006. [Google Scholar]
- 13.Daymont C, Ross ME, Russell Localio A, Fiks AG, Wasserman RC, Grundmeier RW. Automated identification of implausible values in growth data from pediatric electronic health records. J Am Med Inform Assoc. 2017;24(6):1080–1087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Freedman DS, Goodman AB, King RJ, Blanck HM. Tracking of obesity among 2‐to 9‐year‐olds in an electronic heath record database from 2006 to 2018. Obesity Science & Practice. 2020;6(3):300–306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Casey JA, Schwartz BS, Stewart WF, Adler NE. Using Electronic Health Records for Population Health Research: A Review of Methods and Applications. Annu Rev Public Health. 2016;37:61–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Boone-Heinonen J, Tillotson CJ, O’Malley JP, et al. Not so implausible: impact of longitudinal assessment of implausible anthropometric measures on obesity prevalence and weight change in children and adolescents. Ann Epidemiol. 2019;31:69–74 e65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Boone-Heinonen J, Tillotson C, O’Malley J, et al. Characterizing a “big data” cohort of over 200,000 low-income US infants and children for obesity research: The ADVANCE Early Life Cohort. Maternal and child health journal. 2017;21(3):421–431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Center for Chronic Disease Prevention and Health Promotion. Obesity Among WIC-Enrolled Young Children. https://www.cdc.gov/obesity/data/obesity-among-WIC-enrolled-young-children.html. Updated November 22, 2019. Accessed January 22, 2021.
- 19.State of Childhood Obesity. Obesity Rates for Youth Ages 10 to 17. https://stateofchildhoodobesity.org/children1017/. Updated October 2020. Accessed 01/25/2021.
- 20.Georgetown University Center for Children and Families AAoP. Snapshot of Children’s Coverage: How Medicaid, CHIP, and the ACA Cover Children. https://ccf.georgetown.edu/wp-content/uploads/2017/02/North-Carolina-Medicaid-CHIP-new-v1.pdf. Published 2016. Accessed 01/28/2021.
- 21.Kaiser Family Foundation. Health Insurance Coverage of Children 0–18. https://www.kff.org/other/state-indicator/children-0-18/?currentTimeframe=11&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D. Accessed 01/28/2021. [Google Scholar]
- 22.Chaparro MP, Anderson CE, Crespi CM, Whaley SE, Wang MC. The effect of the 2009 WIC food package change on childhood obesity varies by gender and initial weight status in Los Angeles County. Pediatric obesity. 2019;14(9):e12526. [DOI] [PubMed] [Google Scholar]
- 23.Aris IM, Rifas-Shiman SL, Li L-J, et al. Association of Weight for Length vs Body Mass Index During the First 2 Years of Life With Cardiometabolic Risk in Early Adolescence. JAMA Network Open. 2018;1(5):e182460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Perng W, Ringham BM, Glueck DH, et al. An observational cohort study of weight-and length-derived anthropometric indicators with body composition at birth and 5 mo: the Healthy Start study. The American journal of clinical nutrition. 2017;106(2):559–567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Roy SM, Spivack JG, Faith MS, et al. Infant BMI or Weight-for-Length and Obesity Risk in Early Childhood. Pediatrics. 2016;137(5). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Jabakhanji SB, Boland F, Ward M, Biesma R. Body Mass Index Changes in Early Childhood. J Pediatr. 2018;202:106–114. [DOI] [PubMed] [Google Scholar]
- 27.Taveras EM, Rifas-Shiman SL, Belfort MB, Kleinman KP, Oken E, Gillman MW. Weight status in the first 6 months of life and obesity at 3 years of age. Pediatrics. 2009;123(4):1177–1183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Tatem KS, Romo ML, McVeigh KH, et al. Comparing prevalence estimates from population-based surveys to inform surveillance using electronic health records. Preventing chronic disease. 2017;14:E44. [DOI] [PMC free article] [PubMed] [Google Scholar]