Background

The digitization of patient records and adoption of emerging technologies have enhanced patient care in U.S. healthcare but have also introduced vulnerabilities exploited for cyberattacks. Despite advanced technical safeguards, human factors including social engineering and human error remain major contributors to data breaches, costing approximately $9.8 million in 2024, according to IBM. These breaches compromise sensitive patient information, disrupt operations, and erode public trust, highlighting the need for human-centric threat modeling.

Objective/Purpose
This study aims to identify key factors responsible for healthcare data breaches reported in the last decade and evaluate the effectiveness of human-centric threat modeling in reducing cybersecurity risks for healthcare organizations.

Methods
A retrospective analysis was conducted on healthcare data breaches reported between January 2015 and January 2025, sourced from the U.S. Department of Health and Human Services (HHS). Key trends were identified, and findings were enriched with open-source reports to provide insights. Additionally, we assessed the effectiveness of the OCTAVE threat model in mitigating healthcare organizational risks.

Results
Analysis of 5,474 breaches revealed that healthcare providers had the most incidents (3,988 breaches), while cyberattacks on business associates affected the most individuals (301 million). Network servers were the most common breach location (44.67%), followed by email (25.23%). This study’s findings highlight social engineering as primary entry tactics, even for network server and email attacks, which point to human errors contributing significantly to these breaches.

Discussion/Conclusions
Human-centric threat modeling, workforce training, and continuous risk assessments are crucial for protecting and restoring public trust in U.S. healthcare systems.

BACKGROUND: Multi-level interventions (MLIs) are needed to address barriers associated with racial and ethnic minority participation in cancer clinical trials (CCTs).

OBJECTIVE/PURPOSE: To develop and refine digital tools for an MLI aimed at increasing referral and enrollment of minority patients to CCTs at Moffitt Cancer Center (MCC) using user-centered design (UCD).

METHODS: Digital tools were developed to enhance patient education and decision-making, patient referral, portfolio evaluation, eligibility criteria assessment, and enrollment monitoring. Intended users (patients, physicians, CCT coordinators) of each tool were recruited to participate in three 1-hour virtual UCD sessions standardized to include (1) a description of the tool functionality and an initial sketch, (2) a prototype, and (3) the final product. Interview guides were used to elicit feedback on Learner Verification constructs of attraction, comprehension, self-efficacy, persuasion, and on expected changes to facilitate referral/enrollment of patients to CCTs following tool usage.

RESULTS: Thematic findings included interface preferences, confusion regarding data sources, the need for user tutorials, and improvements to data visualizations. Some feedback also led to more considerable modifications, such as changes to the tool's intended user, different ways of referring patients to MCC, and integrating workflows across all digital tools by creating a common landing page. Several users also indicated that tool usage may lead to improvements in monitoring and designing CCTs.

CONCLUSIONS: Findings highlight the value of seeking input to design digital tools that meet user needs and increase potential adoption. Digital tool usage will be evaluated throughout the study to determine if future modifications are needed.

ClinicalTrials.gov has served as the premier repository of clinical studies for 25+ years. It currently provides information for 530K+ global studies covering a range of conditions including the nation’s top health concerns. ClinicalTrials.gov was recently modernized, delivering an improved user experience on an updated platform.

The presentation will highlight modernized ClinicalTrials.gov website features for three key stakeholders seeking information about clinical studies related to current and emerging public health issues:

1. Researchers assessing the current landscape of public health priorities.
2. Healthcare providers seeking research status and findings for emerging public health risks to share with their patients.
3. Individuals seeking research on health conditions affecting a loved one.

This presentation will demonstrate how features on the modernized ClinicalTrials.gov allows users to efficiently review clinical trial records resulting in the easy discoverability of public health information, including:

• Focus Your Search enables users to filter retrieved records using selected search fields and refine their simple search queries.
• Expert Search allows users to search 400+ fields via complex search queries to dig deeper into a particular health area.
• The streamlined Study Records help users more easily find information.

In this presentation, attendees will learn about the importance of user experience (UX) activities conducted throughout the ClinicalTrials.gov modernization. For example, 30+ user surveys were deployed to gain input on Expert Search and the overall download experience. These feedback activities helped prioritize feature development and improve modernized ClinicalTrials.gov website navigation, to better inform users’ understanding of the top impact areas in public health.

Introduction: There has been an expansion of mobile health (mHealth) solutions for prenatal care with limited evidence of efficacy in high-need maternal populations over the last decade, including among Black, Latino, and Asian birthing persons.

Objective: We aim to identify factors associated with eHealth communication, a major functionality of mHealth solutions, among pregnant women at the epicenter of NYC’s maternal health crisis.

Methods: This study presents baseline survey findings from our We Care About Brooklyn trial, which aims to improve postpartum care in a low-resource setting. At baseline, participants were asked, “In the past 12 months, have you used a computer, smartphone, or other electronic device to do any of the following: email providers, check test results, schedule appointments, or use telemedicine?” We use multivariable logistic regression models to assess the association between sociodemographic factors and each outcome.

Result: 60 participants were randomized into our study and completed the baseline survey. Older age, low education, and foreign-born status were all significantly associated with decreased use of eHealth communication. For instance, older and foreign-born pregnant women in our study were less likely to report checking medical test results via a patient portal (aOR = 0.88, CI = [0.79-0.98]; aOR = 0.15, CI = [0.03-0.82]). Similarly, older and less educated women were less likely to report using email to communicate with their healthcare providers (aOR = 0.91, CI = [0.83-0.99]; aOR = 0.22, CI = [0.05-0.86]).

Conclusion: These findings suggest that in addition to age and education level, nativity may influence provider-patient communication patterns, which could have implications for healthcare access and maternal outcomes. This knowledge can increase precision when designing mHealth interventions and allow tailoring for different population segments; for example, relative to younger or US adults, older or foreign-born adults may require more intensive training and support to encourage engagement with mHealth.

Background:
Increasing infectious diseases trends require robust public health surveillance bolstered by health information technology tools. Electronic Case Reporting (eCR) refers to reporting of mandated case data electronically to public health agencies and contains both structured and unstructured/text data.

Objective:
Purpose is to evaluate the Amazon Web Services (AWS) Bedrock generative Artificial Intelligence (AI) tool for its eCR text analysis, in comparison with expert assessment at the Minnesota Department of Health.

Methods:
The first assessment phase comprised of identifying key eCR data fields required for public health. Multiple discussions amongst informatics, epidemiologists and program experts were conducted. Prioritized fields will be utilized in second phase to compare AWS Bedrock data extraction to a manual process.

Results:
Phase 1 yielded a list of elements (n=22) identified as epidemiological priorities and covering a range of reportable conditions. Data fields (n=15) include patient IDs, visit date/time, vital stats, symptoms, medications, sexual partners, pregnancy status, risk factors, travel, exposure data, congregate living, housing status, drug use, incarceration and death date. Contextual criteria (n=7) comprises parsing health encounters/conditions per visit, multiple labs for same disease, different doses of same medication at different intervals, medication ordered, but undispensed, medications embedded in problem list, social issues (food/housing insecurity) as past/present based on dates, travel information (past/present) from various eCR sections.

Discussion:
This research (phases 1 and 2) contributes to understanding incipient issues in eCRs using AI tools to facilitate public health data modernization. As eCR expands from infectious diseases to non-communicable (chronic) conditions, the impact is broader.

Background

This training addresses shelter staff’s needs in identifying and supporting survivors of intimate partner violence (IPV) who may have experienced traumatic brain injuries (TBI). Developed by Banyan, who was contracted to do this work, the training’s primary audience is shelter staff, but it will be accessible through CDC TRAIN. It can benefit a broader range of service providers by filling a critical gap in IPV response and enhancing trauma-informed care by equipping staff with essential knowledge.

Methods

A literature and environmental scan informed the development of a digital training prototype. Structured focus groups with nine IPV shelter staff and nine IPV experts provided feedback on its effectiveness. Additionally, UX research—including individual sessions and a small focus group—assessed usability and accessibility. This multi-method approach ensured that the training was evidence-based and user-centered.

Results / Conclusions

The literature review and environmental scan findings provided best practices for terminology, screening, accommodations, and referrals. Focus group participants emphasized the need for trauma-informed language, appropriate screening tools, and better referral resources. While the UX study confirmed the training’s user-friendliness, participants suggested adding knowledge checks and interactive elements to maintain engagement.

This research highlights the crucial need for specialized training to improve TBI identification among IPV survivors. While the training offers a significant step forward, further research is needed to refine best practices, enhance implementation, and expand IPV response systems to ensure more effective care for this population with complex needs for healing from trauma.

Background: Within electronic health records (EHR), immunizations may only be reflected in the text-based narrative sections. This has become an increasing issue as vaccination reporting requirements have relaxed since the conclusion of the COVID-19 pandemic.

Methods: In response to this, we developed rule-based natural language processing methods to extract vaccination data from clinical notes to identify administration of vaccines and other immunization products for COVID-19, influenza, and respiratory syncytial virus (RSV). The rules were developed using a combination of synthetic data and multiple sources of real-world data, allowing us to account for a range of variations in the presentation of immunization administrations in notes. The methods identify the vaccine type using a dictionary of vaccine types and pattern-matching techniques. Furthermore, we compared our methods to the methodology established by Deady et al. (2021) for detecting influenza vaccines. Improvements over Deady’s method include accounting for grammatical dependencies and negation of vaccination administration verbs, as well as expansion to COVID-19 and RSV immunizations. We assessed concurrence with real-world structured data and compared performance to the method established by Deady et al.

Results: Through a manual review, we created a ground-truth label of 400 notes to calculate precision and recall. The precision was 70%, 91%, and 53% for COVID-19, influenza, and RSV immunizations, respectively. The recall was 97%, 92%, and 96% for COVID-19, influenza, and RSV immunizations, respectively. Compared to Deady et al., our approach achieved an 89-fold improvement in recall in identifying influenza vaccine administration with comparable precision.

Background: Everyday discrimination is associated with worse sleep health, especially in racially and ethnically minoritized groups. Few studies have examined this relationship using real-world sleep data beyond seven days. This study examined associations between everyday discrimination and Fitbit-derived sleep health metrics and explored race and ethnicity as potential moderators.

Methods: Data from All of Us Research Program participants with Fitbit data within one year of survey completion were analyzed (n=15,736). Using the Everyday Discrimination Scale, discrimination was categorized as minimal, low, moderate, and high. Fitbit sleep measures included average total sleep time (TST), TST variability, wake after sleep onset (WASO), and sleep efficiency. Multivariable linear regression models were used, with interaction analyses for race and ethnicity (non-Hispanic White, Black, Asian, Multi-Racial, and Hispanic All Races).

Results: Participants were 81.8% White, 70.6% cisgender women, and 17.8% reported moderate to high discrimination. Compared to White individuals, all racially and ethnically minoritized groups had shorter average TST, with Black individuals being the most affected (29.0 to 40.7 minutes shorter across all discrimination levels). Compared to minimal discrimination, moderate and high levels were associated with 13.5 (95% CI: -17.1, -9.9) and 22.6 minutes (-28.3, -16.8) shorter average TST, and 7.3 (5.6, 8.9) and 11.6 minutes (8.9, 14.2) greater TST variability, respectively. No associations were observed for WASO or sleep efficiency, and no modification by race and ethnicity was found.

Conclusions: Higher everyday discrimination was associated with shorter and more variable sleep. While race and ethnicity did not moderate these relationships, sleep health disparities persisted.

Background:

Men’s health in South Africa remains critically underserved, with men significantly less likely than women to access health services across prevention, treatment, and retention. This disparity contributes to poor HIV outcomes, undiagnosed non-communicable diseases, high suicide rates, and untreated trauma. Mental health is particularly neglected, often masked by substance use, aggression, or emotional withdrawal. Structural drivers including unemployment, poverty, stigma, and rigid gender norms exacerbate these inequities. Yet, few digital solutions have been designed with men’s cultural and psychosocial realities in mind.
Objective: To co-design a gender-sensitive digital intervention addressing men’s mental health, HIV-related stigma, and health-seeking behaviors in the Gauteng and Free State provinces.

Methods:

Between September 2024 and March 2025, a qualitative study using grounded theory methodology engaged 112 men aged 15–49 through focus groups, narrative interviews, and design workshops. Participants included rural and urban men, employed and unemployed, heterosexual and LGBTQI+. Key informants (n=23) included clinicians, traditional leaders, psychologists, and peer mentors. Thematic areas included masculinity, digital behaviors, mental health, HIV stigma, and coping strategies.

Results: Participants expressed a strong preference for anonymous, non-judgmental digital support particularly platforms offering self-guided emotional literacy, relatable content, and peer-led storytelling. Findings informed the design of a digital platform, ManWell, centered on: 1) mental health triage and support, 2) private health symptom checking, and 3) digital storytelling.

Conclusion:

Digital health interventions for men must be trauma-informed, gender-transformative, and culturally grounded. ManWell offers a replicable model for engaging men on their terms restoring agency, trust, and dignity in care.