We analyzed data from the Centers for Disease Control and Prevention (CDC) Pregnancy Risk Assessment Monitoring System (PRAMS) during 2004-2008. PRAMS is an ongoing, population-based surveillance system of maternal behaviors and experiences before, during, and after pregnancy. PRAMS is conducted by state and local health departments in collaboration with CDC. All health departments participating in PRAMS use a standardized data collection methodology developed by CDC (13
). At each site, a monthly stratified sample of 100 to 300 new mothers is selected systematically from recent birth certificates. PRAMS staff at each site mail a self-administered questionnaire to the selected women starting 2 to 3 months after the delivery of a live infant. Women who do not respond to any of the 3 serial mailings are contacted by telephone to complete the survey. To minimize recall bias, efforts to contact women end 9 months after the woman has delivered her baby. Survey data are linked to selected birth certificate data and weighted for sample design, nonresponse, and noncoverage. The weighted data represent all live births delivered in each respective site in the given year.
To minimize nonresponse bias, PRAMS sites were included in the analysis if an overall weighted response rate of 70% or more was achieved for 2004 through 2006 and 65% or more for 2007 through 2008 for each site; these thresholds are established by CDC for published results (14
). The weighted response rate indicates the proportion of women sampled who completed a survey, adjusting for sample design. Our analysis used data from the following 32 states and New York City for 2004-2008, except where noted: Alaska, Arkansas, Colorado, Delaware (2007-2008), Florida (2004-2005), Georgia, Hawaii, Illinois, Louisiana (2004), Maine, Maryland, Massachusetts (2007-2008), Michigan, Minnesota, Mississippi (2004, 2006, 2008), Missouri (2007), Nebraska, New Jersey, New Mexico (2004-2005), New York (excluding New York City), New York City (2004-2007), North Carolina (2004-2005, 2007-2008), Ohio (2005-2008), Oklahoma, Oregon, Rhode Island, South Carolina (2004-2007), Tennessee (2008), Utah, Washington, West Virginia, Wisconsin (2007-2008), and Wyoming (2007-2008). The PRAMS project has been approved by the CDC institutional review board.
Prepregnancy smoking status was ascertained from the PRAMS questionnaire. Among women who reported smoking in the last 2 years, women were asked how many cigarettes they smoked per day on average during the 3 months before pregnancy. Categorical responses were none (0 cigarettes)Prepregnancy smoking status was ascertained from the PRAMS questionnaire. Among women who reported smoking in the last 2 years, women were asked how many cigarettes they smoked per day on average during the 3 months before pregnancy. Categorical responses were none (0 cigarettes), less than 1, 1 to 5, 6 to 10, 11 to 20, 21 to 40, or 41 or more. Women who reported "none" were classified as nonsmokers; others were classified as prepregnancy smokers. For smokers, number of cigarettes smoked per day on average during the 3 months before pregnancy was collapsed into 3 groups: 1) 5 or fewer cigarettes per day (includes <1 cigarette per day), 2) 6 to 20 cigarettes per day, and 3) more than 20 cigarettes per day.
Maternal demographic characteristics included in the bivariate analysis were age, race/ethnicity, education, Medicaid status (proxy for income), parity, pregnancy intention, state of residence, and year of birth. Age, race/ethnicity, education, parity, state of residence, and year of birth were ascertained from the linked birth certificate data, and Medicaid status and pregnancy intention were ascertained from the PRAMS questionnaire. Maternal age was divided into 2 categories, younger adult (18-24 y) and older adult (≥25 y). We were unable to report analysis of women aged 35 years or older because of inadequate sample size in certain racial/ethnic groups. When we ran adjusted relative risks comparing smoking prevalence of women aged 35 years or older with those aged 20 to 24 years, the conclusions by racial/ethnic group were consistent with our results when we grouped women aged 25 years or older. Maternal race/ethnicity was categorized as non-Hispanic black, non-Hispanic white, Hispanic, Alaska Native, American Indian, and Asian/Pacific Islander. Maternal education was categorized as less than 12 years, 12 years, and greater than 12 years. A woman was classified as enrolled in Medicaid if she reported being on Medicaid just before she got pregnant or if Medicaid was used to pay for prenatal care or for her delivery; otherwise, she was classified as not being enrolled in Medicaid. Parity was categorized as no previous live births or 1 or more previous live births. A pregnancy was categorized as unintended if the mother reported that she had wanted to become pregnant later or not at all.
The analysis was conducted using SAS version 9.2 (SAS Institute, Inc, Cary, North Carolina) and SUDAAN version 10 (Research Triangle Institute, Research Triangle Park, North Carolina), to account for the complex survey design of PRAMS. A total of 200,008 records were available for the analysis among the 32 states and NYC live births during 2004 through 2008; singletons and multiples were included. A woman was excluded if her prepregnancy smoking status (n = 3,444, 1.7%), age (n = 12, <0.1%), or race/ethnicity (n = 1,018, 0.5%) was missing. We excluded women aged less than 18 years (n = 7,287, 3.6%) because we could not adequately control for education, which was reported as a categorical variable. We also excluded women reporting a race other than the 6 specified above or mixed race (n = 2,594, 1.3%). The final number of records analyzed was 186,064. At the time of questionnaire completion, the average infant's age was 120 days and ranged from 61 to 270 days.
We examined the distribution of demographic characteristics for the study population by age group (18-24 y and ≥25 y). We calculated the percentage of unintended pregnancies among smokers and nonsmokers. Next, we calculated prepregnancy smoking prevalence and 95% confidence intervals (CIs) by age group overall and separately for each of the racial/ethnic groups. To explore geographic differences, we examined prepregnancy smoking prevalence by age for each state that had an adequate sample size for each racial/ethnic group. We calculated the proportion of smokers in each of the 3 categories of cigarettes smoked per day. We used χ2 tests to examine differences in prevalence estimates by demographic characteristics, state, and the proportion of smokers in each category of cigarettes smoked per day.
For each racial/ethnic group, we modeled the probability of prepregnancy smoking by age group using women aged 25 years or older as the reference population. Unadjusted and adjusted relative risks and 95% CIs were calculated using logistic regression, as described by Bieler et al (15
). Maternal education, Medicaid status, parity, pregnancy intention, state of residence, and year of birth were included in the final adjusted models if they confounded the association between age and smoking status by at least 10%.