Effect of a Paraprofessional Home-Visiting Intervention on American Indian Teen Mothers’ and Infants’ Behavioral Risks: A Randomized Controlled Trial

The trial is a multisite, randomized, parallel-group trial of the Family Spirit intervention plus optimized standard care compared with optimized standard care alone from pregnancy until 3 years postpartum in four tribal communities across three reservations in Arizona. Masked independent evaluators administered primary parenting and child outcome assessments. Comprehensive participant retention strategies and adverse event reporting were employed (28). A detailed description of the methods and design rationale may be found elsewhere (25). Here we report findings for parenting and maternal and early child behavioral outcomes from pregnancy to 12 months postpartum.

The study was approved by relevant tribal, Indian Health Services, and Johns Hopkins University research review boards. An independent data safety and monitoring board oversaw participant safety.

The primary aims of the study were to evaluate whether the Family Spirit intervention plus optimized standard care as compared with optimized standard care alone would significantly 1) improve parenting knowledge, self-efficacy, and behaviors; 2) reduce maternal psychosocial and behavioral risks (drug and alcohol use, depression, conduct problems) that could impede parenting; and 3) improve infant internalizing and externalizing outcomes.

Participants were eligible if they were pregnant and ≤32 weeks’ gestation, 12–19 years of age at conception, American Indian (self-identified), and residing in one of four participating reservation communities. The communities are rural and isolated, with populations ranging from 15,000 to 25,000. Participants provided written informed consent after receiving a complete description of the study. For those under age 18, informed consent was obtained from a parent or guardian and assent from the participant. The trial was conducted between May 2006 and September 2011.

Family Spirit consists of 43 highly structured lessons delivered by Native paraprofessionals. The content targets three domains: parenting skills across early childhood (0–3 years); maternal drug abuse prevention; and maternal life skills and positive psychosocial development. The home visitors deliver lessons one-on-one in participants’ homes using tabletop flip charts. The flip chart is designed so that the participant views illustrated content that often includes a real-life scenario while the home visitor reviews an outline of key points relating to the scenario and illustration (an illustration of the flip chart is provided in the data supplement that accompanies the online edition of this article). The home visitor is trained to use the lesson outline to create a comfortable teaching dialogue, rather than reading points by rote.

As baseline data suggest, teen mother participants were generally living in stressed and unstable home environments and often experienced strong feelings of isolation, helplessness, and distress as they struggled to develop their maternal role. Home visitors were recruited for their strong interpersonal skills, capacity to conduct the intervention, and understanding of participants’ challenges. The intervention’s therapeutic effect was hypothesized to operate through one-on-one teaching of highly structured content by a knowledgeable, empathic Native home visitor. Home visitors’ success in developing a warm, professional relationship was viewed as key to retaining and motivating participants to trust and learn from the curriculum over the long intervention period.

Each home visit was designed to last ≤1 hour, including a brief warm-up conversation, conducting the lesson, a question/answer period, and providing summary handouts. Home visits occurred weekly through the end of pregnancy, biweekly until 4 months postpartum, monthly between 4 and 12 months postpartum, and bimonthly between 12 and 36 months postpartum.

Home visitors received extensive training (>80 hours) in trial protocol and intervention delivery and had to demonstrate mastery (≥85%) of the Family Spirit curriculum through written and oral examinations before delivering any intervention. Supervisors conducted quarterly observations of home visitors administering home visits, rating them on protocol adherence, professionalism, and rapport-building. All intervention sessions were audiorecorded, and a random 20% of recordings were reviewed for protocol adherence.

Optimized standard care consisted of transportation to recommended prenatal and well-baby clinic visits, provision of pamphlets on child care and community resources, and referrals to local services when needed. Optimized standard care was chosen as the control condition because it optimized the standard of care for young mothers and their children within reservation communities, addressed transportation and access barriers to preventive health care, and provided beneficial and ecologically valid services in participants’ settings. By providing optimized standard care to both intervention and control groups, the quality and dose of optimized standard care was controlled so that differences between groups could be validly attributed to the Family Spirit intervention.

After the baseline assessment, eligible participants were randomly assigned in equal numbers to the Family Spirit intervention plus optimized standard care or to optimized standard care alone. Randomization was stratified by site, age (12–15 years versus 16–19 years), and history of previous live births (0 versus ≥1). The data manager created the randomization sequence using Stata, version 9.0 (StataCorp, College Station, Tex.). The study coordinator delivered randomization status by telephone to a staff member who enrolled the participant. Independent evaluators collecting observational and structured interview data were blind to randomization status. Participants and other study personnel were not.

Outcome data were collected at four time points during this study period in participants’ homes: at baseline (≤32 weeks’ gestation) and at 2, 6, and 12 months postpartum. At all four time points, parenting outcomes (parenting knowledge and self-efficacy; maternal acceptance, involvement, and responsivity; and home safety strategies) and mothers’ psychosocial and behavioral status (internalizing problems, externalizing problems, and substance use) were collected through self-report questionnaires, in-person interviews, audio computer-assisted self-interviews, and observational assessments. At 12 months postpartum, children’s psychosocial and behavioral functioning was assessed using the ITSEA. Measures were selected for their wide standard use, strong psychometric properties, and appropriate cross-cultural validity (see Table S1 in the online data supplement).

The primary outcome variable for determining sample size and power at the endpoint of the trial (36 months postpartum) was mother’s effective and competent parenting as measured by the Home Observation for Measurement of the Environment scale (33). This instrument was administered every 6 months from 6 to 36 months postpartum. Assuming 25% attrition, completion of four of six assessments, an alpha of 0.05, and a within-family correlation (r) of 0.5, we needed 160 participants per group (minimum total N=320) in order to detect a meaningful public health effect size of 0.33 with 90% power. This sample size also provided adequate power (≥80%) to detect minimum effect sizes of 0.30 on child developmental outcome measures at 12 months postpartum.

Of the 322 participants, 51 (16%) did not complete the parenting assessments and 66 (21%) did not complete the ITSEA at 12 months. Therefore, a multiple imputation approach was employed to manage missing values (34, 35), using a sequential regression multivariate imputation algorithm implemented via IVEware (36) for SAS. The imputation model included relevant baseline demographic characteristics, total scores on outcomes at each assessment point, and indicator variables for study site and treatment condition. Using these variables, 20 imputed data sets were generated, and results of identical analyses on each imputed data set were combined using Rubin’s established guidelines (34). Intent-to-treat analyses were conducted on all primary and secondary outcome measures. To investigate between-group differences in scalar outcomes, a series of separate analysis of covariance models were fitted for each outcome. Each model included site (treated as a fixed effect) and the following covariates to control for nonequivalence (p<0.1) among the treatment groups at baseline: mother’s total score on the Center for Epidemiologic Studies Depression Scale (CES-D) (37), whether the mother had ever smoked cigarettes, and whether the mother used alcohol during pregnancy. For binary outcome measures, a series of separate logistic regression models were conducted using the same covariates. All models were first applied to the entire sample and then to the subsample of participants who had endorsed prior alcohol or drug use at baseline (N=285, or 88.5% of sample). Analyses were conducted with SAS, version 9.2 (SAS Institute, Cary, N.C.).

Adverse events were recorded by evaluation staff and home visitors and reviewed on a biannual basis by the trial’s data safety and monitoring board. The proportion of adverse events and serious adverse events was similar between groups after accounting for increased contact time within the intervention group.

The outcomes suggest that home visiting by paraprofessionals can address behavioral health disparities associated with teen pregnancy and drug use by supplementing routine prenatal and well-baby care with parenting education to prevent maternal and child behavior problems whose trajectories challenge underresourced reservation communities. Evidence from this trial suggests that the Family Spirit intervention could move children in high-risk settings out of clinically meaningful risk for early child behavior problems, with potential impact on both long-term public health and economic benefits. The fact that twice as many control group infants were clinically at risk for poor emotional and behavioral regulation (21% for externalizing and 27% for internalizing problems) compared with a normative U.S. sample (10%) (13, 46) signals the importance of this type of intervention in similarly stressed populations. The greater benefit afforded infants of substance-using mothers is consistent with previous home-visiting trials that found better response among mothers with “poorer [baseline] psychological resources” (a composite variable including mental health, sense of mastery, and intelligence) (48). These findings suggest that home-visiting interventions may be most critical for young mothers in greater need, particularly those with early initiation of substance use.

In general, the study’s effect sizes are small but consistent with expected effects from a public health intervention, as opposed to a clinical intervention. Because data are reported up to 1 year postpartum in a trial whose endpoint is 3 years postpartum, outcomes should be considered preliminary and viewed as important findings supporting the efficacy of, and the general theory underlying, the intervention. The study has other, more specific limitations. First, the generalizability from this study to the heterogeneous tribal population in the United States is unclear. However, because participants were enrolled across four diverse Native communities, generalizability is likely greater than for studies conducted on a single reservation. Second, mothers who receive the Family Spirit intervention may have altered responses to self-reports based on social desirability. The fact that mothers in both groups reported high rates of substance use across time, including illegal drugs, counters a response bias argument. Some studies have corroborated self-reports with biological data (e.g., cotinine samples) and videotaped observational measures coded by external blind evaluators. These methods were neither culturally acceptable nor financially feasible in this trial. In addition, use of low-cost standardized measures administered by paraprofessionals increases the potential for replication across other low-resource communities. Third, some measurement bias may exist. In spite of challenging home environments, poor living conditions, and high substance use rates, participant scores on the Home Observation for Measurement of the Environment scale, the Achenbach System of Empirically Based Assessment, and the CES-D were better than or comparable to those for U.S. samples of all races overall (49, 50). These self-reports may require additional construct or content validity, as well as normative data for this adolescent population. Finally, because of the large number of dependent variables, the likelihood of significant findings due to chance is increased. However, all significant results and statistical trends are consistent with the study’s theoretical model and hypotheses and extend findings from smaller previous trials (21, 22, 25).