We conducted in-person, semistructured interviews with mothers of infants aged less than 8 months enrolled at Women, Infants, and Children (WIC) Program centers. The WIC Supplemental Nutrition Program is a federally funded program for low-income women who are pregnant, breastfeeding, and/or postpartum and for low-income children aged under 5 years. Study participants were enrolled at 6 sites during the periods June to December 2006, June to September 2007, and July to November 2008. In 2006, 788 mothers were enrolled in WIC centers in Birmingham, Ala, Dallas, Tex, Detroit, Mich, and New Haven, Conn. In 2007, a total of 792 mothers were enrolled in 2 of these sites, Dallas and New Haven, and in 2 new sites, Clarksdale, Miss and Jackson, Miss. In 2008, 775 mothers were enrolled in these same 4 sites. Mothers were eligible to participate in the study if they received benefits from WIC, had an infant aged less than 8 months, and spoke English. We chose WIC centers serving at least 50% African American clients to include a larger percentage of this at-risk population in the study. Two sites were chosen in Mississippi to gain additional data on populations at particular risk for SIDS and with low rates of adherence to sleep practice recommendations, and to obtain baseline data before a federally funded SIDS risk reduction initiative was implemented in the area. A total of 2491 mothers met eligibility criteria and were approached for an interview. Of these, 2411 (97%) consented to participate. A total of 2355 completed the interview, yielding a response rate of 95%.
Research assistants trained by the investigators (A.M. and E.R.C.) conducted all the interviews. Interviewers asked respondents to demonstrate what position they placed their infant in for sleep by using a doll. After the interview, all participants received current Back to Sleep campaign recommendations for safe infant sleep behaviors and a $10 gift certificate. All participants gave informed consent.
The Yale University School of Medicine Human Investigation Committee and the Boston University Medical Campus Institutional Review Board approved the study. In addition, the institutional review boards for the appropriate agencies at each site (eg, county or state departments of public health) approved the study.
Definitions of Predictor and Dependent Variables
We identified 2 predictor variables of interest: 1) maternal ratings of physician qualification to give sleep behavior advice and 2) maternal report of the nature of physician sleep behavior advice received. For the first predictor variable, we collected data on maternal ratings of physician qualification to give advice on each of the 3 recommended sleep behaviors by using the following question stem: “Doctors give advice to parents about different topics. How qualified do you think your baby’s doctor is to give you advice on …” Subsequent phrases included “what position your baby should sleep in,” “whether an adult should share a bed with your baby,” and “whether your baby should use a pacifier when sleeping.” We also asked mothers to rate physician qualification to give advice about 3 other infant care practices: “whether and when to give vaccinations,” “what to do when your baby has a fever,” and “what and how to feed your baby.” The variable for each topic had 2 categories based on responses to a 5-point Likert scale; a high rating of qualification was defined as a 4 or 5 and a low qualification rating defined as ≤3.
For the second predictor variable, mothers were asked about the nature of doctor’s advice regarding “baby’s sleep position,” “sharing bed with baby during sleep,” and “pacifier use during sleep.” Maternal reports of physician advice were coded into 3 categories based on correspondence of the advice with the 2005 AAP recommendations: “concordant” (consistent with AAP recommendations), “discordant” (inconsistent with AAP recommendation) and “no advice” if none was given. In the case of sleep position, concordant means supine-only sleep was recommended. It excludes advice for nonprone sleep, which includes the side position that is no longer recommended.
In addition, we collected data on a previously identified factor associated with whether mothers follow sleep recommendations—who is the mother’s most trusted source of advice regarding infant sleep practices.16
The 4 categories for trusted source of advice were based on the most frequently reported responses in the prior study: “family,” “doctors,” “myself,” and “other.”
The primary dependent variables were based on maternal report of their usage of 3 recommended infant sleep behaviors: supine sleep, no bed sharing during sleep, and pacifier use during sleep. The dependent variables for these analyses were maternal report of the following: 1) “usually placed supine for sleep,” 2) “usually does not share a bed with an adult during sleep,” and 3) “usually use a pacifier during sleep.”
We also collected data on demographic characteristics: maternal age, race/ethnicity, educational level, parity, and infant age. Mothers self-reported their race/ethnicity as African American, Latina, white or other. Mothers who identified themselves as Latina were included in the Latino group regardless of race.
Descriptive statistics including frequencies and percentages were calculated, followed by bivariate and multivariate logistic regression analyses. In bivariate analyses, we calculated unadjusted odds ratios for 3 main dependent measures: usually placed supine for sleep, usually does not share a bed with an adult during sleep, and usually use a pacifier during sleep, with 2 corresponding primary predictor variables for each dependent measure: 1) maternal rating of doctor’s qualification to give advice about that topic and 2) the nature of doctor’s advice regarding that topic. In multivariate logistic regression analyses for the 3 primary dependent measures, covariates were chosen based on bivariate analyses and on our previous research.9,15,17
The maternal rating of physician qualifications, nature of physician advice, and trusted source of advice achieved .05 level of significance in bivariate analyses. Demographic factors proven to be important in previous studies were included independent of their significance in bivariate analyses. The final models for each of the 3 recommended infant sleep behavior dependent measures included mother’s age, self-identified race/ethnicity and level of education, baby’s age, maternal rating of physician qualifications, nature of physician advice, and trusted source of advice for the corresponding behavior, study site, and study year.
We first analyzed data from the 3 years separately. The adjusted analyses yielded similar results, so the analyses presented combine data from all 3 periods. All analyses were conducted with SAS 9.1 (SAS Institute Inc, Cary, NC); α for all tests was 2-sided and set at .05.