Overall, this was an exploratory sequential multiple methods study consisting of semistructured interviews using the critical incident method and a quantitative survey where participants were asked to rate emotional and cognitive outcomes before and after using NBSchat.

The study was reviewed by the University of Utah Institutional Review Board (IRB), which deemed it exempt (IRB_00179621). Interview participants were provided with a concise study summary with a consent cover letter prior to their interviews and were compensated for taking part with US $50 via a secure online funds transfer system. Verbal consent was obtained at the beginning of each interview. Participants were informed that a transcript of their audio recording would be prepared and deidentified for the study. In accordance with the institutional review board protocol, participants were also informed that the audio files and transcripts would be destroyed after the study. Survey participants were provided with a consent cover letter prior to beginning the survey, and those who completed the survey were compensated with US $14.75 for their participation in accordance with the policies of the survey panel provider. Participants were informed that their responses would be kept confidential, that no identifying information would be disclosed, and that results would be reported in aggregate. Chatbot logs contain timestamps and paths through the conversation for each user; they do not contain identifying information.

To ensure that we developed a robust, patient-focused technology that is tailored to parents’ needs, we followed a participatory approach. Guided by principles of health communication [25], an interdisciplinary team that included genetic counselors, informaticists, a labor and delivery nurse, an NBS expert, and a communication scientist developed an initial draft of the chatbot script [26]. In the initial draft, we sought to address information gaps that cause preventable anxiety: not understanding the implications of the result, wondering about the status of other family members and reproductive health, and navigating the next steps in the health care process. The script was then iteratively refined based on the group’s feedback through meetings and offline discussions.

NBSchat was developed using the GARDE-Chat [27] platform developed by investigators at the University of Utah. GARDE-Chat includes a chatbot authoring tool that enables the creation and management of multiple rule-based, large language model (LLM)–based, or hybrid chatbots. LLM-based chatbots, such as ChatGPT, are known to occasionally provide inaccurate information (also known as “hallucinations” [28]) and answer questions that fall outside the predefined boundaries of a chatbot. To avoid these issues, we chose to use a rule-based approach with a predefined conversation script and a fixed set of possible questions and responses that provide accurate and consistent information to parents (Figure 1).

All interviews were audio-recorded and transcribed verbatim by a professional transcription service. We aimed to produce actionable, accurate insights that could be incorporated into the design of NBSchat; therefore, qualitative data collected through the interviews were analyzed using the Rapid and Rigorous Qualitative Data Analysis technique [33,34]. The research team developed a phase 1 matrix in Microsoft Excel that reflected the topics assessed in the interview guide and their preliminary codes, interpreting the data [35]. All members of the research team read the phase 1 matrix and discussed the goals of the analysis. Based on these discussions, authors CG and ACM further reduced the data in the phase 1 matrix and created a phase 2 matrix that provided actionable feedback with supporting quotes addressing the overarching research question. The phase 2 matrix incorporated analysts’ “focused codes [36].” Data reduction and condensation occurred iteratively until the team agreed that the final phase table addressed our overarching research question. Disagreements were resolved through discussion.

A survey instrument was developed using the Qualtrics platform and incorporated 3 attention check items to assess response quality [44-46]. Participants who failed one or more attention checks were removed from the final analysis to minimize the impact of insincere survey takers. Analyses were conducted using complete cases; sample sizes varied slightly by analysis due to item nonresponse. Participant knowledge of genetics and health literacy was assessed; then, participants were presented with a simulated NBS result indicating that a newborn had screened positive for SCT and were asked to complete a modified version of the AURA and FACToR instruments to capture baseline self-efficacy and emotional responses. Following this, participants interacted with NBSchat and were then asked to complete the AURA and FACToR questions to evaluate changes in outcomes. Participants then provided usability and usefulness feedback, followed by demographic information. Given the exploratory mixed methods design, recruitment targets were inflated to account for an anticipated incomplete rate of approximately 30%, with the goal of obtaining at least 100 open-ended responses to support robust and diverse qualitative perspectives and a final analytic sample sufficient to detect small pre-post changes in AURA and the FACToR subscale scores. Missing data were minimal (1 of 250 participants missed one item on a multi-item scale;<0.5%). Scale scores were computed as the mean of completed items. Given the negligible level of missingness, no additional imputation procedures were conducted.

Survey data were analyzed using Microsoft Excel. Descriptive data, including means and standard deviations, were obtained for all measures. The significance of differences in measures before and after using NBSchat was determined using the paired t test, with statistical significance assessed as a P<.05. Open-ended comments were coded into 3 categories: positive comment, negative comment, and neutral comment during a content analysis. After reading all the comments, 2 reviewers agreed upon codes and selected representative quotes. The integration of survey with interview data occurred by comparing themes identified in interviews with patterns observed in survey responses.

In this study, interviews and surveys with new parents demonstrated a clear information gap surrounding SCT results, which can lead to distress and uncertainty. Parents consistently sought reassurance and rapid answers to immediate health concerns. A key finding of the survey is the significant decrease in both distress (negative emotions) and uncertainty following the use of NBSchat. We also observed a small but significant increase in positive emotions. While some parents were content to find out that SCT is not the same as sickle cell disease, many parents reported taking a deeper dive into the menu of questions, with the most common theme being that they learned something new. The result explanation and the list of further resources were parts of NBSchat that filled an information gap. Overall, the study results suggest that NBSchat is a promising digital health tool for scalable and efficient delivery of NBS information to assist in the return of NBS results.

Because trait results represent the largest category of results returned by NBS programs, follow-up teams frequently receive calls from parents seeking clarification. The problem is compounded by delays: results are often mailed, followed by a lag before discussion with the infant’s clinician and an even longer wait for an appointment with a genetic counselor. The goal of NBSchat is not to replace genetic counseling but to supplement it by preparing parents with foundational knowledge and reducing anxiety so they arrive at appointments more informed [47], more confident, and able to ask targeted, personally relevant questions [48,49]. Given genetic counseling workforce shortages [19] and the limited availability of counselors in many regions [20], NBSchat also represents a practical intermediate solution to address gaps in parent communication and support, reflecting a focus on patient-centered care [50].

Previous research has identified significant variation in SCT disclosure to parents during the newborn period [17,21]. NBSchat offers standardized information to all parents, regardless of where screening was performed, and can be tailored to the unique needs of a state’s NBS program. The clinicians in the study were concerned about the communication of trait results across language barriers; in some cases, even with an interpreter. This is an area where deploying a chatbot in the parents’ first language could be very impactful.

The exploration of usability in the survey found the chatbot to be functional and user-friendly, with only minor areas for potential improvement. According to standard SUS interpretation guidelines, a score of 73 places the tool in the “Good” range and corresponds roughly to a B rating, reflecting positive user experience and acceptance.

The survey found that participants’ communication self-efficacy was high before using NBSchat, and there was no significant change after using it. A mean score of 13.5 out of 16 on the AURA scale represents 84% of the maximum, indicating that participants had strong confidence in asking medical questions, with good perceived understanding and a high likelihood of remembering or using the information effectively. Participants reported feeling confident in engaging in health care discussions and seeking clarification. This likely reflects the sample of survey participants and may not represent the communication self-efficacy of parents who receive a screen-positive result for SCT.

NBSchat can be accessed via a web browser on any internet-connected device, such as a smartphone, tablet, or computer. In 2024, 99% of US adults aged 18 to 49 years reported using the internet [37], and the mean age at first birth in the United States was 27.4 in 2022 [51]; thus, it is likely that the majority of infants in the United States will be born to an adult who uses the internet, whether through a smartphone or other internet-connected device.

The alignment between interview themes and survey findings suggests that user experiences identified in the qualitative phase were not isolated but reflected broader user patterns. The results suggest that NBSchat is an effective and highly scalable approach to delivering patient-tailored education to address the unmet informational needs of parents and alleviate emotional distress. These findings align with previous studies, which provide strong evidence demonstrating the effectiveness of health service delivery and education through scripted chatbots in various health domains, with demonstrated benefits in lowering rates of depression and anxiety [24] and in the delivery of pre-test cancer genetic services and genetic testing [23].

Although recent health care chatbots have started to leverage artificial intelligence–based approaches using LLMs [52], rule-based chatbots are simpler to implement, safer (ie, the design team can ensure that the chatbot only provides accurate, compassionate, and culturally adapted information), constrained to only answer questions about the state’s NBS program and the child’s NBS result, and effective for goal-oriented use cases with a clearly defined end point, such as educating users about NBS results and processes [53]. Participants responded positively to having a menu of options to select for further information, and at least some were unsure what to ask without the structured question format, suggesting that they did not perceive a need for an artificial intelligence–based approach. Furthermore, in prior work using a chatbot to provide information about genetic services and genetic test results, few patients asked open-ended questions, indicating that a rule-based approach may have met the patients’ informational needs [54]. Therefore, scripted chatbots, such as NBSchat, are, at least to this time, an adequate and safer approach for chatbot implementation, especially in sensitive topics such as the explanation of NBS results. This approach also enables each state’s follow-up team to tailor the content unique to their program and update it as necessary through a distributed content authoring approach using the GARDE-Chat platform. Nonetheless, we acknowledge that a hybrid approach, using LLMs at end points in the script, could allow parents to explore more nuanced queries. Future work will explore adapting NBSchat to support free-text questions using a carefully constrained LLM approach, such as a retrieval-augmented generation framework grounded exclusively in authoritative NBS content and explicit scope boundaries to balance flexibility with safety.

Limitations of this study include potential selection bias in the parent interviews and survey toward individuals with higher health literacy and genetic knowledge. Participants in the survey had high genetic knowledge on average, with little variation in scores. We also found that 93.2% (n=233) of the participants had home internet service and 99.6% (n=249) had a smartphone, showing a high level of digital access. Additionally, using attention checks may introduce bias toward higher digital literacy respondents, although we minimized this risk by using plain and unambiguous language. Participants could not be blinded to NBSchat usage, which may have changed their information-seeking behavior because they knew they were being “treated.” The survey relied on self-reported measures, which may be affected by social desirability bias. The simulation of delivering hypothetical test results may not reflect real-world outcomes.

While the survey recruited participants from across the United States and attempted to recruit a representative sample across education levels, the study population reflected relatively high literacy and consistent access to digital technologies. Consequently, participants were not proportionally representative of the ancestral and ethnic groups that bear a disproportionate burden of SCT and disease. Future evaluations of NBSchat should focus on recruitment approaches that better reach these populations, including partnering with newborn screening programs, safety-net health systems, and community-based clinics serving predominantly affected families. Future studies testing NBSchat within these settings and populations are needed to more accurately assess its acceptability and effectiveness among those most affected by SCT. The clinical trial would also demonstrate the generalizability and modularity of NBSchat by adapting it to the particular workflow of a state follow-up team.

The findings should also be interpreted in light of important constraints, including the absence of a control group and the use of hypothetical screening scenarios. Accordingly, the observed pre-post changes in emotional measures cannot be interpreted as causal effects and instead represent preliminary signals supporting the feasibility and potential value of NBSchat, which warrant confirmation in controlled, real-world studies.

In this proof-of-concept demonstration, NBSchat was developed in English. Future iterations of NBSchat are planned to include additional language options, informed by community needs and implementation considerations, to enhance accessibility for linguistically diverse families and ensure the delivery of newborn screening information to all populations.

We developed and tested NBSchat, a prototype chatbot to support parents who receive an NBS SCT screen-positive result before speaking with a clinician. Both parents and clinicians reported needing to search multiple sources to understand NBS, suggesting that NBSchat could fill this gap by delivering clear, reliable information when results are received. In a simulated evaluation, parents using NBSchat reported reduced negative emotions and greater reassurance, addressing the distress and uncertainty often experienced while awaiting clinical follow-up. These findings represent preliminary signals of potential benefit and support further evaluation of NBSchat in controlled, real-world clinical settings.