Project DC-HOPE was one project within the NIH-DC Initiative to Reduce Infant Mortality in Minority Populations in the District of Columbia, a collaborative effort involving the Children's National Medical Center, George Washington University, Georgetown University, Howard University, the National Institute of Child Health and Human Development (NICHD)/NIH/DHHS, and RTI International. Project DC-HOPE was a randomized clinical trial targeting smoking (including environmental tobacco smoke exposure), depression, and intimate partner violence (IPV). The primary goal was to estimate whether a multi-modal, integrated counseling and educational intervention reduces smoking and environmental tobacco smoke exposure (ETSE), depression, and IPV (defined as being victimized) among pregnant African American and Latina women. A secondary goal was to estimate whether a clinic-based intervention reducing smoking, ETSE, depression, and IPV in pregnancy would reduce adverse pregnancy outcomes (e.g., prematurity and low birth weight) and lower infant morbidity and mortality.
Outcome measures were selected to address both sets of goals. Medical records of infants and mothers were abstracted to obtain data on pregnancy outcome measures. In addition, an assessment battery of standardized and validated self-report measures of the targeted risks and cotinine levels based on saliva specimens were collected. These assessments were obtained at baseline, 22–26 weeks and 34–38 weeks of gestation, and 8–10 weeks postpartum. Subjects were recruited at six prenatal care clinics in Washington, DC. Women were eligible to participate if they were Washington, D.C. residents, African American or Latino, at least 18 years of age, at 28 weeks gestation or less, and English speaking. These eligibility criteria were verified from clinic records. Institutional review boards at Howard University, RTI International, and NIH approved the study.
To ensure adequate statistical power for testing the hypotheses that multidisciplinary integrated intervention will result in reductions in targeted risk factors and improvements in pregnancy outcomes, determination of sample size requirements was essential. Based on the number of births to African American women in Washington, DC, estimates of the number of women appearing at participating clinics in their first two trimesters of pregnancy, the prevalence of the risk factors estimated from previous studies, the estimated refusal rate, and the 2-year recruitment period, a total of 1,750 pregnant women were expected to enroll. Because of the study design and eligibility criteria, participants identified would have a 100% prevalence of a specific risk factor. Assuming a 20% drop-out rate, and 20% loss to follow-up, 1,050 women were expected to be retained at the end of the follow-up period (525 in each of the care groups). This number, assuming a 5% level of significance, 80% power, would allow the detection of 10–20% reductions in risk-specific factors among those in the intervention group from a 100% prevalence at recruitment time. This number was also sufficient to detect a 25% reduction in prematurity and low birthweight combined in the intervention group as compared to that for the usual care (estimated at 20%).
Project DC-HOPE staff members were grouped into four functional teams: Recruitment Specialists (RSs) conducted participant recruitment and retention of study subjects and collected saliva specimens; Pregnancy Advisors (PAs) delivered the intervention; telephone interviewers administered the evaluation interviews; and abstractors obtained information from medical records.
Recruitment took place between July 9, 2001 and October 31, 2003, four months beyond the original planned 2-year period. Follow-up activities of participants continued until July 31, 2004. The project started at four clinic sites. Because one site closed ten months after the start of the study, a fifth site was added in February 2002, and an additional sixth site was added in May 2003. These steps were taken to ensure the recruitment of the required sample size.
RSs approached women presenting for prenatal care while waiting for their appointments. Clinic logs were monitored closely on a daily basis for both returned participants and new clients. The RSs briefly explained the objectives of the study to the women and provided them with a brochure describing the study and answers to common questions. Interested women gave written consent for screening. Women were screened for socio-demographic eligibility and the presence of one or more of the four targeted risks using an Audio-Computer-Assisted Self Interview (A-CASI). A-CASI methodology allows subjects to listen to digitally recorded questions on headphones that are connected to a laptop computer while the question is simultaneously displayed on the computer screen. As a response choice is heard, it is highlighted on the screen. The subject answers by touching the chosen response option on the screen. A-CASI offers an environment in which sensitive information is easily and privately reported without the need for special computer skills or reading ability. Consequently, risky behaviors are reported more frequently than in more conventional interviewing or self-administered paper and pencil forms [16
The recruitment process was flexible in order to enhance participation. If a woman were reluctant to start the screening process due to time constraints, the RSs would suggest that she complete the screening and recruitment process when she returned to the clinic at a later date.
Women found to be eligible based on A-CASI responses were invited to participate in the study and provided with all information necessary to make an informed and knowledgeable decision. Those interested signed a second form consenting to participate in the randomized trial once they completed the baseline interview.
These consenting participants were asked to provide contact information, including the best phone number for reaching them and the names and telephone numbers of two relatives or friends who did not live with them. Efforts were taken to develop contact procedures from the research staff who maintained confidentiality when communicating with participants outside the clinic setting. Addresses were collected to facilitate tracing efforts, but the women were informed that they would not receive mail from Project DC-HOPE. Since one of the screened risk factors was IPV, staff did not want to raise women's risk for abuse by receiving mail from the study that might be negatively regarded by an abusive partner, or would expose her pregnancy. For similar reasons, women were asked whether or not telephone messages from Project DC-HOPE staff could be left on their telephone answering machines. If not, this was noted in her computerized record accessible by all project teams.
After completing the baseline interview, women were randomly assigned to usual care or intervention group. Site- and risk-specific block randomization of participants was conducted. Women randomized to the intervention group met with trained PAs (in addition to seeing their physician) either immediately before or after a routine prenatal visit and at two postpartum sessions during which they received individualized counseling targeting their area(s) of risk. Women in the intervention group attended on average 3.9 ± 2.8 prenatal sessions, and 53.6% attended at least 4 prenatal sessions. Participants spent an average of 35.3 ± 14.8 minutes per session. Women assigned to the usual care group met with their primary care providers as per standard clinic practice.
Recruitment and retention strategies
Several strategies were implemented in Project DC-HOPE which were designed to promote successful recruitment and retention of study participants. These strategies received careful consideration, taking into account the goals of the study and the type and duration of the intervention. This is important in such a community-based research study in which reliance on clinic infrastructure and personnel influences the success of the study. The strategies included features of the study design, consistent contact with study participants, financial incentives, recruitment training, cooperation from clinic staff, effective tracking of study participants, and continuous monitoring of study progress. A description of each of these elements follows.
As part of the study design, women had to complete the baseline interview before they were randomized and enrolled into the study. Consequently, women were identified who were more likely to complete the study, since there were several opportunities (completing the A-CASI screening, providing main study consent, and completing the baseline interview) leading up to randomization during which they could refuse or be lost to follow-up. This strategy also preserved the balance between the usual care and intervention groups in order to facilitate the intent-to-treat analysis of the impact of the intervention. In fact, the difference in the retention rate between the intervention group (78%) and the usual care group (81%) was not significantly different. A second feature of the study design, which contributed to high retention by reducing participant burden, was scheduling all in-person study activities to coincide with prenatal care visits. This included the collection of saliva, the dispensation of incentives, and the delivery of the intervention sessions.
Critical to retention and the success of a longitudinal study is consistent contact with the study participants and updated contact information [20
]. For Project DC-HOPE, RSs maintained frequent telephone contacts with the participants to remind them of intervention sessions and to reschedule missed appointments. Contact information was updated each time project staff had personal contact with study participants, including when monetary incentives were dispensed and saliva samples were collected, and at each intervention session. Contact information was also updated at the completion of each telephone interview. The four project teams maintained detailed documentation of the time and day of attempted phone calls (both successful and unsuccessful) to determine the best time to reach the participants. Senturia, et al. [21
] found that timing of phone contacts and their intensity are crucial in retaining participants. Thus, great efforts were taken to develop contact procedures for the research staff to maintain confidentiality when communicating with participants outside the clinic setting. In addition, a concerted effort was made by study staff to track hard-to-find participants. The contact information database was checked to obtain the most up-to-date locator information. The subject was then telephoned at her primary and (where necessary) her secondary telephone numbers. If the subject were not reached, Directory Assistance was consulted for a new telephone listing. If a new listing were not found, the clinic staff was consulted for updated or different contact information. In addition, the planned hospital delivery site was contacted to determine if the subject had been admitted.
Financial incentives to compensate participants for their time and effort also contributed to successful recruitment and retention of study participants. All women received $5 for completing the A-CASI screening, a 30-minute telephone card for providing main study consent, and $15 for each telephone interview. Women in the intervention group received $10 for each intervention session and additional $15 and $25 gift certificates for the first and second postpartum intervention sessions, respectively.
The efforts of the RSs and the extensive training they received contributed to successful recruitment. Mostly African American and females, the RSs learned that their verbal and non-verbal behavior was an important part of obtaining cooperation and promoting participant confidence and trust. Rapport was imperative for successful recruitment and was achieved through showing sensitivity to the women and their experiences. Specifically, RSs were taught to be:
• alert, clear spoken, and good listeners;
• positive and assertive, but not aggressive;
• responsive to the woman's reasons for reluctance;
• respectful and culturally sensitive;
• confident, sincere, and spontaneous in their introduction; and
• credible, by knowing the objectives of Project DC-HOPE and the activities required for participation.
Training involved role-playing to address different situations and types of women. Training of new RSs also included the sharing of experiences by the veteran RSs. In addition, RSs discussed recruitment strategy ideas in periodic meetings with the supervisor of data collection and principal investigators.
Cooperation from the clinic staff was another critical strategy for the recruitment and retention efforts. As suggested by Wiemann and colleagues [22
] this [clinic staff] cooperation is facilitated by educating them on the study design and designing the project implementation procedures with the least amount of interference with the daily functioning of the clinic. Before launching the study, the Principal Investigators held meetings with the clinic staff to explain the purpose of the study and describe their role in the study. Their understanding of the study design and sense of involvement with the recruitment and retention process helped maintain the integrity of the research effort and provide the project easy access to the clinic sites. Throughout the duration of the study, project supervisory staff made frequent visits to clinic sites to address problems and to demonstrate willingness to be flexible and modify procedures to adapt to the individual clinic needs. The study also provided partial salary support for the clinic staff who were assuming the extra responsibility of collaborating with the RSs and PAs.
A Data Management System (DMS) was developed to share and track information about study participants by the four functional project teams. The DMS was a PC-based computerized system that allowed for the collection, aggregation, and reporting of the study data. The DMS tracked upcoming events (visits, interviews, incentives, intervention, etc.) for individual study participants and was available to all project teams on computers dedicated to the study at each clinic site, interviewing site, and medical record abstraction site. Reports generated by the DMS made it possible for the project teams to prepare for upcoming events, such as a participant's prenatal care visit or intervention session, or the need to collect a saliva sample or dispense an incentive payment, and were critical in the recruitment and retention efforts. Types of reports included:
• a list of patients previously seen at the clinic with their recruitment status,
• a list of scheduled appointments for a given day with required activities,
• a list of participants with a specific activity due or overdue,
• a report for each study participant summarizing their completed study activities, along with personal information such as contact information, their next prenatal care visit, and pregnancy status.
The Data Coordinating Center (DCC) staff responsible for the implementation of the DMS maintained close communications with the functional project teams, provided quick response to any issues, and immediately resolved any problems the staff had with the system. Senturia, et al. [21
] recommended such a strategy as "essential to achieving nearly complete follow-up within a population historically difficult to follow."
Recruitment and enrollment status reports, first weekly, then bi-weekly, and finally monthly, were used to monitor study progress. These reports provided site-specific information for all study activities, beginning with the number of women approached and screened. In addition to the number of eligible and consenting women, reasons for ineligibility and refusals were also reported. Actions to address recruitment issues were taken accordingly. For women enrolled in the study and randomized, the reports provided the number of women who withdrew from the study or were lost to follow-up. Number of completed intervention sessions, completed telephone interviews for each evaluation time interval, and completed medical record abstractions were also reported. These data provided the study management staff with information at each stage of the study and identified potential problems so that appropriate actions could be taken in a timely fashion, such as the addition of clinic sites to increase recruitment. Site-to-site variability in recruitment was monitored and re-training or increased supervision provided as needed.
The statistical analyses of recruitment and retention data proceeded in three stages. First, we present various screening, eligibility, and enrollment results. Differentials in retention rates, comparing socio-demographic characteristics and risk profiles of women retained with those women not retained, were investigated using Fisher's exact tests and Chi-square tests for categorical variables. The differences in means for continuous variables were tested using t-tests. Retention rates at delivery and at postpartum were computed. Differentials in sample characteristics by clinic site (not shown here) were evaluated. The results indicated that socio-demographic characteristics of participants differed across sites and thus needed to be and were accounted for in the analysis.
In the second stage, we investigated multivariable relationships to identify the various socio-demographic characteristics that contributed to differentials in retention/attrition rates, after controlling for all factors using logistic regression analysis [23
]. A multivariable model is presented with various sets of predictors: (1) socio-demographic characteristics, (2) clinic sites, (3) risk factors at baseline interview, and (4) risk predictors. This model includes fourteen predictors. The number of women not completing the postpartum interview, when divided by the number of predictors is approximately 15. This number is greater than the recommended minimum number of events per predictor. This assures the reliability of the estimates of the regression coefficients and the validity of the statistical inference based on this model [24
In the third stage, we used Classification and Regression Trees (CART) methodology to identify the characteristics of various homogeneous subgroups of women who were likely not to be retained. CART methodology [25
] helped formulate distinct groups of retained and not retained women, thus identifying those characteristics predicting retention status. CART methodology is a non-parametric exploratory procedure that does not require the stringent assumptions of other parametric statistical methods and tests. For more details, the reader may consult various cited references [27