Co-authors affiliated with HARC’s coordinating center conceptualized and designed the observational study, gathered and analyzed data, interpreted study results, and drafted and revised this manuscript. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting observational studies  was used to ensure rigorous reporting of the study (see additional information). National model leadership identified staff to take the lead as their representatives and co-authors for this study. Co-authors representing home visiting models provided study data, interpreted results, and revised the manuscript. Figure 2 illustrates how sample selection and data collection mapped to parts of the Precision Paradigm. As shown, reading from right to left in the figure, study eligibility was based on a model’s intended outcomes, Survey 1 focused on models’ target behaviors, and Surveys 2 and 3 focused on their expectations regarding intervention techniques.
Selection of home visiting models
The intended sample was evidence-based models enrolling families prenatally to promote healthy birth outcomes. We identified models meeting three criteria: designated as evidence-based by the Home Visiting Evidence of Effectiveness (HomVEE) review; implemented in states or tribal communities in the United States; and enrolling families prenatally. Eight models met these criteria. We contacted each model’s national office to ascertain whether the model aimed to prevent premature birth or low birth weight, defined as a birth < 37 weeks gestation and a birth weight < 2500 g. Two of the eight models indicated that promoting healthy birth outcomes was not a central focus. The other six models – Family Spirit, Kentucky’s Health Access Nurturing Development Services (HANDS), Healthy Families America, Minding the Baby, Nurse-Family Partnership, and Parents as Teachers – indicated that improving birth outcomes was one of their intended outcomes. Five of these models agreed to participate in the project; state administrators for HANDS declined due to demands of the COVID-19 pandemic.
Model representatives completed three surveys mapped to the Precision Paradigm. The surveys worked backward, from ascertaining the risks the model aimed to reduce, to the maternal behaviors it aimed to promote to reduce those risks, to the techniques it endorsed visitors to use to promote those behaviors and the emphasis to give to each (Fig. 2). All surveys were developed for this study and have not been previously published elsewhere (see Additional files 1, 2 and 3).
HARC coordinating center investigators distributed each survey to model representatives at the same time, for independent completion within 2–3 weeks. Surveys 2 and 3 were distributed after all models had completed the preceding survey. HARC coordinating center investigators encouraged model representatives to ask for clarification if they were uncertain how to answer a question. Model representatives submitted seven questions. HARC investigators emailed responses to all five models by the next working day.
Survey 1—intended pathways
HARC investigators drew from the literature [23,24,25,26,27] and from relevant American College of Obstetrics and Gynecologists Committee Opinions  to identify modifiable risks for low birth weight and premature birth, and target behaviors to reduce these risks. These are the birth outcomes most often used in home visiting impact studies in the US. While infant mortality is a Sustainable Development Goals indicator , and the US ranks poorly for this indicator , we did not use it in this project because it has not been used in home visiting impact studies in the US and because prematurity is the second leading cause of infant mortality.
To minimize respondent burden, Part 1 of Survey 1 was limited to ten common, diverse, modifiable, evidence-based risk factors that could be reduced through home visiting and that fell within the scope of the current pregnancy. The risks fell into four groups: 1) health care use (inadequate prenatal care); 2) psychosocial well-being (high stress, depression, intimate partner violence); 3) behavioral health (tobacco use, alcohol use, illicit drug use); and 4) biologic risk factors (infection, diabetes, high blood pressure). The survey asked representatives to rate the priority their model gave to reducing each risk. Response choices were: not a priority, low priority, moderate priority, high priority, and not sure. A priority risk was defined as one whose reduction was designated as a low, moderate, or high priority.
Part 2 of the survey focused on 14 behaviors that could be promoted within home visiting for the current pregnancy. We saw these behaviors as falling into four groups: 1) basic health promotion (physical activity, healthy diet, stress reduction activities, social supports); 2) health care use (adherence to prenatal care visit schedule, engagement in substance use treatment, and alerting the prenatal care provider to warning signs; 3) behavioral health (stopping or reducing tobacco use, stopping or reducing alcohol use, stopping or reducing illicit drug use); and 4) specific risk reduction behaviors (condom use, developing a domestic violence safety plan, medication adherence, self-monitoring of physiologic indicators). The survey asked representatives to rate their models’ expectations of home visitors for promoting specific maternal behaviors to reduce each of its priority risks. Response choices were required, recommended but not required, no expectation but compatible with our model, not compatible with our model, and not sure. A target behavior was defined as a behavior the model either required or recommended visitors to promote.
The ten risks and 14 behaviors together defined 41 unique pathways to good birth outcomes (Table 1). The literature recommended some behaviors as a way reduce multiple risks. For example, physical activity is a behavior to reduce high stress, depression, high blood pressure and diabetes. Of note, the literature characterized three risk factors – tobacco use, alcohol use and inadequate prenatal care – not only as risk factors but as behaviors influencing other risk factors. In the same way, we defined these three constructs as both risk factors and maternal behaviors.
At the end of Survey 1, HARC coordinating center investigators used each model’s priority risks and target behaviors to define its intended pathways to good birth outcomes. An intended pathway for a model is one linking a target behavior with a priority risk. Each model could have up to 41 intended pathways; the number and nature of intended pathways depended on the model’s priority risks and target behaviors to reduce those risks.
Survey 2—endorsement of intervention technique categories in intended pathways
Survey 2 asked respondents to rate their models’ stance regarding home visitors’ use of each of 23 technique categories for each of its intended pathways. Response choices were required, recommended but not required, no expectation but compatible with our model, not compatible with our model, and not sure. An endorsed technique category was defined as one that the model required or recommended visitors to use for a specific intended pathway.
The Appendix describes the 23 technique categories. HARC coordinating center investigators defined them by adapting an existing taxonomy of behavior change techniques and by adding techniques commonly used in home visiting but not represented in the existing taxonomy. The existing taxonomy contained 93 techniques grouped into 16 categories and was defined by applying consensus building methods to techniques identified in the behavior change literature . We used technique categories rather than individual techniques to reduce respondent burden. We modified these categories in four ways: 1) split four of the original categories into eight narrower, more homogeneous categories; 2) dropped one of the original categories but assigned some of its techniques to another existing category; 3) added the category, “assess readiness for change,” because it is concordant with a family-centered approach and with theories of behavior change that differentiate motivating, enabling, and reinforcing target behaviors ; and 4) added three categories aligned with the framework of West et al.  to reflect home visiting’s use of referral and coordination.
Survey 3—emphasis in using endorsed technique categories
Survey 3 explored how much models expected home visitors to emphasize technique categories within selected intended pathways. To minimize respondent burden while maximizing the number of comparisons that could be made, the survey’s focus was limited to a subset of pathways defined by behaviors designated as target behaviors by all five models and associated with reducing multiple risks.
Within that subset of pathways, each model’s version of Survey 3 was also limited to the model’s intended pathways as determined by Survey 1 and the technique categories it had endorsed for those pathways in Survey 2. For each pathway-specific set of endorsed technique categories, the model’s representative rated the relative emphasis the model expected visitors to give to each technique category. Response choices were adapted from those of Smith et al.  and ranged from one (low emphasis) to five (high emphasis) and no stance. Response choices two through four were not labeled. A technique category with a rating of five was defined as a high-emphasis technique category.
HARC coordinating center investigators carried out data analyses. After all surveys had been completed, we shared results with model representatives in several iterations, using representatives’ feedback to guide new analyses and to improve the clarity and usefulness of results.
Priority risks, target behaviors and intended pathways
We described the distribution of model responses for each risk. We determined and graphed the number of models designating each of Table 1’s 41 behavioral pathways as an intended pathway.
Stance on technique categories
We calculated the percent distribution of each model’s responses (required, recommended, no expectation but compatible, not compatible, and not sure) for each technique category across all of its intended pathways combined. For all models combined, we calculated the mean of the model-specific percent distribution of responses.
Emphasis on technique categories
We calculated the number of models designating each technique category as a high-emphasis technique category in any of the intended pathways in their version of Survey 3. We elaborated on this for each of four pathways to reduce maternal depression through four target behaviors – physical activity, adherence to the prenatal care visit schedule, stress reduction and social support. For each of these pathways, we calculated the number of models endorsing each technique category at all and as a high-emphasis technique category.
Comparison of models’ priority risks, intended pathways and stance on technique categories
We determined each model’s number of priority risks, target behaviors, and intended pathways. We calculated the number of technique categories each model endorsed, required, recommended, rated as compatible while not specifically endorsed, and rated as not compatible on one or more of the model’s intended pathways. We measured each model’s propensity for explicit endorsement of technique categories by calculating the mean number of categories it endorsed per pathway across its intended pathways. We measured each model’s consistency in endorsing specific technique categories across pathways as the percent of its technique categories that it either always or never endorsed across all its intended pathways. We report the minimum, median and maximum value for all of these model-specific measures.
Discussion and interpretation of results
HARC coordinating center investigators prepared results tables and talking points for three rounds of independent review and written feedback by model representatives followed by group discussion of the collated feedback.