- Email: [email protected]
Increasing Breastfeeding Rates
Women’s Health Issues 17 (2007) 84 –92
INCREASING BREASTFEEDING RATES: Do Changing Demographics Explain Them? Alison Jacknowitz, PhD, MPP* American University, Washington, DC Received 20 July 2006; revised 15 February 2007; accepted 23 February 2007
Objectives: This study examines whether increases in breastfeeding rates between 1991 and 2002 can be attributed to changes in the demographic characteristics of births. The demographic variables investigated include maternal age, maternal education, race/ethnicity, geographic location of birth, and parity. Methods: This study decomposes breastfeeding trends using 1991 through 2002 data from the Ross Laboratories Mothers Survey and birth certificates. Results: Changing birth compositions by maternal age and education explain 9.8% and 11.5% of the increase in breastfeeding initiation rates, respectively. Changing birth compositions by maternal age and education explain 10.2% and 9.0% of increasing breastfeeding rates 6 months after birth, respectively. If changes in the composition of births by race/ethnicity had not occurred, breastfeeding rates would have increased more. Changes in the composition of births by geographic location of birth and parity do not explain any of the increase in breastfeeding rates. Conclusions: Overall, changes in the demographic characteristics of births explain up to approximately 20% of the increasing trends in initial breastfeeding rates and breastfeeding rates 6 months after birth. Although this is a significant amount, it is important to continue investigating which additional factors explain recent breastfeeding increases such as changes in laws and policies, health promotion, the Special Supplemental Program for Women, Infants, and Children, employer support, technological innovation, and attitudes toward breastfeeding.
T
here are well-established short- and long-term health benefits of breastfeeding to children and mothers. Studies in the United States and abroad have found evidence that children who are breastfed have lower rates of urinary and respiratory tract infections, diarrhea, otitis media, bacterial meningitis, bacteremia, and necrotizing enterocolitis (American Academy of Pediatrics, 2005; Leon-Cava, Lutter, Ross & Martin, 2002). Studies also suggest that breastfeeding is beneficial for the mother’s health (American Academy of Pediatrics, 2005; Labbok, 2001). The list of beneficial maternal health outcomes includes lowered risk of breast and ovarian cancers, incidence of longterm osteoporosis, and pregnancy-induced obesity
* Correspondence to: Dr Alison Jacknowitz, American University, 4400 Massachusetts Ave, NW, Washington, DC 20016. E-mail: [email protected] Copyright © 2007 by the Jacobs Institute of Women’s Health. Published by Elsevier Inc.
(American Academy of Pediatrics, 2005; Labbok, 2001). Reflecting research indicating that both children and mothers benefit from breastfeeding, numerous individuals and organizations support and recommend breastfeeding. These individuals and organizations include the US Surgeon General, American Academy of Pediatrics, American Medical Association, American Dietetic Association, American Academy of Family Physicians, and the World Health Organization. Although researchers, public health organizations, and physicians generally agree on the importance of breastfeeding, a sizable percentage of mothers do not breastfeed. In 2002, 29.9% of women in the hospital and 66.8% of mothers 6 months after birth did not breastfeed their children (Abbott Laboratories, 2002). The US Surgeon General states, “The nation must 1049-3867/07 $-See front matter. doi:10.1016/j.whi.2007.02.010
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
address these low breastfeeding rates as a public health challenge and put into place national, culturally appropriate strategies to promote breastfeeding” (United States Department of Health and Human Services [US DHHS], 2000, p. 3). The US Government has selected increasing breastfeeding rates as a Healthy People 2010 objective alongside other national health goals such as decreasing rates of cancer, sexually transmitted diseases, obesity, and food insecurity. In addition, the American Academy of Pediatrics (2005) endorses exclusive breastfeeding (i.e., without supplementation) for approximately 6 months after birth and recommends continued breastfeeding with supplementation until the infant is at least 12 months old. Despite the fact that a share of women do not breastfeed, the likelihood of any breastfeeding (i.e., nonexclusive) initially (i.e., in the hospital) and any breastfeeding 6 months after birth has increased since the early 1990s (Abbott Laboratories, 2002). Between 1991 and 2002, breastfeeding rates for all mothers increased 16.8 percentage points (31.5%) at birth and 15.0 percentage points (82.4%) 6 months after birth (Abbott Laboratories, 2002). Although the literature agrees that breastfeeding rates increased between 1991 and 2002 (e.g., Abbott Laboratories, 2002; Ryan, 1997; Ryan, Zhou, & Acosta, 2002; Wright, 2001), explanations for the upward trend have not been empirically investigated. Two papers (Jacknowitz, 2006; Wright, 2001) offer several possible explanations for the rising breastfeeding rates during this period, including changes in the demographic characteristics of mothers, changes to the Special Supplemental Program for Women, Infants, and Children (WIC), changes in the federal government’s support and promotion of breastfeeding, changes in laws and policies that promote breastfeeding, and changes in workplace characteristics that facilitate working and breastfeeding. However, neither author empirically tests these hypotheses. This paper contributes to the breastfeeding literature by studying 1 potential explanation for the upward trend in breastfeeding rates that has not been empirically tested: changes in the demographic characteristics of births. Evidence suggests that the demographic characteristics of births changed between 1991 and 2002. Shifts in the characteristics of births coupled with the fact that some groups of mothers are more likely to breastfeed than others suggests that changes in the demographic characteristics of births could explain the increasing breastfeeding trends. For example, mothers with more education are more likely to breastfeed, and births to mothers with a college education increased 11.2 percentage points or 31.6% between 1991 and 2002 (National Center for Health Statistics [NCHS], 1993, 2003a). To examine whether increases in breastfeeding rates since 1991 can be attributed to many of the well-known demographic
85
changes in births that have been observed, this study decomposes breastfeeding trends using 1991 through 2002 data from the Ross Laboratories Mothers Survey (RMS) and birth certificates. From a policy perspective, it is most efficient to examine the role of changes in the demographic characteristics of births in breastfeeding trends before implementing policy interventions and evaluations. If demographic changes explain a large majority of the increases in breastfeeding rates, then implementing costly programs may not be an efficient use of resources. Although many programs with the goal of increasing breastfeeding have already been established, the results from this study can be used to shape future breastfeeding research. If changes in the demographic characteristics of births explain some of the increases in breastfeeding, then it is important to control for these demographic factors when evaluating the effectiveness of individual breastfeeding promotion programs. In addition, if these changes do not explain all of the increases in breastfeeding rates, further research is necessary to determine which factors do to ensure the continuing increase of breastfeeding rates.
Methods Data The breastfeeding data used in this study are from aggregate tabulations of the RMS, the only dataset that can produce annual national breastfeeding trends dating back to the early 1990s. RMS, a proprietary survey of the Ross Products Division of Abbott Laboratories, is a large, national mail survey conducted since 1955 to determine patterns of milk feeding from birth to 12 months. Questionnaires are mailed to a probability sample of new mothers selected from a sample frame of names that represent approximately 80 – 85% of all national births. The list includes names from hospital records, county records of birth registrations, photography and diaper services, and newspapers. The samples are very large and have increased over time; approximately 720,000 and 1.4 million questionnaires were mailed in 1991 and 2002, respectively. Two papers summarize other breastfeeding data sources and their advantages and disadvantages relative to the RMS (Grummer-Strawn & Li, 2000; Jacknowitz, 2002). Aggregate breastfeeding estimates by maternal characteristic for 1991 through 2002 are from tabulations available in Abbott Laboratories (2001, 2002). Any breastfeeding rates at the following 2 points in time after birth are available to the public: initiation of breastfeeding and breastfeeding at 6 months. The RMS asks mothers to recall the type of milk her baby was fed in the hospital, at 1 week of age, in the last 30 days, and in the last week. In addition, mothers are asked
86
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92 80 70 NMIHS NHANES
60 RLMS
Rate
50 NSFG
NSFG
40
NSFG FDA-IFPS NHANES
30
NMIHS
20 10 0 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 Year
Figure 1. US breastfeeding initiation rates for all mothers, 1970 – 2002. Figure adapted from Jacknowitz (2002). Data are from the following sources: RLMS data are from Abbott Laboratories (2002); NSFG data are from NCHS (1998, 2006); FDA-IFPS data are from Fein and Roe (1998); NHANES data are from Burstein et al. (2000); and NMIHS data are from Visness and Kennedy (1997). NHANES and NSFG breastfeeding initiations rates are for births over multiple years. In this figure, breastfeeding rates are assigned to the middle year in the case of an estimate over an odd number of years, the earlier year in the case of a 2-year estimate, and the second year in the case of a 4-year estimate.
about their demographic characteristics. Breastfeeding rates stratified by the following demographic characteristics are available: maternal age, maternal education, maternal employment, race/ethnicity, parity, WIC participation, and Census division of residence. The RMS asks whether the infant was low birthweight, but the information is not included in the tabulations in a usable format. Tabulations only include breastfeeding rates for low birthweight infants and all infants (including low birthweight babies); therefore, breastfeeding rates cannot be calculated for non–low-birthweight babies. Figure 1 and other studies illustrate that the annual breastfeeding rates produced with the RMS are comparable to those from other sporadically collected data. Figure 1 shows breastfeeding rates from the RMS, National Health and Nutrition Examination Survey (NHANES), National Maternal and Infant Health Survey (NMIHS), National Survey of Family Growth (NSFG), and the Food and Drug Administration’s Infant Feeding Practices Survey (FDA-IFPS). These comparison data sources are nationally representative, with approximate response rates ⬎70% during this time period. Four of these 5 datasets provide similar estimates of breastfeeding initiation for all mothers, with the only exception of FDA-IFPS. Comparisons of the FDA-IFPS with the NMIHS data indicate that women in the FDA-IFPS sample were more likely to be in middle- and upper-income groups, older, married, and White than the NMIHS sample (Fein & Roe, 1998). These characteristics are associated with higher breastfeeding rates, which is consistent with the higher estimates generated using the FDA-
IFPS. Studies (Hediger, Overpeck, Kuczmarski, & Ruan, 2001; Ryan et al., 1991) that compared breastfeeding estimates from the RMS to those generated using the NHANES and NSFG also found similar breastfeeding rates among the datasets. In addition to breastfeeding statistics, data on the composition of births from 1991–2002 necessary to perform the analysis are from the National Center for Health Statistics’ National Vital Statistics Reports series, which was previously titled the Monthly Vital Statistics Reports series. The National Vital Statistics Reports rely on birth data extracted from 100% of birth certificates registered in the 50 states and the District of Columbia. Because these data are a near census of births for a given year (a small fraction of births are unreported) and include a variety of maternal characteristics that are available in the RMS tabulations using the same definitions as those used by the RMS, they are ideal to use for this study. These data are a near census owing to a small number of unreported births. For example, in 2001, an estimated 99% of births in the United States were reported (NCHS, 2003b). Analytic Approach Standardization and decomposition techniques similar to those described by Das Gupta (1993) and Kitagawa (1955) are used to determine the share of the increase in breastfeeding rates between 1991 and 2002 that are explained by changes in the demographic characteristics of births. For the purposes of this study, these descriptive techniques compare the actual breastfeeding rate with an estimated breastfeeding rate for a given demographic characteristic, which represents what the breastfeeding rate would have been if breastfeeding practices remained at their 1991 values but birth rates by demographic characteristic changed. The estimated breastfeeding rate is calculated holding breastfeeding rates by subgroup constant at their 1991 values and allowing the share of births for a given subgroup to change (equation 1). BFt,S ⫽
N
兺 bf1991,s ⴱ brt,s
(1)
s⫽1
In equation 1, BFt,S is the estimated breastfeeding rate which allows the share of births by demographic characteristic, S, to vary to their values in year t; bf1991,s is the actual breastfeeding rate in 1991 for subgroup s; brt,s is the proportion of births born to mothers of subgroup s in year t; and N is the number of subgroups for the particular demographic characteristic. For example, parity is divided into the subgroups of first births and higher order births. To calculate the 2002 estimated breastfeeding rate the following 2 terms would be aggregated: 1) 1991 breastfeeding rate for first births multiplied by 2002 share of births that are first births and 2) 1991 breastfeeding rate for
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
higher order births multiplied by 2002 share of births that are higher order births. To determine the share, YS, of the increase in breastfeeding rates between 1991 and 2002 that is explained by changes in births by a given demographic characteristic, the difference between the actual aggregate 1991 breastfeeding rate and the 2002 estimated breastfeeding rate is divided by the change in the actual aggregate breastfeeding rates between 1991 and 2002. The calculation of YS for a given demographic characteristic, S, is illustrated in equation 2. YS ⫽
BF2002,S ⫺ bf1991 bf2002 ⫺ bf1991
(2)
In equation 2, bf2002 is the actual aggregate breastfeeding rate in 2002 and bf1991 is the actual aggregate breastfeeding rate in 1991. Their difference is the change in breastfeeding rates between 1991 and 2002. For breastfeeding initiation, this difference is 16.8 percentage points; for breastfeeding at 6 months after birth, it is 15.0 percentage points. If demographic changes explain some of the increase in breastfeeding rates between 1991 and 2002, then the share, YS, would be positive. If the demographic changes explain none of the increase in breastfeeding rates over this time period, then the share, YS, would be close to zero. Finally, if the changes that occurred led to lower overall breastfeeding rates, than the share, YS, would be negative. Demographic Characteristics The demographic characteristics available in the RMS were included in the analysis if 3 criteria were met. First, the demographic characteristics were available in both the RMS and birth certificate data and were in a usable form, which meant that the demographic characteristic was defined in a similar manner in both datasets and information for all subgroups was available. Second, there were sizeable differences in breastfeeding rates between mothers by this characteristic. If differences in breastfeeding by this factor do not exist, then changing the composition of births will not change the estimated 2002 breastfeeding rate derived in equation 1. Third, the composition of births changed by this characteristic between 1991 and 2002. If the composition of births has remained the same then the estimated 2002 breastfeeding rate will be equal to the actual breastfeeding rate in 1991, thus explaining none of the increase in breastfeeding rates. The demographic characteristics available in both datasets in a usable form include maternal age, maternal education, race/ethnicity, parity, and Census division of residence (hereafter referred to as geographic location of birth). Maternal employment and WIC participation are not available in the birth certif-
87
icate data and are therefore excluded from the analysis. Questions on maternal WIC participation were added to the live birth certificate questionnaire in 2003, which is after the time period of interest. To address whether differences exist in breastfeeding rates between women by a given characteristic, this paper turns to breastfeeding trends and the breastfeeding literature. Jacknowitz (2002) identifies breastfeeding correlates by graphing trends in breastfeeding practices by subgroup and conducting a literature review. Studies included in the literature review use multivariate regression techniques to predict correlates of breastfeeding practices. Jacknowitz (2002) finds that of those demographic characteristics available in the RMS, maternal age, maternal education, race/ethnicity, and geographic location of birth are strong correlates of breastfeeding, but parity is not during the time period of interest. Table 1 illustrates findings for demographic characteristics that are related to breastfeeding for 1991 and 2002. Younger mothers are less likely to breastfeed than older mothers. Less educated mothers are less likely to breastfeed than more educated mothers. Non-Hispanic Black mothers are less likely than mothers of other race/ ethnicities to breastfeed. Mothers residing in the Western states (Mountain and Pacific Census regions) have higher breastfeeding rates than other regions, whereas mothers living in the East South Central division have considerably lower breastfeeding rates than all other regions. The third criterion for determining which characteristics to decompose is whether the birth composition by these variables changed. Table 2 illustrates that large changes in the shares of births occurred by maternal age, maternal education, and race/ethnicity. Between 1991 and 2002, births to women ages ⱖ30 increased 6.6 percentage points (21.3%) and births to women with a college education increased 11.2 percentage points (31.6%). Births to Hispanic mothers, who are more likely to breastfeed than non-Hispanic Black mothers, increased 6.6 percentage points. At the same time births to non-Hispanic White mothers, who are also more likely to breastfeed than non-Hispanic Black mothers, declined by 5.9 percentage points. Although race/ethnicity is a strong correlate of breastfeeding, these changes considered together suggest that compositional changes in the race/ethnicity of births may not influence overall breastfeeding rates because, although race/ethnicity is associated with breastfeeding rates, these 2 effects may cancel each other out. The 1 demographic variable that did not undergo compositional changes in births and is excluded from the analysis is geographic location of birth. Hence, this exercise is performed for all years between 1991 and 2002 for breastfeeding initiation and duration for maternal age, maternal education, and
88
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
Table 1. US Breastfeeding Rates by Selected Characteristics, 1991 and 2002 Breastfeeding Rate in Hospital
Maternal age (yrs) ⬍20 20–24 25–29 30–34 ⱖ35 Maternal education Grade school High school Non-college College Unknown Race/ethnicity Hispanic Non-Hispanic, Black Non-Hispanic, White Other Census division New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific
Breastfeeding Rate 6 Months After Birth
1991
2002
Change
1991
2002
Change
32.2 45.7 57.8 65.2 66.7
56.2 66.0 73.4 76.4 74.1
24.0 20.3 15.6 11.2 7.4
5.9 11.4 19.7 27.3 32.9
16.7 25.4 35.7 42.3 43.1
10.8 14.0 16.0 15.0 10.2
35.2 42.9 42.6 70.2 N/A
55.1 60.7 60.5 81.2 N/A
19.9 17.8 17.9 11.0
12.6 12.2 12.2 27.7 N/A
27.1 23.4 23.5 44.6 N/A
14.5 11.2 11.3 16.9
51.8 25.8 59.2 N/A
70.7 53.9 73.4 N/A
18.9 28.1 14.2
14.8 6.6 21.0 N/A
32.7 19.2 36.0 N/A
17.9 12.6 15.0
55.9 48.3 48.8 56.4 46.4 37.5 48.4 70.0 69.5
73.3 65.5 66.7 73.1 67.5 57.0 64.9 81.0 81.5
17.4 17.2 17.9 16.7 21.1 19.5 16.5 11.0 12.0
19.7 16.8 16.6 18.9 14.5 10.7 14.0 28.0 25.7
37.0 33.2 28.7 35.2 31.2 22.0 27.3 40.5 43.4
17.3 16.4 12.1 16.3 16.7 11.3 13.3 12.5 17.7
Notes: Data are from Abbott Laboratories (2001, 2002). Breastfeeding rates are unavailable for mothers with unknown education or other race/ethnicity.
race/ethnicity. The subgroups of maternal age at birth include ⬍20 years, 20 –24 years, 25–29 years, 30 –34 years, and ⱖ35 years. These groups and those for maternal education and race/ethnicity are determined by the RMS tabulations. Maternal education subgroups include grade school, high school, non-college, college, and unknown. Finally, the subgroups for race/ethnicity are Hispanic, non-Hispanic Black, nonHispanic White, and other. In addition, decompositions are performed for both breastfeeding initiation and 6 months after birth because differences in breastfeeding rates between subgroups may differ for the 2 outcomes (Table 1). For example, there are larger differences in breastfeeding initiation between more educated and less educated mothers than in breastfeeding 6 months after birth. Mothers with unknown education and other race/ ethnicity are assigned the mean breastfeeding rate for all mothers. To test whether this is a reasonable strategy, 1991 and 2002 actual breastfeeding rates in the hospital and 6 months after birth are compared with ones that weight each subgroup’s breastfeeding rate by the proportion of births to that given subgroup of a demographic variable. If assigning the other and unknown subgroups the mean breastfeeding rate for all mothers is a reasonable strategy, we would expect
the actual and weighted breastfeeding rates to be very close. Indeed, the estimated breastfeeding rates (1991 and 2002 initial and 6 months after birth by maternal education and race/ethnicity) are within 1 percentage point of those produced by the RMS. Microsoft Excel is used to organize the data and perform the analysis.
Results Figures 2 and 3 graph the actual breastfeeding rates and the estimated breastfeeding rates by year in the hospital and 6 months after birth, respectively. If changes in birth rates by a given demographic characteristic explains any of the increasing breastfeeding rate, we would expect to see the estimated breastfeeding rates for that demographic characteristic increase over time. Therefore, the closer the estimated breastfeeding rates for a given characteristic and the actual breastfeeding rate, the larger share of the increase in breastfeeding rates that changes in the composition of births for that characteristic explains. Both figures show that the estimated breastfeeding rates for maternal education and age increase slightly and gradually over time, suggesting that changes in the composition
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92 Table 2. Composition of US Births by Selected Characteristics, 1991 and 2002
80 70 60
Change
12.9 26.5 29.7 21.5 9.4
10.8 25.4 26.4 23.7 13.8
⫺2.2 ⫺1.1 ⫺3.3 2.1 4.4
6.1 16.5 34.9 35.5 7.1
6.0 15.3 30.7 46.7 1.3
⫺0.1 ⫺1.2 ⫺4.2 11.2 ⫺5.7
15.2 16.2 63.0 5.6
21.8 14.4 57.1 6.7
6.6 ⫺1.8 ⫺5.9 1.1
4.7 14.2 16.2 6.5 16.7 5.7 11.5 5.9 18.6 4,110,907
4.2 12.7 15.2 6.6 18.0 5.8 13.1 7.6 16.9 4,021,726
⫺0.5 ⫺1.5 ⫺1.0 0.1 1.3 0.1 1.5 1.7 ⫺1.6
Notes: The 1991 birth data are from NCHS (1993) and the 2002 birth data are from NCHS (2003a). The 1991 unknown education category is primarily comprised of births from Washington and New York states (excluding New York City), which did not collect education information; 16,341 of the 1991 other race/ethnicity births are from New Hampshire, which did not collect race/ethnicity data.
of births by these 2 characteristics explain some of the increasing trends in initial breastfeeding rates and breastfeeding rates 6 months after birth. In contrast, the estimated breastfeeding rates for race/ethnicity remain fairly constant over time, suggesting that changes in the composition of births by race/ethnicity do not explain the increases in breastfeeding over this time period. Table 3 takes the analysis further and calculates the share of the increase in breastfeeding rates between 1991 and 2002 that changes in the composition of births explain. Table 3 shows that changes in the share of births by maternal age explain 9.8% of the increase in initial breastfeeding rates. The change in the share of births by maternal education explains 11.5% of the increasing initial breastfeeding rates. As expected, changes in shares of births to mothers of different race/ethnicity explain none of the increase in in-hospital breastfeeding rates since 1991. In fact, applying the 2002 composition of births by race/ ethnicity to the appropriate 1991 breastfeeding rates
50
Actual
40
Maternal age
30
Maternal education
20
Race/ethnicity
10 0 1991
1993
1995 1997 Year
1999
2001
Figure 2. Decomposition of US initiation breastfeeding rates, 1991– 2002. Maternal age, maternal education, and race/ethnicity breastfeeding rates are calculated by weighting the 1991 breastfeeding rate by the year’s share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
results in a lower overall initial breastfeeding rate, implying that initial breastfeeding rates would have been higher if compositional changes in births by race/ethnicity had not occurred. Similar to the findings for the initiation of breastfeeding, Table 3 illustrates that changes in the composition of births explain some of the increase in breastfeeding rates 6 months after birth since 1991. The shift of births by maternal age explains 10.2% of the increase in breastfeeding rates 6 months after birth. The changes in the birth shares by maternal education explain 9.0% of the increasing breastfeeding rates 6 months after birth. This statistic is approximately 2.5 percentage points lower than the number for the initiation of breastfeeding rates. This is because differences in breastfeeding rates between the most and least educated mothers are larger for initiation (on average 30.2 percentage points) than duration (on average 16.5 percentage points). Finally, changes in the racial/ethnic composition of births account for 35 30 25 Rate
Maternal age (yrs) ⬍20 20–24 25–29 30–34 ⱖ35 Maternal education Grade school High school Non-college College Unknown Race/ethnicity Hispanic Non-Hispanic, Black Non-Hispanic, White Other Census division New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific Total births
2002
Rate
Percentage of Births 1991
89
Actual
20
Maternal age 15
Maternal education
10
Race/ethnicity
5 0 1991
1993
1995 1997 Year
1999
2001
Figure 3. Decomposition of US breastfeeding rates 6 months after birth, 1991–2002. Maternal age, maternal education, and race/ ethnicity breastfeeding rates are calculated by weighting the 1991 breastfeeding rate by the year’s share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
90
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
Table 3. Decomposition of US Breastfeeding Rates Breastfeeding Rate
Actual 1991 breastfeeding rate Actual 2002 breastfeeding rate Change in breastfeeding rate Maternal age (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained Maternal education (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained Race/ethnicity (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained
In Hospital
Six Months After Birth
53.3 70.1 16.8
18.2 33.2 15.0
54.9 1.6 9.8
19.7 1.5 10.2
55.2 1.9 11.5
19.5 1.3 9.0
52.4 ⫺0.9 ⫺5.4
17.4 ⫺0.8 ⫺5.4
Notes: Estimated 2002 breastfeeding rate is calculated by weighting the 1991 breastfeeding rate by the 2002 share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
essentially none of the increasing breastfeeding rates. Similar to the case with breastfeeding initiation, applying the 2002 composition of births by race/ethnicity to the appropriate 1991 breastfeeding rates results in a lower overall breastfeeding rate 6 months after birth.
Discussion This paper examines whether increases in breastfeeding rates since 1991 can be attributed to changes in key demographic characteristics of births. To answer the research question, this study decomposes breastfeeding trends using 1991 through 2002 data from the RMS and birth certificate data. Findings suggest that changes in the composition of births with respect to maternal age and education explain up to approximately 20% of the upward trend in initial breastfeeding rates and breastfeeding rates 6 months after birth. Changing birth compositions by maternal age and education explain 9.8% and 11.5% of the increase in breastfeeding initiation rates, respectively. Changing birth compositions by maternal age and education explain 10.2% and 9.0% of increasing breastfeeding rates 6 months after birth, respectively. Geographic location of birth and parity do not explain any of the increase in breastfeeding rates. Changes in the composition of births by race/ ethnicity lead to lower breastfeeding rates in the hospital and 6 months after birth. Although there is an increase in the percent of births to Hispanic women, the breastfeeding rate decreases because this increase is offset by a decrease in the percent of births to non-Hispanic White women who have higher breastfeeding rates than Hispanic women. If the increase in
the percent of births to Hispanic women had been accompanied by a decrease in the percent of births to non-Hispanic Black women, the overall breastfeeding rate would have increased. Limitations of this study exist, which are primarily data driven. First, the variables available in the RMS and birth certificate data include the majority of key demographic variables, but are restricted. The birth certificate data do not contain information on WIC participation or maternal employment. In addition, the aggregate RMS data do not stratify breastfeeding rates after breastfeeding by birthweight status. However, none of these variables are likely to explain the increase in breastfeeding trends over the time period of interest. WIC participation among eligible infants has increased (United States Department of Agriculture, 2006); however, evidence suggests that WIC participation is negatively related to breastfeeding practices (Bitler & Currie, 2005; Ryan & Zhou, 2006). Further, maternal employment rates among women with young children ⬍3 have increased (United States Department of Labor, 2006) and studies indicate that maternal employment is negatively related to breastfeeding (Jacknowitz, 2002). The research on the relationship between low birthweight status and breastfeeding is mixed between detecting no relationship between the 2 variables and a negative relationship (e.g., Chatterji & Brooks-Gunn, 2004; Ryan & Zhou, 2006). However, even if the relationship were negative, with an increasing percent of low-birthweight infants (NCHS, 2003a), this demographic characteristic would not explain the increase in breastfeeding rates. The second primary limitation of the study is that the aggregate nature of the data does not allow for
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
methods that account for the associations between the independent variables such as changing the demographic characteristics sequentially or using multivariate regression analysis. For example, maternal age at birth and maternal education at birth are likely correlated as more educated women tend to have children later in life. Hence, the estimate of 20% of the increase in breastfeeding explained by maternal education and age is overestimated as a result of this positive correlation. A final limitation is that, although breastfeeding rates from the RMS are comparable to those from other nationally representative datasets, they represent 80 – 85% of births or fewer depending on the response rate for the year. Thus, the breastfeeding rates are not completely comparable to the birth data, which represent a near census of US births. Although there are limitations, this study does contribute to the literature by starting to investigate the factors that explain recent trends in breastfeeding practices. Given the health benefits of breastfeeding to both mothers and children, it is important to identify and understand the factors explaining these increasing breastfeeding rates to ensure their continued growth. Although this study does explain a sizeable share of the increase in breastfeeding rates, it is important to continue investigating which additional factors explain recent breastfeeding increases. Plausible factors to investigate include changes in laws and policies, health promotion, the WIC Program, employer support, and technological innovation. A final promising explanation for the increasing breastfeeding trends is changes in attitudes toward breastfeeding practices. Given that changes in the demographic composition of births do not explain the majority of the increase in breastfeeding trends, it is worthwhile to pursue and evaluate current and additional programs to improve breastfeeding. However, the findings do highlight the importance of controlling for demographic variables when examining the effects of programs to improve breastfeeding rates after breastfeeding. Finally, the study does raise the issue of the importance of considering whether demographic changes influence health behaviors before considering new policies and programs. If demographic changes do explain a majority of trends in a health behavior, the implementation of an intervention may not be an efficient use of resources.
Acknowledgments I thank Julie DaVanzo, Steven Haider, Rebecca Kilburn, Elizabeth Peters, and Robert Schoeni for helpful comments; Steven Putansu provided excellent research assistance. Financial support from the RAND Graduate School Hagopian Dissertation Award is gratefully acknowledged. This work reflects only the opinions of the author.
91
References Abbott Laboratories. (2001). Breastfeeding trends through 2001. Columbus, OH: Ross Products Division. Abbott Laboratories. (2002). Breastfeeding trends—2002. Columbus, OH: Ross Products Division. American Academy of Pediatrics. (2005). Breastfeeding and the use of human milk. Pediatrics, 115, 496 –506. Bitler, M. P., & Currie, J. (2005). Does WIC work? The effects of WIC on pregnancy and birth outcomes. Journal of Policy Analysis and Management, 24, 73–91. Burstein, N. R., Fox, M. K., Hiller, J. B., Kornfeld, R., Lam, K., Price, C., et al. (2000). WIC general analysis project: Profile of WIC children. Cambridge, MA: Abt Associates. Chatterji, P., & Brooks-Gunn, J. (2004). WIC participation, breastfeeding practices, and well-child care among unmarried, lowincome mothers. American Journal of Public Health, 94, 1324 –1327. Das Gupta, P. (1993). Standardization and decomposition of rates: A user’s manual (Current Population Reports Series, P23-186). Washington, DC: US Department of Commerce, US Bureau of the Census. Fein, S. B., & Roe, B. (1998). The effect of work status on initiation and duration of breast-feeding. American Journal of Public Health, 88, 1042–1046. Grummer-Strawn, L. M., & Li, R. (2000). US national surveillance of breastfeeding behavior. Journal of Human Lactation, 16, 283–290. Hediger, M. L., Overpeck, M. D., Kuczmarski, R. J., & Ruan, W. J. (2001). Association between infant breastfeeding and overweight in young children. The Journal of the American Medical Association, 285, 2453–2460. Jacknowitz, A. (2002). Estimating eligibility for WIC: The role of breastfeeding [working paper]. Santa Monica, CA: RAND Graduate School. Jacknowitz, A. (2006). Understanding gains in breastfeeding rates: Promising policies and programs. Women’s Health Issues, 16, 101–103. Kitagawa, E. M. (1955). Components of a difference between two rates. Journal of the American Statistical Association, 50, 1168 –1194. Labbok, M. H. (2001). Effects of breastfeeding on the mother. Pediatric Clinics of North America, 48, 143–158. Leon-Cava, N., Lutter, C., Ross, J., & Martin, L. (2002). Quantifying the benefits of breastfeeding: A summary of the evidence. Washington, DC: Pan American Health Organization. National Center for Health Statistics (NCHS). (1993). Advance report of final natality statistics, 1991 (Monthly Vital Statistics Reports, Vol. 42, No. 3). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (1998). Health, United States, 1998 with socioeconomic status and health chartbook (US DHHS Publication No. PHS 98-1232). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2003a). Births: Final data for 2002 (National Vital Statistics Reports, Vol. 52, No. 10). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2003b). Vital statistics of the United States, 2001. Volume 1: Natality technical appendix. Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2006). Health, United States, 2006 with chartbook on trends in the health of Americans (US DHHS Publication No. PHS 2006-1232). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. Ryan, A. S. (1997). The resurgence of breastfeeding in the United States. Pediatrics, 99, 12. Ryan, A. S., Pratt, W. F., Wysong, J. L., Lewandowski, G., McNally, J. W., & Krieger, F. W. (1991). A comparison of breast-feeding
92
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
data from the National Surveys of Family Growth and the Ross Laboratories Mothers Survey. American Journal of Public Health, 81, 1049 –1052. Ryan A. S., & Zhou, W. (2006). Lower breastfeeding rates persist among the Special Supplemental Nutrition Program for Women, Infants, and Children participants, 1978 –2003. Pediatrics, 117, 1136 –1146. Ryan, A. S., Zhou, W., & Acosta, A. (2002). Breastfeeding continues to increase into the new millennium. Pediatrics, 110, 1103–1109. United States Department of Agriculture. (2006). WIC program coverage: How many eligible individuals participated in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC): 1994 to 2003? Alexandria, VA: US Department of Agriculture, Food and Nutrition Service. United States Department of Health and Human Services (US DHHS). (2000). HHS blueprint for action on breastfeeding. Washington, DC: US Department of Health and Human Services, Office of Women’s Health. United States Department of Labor. (2006). Women in the labor force: A databook (Report 996). Washington, DC: US Department of Labor, Bureau of Labor Statistics.
Visness, C. M., & Kennedy, K. I. (1997). Maternal employment and breast-feeding: Findings from the 1988 National Maternal and Infant Health Survey. American Journal of Public Health, 87, 945–950. Wright, A. L. (2001). The rise of breastfeeding in the United States. Pediatric Clinics of North America, 48, 1–12.
Author Description Alison Jacknowitz, PhD, MPP, is an Assistant Professor of Public Administration and Policy at American University. Her interests include child and family policy, child well-being, poverty, and public assistance programs. She has written several papers on the topic of breastfeeding including examining the effects of welfare work requirements and workplace characteristics on breastfeeding decisions.
INCREASING BREASTFEEDING RATES: Do Changing Demographics Explain Them? Alison Jacknowitz, PhD, MPP* American University, Washington, DC Received 20 July 2006; revised 15 February 2007; accepted 23 February 2007
Objectives: This study examines whether increases in breastfeeding rates between 1991 and 2002 can be attributed to changes in the demographic characteristics of births. The demographic variables investigated include maternal age, maternal education, race/ethnicity, geographic location of birth, and parity. Methods: This study decomposes breastfeeding trends using 1991 through 2002 data from the Ross Laboratories Mothers Survey and birth certificates. Results: Changing birth compositions by maternal age and education explain 9.8% and 11.5% of the increase in breastfeeding initiation rates, respectively. Changing birth compositions by maternal age and education explain 10.2% and 9.0% of increasing breastfeeding rates 6 months after birth, respectively. If changes in the composition of births by race/ethnicity had not occurred, breastfeeding rates would have increased more. Changes in the composition of births by geographic location of birth and parity do not explain any of the increase in breastfeeding rates. Conclusions: Overall, changes in the demographic characteristics of births explain up to approximately 20% of the increasing trends in initial breastfeeding rates and breastfeeding rates 6 months after birth. Although this is a significant amount, it is important to continue investigating which additional factors explain recent breastfeeding increases such as changes in laws and policies, health promotion, the Special Supplemental Program for Women, Infants, and Children, employer support, technological innovation, and attitudes toward breastfeeding.
T
here are well-established short- and long-term health benefits of breastfeeding to children and mothers. Studies in the United States and abroad have found evidence that children who are breastfed have lower rates of urinary and respiratory tract infections, diarrhea, otitis media, bacterial meningitis, bacteremia, and necrotizing enterocolitis (American Academy of Pediatrics, 2005; Leon-Cava, Lutter, Ross & Martin, 2002). Studies also suggest that breastfeeding is beneficial for the mother’s health (American Academy of Pediatrics, 2005; Labbok, 2001). The list of beneficial maternal health outcomes includes lowered risk of breast and ovarian cancers, incidence of longterm osteoporosis, and pregnancy-induced obesity
* Correspondence to: Dr Alison Jacknowitz, American University, 4400 Massachusetts Ave, NW, Washington, DC 20016. E-mail: [email protected] Copyright © 2007 by the Jacobs Institute of Women’s Health. Published by Elsevier Inc.
(American Academy of Pediatrics, 2005; Labbok, 2001). Reflecting research indicating that both children and mothers benefit from breastfeeding, numerous individuals and organizations support and recommend breastfeeding. These individuals and organizations include the US Surgeon General, American Academy of Pediatrics, American Medical Association, American Dietetic Association, American Academy of Family Physicians, and the World Health Organization. Although researchers, public health organizations, and physicians generally agree on the importance of breastfeeding, a sizable percentage of mothers do not breastfeed. In 2002, 29.9% of women in the hospital and 66.8% of mothers 6 months after birth did not breastfeed their children (Abbott Laboratories, 2002). The US Surgeon General states, “The nation must 1049-3867/07 $-See front matter. doi:10.1016/j.whi.2007.02.010
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
address these low breastfeeding rates as a public health challenge and put into place national, culturally appropriate strategies to promote breastfeeding” (United States Department of Health and Human Services [US DHHS], 2000, p. 3). The US Government has selected increasing breastfeeding rates as a Healthy People 2010 objective alongside other national health goals such as decreasing rates of cancer, sexually transmitted diseases, obesity, and food insecurity. In addition, the American Academy of Pediatrics (2005) endorses exclusive breastfeeding (i.e., without supplementation) for approximately 6 months after birth and recommends continued breastfeeding with supplementation until the infant is at least 12 months old. Despite the fact that a share of women do not breastfeed, the likelihood of any breastfeeding (i.e., nonexclusive) initially (i.e., in the hospital) and any breastfeeding 6 months after birth has increased since the early 1990s (Abbott Laboratories, 2002). Between 1991 and 2002, breastfeeding rates for all mothers increased 16.8 percentage points (31.5%) at birth and 15.0 percentage points (82.4%) 6 months after birth (Abbott Laboratories, 2002). Although the literature agrees that breastfeeding rates increased between 1991 and 2002 (e.g., Abbott Laboratories, 2002; Ryan, 1997; Ryan, Zhou, & Acosta, 2002; Wright, 2001), explanations for the upward trend have not been empirically investigated. Two papers (Jacknowitz, 2006; Wright, 2001) offer several possible explanations for the rising breastfeeding rates during this period, including changes in the demographic characteristics of mothers, changes to the Special Supplemental Program for Women, Infants, and Children (WIC), changes in the federal government’s support and promotion of breastfeeding, changes in laws and policies that promote breastfeeding, and changes in workplace characteristics that facilitate working and breastfeeding. However, neither author empirically tests these hypotheses. This paper contributes to the breastfeeding literature by studying 1 potential explanation for the upward trend in breastfeeding rates that has not been empirically tested: changes in the demographic characteristics of births. Evidence suggests that the demographic characteristics of births changed between 1991 and 2002. Shifts in the characteristics of births coupled with the fact that some groups of mothers are more likely to breastfeed than others suggests that changes in the demographic characteristics of births could explain the increasing breastfeeding trends. For example, mothers with more education are more likely to breastfeed, and births to mothers with a college education increased 11.2 percentage points or 31.6% between 1991 and 2002 (National Center for Health Statistics [NCHS], 1993, 2003a). To examine whether increases in breastfeeding rates since 1991 can be attributed to many of the well-known demographic
85
changes in births that have been observed, this study decomposes breastfeeding trends using 1991 through 2002 data from the Ross Laboratories Mothers Survey (RMS) and birth certificates. From a policy perspective, it is most efficient to examine the role of changes in the demographic characteristics of births in breastfeeding trends before implementing policy interventions and evaluations. If demographic changes explain a large majority of the increases in breastfeeding rates, then implementing costly programs may not be an efficient use of resources. Although many programs with the goal of increasing breastfeeding have already been established, the results from this study can be used to shape future breastfeeding research. If changes in the demographic characteristics of births explain some of the increases in breastfeeding, then it is important to control for these demographic factors when evaluating the effectiveness of individual breastfeeding promotion programs. In addition, if these changes do not explain all of the increases in breastfeeding rates, further research is necessary to determine which factors do to ensure the continuing increase of breastfeeding rates.
Methods Data The breastfeeding data used in this study are from aggregate tabulations of the RMS, the only dataset that can produce annual national breastfeeding trends dating back to the early 1990s. RMS, a proprietary survey of the Ross Products Division of Abbott Laboratories, is a large, national mail survey conducted since 1955 to determine patterns of milk feeding from birth to 12 months. Questionnaires are mailed to a probability sample of new mothers selected from a sample frame of names that represent approximately 80 – 85% of all national births. The list includes names from hospital records, county records of birth registrations, photography and diaper services, and newspapers. The samples are very large and have increased over time; approximately 720,000 and 1.4 million questionnaires were mailed in 1991 and 2002, respectively. Two papers summarize other breastfeeding data sources and their advantages and disadvantages relative to the RMS (Grummer-Strawn & Li, 2000; Jacknowitz, 2002). Aggregate breastfeeding estimates by maternal characteristic for 1991 through 2002 are from tabulations available in Abbott Laboratories (2001, 2002). Any breastfeeding rates at the following 2 points in time after birth are available to the public: initiation of breastfeeding and breastfeeding at 6 months. The RMS asks mothers to recall the type of milk her baby was fed in the hospital, at 1 week of age, in the last 30 days, and in the last week. In addition, mothers are asked
86
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92 80 70 NMIHS NHANES
60 RLMS
Rate
50 NSFG
NSFG
40
NSFG FDA-IFPS NHANES
30
NMIHS
20 10 0 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 Year
Figure 1. US breastfeeding initiation rates for all mothers, 1970 – 2002. Figure adapted from Jacknowitz (2002). Data are from the following sources: RLMS data are from Abbott Laboratories (2002); NSFG data are from NCHS (1998, 2006); FDA-IFPS data are from Fein and Roe (1998); NHANES data are from Burstein et al. (2000); and NMIHS data are from Visness and Kennedy (1997). NHANES and NSFG breastfeeding initiations rates are for births over multiple years. In this figure, breastfeeding rates are assigned to the middle year in the case of an estimate over an odd number of years, the earlier year in the case of a 2-year estimate, and the second year in the case of a 4-year estimate.
about their demographic characteristics. Breastfeeding rates stratified by the following demographic characteristics are available: maternal age, maternal education, maternal employment, race/ethnicity, parity, WIC participation, and Census division of residence. The RMS asks whether the infant was low birthweight, but the information is not included in the tabulations in a usable format. Tabulations only include breastfeeding rates for low birthweight infants and all infants (including low birthweight babies); therefore, breastfeeding rates cannot be calculated for non–low-birthweight babies. Figure 1 and other studies illustrate that the annual breastfeeding rates produced with the RMS are comparable to those from other sporadically collected data. Figure 1 shows breastfeeding rates from the RMS, National Health and Nutrition Examination Survey (NHANES), National Maternal and Infant Health Survey (NMIHS), National Survey of Family Growth (NSFG), and the Food and Drug Administration’s Infant Feeding Practices Survey (FDA-IFPS). These comparison data sources are nationally representative, with approximate response rates ⬎70% during this time period. Four of these 5 datasets provide similar estimates of breastfeeding initiation for all mothers, with the only exception of FDA-IFPS. Comparisons of the FDA-IFPS with the NMIHS data indicate that women in the FDA-IFPS sample were more likely to be in middle- and upper-income groups, older, married, and White than the NMIHS sample (Fein & Roe, 1998). These characteristics are associated with higher breastfeeding rates, which is consistent with the higher estimates generated using the FDA-
IFPS. Studies (Hediger, Overpeck, Kuczmarski, & Ruan, 2001; Ryan et al., 1991) that compared breastfeeding estimates from the RMS to those generated using the NHANES and NSFG also found similar breastfeeding rates among the datasets. In addition to breastfeeding statistics, data on the composition of births from 1991–2002 necessary to perform the analysis are from the National Center for Health Statistics’ National Vital Statistics Reports series, which was previously titled the Monthly Vital Statistics Reports series. The National Vital Statistics Reports rely on birth data extracted from 100% of birth certificates registered in the 50 states and the District of Columbia. Because these data are a near census of births for a given year (a small fraction of births are unreported) and include a variety of maternal characteristics that are available in the RMS tabulations using the same definitions as those used by the RMS, they are ideal to use for this study. These data are a near census owing to a small number of unreported births. For example, in 2001, an estimated 99% of births in the United States were reported (NCHS, 2003b). Analytic Approach Standardization and decomposition techniques similar to those described by Das Gupta (1993) and Kitagawa (1955) are used to determine the share of the increase in breastfeeding rates between 1991 and 2002 that are explained by changes in the demographic characteristics of births. For the purposes of this study, these descriptive techniques compare the actual breastfeeding rate with an estimated breastfeeding rate for a given demographic characteristic, which represents what the breastfeeding rate would have been if breastfeeding practices remained at their 1991 values but birth rates by demographic characteristic changed. The estimated breastfeeding rate is calculated holding breastfeeding rates by subgroup constant at their 1991 values and allowing the share of births for a given subgroup to change (equation 1). BFt,S ⫽
N
兺 bf1991,s ⴱ brt,s
(1)
s⫽1
In equation 1, BFt,S is the estimated breastfeeding rate which allows the share of births by demographic characteristic, S, to vary to their values in year t; bf1991,s is the actual breastfeeding rate in 1991 for subgroup s; brt,s is the proportion of births born to mothers of subgroup s in year t; and N is the number of subgroups for the particular demographic characteristic. For example, parity is divided into the subgroups of first births and higher order births. To calculate the 2002 estimated breastfeeding rate the following 2 terms would be aggregated: 1) 1991 breastfeeding rate for first births multiplied by 2002 share of births that are first births and 2) 1991 breastfeeding rate for
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
higher order births multiplied by 2002 share of births that are higher order births. To determine the share, YS, of the increase in breastfeeding rates between 1991 and 2002 that is explained by changes in births by a given demographic characteristic, the difference between the actual aggregate 1991 breastfeeding rate and the 2002 estimated breastfeeding rate is divided by the change in the actual aggregate breastfeeding rates between 1991 and 2002. The calculation of YS for a given demographic characteristic, S, is illustrated in equation 2. YS ⫽
BF2002,S ⫺ bf1991 bf2002 ⫺ bf1991
(2)
In equation 2, bf2002 is the actual aggregate breastfeeding rate in 2002 and bf1991 is the actual aggregate breastfeeding rate in 1991. Their difference is the change in breastfeeding rates between 1991 and 2002. For breastfeeding initiation, this difference is 16.8 percentage points; for breastfeeding at 6 months after birth, it is 15.0 percentage points. If demographic changes explain some of the increase in breastfeeding rates between 1991 and 2002, then the share, YS, would be positive. If the demographic changes explain none of the increase in breastfeeding rates over this time period, then the share, YS, would be close to zero. Finally, if the changes that occurred led to lower overall breastfeeding rates, than the share, YS, would be negative. Demographic Characteristics The demographic characteristics available in the RMS were included in the analysis if 3 criteria were met. First, the demographic characteristics were available in both the RMS and birth certificate data and were in a usable form, which meant that the demographic characteristic was defined in a similar manner in both datasets and information for all subgroups was available. Second, there were sizeable differences in breastfeeding rates between mothers by this characteristic. If differences in breastfeeding by this factor do not exist, then changing the composition of births will not change the estimated 2002 breastfeeding rate derived in equation 1. Third, the composition of births changed by this characteristic between 1991 and 2002. If the composition of births has remained the same then the estimated 2002 breastfeeding rate will be equal to the actual breastfeeding rate in 1991, thus explaining none of the increase in breastfeeding rates. The demographic characteristics available in both datasets in a usable form include maternal age, maternal education, race/ethnicity, parity, and Census division of residence (hereafter referred to as geographic location of birth). Maternal employment and WIC participation are not available in the birth certif-
87
icate data and are therefore excluded from the analysis. Questions on maternal WIC participation were added to the live birth certificate questionnaire in 2003, which is after the time period of interest. To address whether differences exist in breastfeeding rates between women by a given characteristic, this paper turns to breastfeeding trends and the breastfeeding literature. Jacknowitz (2002) identifies breastfeeding correlates by graphing trends in breastfeeding practices by subgroup and conducting a literature review. Studies included in the literature review use multivariate regression techniques to predict correlates of breastfeeding practices. Jacknowitz (2002) finds that of those demographic characteristics available in the RMS, maternal age, maternal education, race/ethnicity, and geographic location of birth are strong correlates of breastfeeding, but parity is not during the time period of interest. Table 1 illustrates findings for demographic characteristics that are related to breastfeeding for 1991 and 2002. Younger mothers are less likely to breastfeed than older mothers. Less educated mothers are less likely to breastfeed than more educated mothers. Non-Hispanic Black mothers are less likely than mothers of other race/ ethnicities to breastfeed. Mothers residing in the Western states (Mountain and Pacific Census regions) have higher breastfeeding rates than other regions, whereas mothers living in the East South Central division have considerably lower breastfeeding rates than all other regions. The third criterion for determining which characteristics to decompose is whether the birth composition by these variables changed. Table 2 illustrates that large changes in the shares of births occurred by maternal age, maternal education, and race/ethnicity. Between 1991 and 2002, births to women ages ⱖ30 increased 6.6 percentage points (21.3%) and births to women with a college education increased 11.2 percentage points (31.6%). Births to Hispanic mothers, who are more likely to breastfeed than non-Hispanic Black mothers, increased 6.6 percentage points. At the same time births to non-Hispanic White mothers, who are also more likely to breastfeed than non-Hispanic Black mothers, declined by 5.9 percentage points. Although race/ethnicity is a strong correlate of breastfeeding, these changes considered together suggest that compositional changes in the race/ethnicity of births may not influence overall breastfeeding rates because, although race/ethnicity is associated with breastfeeding rates, these 2 effects may cancel each other out. The 1 demographic variable that did not undergo compositional changes in births and is excluded from the analysis is geographic location of birth. Hence, this exercise is performed for all years between 1991 and 2002 for breastfeeding initiation and duration for maternal age, maternal education, and
88
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
Table 1. US Breastfeeding Rates by Selected Characteristics, 1991 and 2002 Breastfeeding Rate in Hospital
Maternal age (yrs) ⬍20 20–24 25–29 30–34 ⱖ35 Maternal education Grade school High school Non-college College Unknown Race/ethnicity Hispanic Non-Hispanic, Black Non-Hispanic, White Other Census division New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific
Breastfeeding Rate 6 Months After Birth
1991
2002
Change
1991
2002
Change
32.2 45.7 57.8 65.2 66.7
56.2 66.0 73.4 76.4 74.1
24.0 20.3 15.6 11.2 7.4
5.9 11.4 19.7 27.3 32.9
16.7 25.4 35.7 42.3 43.1
10.8 14.0 16.0 15.0 10.2
35.2 42.9 42.6 70.2 N/A
55.1 60.7 60.5 81.2 N/A
19.9 17.8 17.9 11.0
12.6 12.2 12.2 27.7 N/A
27.1 23.4 23.5 44.6 N/A
14.5 11.2 11.3 16.9
51.8 25.8 59.2 N/A
70.7 53.9 73.4 N/A
18.9 28.1 14.2
14.8 6.6 21.0 N/A
32.7 19.2 36.0 N/A
17.9 12.6 15.0
55.9 48.3 48.8 56.4 46.4 37.5 48.4 70.0 69.5
73.3 65.5 66.7 73.1 67.5 57.0 64.9 81.0 81.5
17.4 17.2 17.9 16.7 21.1 19.5 16.5 11.0 12.0
19.7 16.8 16.6 18.9 14.5 10.7 14.0 28.0 25.7
37.0 33.2 28.7 35.2 31.2 22.0 27.3 40.5 43.4
17.3 16.4 12.1 16.3 16.7 11.3 13.3 12.5 17.7
Notes: Data are from Abbott Laboratories (2001, 2002). Breastfeeding rates are unavailable for mothers with unknown education or other race/ethnicity.
race/ethnicity. The subgroups of maternal age at birth include ⬍20 years, 20 –24 years, 25–29 years, 30 –34 years, and ⱖ35 years. These groups and those for maternal education and race/ethnicity are determined by the RMS tabulations. Maternal education subgroups include grade school, high school, non-college, college, and unknown. Finally, the subgroups for race/ethnicity are Hispanic, non-Hispanic Black, nonHispanic White, and other. In addition, decompositions are performed for both breastfeeding initiation and 6 months after birth because differences in breastfeeding rates between subgroups may differ for the 2 outcomes (Table 1). For example, there are larger differences in breastfeeding initiation between more educated and less educated mothers than in breastfeeding 6 months after birth. Mothers with unknown education and other race/ ethnicity are assigned the mean breastfeeding rate for all mothers. To test whether this is a reasonable strategy, 1991 and 2002 actual breastfeeding rates in the hospital and 6 months after birth are compared with ones that weight each subgroup’s breastfeeding rate by the proportion of births to that given subgroup of a demographic variable. If assigning the other and unknown subgroups the mean breastfeeding rate for all mothers is a reasonable strategy, we would expect
the actual and weighted breastfeeding rates to be very close. Indeed, the estimated breastfeeding rates (1991 and 2002 initial and 6 months after birth by maternal education and race/ethnicity) are within 1 percentage point of those produced by the RMS. Microsoft Excel is used to organize the data and perform the analysis.
Results Figures 2 and 3 graph the actual breastfeeding rates and the estimated breastfeeding rates by year in the hospital and 6 months after birth, respectively. If changes in birth rates by a given demographic characteristic explains any of the increasing breastfeeding rate, we would expect to see the estimated breastfeeding rates for that demographic characteristic increase over time. Therefore, the closer the estimated breastfeeding rates for a given characteristic and the actual breastfeeding rate, the larger share of the increase in breastfeeding rates that changes in the composition of births for that characteristic explains. Both figures show that the estimated breastfeeding rates for maternal education and age increase slightly and gradually over time, suggesting that changes in the composition
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92 Table 2. Composition of US Births by Selected Characteristics, 1991 and 2002
80 70 60
Change
12.9 26.5 29.7 21.5 9.4
10.8 25.4 26.4 23.7 13.8
⫺2.2 ⫺1.1 ⫺3.3 2.1 4.4
6.1 16.5 34.9 35.5 7.1
6.0 15.3 30.7 46.7 1.3
⫺0.1 ⫺1.2 ⫺4.2 11.2 ⫺5.7
15.2 16.2 63.0 5.6
21.8 14.4 57.1 6.7
6.6 ⫺1.8 ⫺5.9 1.1
4.7 14.2 16.2 6.5 16.7 5.7 11.5 5.9 18.6 4,110,907
4.2 12.7 15.2 6.6 18.0 5.8 13.1 7.6 16.9 4,021,726
⫺0.5 ⫺1.5 ⫺1.0 0.1 1.3 0.1 1.5 1.7 ⫺1.6
Notes: The 1991 birth data are from NCHS (1993) and the 2002 birth data are from NCHS (2003a). The 1991 unknown education category is primarily comprised of births from Washington and New York states (excluding New York City), which did not collect education information; 16,341 of the 1991 other race/ethnicity births are from New Hampshire, which did not collect race/ethnicity data.
of births by these 2 characteristics explain some of the increasing trends in initial breastfeeding rates and breastfeeding rates 6 months after birth. In contrast, the estimated breastfeeding rates for race/ethnicity remain fairly constant over time, suggesting that changes in the composition of births by race/ethnicity do not explain the increases in breastfeeding over this time period. Table 3 takes the analysis further and calculates the share of the increase in breastfeeding rates between 1991 and 2002 that changes in the composition of births explain. Table 3 shows that changes in the share of births by maternal age explain 9.8% of the increase in initial breastfeeding rates. The change in the share of births by maternal education explains 11.5% of the increasing initial breastfeeding rates. As expected, changes in shares of births to mothers of different race/ethnicity explain none of the increase in in-hospital breastfeeding rates since 1991. In fact, applying the 2002 composition of births by race/ ethnicity to the appropriate 1991 breastfeeding rates
50
Actual
40
Maternal age
30
Maternal education
20
Race/ethnicity
10 0 1991
1993
1995 1997 Year
1999
2001
Figure 2. Decomposition of US initiation breastfeeding rates, 1991– 2002. Maternal age, maternal education, and race/ethnicity breastfeeding rates are calculated by weighting the 1991 breastfeeding rate by the year’s share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
results in a lower overall initial breastfeeding rate, implying that initial breastfeeding rates would have been higher if compositional changes in births by race/ethnicity had not occurred. Similar to the findings for the initiation of breastfeeding, Table 3 illustrates that changes in the composition of births explain some of the increase in breastfeeding rates 6 months after birth since 1991. The shift of births by maternal age explains 10.2% of the increase in breastfeeding rates 6 months after birth. The changes in the birth shares by maternal education explain 9.0% of the increasing breastfeeding rates 6 months after birth. This statistic is approximately 2.5 percentage points lower than the number for the initiation of breastfeeding rates. This is because differences in breastfeeding rates between the most and least educated mothers are larger for initiation (on average 30.2 percentage points) than duration (on average 16.5 percentage points). Finally, changes in the racial/ethnic composition of births account for 35 30 25 Rate
Maternal age (yrs) ⬍20 20–24 25–29 30–34 ⱖ35 Maternal education Grade school High school Non-college College Unknown Race/ethnicity Hispanic Non-Hispanic, Black Non-Hispanic, White Other Census division New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific Total births
2002
Rate
Percentage of Births 1991
89
Actual
20
Maternal age 15
Maternal education
10
Race/ethnicity
5 0 1991
1993
1995 1997 Year
1999
2001
Figure 3. Decomposition of US breastfeeding rates 6 months after birth, 1991–2002. Maternal age, maternal education, and race/ ethnicity breastfeeding rates are calculated by weighting the 1991 breastfeeding rate by the year’s share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
90
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
Table 3. Decomposition of US Breastfeeding Rates Breastfeeding Rate
Actual 1991 breastfeeding rate Actual 2002 breastfeeding rate Change in breastfeeding rate Maternal age (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained Maternal education (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained Race/ethnicity (1) Estimated 2002 breastfeeding rate (2) Difference between actual 1991 and estimated 2002 breastfeeding rates (3) Percent of breastfeeding increase explained
In Hospital
Six Months After Birth
53.3 70.1 16.8
18.2 33.2 15.0
54.9 1.6 9.8
19.7 1.5 10.2
55.2 1.9 11.5
19.5 1.3 9.0
52.4 ⫺0.9 ⫺5.4
17.4 ⫺0.8 ⫺5.4
Notes: Estimated 2002 breastfeeding rate is calculated by weighting the 1991 breastfeeding rate by the 2002 share of births to each demographic subgroup. Mothers with unknown education are assigned the breastfeeding rate for all mothers. Mothers with other race/ethnicity are also assigned the breastfeeding rate for all mothers.
essentially none of the increasing breastfeeding rates. Similar to the case with breastfeeding initiation, applying the 2002 composition of births by race/ethnicity to the appropriate 1991 breastfeeding rates results in a lower overall breastfeeding rate 6 months after birth.
Discussion This paper examines whether increases in breastfeeding rates since 1991 can be attributed to changes in key demographic characteristics of births. To answer the research question, this study decomposes breastfeeding trends using 1991 through 2002 data from the RMS and birth certificate data. Findings suggest that changes in the composition of births with respect to maternal age and education explain up to approximately 20% of the upward trend in initial breastfeeding rates and breastfeeding rates 6 months after birth. Changing birth compositions by maternal age and education explain 9.8% and 11.5% of the increase in breastfeeding initiation rates, respectively. Changing birth compositions by maternal age and education explain 10.2% and 9.0% of increasing breastfeeding rates 6 months after birth, respectively. Geographic location of birth and parity do not explain any of the increase in breastfeeding rates. Changes in the composition of births by race/ ethnicity lead to lower breastfeeding rates in the hospital and 6 months after birth. Although there is an increase in the percent of births to Hispanic women, the breastfeeding rate decreases because this increase is offset by a decrease in the percent of births to non-Hispanic White women who have higher breastfeeding rates than Hispanic women. If the increase in
the percent of births to Hispanic women had been accompanied by a decrease in the percent of births to non-Hispanic Black women, the overall breastfeeding rate would have increased. Limitations of this study exist, which are primarily data driven. First, the variables available in the RMS and birth certificate data include the majority of key demographic variables, but are restricted. The birth certificate data do not contain information on WIC participation or maternal employment. In addition, the aggregate RMS data do not stratify breastfeeding rates after breastfeeding by birthweight status. However, none of these variables are likely to explain the increase in breastfeeding trends over the time period of interest. WIC participation among eligible infants has increased (United States Department of Agriculture, 2006); however, evidence suggests that WIC participation is negatively related to breastfeeding practices (Bitler & Currie, 2005; Ryan & Zhou, 2006). Further, maternal employment rates among women with young children ⬍3 have increased (United States Department of Labor, 2006) and studies indicate that maternal employment is negatively related to breastfeeding (Jacknowitz, 2002). The research on the relationship between low birthweight status and breastfeeding is mixed between detecting no relationship between the 2 variables and a negative relationship (e.g., Chatterji & Brooks-Gunn, 2004; Ryan & Zhou, 2006). However, even if the relationship were negative, with an increasing percent of low-birthweight infants (NCHS, 2003a), this demographic characteristic would not explain the increase in breastfeeding rates. The second primary limitation of the study is that the aggregate nature of the data does not allow for
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
methods that account for the associations between the independent variables such as changing the demographic characteristics sequentially or using multivariate regression analysis. For example, maternal age at birth and maternal education at birth are likely correlated as more educated women tend to have children later in life. Hence, the estimate of 20% of the increase in breastfeeding explained by maternal education and age is overestimated as a result of this positive correlation. A final limitation is that, although breastfeeding rates from the RMS are comparable to those from other nationally representative datasets, they represent 80 – 85% of births or fewer depending on the response rate for the year. Thus, the breastfeeding rates are not completely comparable to the birth data, which represent a near census of US births. Although there are limitations, this study does contribute to the literature by starting to investigate the factors that explain recent trends in breastfeeding practices. Given the health benefits of breastfeeding to both mothers and children, it is important to identify and understand the factors explaining these increasing breastfeeding rates to ensure their continued growth. Although this study does explain a sizeable share of the increase in breastfeeding rates, it is important to continue investigating which additional factors explain recent breastfeeding increases. Plausible factors to investigate include changes in laws and policies, health promotion, the WIC Program, employer support, and technological innovation. A final promising explanation for the increasing breastfeeding trends is changes in attitudes toward breastfeeding practices. Given that changes in the demographic composition of births do not explain the majority of the increase in breastfeeding trends, it is worthwhile to pursue and evaluate current and additional programs to improve breastfeeding. However, the findings do highlight the importance of controlling for demographic variables when examining the effects of programs to improve breastfeeding rates after breastfeeding. Finally, the study does raise the issue of the importance of considering whether demographic changes influence health behaviors before considering new policies and programs. If demographic changes do explain a majority of trends in a health behavior, the implementation of an intervention may not be an efficient use of resources.
Acknowledgments I thank Julie DaVanzo, Steven Haider, Rebecca Kilburn, Elizabeth Peters, and Robert Schoeni for helpful comments; Steven Putansu provided excellent research assistance. Financial support from the RAND Graduate School Hagopian Dissertation Award is gratefully acknowledged. This work reflects only the opinions of the author.
91
References Abbott Laboratories. (2001). Breastfeeding trends through 2001. Columbus, OH: Ross Products Division. Abbott Laboratories. (2002). Breastfeeding trends—2002. Columbus, OH: Ross Products Division. American Academy of Pediatrics. (2005). Breastfeeding and the use of human milk. Pediatrics, 115, 496 –506. Bitler, M. P., & Currie, J. (2005). Does WIC work? The effects of WIC on pregnancy and birth outcomes. Journal of Policy Analysis and Management, 24, 73–91. Burstein, N. R., Fox, M. K., Hiller, J. B., Kornfeld, R., Lam, K., Price, C., et al. (2000). WIC general analysis project: Profile of WIC children. Cambridge, MA: Abt Associates. Chatterji, P., & Brooks-Gunn, J. (2004). WIC participation, breastfeeding practices, and well-child care among unmarried, lowincome mothers. American Journal of Public Health, 94, 1324 –1327. Das Gupta, P. (1993). Standardization and decomposition of rates: A user’s manual (Current Population Reports Series, P23-186). Washington, DC: US Department of Commerce, US Bureau of the Census. Fein, S. B., & Roe, B. (1998). The effect of work status on initiation and duration of breast-feeding. American Journal of Public Health, 88, 1042–1046. Grummer-Strawn, L. M., & Li, R. (2000). US national surveillance of breastfeeding behavior. Journal of Human Lactation, 16, 283–290. Hediger, M. L., Overpeck, M. D., Kuczmarski, R. J., & Ruan, W. J. (2001). Association between infant breastfeeding and overweight in young children. The Journal of the American Medical Association, 285, 2453–2460. Jacknowitz, A. (2002). Estimating eligibility for WIC: The role of breastfeeding [working paper]. Santa Monica, CA: RAND Graduate School. Jacknowitz, A. (2006). Understanding gains in breastfeeding rates: Promising policies and programs. Women’s Health Issues, 16, 101–103. Kitagawa, E. M. (1955). Components of a difference between two rates. Journal of the American Statistical Association, 50, 1168 –1194. Labbok, M. H. (2001). Effects of breastfeeding on the mother. Pediatric Clinics of North America, 48, 143–158. Leon-Cava, N., Lutter, C., Ross, J., & Martin, L. (2002). Quantifying the benefits of breastfeeding: A summary of the evidence. Washington, DC: Pan American Health Organization. National Center for Health Statistics (NCHS). (1993). Advance report of final natality statistics, 1991 (Monthly Vital Statistics Reports, Vol. 42, No. 3). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (1998). Health, United States, 1998 with socioeconomic status and health chartbook (US DHHS Publication No. PHS 98-1232). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2003a). Births: Final data for 2002 (National Vital Statistics Reports, Vol. 52, No. 10). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2003b). Vital statistics of the United States, 2001. Volume 1: Natality technical appendix. Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. National Center for Health Statistics (NCHS). (2006). Health, United States, 2006 with chartbook on trends in the health of Americans (US DHHS Publication No. PHS 2006-1232). Hyattsville, MD: US Department of Health and Human Services, National Center for Health Statistics. Ryan, A. S. (1997). The resurgence of breastfeeding in the United States. Pediatrics, 99, 12. Ryan, A. S., Pratt, W. F., Wysong, J. L., Lewandowski, G., McNally, J. W., & Krieger, F. W. (1991). A comparison of breast-feeding
92
A. Jacknowitz / Women’s Health Issues 17 (2007) 84 –92
data from the National Surveys of Family Growth and the Ross Laboratories Mothers Survey. American Journal of Public Health, 81, 1049 –1052. Ryan A. S., & Zhou, W. (2006). Lower breastfeeding rates persist among the Special Supplemental Nutrition Program for Women, Infants, and Children participants, 1978 –2003. Pediatrics, 117, 1136 –1146. Ryan, A. S., Zhou, W., & Acosta, A. (2002). Breastfeeding continues to increase into the new millennium. Pediatrics, 110, 1103–1109. United States Department of Agriculture. (2006). WIC program coverage: How many eligible individuals participated in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC): 1994 to 2003? Alexandria, VA: US Department of Agriculture, Food and Nutrition Service. United States Department of Health and Human Services (US DHHS). (2000). HHS blueprint for action on breastfeeding. Washington, DC: US Department of Health and Human Services, Office of Women’s Health. United States Department of Labor. (2006). Women in the labor force: A databook (Report 996). Washington, DC: US Department of Labor, Bureau of Labor Statistics.
Visness, C. M., & Kennedy, K. I. (1997). Maternal employment and breast-feeding: Findings from the 1988 National Maternal and Infant Health Survey. American Journal of Public Health, 87, 945–950. Wright, A. L. (2001). The rise of breastfeeding in the United States. Pediatric Clinics of North America, 48, 1–12.
Author Description Alison Jacknowitz, PhD, MPP, is an Assistant Professor of Public Administration and Policy at American University. Her interests include child and family policy, child well-being, poverty, and public assistance programs. She has written several papers on the topic of breastfeeding including examining the effects of welfare work requirements and workplace characteristics on breastfeeding decisions.