The NIH-DC Initiative to Reduce Infant Mortality in Minority Populations is a long-standing and ongoing research collaboration between 4 major academic institutions in Washington, DC (the Children's National Medical Center, Georgetown University, George Washington University, and Howard University); the Eunice Kennedy Shriver National Institute of Child Health and Human Development; the National Center for Minority Health and Health Disparities; and RTI International. As part of this collaboration, we undertook a randomized clinical trial (“Healthy Outcomes of Pregnancy Education,” or DC-HOPE) to evaluate the benefits of an integrated behavioral and psychosocial intervention delivered during routine prenatal care. (Details on or copies of the training materials, instruments, and the DC-HOPE intervention are available from the authors.)
In this randomized study, we assessed a clinic-based intervention focusing on 4 psychosocial and behavioral risk factors (cigarette smoking, secondhand smoke exposure, depression, and IPV) associated with poor pregnancy outcomes. A site- and risk-specific block randomization design was used, and women were randomized to an intervention group or a usual care group. Interviewers collecting outcome information were unaware of participants’ group assignments. Primary and secondary study hypotheses were specified in advance.
Screening, Recruitment, and Retention
Details of the study screening, recruitment, and retention methods (including the methods used in calculating the required sample size) have been previously described,49
and only a brief description is provided here. Determination of sample size requirements was essential to ensure that the necessary statistical power was available to test the hypothesis that the multidisciplinary integrated intervention would produce reductions in the risk factors targeted. According to conventional assumptions, the study required that data be available from 1050 women (equally divided between the 2 randomized groups) to allow detection of a 10% to 20% reduction in risk factors (present in 100% of women at the time of recruitment) among those in the intervention group.
Between July 2001 and October 2003, we collaborated with 6 prenatal care sites in Washington, DC, serving predominantly African American populations to conduct a 2-stage consent and enrollment process. Women were approached while they were waiting for routine prenatal care appointments and invited to immediately complete a 10-minute audio computer-assisted self-interview (ACASI) to assess their potential eligibility for the study. Approximately 9 days after completing the screening interview, women were administered a baseline telephone interview during which more detailed information was gathered; this information served as the basis for behavioral change comparisons at follow-up. At the conclusion of this interview, each eligible consenting woman was randomized into the intervention or usual care group.
Eligible women were at least 18 years old, were no more than 28 weeks pregnant, were African American or Latina, and had at least one of the 4 risk factors of interest. Follow-up data collection telephone interviews were conducted during the second and third trimesters of pregnancy (22–26 and 34–38 weeks estimated gestational age, respectively). Intervention and follow-up activities continued until July 2004.
Participants were compensated for their time and effort; they received $5 for completing the ACASI screening, $15 for completing each telephone interview, and a 30-minute telephone card after providing consent. Women in the intervention group received an additional $10 at each intervention session conducted during their pregnancy.
Women who reported suicidal ideation during screening or follow-up interviews were immediately referred to an emergency mental health consultation team. The mental health consultant contacted and evaluated each such participant for actual suicidality. Referrals were made as needed, and women found to be potentially suicidal (n=10) were excluded from the study.
The behavioral intervention has been described in detail,50
and only a brief description is provided here. The intervention, based on the behavioral change literature specific to the psychosocial and behavioral risks targeted, was designed to be delivered in prenatal care clinics. To assess smoking and secondhand smoke exposure, the DC-HOPE project incorporated elements from the successful Smoking Cessation or Reduction in Pregnancy Program Treatment (SCRIPT) trial,51
the transtheoretical model of behavior change, the “pathways to change” self-help manual, and cognitive–behavioral therapy.52
Similarly, we adapted a group cognitive–behavioral therapy treatment developed by Miranda and Munoz53
that had been used successfully with low-income, ethnic minority obstetrics and gynecology patients suffering from depression. A single-visit intervention developed by Parker et al.54
for women experiencing IPV was modified to provide ongoing evaluation, guidance, and feedback throughout pregnancy.
Eight prenatal sessions were required to deliver the complete intervention, but a minimum of 4 sessions was deemed “adequate” on the basis of the amount of material that could be covered in that number of sessions. Individualized counseling provided an integrated and tailored approach to the multiple risks reported by each woman; intervention sessions occurred immediately before or after routine prenatal care, for an average of 35 minutes and no more than 45 minutes. Participants were presented with material during intervention sessions that required their active engagement with tasks such as identifying smoking triggers, developing strategies for mood management, and conducting a danger assessment.
In each session, the interventionists (referred to as “pregnancy advisors”) and participants developed a plan for “homework” designed to reinforce the intervention in the woman's real-life circumstances. The majority of the pregnancy advisors had master's degrees in counseling disciplines such as psychology as well as experience in interpersonal counseling, health education, or behavior change; 2 were registered nurses. All received 3 weeks of intensive training in addition to ongoing supervision to ensure their familiarity with, and fidelity to, the integrated intervention. Pregnancy advisors were trained to address multiple risk factors in an integrated fashion that was responsive to each woman's circumstances and to promote an overall sense of mastery and self-efficacy based on an integrated approach to these behaviors. A single pregnancy advisor was assigned to each clinic to provide consistency for participating women.
Validated instruments assessing cigarette smoking, secondhand smoke exposure, depression, and IPV were used for the ACASI screening and the baseline and follow-up telephone assessments.55–60
In addition to standard sociodemographic data and information on the 4 risk factors, the baseline telephone interview collected information regarding drug and alcohol use, relevant obstetric history, and the participant's attitude toward her current pregnancy. As mentioned, telephone interviewers (and their supervisors) were unaware of the participant's randomization group. (Details regarding study instrumentation are available on request from the authors.)
We used an intention-to-treat approach to evaluate the overall benefits of the intervention and to avoid bias. Participants were analyzed according to their randomized care group assignment, regardless of actual receipt of intervention or withdrawal from the study.61
For the purposes of the intention-to-treat analysis, multiple imputation methodology was used to estimate missing data using the sequential regression imputation method described by Raghunathan et al.62
A linear, logistic, or polytomous regression model was used to impute continuous, binary, or categorical missing values. Imputation was performed multiple times to create 5 complete data sets; variability among the imputed data sets reflected the uncertainty inherent in predicting unknown values. IVEware imputation and variance estimation software was used in completing these procedures.63
In conducting our analyses, we used standard statistical techniques to separately assess each of the 5 imputed data sets. We then combined the results via the MIANALYZE procedure in SAS (SAS Institute, Cary, NC) to obtain the overall inference for the multiply imputed data. The final parameter estimates were averages of the estimates from the 5 imputed data sets, and the associated standard errors accounted for both within- and between-imputation variance.
As a means of assessing the effects of the intervention, 2 approaches were used to quantify behavioral changes occurring during pregnancy and to compare the 2 randomized groups at baseline, the first follow-up (second trimester), and the second follow-up (third trimester). The first approach contrasted the distribution of number of risks as reported by women in the 2 groups at each time point. The second quantified within-person change over time to assess the benefits of the intervention in line with the method used to evaluate the Colorado Prenatal Plus Program.48
According to this integrated approach, each woman was characterized with respect to number of risks at the baseline interview. This number could range from 0 (for women whose reported risks at screening were not present when they were assessed at baseline, either because of behavioral change or because of the change in instrumentation) to 4 (for those women reporting active smoking, secondhand smoke exposure, depression, and IPV).
At the eligibility screening, women who reported that they had smoked in the 6 months preceding their pregnancy or since learning they were pregnant were defined as smokers; at study enrollment, women were defined as smokers if they had smoked a puff of a cigarette in the previous week. At the eligibility screening secondhand smoke exposure was defined as exposure to smokers at home, in the same room, or in a car, whereas at enrollment it was defined as having been in the same room or area where someone was smoking in the past week.
The Beck Depression Inventory II was used to evaluate depression at the eligibility screening, and the Hopkins Symptom Checklist was used to assess depression at study enrollment (women who had a score above 0.75 on the latter instrument were categorized as depressed). As a means of assessing IPV at the eligibility screening, women were asked whether a current or previous partner, boyfriend, or husband or the baby's father had pushed, shoved, slapped, kicked, or physically hurt them or forced them to have sexual intercourse in the past year; they were also asked whether they were afraid of their current partner. At study enrollment, IPV was confirmed if a woman reported having been subjected to any of the actions addressed in the revised Conflict Tactics Scale at least once by her partner in the past year.
Next, each woman was characterized in one of 3 ways on the basis of data provided in the follow-up telephone interviews conducted during the third trimester. That is, women were classified as (1) having resolved all of their risks, (2) having resolved some but not all of their risks, or (3) having resolved none of their risks or having increased their number of risks. A woman was defined as having “resolved” a risk if she had eliminated one of the behaviors or exposures of interest (i.e., she had quit smoking, she had not been exposed to secondhand smoke, her depression scores were below the clinical cutoff, or she had experienced no IPV episodes).
The primary measure of intervention effectiveness was the comparison of the proportion of women in the 2 randomized groups who had resolved all or some of their risks and the proportion of women who had resolved none of their risks. In addition, we compared the proportion of participants in the intervention and usual care groups who had resolved all of their risks and the proportion who had resolved none of their risks.
Basic descriptive statistics were used to gather information on the characteristics of the participants, both overall and stratified by randomization assignment. Relevant comparisons were conducted via general linear models (for continuous variables) and generalized linear models (for categorical variables) (). The GLM and GENMOD procedures in SAS were used in all of the analyses.
Participant Characteristics: DC-HOPE Project, 2001–2004
As a means of evaluating the impact of the integrated intervention on risk reduction between the baseline and second follow-up interviews, we used logistic regression procedures to model the probability of women resolving all versus none of their risks and the probability of their resolving at least one versus none of their risks (). We used the LOGISTIC procedure in SAS to conduct analyses focusing on the individual imputations. Control variables included maternal age, education, marital status, employment status, enrollment in federal assistance programs (Medicaid and the Special Supplemental Nutrition Program for Women, Infants, and Children), and drug and alcohol use at baseline; previous premature delivery, pregnancy loss, or live birth; early initiation of prenatal care; gestational age at baseline; smoking, secondhand smoke exposure, depression, and IPV at the time of the ACASI screening; and total number of risks at baseline.
Changes in Risk Status From Baseline to Second Follow-Up: DC-HOPE Project, 2001–2004
Preliminary analyses identified no covariates that differed significantly between the 2 randomized groups (). Nonetheless, these variables descriptive of socioeconomic status and previous pregnancy outcomes were included as covariates so that their cumulative effects could be accounted for in the final model.