Individual-level and state-level data drawn from several sources were used for these analyses. For individual infant data we accessed the 2009 National Immunization Survey (NIS).13
The NIS is obtained annually from a nationally representative sample (n
=25,241 children) by the CDC. It is a list-assisted random-digit-dialing telephone survey followed by a mail survey to participating children's immunization providers. It provides estimates of immunization coverage, weighted to represent children between the ages of 19 and 35 months, nationally, and by region, state, and select metropolitan areas using a complex sampling design.13
Questions about breastfeeding were included starting in 2001. The 2009 dataset includes infants born between February 2006 and May 2008. We eliminated infants whose mothers were 18 years old or under because the worksite statutes were not thought to affect their breastfeeding decisions. Furthermore, we omitted infants who by the time of the survey had relocated to a different state, resulting in 16,145 infants for analysis. Maternal variables included sociodemographic characteristics associated with breastfeeding (e.g., age, race), parity, and infant gender. Data on maternal birthplace (e.g., not born in the United States) were not available. We used whether the interview was conducted in Spanish as a proxy for non-U.S. maternal birthplace. Given the exploratory nature of this analysis we used the data as provided by the CDC, not modifying or collapsing any variable, except for the outcome variable, breastfeeding duration, recoded from continuous (days) to categorical (e.g., <6 months or ≥6 months).
The main exposure of interest was the presence of a state statute about worksite breastfeeding.7
This variable () was coded into three categories: No statute in place by 2008 (n
=32), break time and/or pumping site encouraged (n
=6), or break time and/or pumping site required (n
=15). For the purposes of these analyses the District of Columbia was considered the 51st
state. We also categorized the length of time the law had been in effect similar to the methodology of Kogan et al.14
Because we did not have month of birth we created three categories of length based on year only: Prior to 2005 (n
=11), between January 1, 2005 and January 1, 2008 (n
=8), and none or after January 1, 2008 (n
Other data in were drawn from publicly available sources and included measures at the institutional level (e.g., hospitals) and system/policy level. Each state's average Maternity Practices in Infant Nutrition and Care (mPINC) in 2007 score was included to represent the influence of hospital practices on initiation of breastfeeding.9
(2009 mPINC scores were not available.) The mPINC is a biennial CDC-mailed survey of all U.S. state and territorial hospitals and birth centers with registered maternity beds.15
Begun in 2007, each hospital reports on 34 items representing various practices that may affect infant feeding. Over 80% returned the 2007 survey representing 2,687 facilities. Facility responses were tallied into subscales and total score (range 0–100).16
The average 2007 mPINC score was 65.9
We included the proportion of births at Baby-Friendly Hospital Initiative (BFHI)–designated hospitals as an additional institutional-level characteristic. Facilities are designated BFHI after a rigorous external review process that affirms full implementation of specific breastfeeding friendly practices.17
We included this item given the Healthy People 2020 goal to increase the proportion of births occurring at BFHI designated hospitals (from 2007 baseline of 2.9% to 8.1%).4
Policy and system-level variables were obtained from the National Conference of State Legislatures7
including other statutes pertinent to breastfeeding such as employment discrimination and protecting breastfeeding in public.
The final system level characteristic included was the presence of a State Breastfeeding Coalition (). Although by 2010 all states had a coalition, they were established at varying times during the prior two decades. To be consistent we classified their founding dates (provided by the United States Breastfeeding Committee; personal communication, 2011, Megan Renner) in a manner similar to how the state statute dates were coded (pre-2005; 2005–2007; 2008 or later).
Because our interest was whether the presence of state worksite statutes increased breastfeeding duration, our main outcome variable—any breastfeeding at 6 months—was modeled to examine who was still breastfeeding (rather than who had stopped). We selected 6-month breastfeeding duration because it is a common maternal goal, it is beyond the time period when most women return to work, and it is consistent with the Healthy People 2020 objectives.
In our analyses we first modeled the prevalence of breastfeeding at 6 months by states with different types of statutes. We then ran six logistic regression models of any breastfeeding at 6 months, first including just the presence of the type of statute, then adding length of time the statute had been in effect, other state statutes pertinent to supporting lactating women, when state coalition was founded, and lastly variables depicting state implementation of breastfeeding supportive hospital practices. The final model included individual-level maternal and infant covariates.