Trends in State-Specific Prevalence of Overweight and Underweight in 2- Through 4-Year-Old Children From Low-Income Families From 1989 Through 2000

Objectives  To document overweight and underweight state-specific prevalence and examine trends among 2- through 4-year-old children from low-income families.

Methods  State-specific and overall overweight and underweight prevalence for 1989, 1994, and 2000 and trend analyses during the study period are documented. Overweight was defined as a sex-specific body mass index (BMI) for age in the 95th percentile or higher and underweight as a sex-specific BMI for age in less than the fifth percentile on the 2000 Centers for Disease Control and Prevention (CDC) growth charts. These analyses are based on one randomly selected record per child per year for 30 states consistently participating in the CDC Pediatric Nutrition Surveillance System in 1989, 1994, and 2000. Prevalence in 1989 and 1994 is adjusted to state-specific age and race/ethnicity distribution of the population in 2000. Overweight and underweight prevalence were categorized as 5% or less, more than 5% to 10%, more than 10% to 15%, more than 15% to 20%, and more than 20%.

Results  The number of states that reported overweight prevalence of more than 10% increased from 11 in 1989 to 28 in 2000. Underweight decreased during the study period: 9 states in 1989 and 23 states in 2000 had a prevalence of 5% or less. No geographic predominance was apparent. Trend analyses showed significant increases in overweight in 30 states (P < .01) and decreases in underweight in 26 states (P < .05).

Conclusions  Overweight is increasing and underweight is decreasing in our study population. We need to expand prevention and intervention efforts to reverse the rising trend of overweight in the United States.

Overweight is one of the most prevalent nutritional problems that affects US children, and its magnitude makes it a major public health concern. Overweight children are at risk for serious health consequences during childhood^1,2: increased risk for adverse lipid, insulin, and blood pressure levels³; type 2 diabetes mellitus⁴; hepatic steatosis⁵; cholelithiasis⁶; and disorders such as sleep apnea.⁷

Overweight among children increases with advancing age,^8,9 increasing morbidity in adulthood.^1,2 Adults who were overweight adolescents are at increased risk for coronary heart disease and atherosclerosis compared with those who were not overweight adolescents.¹⁰ Men who were overweight adolescents also have an increased risk of mortality from coronary heart disease, atherosclerotic cerebrovascular disease, and colorectal cancer.¹⁰ Being overweight during adolescence and young adulthood is associated with social and economic outcomes, including lower educational attainment, income earning capacity, and likelihood of being married.¹¹

National data show an increasing secular trend in overweight prevalence. Among 2- to 5-year-old boys, overweight prevalence (body mass index [BMI] for sex and age in the 95th percentile or higher) increased from 6.1% in the 1988-1994 National Health and Nutrition Examination Survey (NHANES) to 9.9% in the 1999-2000 NHANES.¹² Among girls in this age group, overweight prevalence increased from 8.2% in the 1988-1994 NHANES to 11.0% in the 1999-2000 NHANES.¹² The highest overweight prevalences among 2- to 5-year-old children in the 1999-2000 survey were for Mexican American boys and non-Hispanic white girls.¹²

The Centers for Disease Control and Prevention (CDC) Pediatric Nutrition Surveillance System (PedNSS) of US children from low-income families documented an increase in overweight prevalence (>95th percentile weight for height, 1977 National Center for Health Statistics [NCHS] growth reference) among 2- through 4-year-old children. Among boys, overweight increased from 6.6% in 1983 to 8.3% in 1995; among girls, overweight increased from 7.2% in 1983 to 9.0% in 1995.¹³

A recent US Department of Agriculture report on overweight prevalence, based on 1 month of data for each state for children enrolled in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), found that 11% in 1992 and 13.2% in 1998 were overweight.¹⁴ This report provides useful national and state-specific data on low-income children not available in the peer-reviewed literature. The analyses excluded children with missing height, weight, and sex data but did not state whether children with biologically implausible anthropometric measurements, likely from measurement or recording errors, were excluded.

In contrast to the high and increasing overweight prevalence, underweight prevalence was low and declined over time. Underweight indicates undernutrition due to food shortage or a medical condition interfering with weight gain. The Pediatric Nutrition Surveillance 2001 Report¹⁵ documented a 5.4% underweight prevalence among low-income children from birth through 4 years, a steady decline from 6.9% in 1989. Although a few states may have a higher underweight prevalence, it is not a major public health problem.

Our goal was to examine changes in overweight and underweight prevalence by state among 2- through 4-year-old, low-income children participating in federally funded programs that submit data to PedNSS. To our best knowledge, these state-specific analyses are not available in the peer-reviewed literature. Two important reasons for examining overweight and underweight prevalence by state are to look for geographic differences to stimulate investigation as to why they exist. These findings can be used to target intervention programs, promote further research to understand the differences among states, and identify how states achieve low prevalence.

We used CDC PedNSS cross-sectional data for our analyses. In 1973, PedNSS began monitoring growth indicators, hematologic test results, and breastfeeding practices for low-income infants and children who participate in publicly funded maternal and child nutrition and health programs. The primary source of these data is WIC: in 2001, 82% of the national PedNSS database was derived from WIC records.¹⁵ The PedNSS is the only source of state-based surveillance data on overweight and underweight among low-income young children reportable at the state and local level.

We documented overweight and underweight prevalence and conducted a trend analysis for states that participated in PedNSS in 1989, 1994, and 2000. This combination of years allowed for the maximum number of states to be included in our 12-year trend analyses: 33 units, including 29 states, the District of Columbia, Puerto Rico, the Navajo Nation, and the Inter Tribal Council of Arizona (ITCA), consistently participated. We excluded 3 states with incomplete or questionable data for at least 1 of the years: one state had a 1725% increase in participants between 1989 and 2000, indicating the inclusion of additional agencies, WIC priority levels, or both; one state had only 14% of 1989 child visit records in 2000, indicating incomplete 2000 data; and one state had inaccurate height data for 2000 because of rounding errors. To simplify terms, we refer to each of the 30 units as states.

Height and weight (directly measured) and age and race/ethnicity data were recorded into an automated computer system during WIC clinic visits. These data were aggregated by states and submitted to the CDC for inclusion in the PedNSS database. In PedNSS, standardized protocols are used for measuring height and weight: for children 2 years or older, standing height is measured to the nearest 0.1 cm or in using a stadiometer^16,17 and weight is measured to the nearest 0.1 kg or ¼ lb using a pediatric or beam balance scale.^16,17 Age at the clinic visit was calculated as the difference in years and months between date of visit and date of birth. Race/ethnicity was reported by the parent or caretaker.

We based weight status on BMI (weight in kilograms divided by the square of height in meters) because our analyses included children who were 2 years or older, the youngest age for which BMI assessment is appropriate. We defined overweight as recommended by the Expert Committee of Pediatric Obesity convened by the Maternal and Child Health Bureau: a sex-specific BMI for age in the 95th percentile or higher.¹⁸ Underweight was defined as a sex-specific BMI for age in the less than fifth percentile.¹⁹ Percentile cutoffs for BMI were based on the 2000 CDC growth reference, which for 2- through 4-year-old children is based on the distribution of US children between 1971 and 1994 (NHANES I through NHANES III).²⁰ Although this population was not based on children exclusively without health problems, it represents children before the epidemic of obesity became apparent. With these cutoffs, we would expect 5% of children to be underweight and 5% to be overweight.²¹

We randomly selected data from 1 clinic visit per year for each 2- through 4-year-old child for each state for each year of study based on the child identifier and date of visit. Individual identifiers were stripped from the data files before transmission to the researchers, so these analyses were exempt from human subjects review at the CDC. To ensure that only biologically plausible values were included, we excluded children with BMI-for-age z scores of less than −4.0 or greater than 5.0, the same as the exclusion criteria suggested by the World Health Organization for weight-for-height z score.¹⁹ Using these cutoffs seemed reasonable, because Mei et al²² documented similar sensitivity and specificity of BMI for age and weight for height in classifying body fatness of children. Biologically implausible values were minimal (<1%) and ranged from 0.04% to 0.68%, varying by year and state. We also examined the standard deviations of the z scores to verify data quality. The z score standard deviations are expected to be close to 1.0; those greater than 1.20 are questionable; and those 1.40 or greater are suspect of measurement, calculation, or recording inaccuracies.¹⁷ The BMI-for-age z score standard deviation ranges varied by year: of 30 states, 21 in 1989, 25 in 1994, and 23 in 2000 had values less than 1.20. Data from Puerto Rico were of concern because of standard deviations of BMI-for-age z score values of 1.40 or higher for all 3 years (1.49 in 1989, 1.44 in 1994, and 1.47 in 2000). Although data from Puerto Rico could have had measurement, calculation, or recording inaccuracies, the prevalences of both overweight and underweight (>10%) were high.

We calculated the unadjusted overweight and underweight prevalence for each state and the total study population for each of the 3 study years. Next, we stratified state and total study population data by sex, single-year age group, and race/ethnicity and calculated overweight and underweight prevalence to ascertain whether changes in prevalence could be attributed to changes in demographic distribution. For the total study population, we used χ² tests for trend to identify statistically significant changes in sex, single-year age group, and race/ethnicity. For the state-specific race/ethnicity analyses, we included only white Americans, African Americans, and Hispanic Americans and limited our analyses to 24 states, because the other 6 had sample sizes of less than 100 for these categories. In most states, other racial/ethnic groups had even smaller sample sizes, making point estimates of prevalence of overweight and underweight too unstable to report. To examine the effect of differences in age and race/ethnicity composition among states on differences in state-specific prevalence, we compared the variance in overweight prevalence of the states in 2000 before and after adjustment with the age and racial/ethnic composition of overall data.

To account for observed changes in age and race/ethnicity characteristics in each state between 1989 and 2000, we adjusted overweight and underweight prevalence for 1989 and 1994 in each state to that state’s age and race/ethnicity distribution in 2000. Because sex distribution did not change by state during the reporting years, we did not adjust for it. After adjustment, we used χ² contingency table tests to document whether statistically significant differences occurred in overweight and underweight prevalence among groups within sex, age, and racial/ethnic categories for the overall population. To document whether observed changes in the prevalence of overweight and underweight were statistically significant (P < .05) in each state, we conducted state-specific χ² tests for trend, using the adjusted prevalence data for 1989 and 1994 and the unadjusted data for 2000. In addition, we calculated prevalence change per year ±95% confidence intervals during the study periods. State overweight and underweight prevalences were classified into 5 categories (≤5%, >5% to 10%, >10% to 15%, >15% to 20%, and >20%) and displayed by state on US maps.

Table 1 gives the sample size and characteristics of the study population for each state for 2000. The number of participating children varied widely among states. Sex distributions were close to 50%; age distributions were relatively similar, with a slight predominance of 2-year-old children. However, Puerto Rico had substantially more 2-year-old and fewer 4-year-old children. Racial/ethnic distributions varied widely: Vermont was 97.1% white, the District of Columbia was 76.0% African American, Puerto Rico was 99.0% Hispanic American, the Navajo Nation was 97.9% American Indian, and Hawaii was 74.7% Asian/Pacific Islander. Other states had less single race/ethnicity dominance.

The total study population more than doubled during the study period, increasing from 953 375 children in 1989 to 2 209 444 children in 2000 (data not shown). The distribution of race/ethnicity changed substantially during this period: the proportion of whites, African Americans, American Indians, and Asians decreased and the proportion of Hispanics increased (χ² for trend = 38 942.76, P < .001). The age distribution shifted significantly; the major changes were a decrease in 2-year-old children and an increase in 4-year-old children (χ² for trend = 4452.72, P < .001). Although the trend in sex distribution during the study period changed significantly (χ² for trend = 11.04, P = .004), the actual change was very small and statistically significant only because of the extremely large sample size. In 1989, the overall population was 50.5% male; in 1994, 50.7% male; and in 2000, 50.5% male; thus, we did not adjust for sex in our analyses.

The overall prevalence of overweight increased significantly in all sex, age, and race/ethnicity categories between 1989 and 2000 (Table 2). Boys had a higher overweight prevalence than girls. Overall, overweight increased with increasing age. Overall, Hispanics had the highest overweight prevalence followed by American Indians, Asian/Pacific Islanders, African Americans, and whites.

To examine the state-specific data, we first stratified the data to look for differences in overweight prevalence by sex, age, and race/ethnicity (data not shown). More states had a greater number of boys than girls in the higher categories (>10%). For boys in 1989 and 2000, 11 and 28 of 30 states had a prevalence of more than 10%, respectively; for girls in 1989 and 2000, only 7 and 24 of 30 states had a prevalence of more than 10%, respectively. In these states, boys contributed more to overweight prevalence than girls, because the proportions of boys and girls in each state were approximately equivalent. Age differences in overweight prevalence were less clear. In general, more states had an overweight prevalence of more than 10% among 3- and 4-year-old children than among 2-year-old children. Racial/ethnic differences in overweight prevalence were apparent. Because sample sizes were smaller than 100 for some racial groups in some states, we limited our stratified comparisons to 24 states. Commonly, Hispanics had an overweight prevalence of more than 10%. In 18 states in 1989 and in 22 states in 2000, Hispanics had an overweight prevalence of more than 10%; in 15 states in 2000, Hispanics had a prevalence of more than 15%. In contrast, a prevalence of more than 10% was observed less frequently among both African Americans (4 states in 1989 and 14 in 2000) and whites (1 state in 1989 and 15 in 2000).

Age and race/ethnicity differences accounted for only a small difference in overweight prevalence among states. We compared the variance in overweight prevalence in 2000 for each state before and after adjusting for age and racial/ethnic overall composition. For age, the variance after adjustment was 0.7% higher than before adjustment; for race/ethnicity, the variance after adjustment was 11.05% lower than before adjustment, indicating that the effect of age variation on prevalence was minor (<1%) and that different racial/ethnic composition explained only approximately 11% of the difference in overweight prevalence among states (data not shown).

Age- and race/ethnicity-adjusted prevalences of overweight by category are shown in Figure 1; prevalence data, including changes, are presented in Table 3. No geographic concentration in overweight prevalence occurred, and prevalence increased in all parts of the country during the study period (Figure 1). Only Utah in 1989 met the expected 5% overweight prevalence. By 2000, 28 of 30 states had an overweight prevalence of more than 10%, whereas in 1989 and 1994 only 11 and 12 of 30 states, respectively, were in this category. The major shift to higher overweight prevalence occurred between 1994 and 2000; between 1989 and 2000, the change for each state was statistically significant (P<.01). For the total study population for each study year, significant differences occurred in overweight prevalence between boys and girls; among 2-, 3-, and 4-year-old children; and among whites, African Americans, and Hispanics (P < .001) (data not shown).

Between 1989 and 1994, overweight prevalence per year overall and for each state except Arkansas, Nebraska, and Oregon changed significantly (Table 3). Most states showed a significant increase, but Arizona, California, Kansas, Maryland, Missouri, New Mexico, and ITCA showed a significant decrease. Between 1994 and 2000, overweight prevalence overall and for each state changed significantly (Table 3); except for the District of Columbia, all prevalences increased. Overweight prevalence per year overall and for each state except the District of Columbia, Hawaii, Kentucky, Louisiana, New York, Utah, and the Navajo Nation increased significantly more between 1994 and 2000 than between 1989 and 1994, showing that, in general, the increase in overweight prevalence is accelerating (eg, overall increase in prevalence per year of 0.09% ± 0% between 1989 and 1994 vs 0.40% ± 0% between 1994 and 2000). For the entire study period, overall and all state values for overweight prevalence increased significantly except for the District of Columbia, for which they significantly decreased (Table 3).

We first stratified the data to look for differences in underweight prevalence by sex, age, and race/ethnicity (data not shown). Underweight prevalences for boys and girls were similar; those for 2-year-old children were highest and those for 3- and 4-year-old children were similar. For 24 states with racial/ethnic sample sizes per year of more than 100 (which would be considered large enough to provide stable estimates of prevalence), African American children had the highest underweight prevalence, followed by white and Hispanic children; however, in 2000, white and Hispanic children had similar low underweight prevalence. In 1989, African American children in 23 states, white children in 18 states, and Hispanic children in 8 states had underweight prevalences of more than 5%. A slightly different pattern emerged in 2000: African American children in 18 states, white children in 4 states, and Hispanic children in 3 states had an underweight prevalence of more than 5%.

Figure 2 shows adjusted underweight prevalence in categories for 1989 and 1994 and the unadjusted prevalence for 2000; prevalence data, including changes, are given in Table 4. Between 1989 and 2000, there was a major shift downward to the lowest prevalence category: in 1989, 9 of 30 states had an underweight prevalence of 5% or less, whereas by 2000, 23 of 30 states had a prevalence of 5% or less. The categorical map data show that major decreases occurred in each state between 1994 and 2000. Trend analyses indicated that between 1989 and 2000, changes in underweight prevalence were statistically significant (P<.05) in each state except the District of Columbia (χ² for trend = 1.08, P = .30), North Dakota (χ² for trend = 1.25, P = .26), Puerto Rico (χ² for trend = 0.26, P = .60), and the Navajo Nation (χ² for trend = 2.86, P = .09). No geographic area had a predominance of underweight. The adjusted underweight prevalence in the overall population declined from 6.8% in 1989 to 5.7% in 1994 and to 4.6% in 2000. For each year of study, significant differences in prevalence occurred between boys and girls; among 2-, 3-, and 4-year-old children; and among whites, African Americans, and Hispanics (P<.001) (data not shown).

Slightly different changes occurred in underweight prevalence per year (Table 4). For the overall study population, changes were significant and significantly different from each other, with the greater decrease between 1989 and 1994 than between 1994 and 2000(–0.22% ± 0.00% and –0.19% ± 0.00, respectively). A greater decrease in underweight prevalence per year between 1989 and 1994 than between 1994 and 2000 occurred in 20 of 30 states, suggesting a deceleration in the decrease of underweight during the study period. New Mexico had the greatest decrease between 1989 and 1994 (–2.06% ± 0.05% per year) and during the entire study period (–0.92% ± 0.02% per year). Hawaii also had a large reduction during the study period (–0.74% ± 0.04% per year). Iowa and Idaho had similar changes in prevalence per year for both periods. For the entire study period, overall and state-specific underweight prevalence significantly decreased per year except for the District of Columbia, which did not change (–0.04% ± 0.04%).

Overweight is pervasive throughout the United States in low-income children 2 through 4 years who participated in publicly funded health and nutrition programs, and overweight prevalence increased during our study period. By 2000, 28 of the 30 states studied had an overweight prevalence of more than 10%, more than twice the expected rate of 5%. In contrast, in 1989 and 1994 only 11 and 12 of the 30 states, respectively, were at this level. In 2000, 3 states had overweight prevalences of more than 15% to 20% and 2 had prevalences of more than 20%, which is 3 to more than 4 times greater than expected. For the overall population and 23 of 30 states, prevalence change per year was greater between 1994 and 2000 than between 1989 and 1994, indicating an acceleration.

In contrast to Mei et al¹³ and Ogden et al,¹² we found a higher overweight prevalence among boys than girls. In 1989 and 2000, 29 of 30 states had a higher prevalence among boys than girls: the higher prevalence among girls was found only in ITCA in 1989 and in the Navajo Nation in 2000. Mei et al¹³ also used PedNSS data but analyzed the ages of birth to 5 years, whereas we analyzed the ages of 2 through 4 years. Ogden et al¹² found that 2- to 5-year-old boys initially had a higher prevalence than girls, yet there was a decrease in the difference between the overweight prevalence for boys and girls between the 1989-1994 and 1999-2000 surveys. A possible explanation is the different growth references used for analyses.¹⁴ Using the same data, the US Department of Agriculture reported that overweight prevalence based on weight for height calculated from analyses using the 2000 CDC growth reference was 2.4% higher in boys and 0.7% lower in girls than for analyses using the 1977 NCHS growth reference.¹⁴ Mei et al¹³ used the 1977 NCHS growth reference for their analyses. The 2000 CDC growth charts are based on a larger sample of children, and better curve smoothing was used to make the z score and percentile values completely interchangeable,²⁰ making analyses of these parameters more comparable.

Our findings on racial/ethnic differences in overweight prevalence are similar to those reported for the 1999-2000 NHANES¹²: Hispanics had the highest prevalence. Although we did not have sample sizes adequate for accurate estimates of overweight prevalence among American Indians for each state, in our overall study population the group prevalence of 17.1% in 2000 was similar to that for Hispanics. Other researchers also reported that American Indian children have a high overweight prevalence.^23,24

Underweight prevalence decreased substantially between 1989 and 2000, with the greatest decrease per year occurring between 1989 and 1994; overall, underweight prevalence decreased from 6.8% in 1989 to 5.7% in 1994 and 4.6% in 2000. Our observed decrease in underweight is consistent with the overall national 2001 PedNSS report of a decrease in underweight prevalence among newborns to 5-year-olds from 6.9% in 1992 to 5.4% in 2001.¹⁵ The low prevalence and decreasing rate of underweight indicate that undernutrition or medical conditions that reduce weight status are not common in our overall population. However, our finding that in 2000, 18 of 24 states reported an underweight prevalence greater than the 5% expected rate among African American children requires attention.

Our data set did not include variables on diet or physical activity. High birth weight, a known risk factor for overweight among children, was not likely a major contributor to the increases in overweight prevalence, because it decreased slightly from 8.4% in 1992 to 7.9% in 2001.¹⁵ It is also unlikely that WIC participation contributed to the increase in overweight prevalence. Analyses of NHANES III data documented insignificant differences in weight status between WIC and non-WIC participants.²⁵ In addition, national data representative of the US population also showed increases in overweight prevalence over time, indicating that overweight is a national problem, not a problem exclusively associated with publicly funded programs or low income.

The decrease in underweight prevalence during our study period could possibly be a true decrease attributable to more comprehensive WIC services, including more aggressive interventions and referral of destitute families to emergency food sources, although no data document this. Another possibility is that the proportion of children at high nutritional or medical risk was reduced. The primary source of PedNSS data is WIC, and national WIC data showed a 128% increase in the participation of 1-year-old to younger than 5-year-old children between 1988 and 1997,²⁶ but we did not have access to the data to compare the WIC risk factors of the children who participated in WIC in our data set to determine whether the population had changed. However, increased WIC participation during our study period was not associated with a simultaneous decrease in overweight prevalence, even though being overweight is considered to be an anthropometric risk factor for inclusion in WIC. A final possible explanation for the decrease in underweight is the upward shift in the distribution of BMI.

The strength of our study is that we had large sample sizes. Our findings are unique because it is the first presentation of state-based prevalence data on overweight and underweight among young children in the United States, albeit limited to low-income participants of public health programs. Our study was limited because several states did not contribute to PedNSS during the 3 years of our study and not all low-income preschool children participate in WIC. However, despite these differences, our overall overweight prevalence of 13.7% for 2000 is only 1.3% higher than the 12.4% reported for the national PedNSS data for 2000 among 2- through 4-year-old children.¹⁵ Because not all states participated, we could not identify geographic areas that may have had higher prevalence.

Overweight is a serious and increasingly important public health problem in our population of 2- through 4-year-old, low-income children who participate in publicly funded health and nutrition programs. Underweight became a less serious problem between 1989 and 2000 in this population but remains a concern for African American children. This concern should be addressed in programs that provide assistance to these children. Clearly, we need to continue efforts to treat and prevent underweight and expand efforts to treat and prevent overweight in children to avoid health risks during childhood, overweight tracking to adulthood, and adult morbidity. WIC clinics are opportune, logical sites for these interventions.

Article

National studies have documented the prevalence of overweight and underweight and the trends over time, yet most public health intervention efforts are either state or community based. State-specific prevalence and trends for overweight and underweight among children are not available in the peer-reviewed literature.

This study provides state-specific overweight and underweight prevalence for 1989, 1999, and 2000 and analyses of their trends. The increasing prevalence of overweight and decreasing prevalence of underweight emphasize the urgent need to implement interventions to attenuate the increase in overweight.

Correspondence: Bettylou Sherry, PhD, RD, Maternal and Child Nutrition Branch, Division of Nutrition and Physical Activity, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Mailstop K-25, 4770 Buford Highway NE, Atlanta, GA 30341-3717 (bsherry@cdc.gov).

Accepted for Publication: June 11, 2004.