This study was registered at ClinicalTrials.gov on August 24, 2021, before study initiation. The full study protocol and study consent document are publicly available within the ClinicalTrials.gov study record. All study procedures and materials were reviewed and approved by the University Institutional Review Board (STUDY00006555). Deidentified data were coded by participant number and identifying information was securely stored separately. There were no costs to participants for participating in the study. They received payments for completing surveys (US $40 at baseline, US $20 at 6 weeks, and US $40 at 12 weeks).

This study involved a 3-arm parallel RCT with equal allocation to the following conditions: (1) FMF Connect+Coaching, (2) FMF Connect without coaching, and (3) waitlist control. Quantitative survey data were collected at 3 time points—baseline, 6 weeks, and 12 weeks. The 6-week time point involved an abbreviated battery designed to allow for tests of statistical mediation hypotheses. App usage was recorded through Amazon Web Services (AWS) during the study period. After 12 weeks, study participants in the waitlist control group received the FMF Connect app. No changes were made to the trial design after initiation.

The participant population included caregivers of children (aged 3‐12 y) with FASD or PAE who live in the United States. To be eligible, caregivers needed to be a biological parent or other primary caregiver (eg, foster or adoptive parent, relative, or legal guardian) who was at least 18 years old, living in the United States, and had a smartphone or iPad (Apple Inc) with iOS operating system. Caregivers needed to have a child between the ages of 3 and 12 years with a diagnosis of FASD or confirmed PAE. As part of study registration, caregivers completed a survey about the child’s PAE and uploaded diagnostic records, if available. If there was more than 1 child in the family with FASD or PAE in this age range, the caregiver was instructed during screening to choose 1 child as the target of study surveys. The child needed to have lived with the caregiver for at least 4 months and was expected to remain in the home for at least 1 year. Exclusion criteria included the caregiver not being fluent in English (the FMF Connect app and measures were only available in English). Caregivers were also excluded if there was another caregiver of the same child or living in the home who was already enrolled in the study (couples are excluded to prevent dependence in the data), the caregiver participated in a previous trial of the FMF Connect app as part of earlier development phases of this project [21-23], or the family had previously received or was currently receiving the specialist-led standard FMF Program.

Recruitment, informed consent, and data collection procedures were conducted online. Study recruitment strategy and materials were developed through consultation with the University’s Clinical and Translational Science Institute, with considerations to engage a diverse participant pool. These considerations included materials representing diverse caregivers and families across race, ethnicity, age, and family structures and strategies to leverage partnerships and social media (ie, Facebook [Meta] and Twitter [subsequently rebranded as X]) outlets across the United States. Participants were recruited through a variety of mechanisms by our study team and collaborators within the Collaborative Initiative on FASD (CIFASD; [31]) and community partners. These included disseminating recruitment materials on social media, through national and regional FASD listserves (eg, FASD United’s Weekly Round-up), at relevant conferences, and to participants or patients with whom the study team or collaborators had routine access and/or had agreed to contact about new research opportunities. We aimed to enroll about 150 participants in this study. Power analyses were conducted before recruitment using GPower3. This analysis determined that with a sample size of 120 and assuming α=.05, there will be a power of 0.97 to detect medium effects (f²=0.15) and power >0.99 to detect large effects (f²=0.35). We anticipated about 20% attribution based on the previous feasibility trial [21].

Study recruitment materials included links to the study website [32], where potential participants could access study information and link to the study’s consenting and screening module. All surveys were administered via REDCap (Research Electronic Data Capture), hosted by the University of Rochester [33,34], and REDCap’s eConsent package was used to document participants’ informed consent. Participants had the option to review study consent information by video or text and were required to answer several questions to ensure comprehension. Participants were also offered the opportunity to ask questions via direct contact with the study coordinator before proceeding to other parts of the survey.

After completing informed consent, participants were directed in REDCap to answer screening questions to assess for study eligibility, including information about PAE, FASD diagnostic records, and demographic information. The study team determined eligibility using a study-created eligibility verification form in REDCap to carefully review the criteria and records provided. No evidence of bots or fraudulent participants was detected. A globally unique identifier [35] was then assigned to each eligible caregiver participant and child to facilitate data sharing within CIFASD or future studies. Eligible participants then received an email inviting them to complete baseline study surveys.

Table 1 lists all study surveys administered per timepoint. Primary study outcomes included child behavior problems (the Eyberg Child Behavior Inventory [ECBI]), attributions of child behavior (the Reasons for Children’s Behavior Scale [RCB]), parenting self-efficacy and satisfaction (the Parenting Sense of Competence Scale [PSOC]), family needs met (the Family Needs Met [FNM] Questionnaire), and caregiver FASD knowledge (the FASD Knowledge and Advocacy [K&A] Questionnaire). Secondary outcomes included child adaptive functioning (the Everyday Life Scale [ELS]) and caregiver use of self-care. We also examined participant ratings of app quality (the User Version of the Mobile Health Application Rating Scale [uMARS]) and patterns of app usage (from AWS). Additional measures of family background and services were included to characterize the sample. There were no changes in measurement after study initiation.

The ECBI [36] is a 36-item caregiver-report scale measuring frequency (Intensity scale) and perception (Problem scale) of disruptive behavior in children aged 2-16 years. Items are rated on a 7-point scale indicating frequency of the behavior. Scores are presented as T-scores (mean 50, SD 10), with higher scores indicating more frequent behavior problems (Intensity) or greater caregiver perception that behavior is problematic. T-scores of ≥60 are viewed as clinically elevated. Internal consistency in the current sample was high (baseline: α=.90 for Intensity and α=.88 for Problem, follow-up: α=.92 for Intensity and α=.91 for Problem).

The RCB [37] is a 30-item scale assessing caregiver attributions for children’s behavior. The RCB includes 7 subscales measuring neurodevelopmental and willful attributions across multiple areas of behavior, including task completion, sensory seeking and avoiding, and dysregulated or disruptive behavior. Items are rated on a 6-point Likert scale ranging from strongly disagree to strongly agree. Higher scores indicate greater agreement with the attribution. Internal consistency across subscales in the current sample was adequate to high (baseline ranging from α=.68 to α=.92, follow-up ranging from α=.67 to α=.95).

The PSOC [38] is a 16-item self-report measure of the caregiver’s sense of parenting efficacy (ie, perceived competence, problem-solving ability, and capability) and satisfaction (ie, the extent to which the individual enjoys the parenting role or experiences parenting frustration and anxiety). The items in the PSOC are answered on a 6-point scale ranging from “strongly disagree” to “strongly agree.” Higher scores on the Efficacy scale indicate lower efficacy, and higher scores on the Satisfaction scale indicate higher satisfaction. Internal consistency in the current sample was high (baseline: α=.80 for satisfaction and α=.80 for Efficacy, follow-up: α=.84 for Satisfaction and α=.80 for Efficacy).

The FMN Questionnaire [39] includes 19 items relating to potential needs of caregivers of children with FASD or PAE and assesses the degree to which caregivers perceive these needs have been met. It is based on a measure developed for traumatic brain injury [40] and has been used in several previous studies testing FMF and derivative products [21,24,27,28]. Each need is rated on a 4-point scale (1=not at all met and 4=a great deal met) and is scored by computing the average response across items. Higher scores reflect caregivers’ perceptions of more needs being met. Internal consistency in the current sample was high (baseline α=.90 and follow-up α=.93).

The K&A Questionnaire [39] was developed for use in the FMF Program. The version of the K&A used in this study was adapted to emphasize the content most relevant to the FMF Connect app and was informed by response patterns in the larger feasibility trial [21]. It includes 28 core items testing knowledge of FASD, advocacy, and key FMF Program principles. It is scored by calculating the number of items answered correctly. Furthermore, 8 items are rated on a true or false scale, and 20 items are multiple choice. For this study, a total score was computed by summing correct answers. Higher scores indicate higher levels of knowledge of FASD and advocacy. Internal consistency was not calculated for this measure as it measures content mastery targeted by the app versus a unified construct.

The ELS was developed for this study to measure caregiver-reported child adaptive functioning. The ELS includes 20 items that are each rated twice by the caregiver, yielding 2 scales, Responsibility and Smoothness. Sample items include “getting ready in the morning,” “making a transition between tasks,” and “calming down when upset.” The Responsibility scale, rated on a 5-point Likert scale, assesses to what degree the child and/or adult is responsible for the behavior. Lower scores reflect that the adult takes full or most responsibility for the behavior, whereas higher scores reflect that the child takes greater responsibility and independence in the behavior. The Smoothness scale, rated on a 5-point Likert scale, assesses how smoothly the behavior goes, ranging from “rarely goes well” to “almost always goes well.” Higher scores reflect greater smoothness. Internal consistency in this study’s sample was high (baseline: α=.92 for Responsibility and α=.88 for Smoothness, follow-up: α=.94 for Responsibility and α=.90 for Smoothness).

The Self-care Assessment (SCA) [24,41] was administered at follow-up. Caregivers were asked how their self-care changed over the past 3 months, rated on a 5-point scale (1=a lot less self-care, 2=somewhat less self-care, 3=no change in level of self-care, 4=somewhat more self-care, and 5=a lot more self-care).

The uMARS [42] is a 26-item survey measuring user perception of the quality of mHealth apps. It includes 6 subscales—Engagement, Functionality, Aesthetics, Information, Subjective Quality, and Perceived Impact, and a total App Quality score (summary of first 4 subscales). The uMARS was completed at follow-up only. Caregivers rated the items on a 5-point scale that was specific to each question; higher scores indicate more positive impressions of the app. Internal consistency in this study’s sample was adequate (α=.62 for Engagement, α=.76 for Functionality, α=.82 for Aesthetics, α=.66 for Information, α=.72 for Subjective Quality, α=.93 for Perceived Impact, and α=.93 for App Quality).

Randomization was conducted using REDCap’s Randomization package. The allocation sequence was developed by a biostatistician at the University not affiliated with the project to randomly allocate participants equally across the 3 study arms. No stratification or blocking was used. The University Research IT department implemented the allocation table in the REDCap Randomization package, and the research team did not have access to the sequence. Once the study coordinator determined a participant was eligible and baseline measures were complete, the study randomization form would determine the participant’s group status. The coordinator then informed the participant of their group status and sent instructions on next steps (ie, how to install the app or wait time) via email. Participants in the intervention groups were given a username and password and were given instructions on how to install the app from the Apple App Store. The research assistant processing survey data and payments could not access the REDCap Randomization package and was blind to group status.

Data analyses were conducted using SPSS (version 29.0; IBM Corp) and Mplus 8 (Muthén & Muthén). All variables were assessed for outliers, skew, and kurtosis. No extreme outliers or skew or kurtosis were detected. Missing data in SPSS were handled using multiple imputation. Missing data in Mplus used Full Information Maximum Likelihood (FIML).

Demographic variables were analyzed using descriptive statistics in SPSS. Group differences in demographic variables were assessed using 1-way ANOVAs for continuous variables and Pearson chi-square analysis for categorical variables.

Attrition analysis involved analyzing demographic differences at baseline, 6 weeks, and 12 weeks between those who did complete each timepoint and those who did not. For demographics, these differences were conducted using 1-way ANOVAs including all 3 timepoints. Pearson correlations (r) were used to detect relationships between baseline variables, demographic variables (caregiver age and child age), and attrition at all timepoints. App usage and perceived app quality were assessed using descriptive statistics and independent samples t tests grouped by FMF Connect+Coaching and FMF Connect.

Intervention change analyses were intent-to-treat, including all participants assigned to their respective randomly assigned groups whether or not they completed the FMF Connect program, and involved 2-sided hypothesis testing. Analyses used linear mixed modeling in SPSS for variables collected at the 3 timepoints (RCB, ELS, and PSOC). False discovery rate (FDR) corrections for all P values are reported. Simultaneous linear regression was used in Mplus for variables collected at baseline and 12-week follow-up. For regressions, group status and baseline levels were entered in step 1. All regressions were predicting outcomes at 12-week follow-up. FDR correction was used and reported for all regressions. Cohen d effect sizes were calculated for each outcome to demonstrate the magnitude of the intervention change. Additionally, simultaneous linear regressions were carried out to assess whether app usage within the FMF group predicted 6-week change scores and 12-week change scores.

Point-biserial correlations between group (FMF Connect vs WLC) and change scores (6-wk timepoint−baseline) were conducted to see if mediation was plausible. For those significant correlations, mediation analyses were conducted in Mplus using change scores as the mediator and the ECBI (child behavior) and PSOC (parenting efficacy or satisfaction) as the outcomes, controlling for baseline levels.

Recruitment for this study began on January 6, 2022, with enrollment ending on August 31, 2022. The study was completed on December 31, 2022, after all participants had the opportunity to complete follow-up surveys. Baseline sample demographics can be found in Table 2. No harms or unintended effects were identified for participants.

Baseline correlations by group can be found in Multimedia Appendix 1.

There were no significant differences between the FMF Connect+Coaching, FMF Connect, and WLC groups on any continuous variables (age, income, and household structure), nor were there significant group differences among the majority of categorical variables (gender, race, ethnicity, and relationship to child). There was a significant group difference on variables characterizing caregiver marital status. However, given the small size of marital status subgroups, this variable was not controlled for in further analyses. Given the absence of group differences, no other demographic factors were controlled for.

The trial CONSORT (Consolidated Standards of Reporting Trials) diagram is presented in Figure 2. There were no significant differences in demographic characteristics (caregiver’s gender, caregiver’s race or ethnicity, caregiver’s education level, comfort with technology, knowledge of FASD, income, child’s gender, child’s race or ethnicity, and caregiver type) between those who did and did not complete data collection at the 6-week or 12-week timepoints. Caregiver’s age and child’s age were not correlated with attrition. t-tests resulted in no significant difference in attrition by number of children in the home.

Differential attrition was also assessed for in baseline levels of functioning across groups. Greater attrition was identified among caregivers who, at baseline, reported lower child responsibility for daily tasks on the ELS for completion at 6 weeks (t₁₁₇=−2.85; P=.003) and 12 weeks (t₁₁₇=−2.69; P=.004). No other baseline scores were found to correlate significantly to attrition.

Table 3 provides means, SDs, and score ranges on the uMARS measure of perceived app quality by users in the 2 intervention groups with data at 12 weeks. Consistent with a past feasibility trial [21], users reported high perceived quality and satisfaction with the app, with nearly all scores averaging at or above 4 (maximum 5). The quality of information provided remains the highest rated domain across groups, consistent with the goal of the FMF Connect app. This scale alone was significantly different across groups. The FMF Connect+Coaching group reported slightly lower mean ratings on quality of information than did the FMF Connect app group.

App usage in the FMF Connect groups was measured through number of app openings, number of modules completed, number of fact sheet openings, and number of videos watched. Table 4 provides means, SDs, and ranges for participants in each of the intervention groups who registered usage in the app (which is higher than the number who completed follow-up surveys).

For those assigned to coaching, participants messaged their coach an average of 2.95 (SD 2.72) times. The number of messages sent from the participant to their coach ranged from 0 to 10, with 10 participants never using the coaching feature. Coaches messaged participants an average of 6.55 (SD 3.61) times. The number of messages sent from coaches to the participant ranged from 1 to 17.

App usage was not significantly different between the FMF Connect+Coaching and FMF Connect only groups on any of the usage measures, which suggests coaching did not significantly impact app engagement as hypothesized. Pearson r correlations were used to see if demographic and baseline data were related to app usage across FMF Connect groups. In terms of demographic factors, older caregivers opened the app (r=0.39, P<.001) and fact sheets (r=0.31, P=.009) more than younger caregivers. Caregivers who had less comfort with technology opened fewer PDFs in the Library (r=−0.23, P=.048). No other demographic factors were associated with app usage.

In terms of baseline measures, caregivers with lower parenting efficacy opened the app (r=−0.31, P=.007) and fact sheets (r=−0.30, P=.01) more. Caregivers with fewer family needs met opened the app (r=−0.25, P=.03) and fact sheets (r=−0.25, P=.03) more than families with greater needs met at baseline. Caregivers with lower baseline FASD knowledge (K&A) finished fewer Learning Modules (r=0.24, P=.04). Caregivers who rated everyday life tasks as less smooth (ELS) opened the app (r=−0.27, P=.02) and fact sheets (r=−0.27, P=.02) more. Caregivers with higher sensory seeking attributions (RCB) opened the app (r=−0.26, P=.03) and fact sheets (r=−0.26, P=.03) less often. Those with higher ability-based task attributions watched fewer videos (r=−0.23, P=.046). In summary, these score patterns suggest that caregivers reporting greater difficulties (eg, less smooth daily tasks, lower efficacy, and fewer needs met) generally engaged with the app more than those with relatively fewer difficulties.

The FMF Connect+Coaching and FMF Connect groups were aggregated for analyses examining intervention change, given minimal group differences across demographic characteristics, baseline and follow-up measures, and app usage. Analyses conducted with 3 separate groups, as documented on clinicaltrials.gov, can be found in Multimedia Appendix 2. These analyses did not use intent-to-treat or multiple imputation methodologies; therefore, they should be interpreted with caution.

Table 5 reports linear mixed models for all intervention outcomes with 3 time points (RCB, PSOC, and ELS). Pairwise comparisons for significant models are further discussed in this study.

There was a significant group×timepoint interaction on the RCB Task Willful measure (F₂=4.60, P=.01; Figure 3). Pairwise comparisons indicate significant group differences at the 12-week timepoint (mean difference=0.86, P=.002), with the FMF Connect group demonstrating lower endorsement of task willful attributions (mean 9.63, SE 0.30) compared with the control group (mean 10.13, SE 0.43). Interactions are no longer statistically significant when applying FDR correction.

No differences were detected for parenting efficacy and satisfaction or child adaptive behavior.

Linear regression analyses were conducted for all outcome measures collected at baseline and 12-week follow-up timepoints only (ECBI, FNM, K&A, and SCA). Baseline and FMF Connect group status were entered into the regression model simultaneously. The dependent variable entered was the measure at the 12-week follow-up to determine treatment effects (Table 6). Covariates were not included in analyses as there were minimal correlations or group differences on any of the demographic variables assessed (Table 2). Significant findings are summarized further in this study. FDR corrections for all P values are reported.

Group status (FMF Connect: n=80 and control: n=39) significantly predicted the FNM summary score at 12 weeks (b=0.21, P=.04), when accounting for scores at baseline and accounting for the interaction. The FMF Connect group showed significantly higher scores on the FNM post intervention (mean 47.56, SD 11.06) compared with WLC (mean 42.30, SD 11.29), indicating more family needs met. This prediction approaches significance when adjusting for multiple comparisons (P=.06).

Group status (FMF Connect: n=80 and control: n=39) significantly predicted the K&A summary score at 12 weeks (b=0.18, P=.04), when accounting for scores at baseline. The FMF Connect group showed significantly higher scores on the K&A post intervention (mean 21.28, SD 3.28) compared with the WLC group (mean 20.33, SD 2.69), meaning they acquired greater knowledge about FASD, advocating for their child, and core FMF Program concepts. This prediction approaches significance when adjusting for multiple comparisons (P=.06).

Group status (FMF Connect: n=80 and control: n=39) significantly predicted the SCA score at 12 weeks (b=0.17, P=.048), when accounting for scores at baseline. The FMF Connect group showed significantly higher scores on the SCA post intervention (mean 5.38, SD 2.46) compared with the WLC group (mean 4.401, SD 2.59), meaning they acquired more self-care skills compared with waitlist controls. This prediction approaches significance when adjusting for multiple comparisons (P=.07).

Group status did not predict child behavior (ECBI) at 12 weeks when controlling for scores at baseline.

Learning module completion and app opening were entered into a simultaneous regression model to predict change scores at the 6-week and 12-week timepoints (Multimedia Appendix 3). Learning module completion (b=−0.20, P=.02) and app openings (b=0.79, P=.02) had an effect on PSOC Efficacy change scores over 12 weeks, in that those who completed more learning modules and opened the app more had greater levels of parenting efficacy after 12 weeks. Caregivers who opened the app more showed more improvement on RCB Emotion Seeking at the 12-week timepoint (b=0.76, P=.03) and PSOC Efficacy at the 6-week timepoint (b=0.35, P=.002).

Point-biserial correlations between group (FMF Connect vs WLC) and change scores (6-week timepoint−baseline) were conducted to see if mediation was plausible. Group status was correlated with 6-week change scores on the RCB Disruptive Behavior (r=−0.24, P=.03) and RCB Task Willful (r=−0.26, P=.02) scales. Mediation analyses were conducted in Mplus using the model illustrated (Figure 4) looking at both the RCB Disruptive Behavior and Task Willful scores as the mediator and the ECBI and PSOC as the outcome. No significant mediation was found.

Families raising children with FASD experience poor access to evidence-based care in their communities given multiple systems-level barriers. These include limited professional training in FASD, long waitlists and provider shortages, and pervasive stigma [9,12,43]. Complementing efforts to build workforce capacity, this study tested the efficacy of FMF Connect, a first of its kind—a caregiver self-directed mHealth intervention. Results from this RCT provide initial support for the efficacy of the FMF Connect app for caregivers raising preschool and school-aged children with FASD or PAE. Relative to the waitlist group, groups that received the FMF Connect app had greater increases in FASD knowledge, reductions in willful attributions of child behavior, improved self-care practices, and more family needs met. These changes in caregiver outcomes are notable given the low intensity, self-directed nature of the intervention. Contrary to hypotheses, no significant group differences were found for child outcomes assessed. Consistent with our previous feasibility trial [21], caregivers gave high ratings for app quality, with the quality of information rated highest among indicators. App usage was relatively greater for caregivers reporting more difficulties with parenting efficacy, completing daily tasks with their children, and meeting family needs. Higher app usage was also related to greater changes in parenting efficacy scores across the 12-week intervention period.

The systematic development and testing of the FMF Connect app is particularly notable given that a 2024 review of parenting apps found that none of the available apps in app stores had been empirically tested [13]. The FMF Connect app is also part of a growing continuum of products, all derived from the standard FMF Program to meet the varying needs of children and youth of different ages, the families who care for them, and different settings in which they are treated. Similar to the standard specialist-led FMF Program, the FMF Connect app is related to important changes in caregiver and family functioning [24,27,28]. The magnitude of effect sizes for the FMF Connect app (small-medium) is generally smaller than the standard FMF Program (many medium-large), which is not surprising given the difference in duration (12 wk vs 9 mo) and format (self-directed vs specialist implemented) of the interventions. Dosage of content is also much more variable with the FMF Connect app (intervention groups averaging 56 of 131 core Learning Module sections; range 0‐148, including repeated content); past trials with the FMF Program have had nearly all families complete the full program. Across studies, the FMF Program and FMF Connect app result in consistent improvements in FASD knowledge and family needs met. Changes in caregiver attributions of child behavior have also been found across studies, although measurement approaches have been variable. Interestingly, studies testing the standard FMF Program have found medium to large effects in parenting efficacy, whereas this study did not find significant group differences after access to the FMF Connect app. However, higher levels of app usage did relate to greater change in parenting efficacy scores, suggesting greater dosage of content is important for this outcome. Caregivers across studies have reported a high level of satisfaction with both interventions. While the magnitude of effects for the FMF Connect app is more modest, the potential for scalability creates the possibility of a broad public health impact. This impact and reach could potentially be leveraged as additional products in the FMF continuum of care are developed, continuing to increase accessibility and capacity to provide evidence-based care.

In contrast to our hypothesis, text-based coaching did not result in improved app usage or outcomes for families. Across a broad range of mHealth apps designed to manage health conditions in adults, a positive correlation has been found between app engagement and the level of personal support provided [44]. However, there is a high level of heterogeneity in methods and outcomes for mHealth interventions using coaching, and many studies did not directly assess the effect of coaching on outcomes [45]. Another possible explanation for the pattern of findings in this study is that user engagement in coaching was lower than anticipated. Study team observations noted that a few users reported difficulty finding or navigating the coaching messaging component. It is also possible that the users turned off app notifications that would have alerted them to new messages or were not interested in this feature. A future direction could explore whether peer coaches (vs graduate students) increase engagement.

We also predicted that more app usage would relate to greater improvement in outcomes. However, this hypothesis was not generally supported; usage did not correlate with greater effects for the majority of outcomes, with the exception of parenting efficacy noted above. Possibly, there is a threshold effect after accessing a certain amount (or type) of content. Alternatively, change may be related to more individualized patterns of usage or needs based on caregiver and family circumstances. Future larger-scale investigations may allow for greater power and ability to use machine learning techniques to explore usage patterns in more nuanced ways.

Finally, based on the theorized logic model for the FMF Program, this study hypothesized that caregiver attributions of behavior would mediate changes in caregiver efficacy and satisfaction and child behavior outcomes. Results of this study did not support this hypothesis. While the intervention group was correlated with 6-week change on 2 scales, mediation was not significant. In addition, the app was not associated with change in parenting efficacy, satisfaction, or child behavior over the 12-week evaluation period. Given intervention effects are seen on these outcomes in the longer duration follow-ups for the specialist-led FMF program [24,27,28], it is possible that more intensive intervention and/or follow-up is needed for these outcomes.

Results from this study should be interpreted in the context of several notable strengths and limitations. First, this study empirically tested one of the first stand-alone, self-directed mHealth interventions targeting parenting. Although the FMF Connect app is tailored to caregivers raising children with FASD, findings from this study have broader relevance for parenting programs and use of mHealth interventions more generally. Much of the content and form of FMF Connect could be adapted and trialed with other related populations. The FMF Connect app was rigorously developed and tested using a systematic user-centered design approach [21-23]. The app was derived from an evidence-based intervention with a strong theoretical foundation and adapted for self-directed administration on a mobile platform. Our approach is consistent with best practices in mental health app development and broader principles of user-centered design [20,46].

The generalizability of study findings may well be impacted by characteristics of study participants who enrolled in the study and other design features. Although careful attention was paid to participant reach in previous feasibility trials [21] and efforts were made to recruit a diverse participant pool, enrolled participants were largely White, relatively wealthy, adoptive mothers. Caregiver education and income were higher than national averages. Our study was also limited to participants with Apple iOS devices, which may have influenced patterns of participant demographics. For example, commercial survey data indicate iOS users tend to have higher average income, longer duration of smartphone use per day, and be younger in age than Android users [47]. The decision to limit the study to iOS devices was made given poorer stability of our Android prototypes across user devices in previous feasibility studies [21,22]. This decision likely eliminated about 40% of potential participants based on rates of Android users in our past studies and current market share in the United States [47]. Expanding the app to the Android platform and further engagement or co-design work with demographic groups not well-represented in this study are important future steps to address this limitation. Since this trial, we have subsequently rebuilt the app on Flutter (Google), which allows seamless development and refinement across iOS, Android, and web platforms. We are also in the process of commercializing the app for broad distribution. Some families may not be interested in participating in research due to barriers or mistrust but might be more likely to download the app from the app store and use it outside of the research context. It is possible these steps will increase the diversity of users. Additionally, targeted outreach and collaboration with community organizations serving underrepresented populations could help address this generalizability gap.

As with most behavioral intervention studies, participants were aware of their treatment condition, which could have potentially biased follow-up responses. FMF Connect targeted multiple child and family outcomes. As a result, multiple analyses were run to assess the range of possible intervention effects targeted. The number of analyses may have increased the likelihood of false positive intervention effects, as suggested by FDR corrections.

This study provided initial evidence for the efficacy of the FMF Connect app for targeted caregiver outcomes, with effect sizes in the small to medium range. As an mHealth app, the FMF Connect intervention has potential for scalability and accessibility. Even with modest effect sizes, this potential for scalability could result in a sizeable public health impact, especially for families who may have difficulty accessing evidence-based resources in their communities or have other barriers to care. The FMF Connect app is not intended to replace diagnostic and treatment services delivered by providers. Yet, given the severe lack of FASD-informed providers in the United States, this resource can help provide families with easy access to evidence-based information and ideas for how to access and advocate for services in their communities. Beyond that, through the Family Forum, the app can provide connections with other families with similar experiences, providing peer support.

Building on this work, our team is developing and testing a new program called FMF Connect Pro to further expand the continuum of care for families based on the FMF Program. This effort aims to train mental health providers in how to screen for PAE, diagnose FASD (using Neurobehavioral Disorder Associated with Prenatal Alcohol Exposure in the DSM-5 [Diagnostic and Statistical Manual of Mental Disorders {Fifth Edition}] [3]), and support families using the FMF Connect app within their clinical practice. This middle tier between the self-directed FMF Connect app and the standard specialist-led FMF Program has the potential to increase identification of FASD in settings with increased prevalence [48,49] and enhance access to FASD-informed treatment for families. It could also provide clinicians with easy-to-use tools and a lower training investment option to gain competencies with this population across a range of mental health settings. It is also possible that caregiver-professional relationships may increase engagement and use with families, especially those less likely to use technology on their own. Engagement with FMF Connect Pro could also increase future provider uptake of the standard specialist-led FMF Program, which may often be the best fit for higher-acuity children and families.