Does Greater Exposure to WIC Affect Maternal Behavior and Improve Infant Health Evidence from the Pr

1
Does Greater Exposure to WIC Affect Maternal
Behavior and Improve Infant Health? Evidence from
the Pregnancy Nutrition Surveillance System.

• Ted Joyce
• Cristina Yunzal
• Supported by a grant from the Institute for
  Research on Poverty and the United States
  Department of Agriculture (IRP-USDA)
• Preliminary results. Not for citation.

2

• WIC works, perhaps better than any other
  government program in history.
• USDA Secretary Dan Glickman, 1997
• The WIC Program results in significant Medicaid
  savings that far outweigh the programs costs by
  a ratio of 3 to 1.
• - HHS Secretary Louis Sullivan, 1995

3
What is WIC?

• The Special Supplemental Nutrition Program for
  Women, Infants, and Children
• Piloted by Congress in 1972, permanently
  authorized in 1974
• Sources http//www.fns.usda.gov/wic/aboutwic/wica
  taglance.htm
• http//www.nutritionnc.com/wic/his
  tory.htm

4
Target Population

• Low-income, nutritionally at risk
• Pregnant women (through pregnancy and up to 6
  weeks after birth or after pregnancy ends).
• Breastfeeding women (up to infants 1st
  birthday) 
• Nonbreastfeeding postpartum women (up to 6 months
  after the birth of an infant or after pregnancy
  ends) 
• Infants (up to 1st birthday). WIC serves 45
  percent of all infants born in the United
  States. 
• Children up to their 5th birthday.
• Source http//www.fns.usda.gov/wic/aboutwic/wicat
  aglance.htm

5
WIC Benefits

• Nutritional supplementation via checks/vouchers
  to purchase specific foods that are high in
  protein, calcium, iron, and vitamins A and C
• Nutrition education and counseling 
• Screening and referrals to other health, welfare
  and social services
• Source http//www.fns.usda.gov/wic/aboutwic/wicat
  aglance.htm

6
Funding and participation

• Overseen by the USDA and administered by 90 WIC
  state agencies.
• Not an entitlement program Congress authorizes a
  specific amount of funds each year for it.
• However, pregnant women receive highest admission
  priority
• In North Carolina, all eligible applicants from
  1995 onwards were admitted into the program
• 2005 funding 5.16 billion
• 2005 participation 8 million, from 88,000 in
  1974
• Source http//www.fns.usda.gov/wic/aboutwic/wicat
  aglance.htm

7
Study objective

• Estimate the impact of North Carolinas prenatal
  WIC program on singleton birth outcomes in
  1996-2003

8
Why focus on birth weight?

• LBW significant factor in infant mortality and
  morbidity
• In 2002, infant mortality in U.S. was 7 per 1,000
  live births
• Among low birth weight infants (lt2,500 grams),
  rate was 60.3 per 1,000 live births
• Birth weight well-documented on birth certificates

9
Low birth weight trends, selected years
10
Causes of low birth weight

• 1. Short gestational period (born preterm)
• Preterm infants are those born before 37 weeks of
  gestation
• 2. Intrauterine growth retardation (IUGR)
• Infants who have low birth weights for their
  gestational age

11
WIC Works Literature General Accounting
Office (GAO) Report (1992)

• Combined results of 17 existing WIC studies to
  estimate WICs effect on reducing incidence of
  low and very low birth weight (LBW is lt2500
  grams VLBW is lt1500 grams).
• Prenatal WIC participation reduced LBW by 25 and
  VLBW by 44
• Each dollar spent on WIC for pregnant women saved
  3.50 in medical disability costs

12
WIC Works Literature Devaney, Bilheimer, and
Schore (1992)
13
WIC Works Literature Gordon Nelson (1995)
14
WIC Works Literature Bitler Currie (2005)
15
WIC Works Criticism Besharov Germanis (2001)

• Review existing WIC studies and conclude that
  findings on birth outcomes are unreliable. Most
  research poorly designed and suffer from
  selection bias, simultaneity bias, and lack of
  generalizeability.
• Argue that the program has suffered from
  eligibility creep. Applicants must be income
  eligible and be at nutrition risk.
• Suggest that nutritional risk is assumed if
  income criteria are met. States have used
  generous cut-off points and loosely defined risk
  criteria.

16
WIC Works Criticism GAO (2001)

• Study of six local WIC agencies found that
  individual nutrition education sessions did not
  last long, ranging from 4-17 minutes.
• Only two nutrition education contacts required
  per 6-month certification period.
• It is difficult to help prevent numerous
  nutrition-related diseases with a few brief
  nutrition education sessions.

17
WIC Works Criticism Joyce, Gibson, Colman
(2005)
18
Clinical literatureRush, Stein Susser (1980)

• Randomized controlled trial of women attending
  clinic in a large black community in NYC
  (1970-73).
• Target population selected on grounds that the
  lower the expected birth weight, the more likely
  a better diet would help
• Women lt 140 lbs at conception who fulfilled at
  least one of the following
• Low prepregnant weight (lt110 lbs) at conception
• Low weight gain up to recruitment
• At least one previous infant of low birth weight
• Protein intake lt50 grams in 24 hours before
  registration (24-hour recall)

19
Clinical literatureRush, Stein Susser (1980),
cont.

• Study groups
• Supplement Two 8-oz cans of beverage daily (40
  grams of protein, 470 calories)
• Complement Two 8-oz cans of beverage daily (6
  grams of protein, 322 calories)
• Control Regular clinic care
• Results
• Significant effects on weight gain limited to
  those recruited early in pregnancy (lt15 weeks)
• However, no impact on birth weight for this
  cohort
• Among full-term infants, mean birth weight not
  significantly related to treatment in any cohort
• Among preterm infants, mean birth weight lower in
  Supplement than in Complement in 3 of 4 cohorts,
  lower than Control

20
Clinical literatureMetcoff, et al (1985)

• Impact of WIC supplement on birth weight tested
  in a randomized, controlled trial involving a
  population of women in midpregnancy at high risk
  of having low birth weight infants.
• No significant effect on birth weight when
  midpregnancy weight added as a covariate
• However, significant effect on smokers ( gt 10
  cigarettes/day)

21
Clinical literature

• Goldenberg Rouse (NEJM, 1998)
• Most interventions designed to prevent preterm
  birth do not work, and the few that do are not
  universally effective and are applicable only to
  a small percentage of women at risk for preterm
  birth.
• Ancel (EJOG, 2004)
• Etiology of preterm delivery remains poorly
  understood.
• little support for the efficacy of nutritional
  supplements, more intense prenatal monitoring,
  and/or social support and limitation of risky
  behaviour in reducing preterm deliveries.

22
Expectations given the literature

• For a program to impact birth weight, must either
  lengthen gestation or prevent fetal growth
  retardation, or both.
• Little evidence in clinical literature to
  indicate that supplemental and referral programs
  like WIC can lengthen gestation. Any impact of
  WIC on gestation may be spurious and suggestive
  of omitted variable bias.
• Impact on fetal growth is more plausible through
  nutritional supplementation-gtmaternal weight gain

23
Our Data

• North Carolina Pregnancy Nutrition Surveillance
  System (PNSS), 1996-2003
• Dataset combines rich administrative data from
  the WIC program linked to birth certificates
• More than 400,000 women in the database
• Size of sample allows us to stratify by race,
  timing of prenatal care, etc. we can estimate
  fully interacted models
• All women in database are eligible for WIC and
  have chosen to participate
• Can compare outcomes for women who enrolled while
  pregnant with women who did not join WIC until
  the postpartum period

24
WIC Eligibility

• Income lt 185 percent of U.S. Poverty Income
  Guidelines. (Food Stamp/Medicaid/Temporary
  Assistance for Needy Families recipients
  adjunctively eligible.)
• State residency
• Applicants determined to be at nutrition risk
  by a health professional.
• Medically-based risks such as anemia,
  underweight, overweight, history of pregnancy
  complications, or poor pregnancy outcomes.
• Dietary risks, such as failure to meet the
  dietary guidelines or inappropriate nutrition
  practices
• Source http//www.fns.usda.gov/wic/aboutwic/wicat
  aglance.htm

25

• Selected Characteristics, North Carolina WIC
  Participantsa with Singleton Births, 1996-2003
• aExcluding prenatal women who do not return for a
  follow-up visit

26
Race/Ethnicity
27
Mothers age
28
Education (Mothers gt 20 years old)
29
Prenatal care use
30
Parity
31
Program Evaluation

• Typical model
• Hit a0 a1WICit Xitß S?tYrt eit,
• where Hit infant/maternal outcome
• WICit dummy variable indicating WIC
  participation
• Xit matrix of other relevant
  variables, including demographic characteristics
  of the mother
• Yrt year indicator
• ei error term, E(ei)0
• a1 average treatment effect of WIC on
  health outcome

32
Selection problem

• Decision to participate in WIC likely based on
  factors that also affect Hit.
• Women who enlist in WIC may be healthier and/or
  more motivated to have healthy pregnancies a1
  may overstate impact of WIC
• Women may enroll in WIC because they are less
  healthy and desire assistance a1 may
  underestimate impact of WIC
• Even after controlling for variables that
  simultaneously impact Hit and WICit, cannot
  guarantee that participants and nonparticipants
  are similar on unobservables.
• E(eitWICit) ? 0 and the estimator identifies
  (sans other regressors)
• E(?1) a1 E(eitWICit1) E(eitWICit0)

33
Addressing selection bias

• Sample consists only of WIC women, differentiated
  by trimester of pregnancy when they enrolled.
  Comparison group is women who only enlist after
  giving birth.
• In lieu of usual WIC vs. non-WIC comparison, all
  women in sample know of and have chosen to
  participate in WIC
• Focus on singleton births we exclude prenatal
  enrollees who do not return for a WIC postpartum
  visit
• Use large sample to stratify into more
  homogeneous subgroups

34
Birth outcomes Means
35
Birth outcomes Coefficients on WIC
36
Birth outcomes Coefficients on WIC
37
Birth outcomes Coefficients on WIC
38
Comments on prenatal WIC coefficients

• WIC coefficient for birth weight unadjusted for
  gestation (87 grams) fall in range of findings
  by Devaney, et al (51-117). Larger than Gordon
  Nelson (67.9), and Bitler Currie (63.65).
• Marginal probit effect on LBW (-3.2) similar to
  Devaney (-5.1 to 2.2) and slightly larger than
  Bitler Curries (OR 0.726 LBW incidence in
  our sample is 9.1).
• Marginal probit effect on preterm birth (-3.9)
  also similar to Devaney (-6.3 to 2.3) and
  larger than Bitler Currie (OR 0.708 preterm
  incidence in our sample is 10.6).

39
Our model

• Hit a0 a1WIC1it a2WIC2it a3WIC3it Xitß
  S?tYrt eit,
• where
• WIC1it first-trimester WIC enrollment
• WIC2it second-trimester WIC enrollment
• WIC3it third-trimester WIC enrollment

40
Our model, cont.

• We expect that the earlier a woman enlists in
  WIC, the more favorable the outcome a1gta2gta3
• Following the clinical literature, we assume that
  WIC cannot affect preterm birth/length of
  gestation. Look for WIC impact that is primarily
  driven by gestational age.

41
Birth outcomes Birth weight (grams)
42
Birth outcomes Birth weight, adjusted for
gestation
43
Birth outcomes Length of gestation (weeks)
44
Birth outcomes Low birth weight (lt2500 grams)
45
Birth outcomes Small for gestational age
46
Birth outcomes Preterm birth (lt37 weeks)
47
Birth outcomes Term low birth weight (gt37
weeks lt2500 grams)
48
Comments on WIC trimester coefficients

• WIC coefficients for birth outcomes that are
  unadjusted for gestational age have relative
  magnitudes a1lta2lta3. This suggests that WICs
  impact is partly driven by the already lengthier
  gestations of third-trimester enrollees.
• Coefficients for birth outcomes adjusted for
  gestational age (except term LBW) have the
  expected relative magnitudes a1gta2gta3. For blacks
  and Hispanics, significant results limited to 1st
  and/or 2nd trimester.

49
Maternal Inputs

• Smoking cessation
• Weight gain
• Breastfeeding

50
Smoking during pregnancy(Birth certificates)
51
Variables for generating smoking indicators
52
Variables for generating smoking indicators
53
Variables for generating smoking indicators
54
Maternal smoking rates by race/ethnicity
(all questions)
55
Maternal smoking rates by WIC Participation
56
Smoking quit rates, by WIC participation
57
Smoking quit rates, by WIC participation
58
Maternal Inputs Smoking cessation during
pregnancy (all questions prenatal women only)
59
Maternal Inputs Pregnancy weight gain
60
Maternal Inputs Breastfeeding initiation
61
Maternal Inputs Breastfeeding initiation, four
weeks after birth
62
Birth outcomes Birth weight, adjusted for
gestation (early care, first births)
63
Birth outcomes Small for gestational age (early
care, first births)
64
Birth outcomes Term low birth weight (early
care, first births)
65
Maternal Inputs Tried to quit smoking upon
learning of pregnancy
(early care, first births prenatal
women only)
66
Conclusions

• WICs impact on birth weight smaller than implied
  in previous research. Results partly driven by
  third-trimester enrollment of women already
  experiencing longer gestations.
• Small effects on birth outcomes adjusted for
  gestation. Largest effects among whites. Effects
  among blacks and Hispanics limited to 1st and/or
  2nd trimesters.
• Weight gain effects limited to first- and
  second-trimester enrollees.
• Unreliably reported rates of smoking and smoking
  cessation among WIC women, particularly
  postpartum women, may confound results.
• Impact on breastfeeding ranges from negative
  (among whites and blacks) to substantial (among
  Hispanics).