Winter 2024 Tier 2 pilot grant awardees share their final project findings

Investigators
Gregory C. Valentine, Department of Pediatrics
Kristopher Kerns, Department of Periodontics
Benjamin Shayo, Baylor College of Medicine Children’s Foundation-Malawi
Jessie Mlotha-Namarika, Baylor College of Medicine Children’s Foundation-Malawi
Joseph Mhango, Baylor College of Medicine Children’s Foundation-Malawi
Peter Milgrom, Department of Oral Health Sciences
Kathryn Gray, Department of Obstetrics & Gynecology
André Ritter, School of Dentistry
Maxim Seferovic, Baylor College of Medicine
Theo Bammler, Department of Environmental & Occupational Health Sciences
Tessa Rue, Department of Biostatistics

Project summary
Periodontal disease (including gingivitis and periodontitis) is an established risk factor for adverse pregnancy outcomes (APOs) including preterm birth (PTB), low birthweight offspring, pre-eclampsia, eclampsia, gestational hypertension, stillbirth and miscarriage. Xylitol, a naturally-occurring sugar substitute commonly used in many chewing gum products, prevents growth of bacteria, including those strongly associated with the development of periodontal disease in vitro. We previously completed a cluster-randomized trial in Malawi to test the efficacy of xylitol gum in preventing maternal periodontitis and subsequent PTB. Among N=9,670 participants, xylitol gum use significantly reduced maternal periodontal disease compared to control (standard care including dental visits but without chewing gum; p=0.01 vs p=0.23). Xylitol gum also significantly reduced rates of PTB (12.6% vs. 16.5%; adjusted relative risk (aRR) 0.76, 95% CI 0.59-0.99). Thus, xylitol chewing gum may be a safe affordable and readily accessible, over the counter intervention that aims to reduce health inequities associated with APOs, especially in low- and middle-income countries (LMICs).

While our initial study showed efficacy of xylitol in reducing PTB, the current study seeks to determine the impact of xylitol containing chewing gum by rigorously characterizing the site of periodontal diseases, including the oral microbiome and gingival tissues, and distant/systemic effects on the vaginal microbiome. Through the funding from the Population Health Initiative Tier 2 funding mechanism, we have successfully enrolled and completed an individually randomized, placebo-controlled, triple blinded study design comparing xylitol-containing chewing gum (~6 g/day, 100% xylitol concentration; n=25) vs placebo gum (sorbitol gum base without xylitol; n=25). Through this trial, we have collected biospecimens pertaining to subgingival plaque (n=150), gingival crevicular fluid (n=150), vaginal swabs (n=150) at 3 time points in pregnancy including 3 million dollars in funding on this topic.

Investigators
Sharon Laing, UW Tacoma Nursing & Healthcare Leadership
Zaher Kmail, UW Tacoma Interdisciplinary Arts & Sciences
Kathleen Shannon-Dorcy, Fred Hutchinson Cancer Center
Adam Heath, Community Health Center of Snohomish County

Project summary
A team of researchers from the University of Washington (UW) and Fred Hutchinson Cancer Research Center partnered with a regional community health center to address workforce well-being. The proposed collaborative served three purposes: (1) Assess burnout occurrence among direct clinical service staff at the center. (2) Investigate existing evidence-based strategies that can reduce burnout and improve employee well-being. (3) Conduct listening sessions with center staff to ascertain the feasibility of evidence-based strategies to reduce burnout and improve employee well-being.

Phase I Needs Assessment entailed the analysis of survey responses from 94 health center staff. The findings showed strong results for the center on job meaningfulness, a workplace culture valuing diversity, strong work teams and favorable job satisfaction. These results are consistent with the mission and value-orientation of the center. For areas of growth, not unlike most healthcare organizations, staff wish to see improved work-life balance, additional resources to support patients’ needs, a reduction in workload and health center processes (Electronic Medical Records) that are more conducive to job efficiency. All the aforementioned factors if not addressed, can contribute to heightened staff burnout at the center. Results also indicated that several health professions appear to be more prone to burnout experiences, and they include medical providers, dental staff, administrative support staff, clinic leadership staff and behavioral health staff.

Phase II Focus Group Sessions. During this study phase, investigators met with 22 healthcare providers. The groups comprised (1) Medical Providers, including physicians, physician assistants, nurse practitioners and other primary care providers (n=7); Medical Support Staff, including medical assistants and clinic support personnel (n=3); Dental Staff, including managers and front desk staff across dental clinics (n=6) and Nursing Staff which included 6 registered nurses. The focus group sessions were principally designed to ascertain evidence-based practices that are deemed to be feasible to remediate staff burnout.

We assessed feasibility by asking respondents to address three priorities for each intervention (1) acceptability of the intervention, (2) usefulness and (3) ease of implementation. Among the evidence-based interventions presented to staff, four emerged as potentially feasible measures to improve well-being and reduce burnout. Workflow Redesign and Advanced Team-Based Care and these were consistently ranked in the top 3 interventions of importance to improve staff well-being and reduce burnout. For Workflow Redesign, staff noted the need for clear communication of expectations from leadership; they stressed the importance of active staff engagement in the implementation process and emphasized the need for Workflow Redesign implementation to be standardized to ensure consistency across the different clinics. For Advanced Team-Based Care, the approach is valued for distributing workload, improving coordination, and supporting staff well-being. The next measures deemed to be potentially feasible to improve burnout and promote staff well-being are Person-Centered Leadership Training and Flexible Work Schedules. For Person-Centered Leadership Training, staff note that training managers and leaders to be person-centered is highly acceptable and essential for fostering supportive work environments. The Flexible Work Schedules strategy is strongly endorsed as a protective factor against burnout and a tool for improving work-life balance.

Investigators queried about the ways to improve the Electronic Medical Records (EMR) process for staff consumption and engagement; this information was deemed valuable to center leadership. Respondents noted that it is essential that the center aligns its EMR workflows with care delivery to reduce job demands. The most urgent issues to address for supporting well-being are that of improving system speed and reliability, integrating external systems of records more efficiently, and streamlining workflow documentation activities.

As a team, our investigation concluded that the healthcare organization demonstrates a deep commitment to meeting the unique needs of its patient population. However, additional work may be required to implement the evidence-based practices that were deemed acceptable and feasible by center staff.

Investigators
Kelli N. O’Laughlin, Department of Emergency Medicine
Monisha Sharma, Department of Global Health
Jacob Oluma, Medical Teams International
Timothy Muwonge, Infectious Diseases Institute, Makerere University

Project summary
The overarching goal of the study Differentiated service delivery for HIV: community ART delivery preferences among people living with HIV in refugee settlements in Uganda, also known as “ART2FIT”, was to generate evidence to inform the optimization of differentiated service delivery for HIV care in humanitarian contexts.

To contribute to this goal, we leveraged the findings from prior University of Washington Population Health Initiative-funded qualitative research on barriers to community ART delivery participation and implementation in this setting to design a discrete choice experiment. The discrete choice experiment was administered to people living with HIV (PWH) in refugee settlements in Uganda to pursue two interconnected aims. First, we sought to identify attributes (characteristics) of community ART delivery, e.g., ‘delivery location’ or ‘provider type’ that drive decision-making around community ART delivery uptake by PWH and elucidate the levels that ART delivery attributes should assume to meet the needs of humanitarian populations, e.g., for the attribute ‘delivery location’, would PWH prefer to receive their ART at ‘a pharmacy’, ‘a school’, and ‘an expert client home’, or ‘their home’? Second, the study sought to explore community ART delivery preference heterogeneity, i.e., assess whether distinct preference profiles can be identified, representing groups of PWH with similar preferences.

To achieve these aims, as part of the discrete choice experiment, PWH were presented with a series of choice tasks. In each choice task, PWH were asked to choose between two hypothetical community ART delivery models varying in delivery location, delivery structure (group versus individual), provider type, dispensing interval, delivery timing and service duration. For each choice task, they were also given the option to choose ‘neither’ and opt-out from participating in community ART delivery if these were the only two options available. Between January 21, 2026 and April 30, 2026, 869 PWH in five refugee settlements in Uganda completed the discrete choice experiment as well as an intake survey capturing demographic characteristics. We estimated preference weights using a Hierarchical Bayes model and assessed preference heterogeneity in a latent class analysis.

Overall, the study captured the degree to which community models are preferred over opting out, revealed the characteristics most important to PWH in deciding whether they want to participate in community ART delivery, identified the most preferred community ART delivery model designs, provided insight into the extent to which PWH are willing to pay for the model of their choice, provided an indication of potential uptake if specific models were offered and shed a light on preference heterogeneity. Detailed analyses and findings are currently being prepared for dissemination through peer-reviewed publications.

A portion of the funding for this award came via a partnership with the UW Global, which seeks to enhance the UW’s international engagement and reach.

Investigators
N. Jeanie Santaularia, Department of Epidemiology
Gabriela Bustamante, Escuela de Salud Pública, Universidad San Francisco de Quito
Paulina Ponce, Fundación Azulado
Carmen Gonzalez, Department of Communication
India Ornelas, Department of Health Systems and Population Health

Project summary
Child sexual abuse (CSA) is a major global public health problem affecting millions of children worldwide. All children are at risk of CSA, however, there may be more vulnerable groups. Interventions to prevent CSA in Latin America are limited. Fundación Azulado developed the “Mi Escudo” Family Kit, an educational intervention designed for children ages 5–12 and their caregivers to promote body autonomy, recognition of unsafe situations, identification of trusted adults, and communication about abuse prevention. The overall goal of our project was to generate proof-of-concept data that establishes preliminary efficacy of this kit in raising awareness and prevention of CSA across three regions in Ecuador (highlands,coast, and amazon). The aims were to: 1) assess changes and retention in knowledge and attitudes regarding CSA protection skills among children ages 5 to 12 after interacting with Kit Mi Escudo, 2) evaluate differences in post-intervention knowledge and attitudes among children of diverse SES, and 3) identify barriers and facilitators of Kit Mi Escudo and its implementation through in depth interviews with children and their caregivers post-intervention.

We worked closely with our community partner Fundación Azulado to receive input and feedback on all aspects of the project. We had two graduate students at University of Washington and one student at Universidad San Francisco de Quito assisting on the project. We met biweekly as a research team to check in and make progress. Aims 1 & 2. The project began following approval from the ethics committee at Universidad San Francisco de Quito and University of Washington as well as approval from the Ecuadorian Ministry of Education. Schools were recruited with support from Fundación Azulado, the project’s community partner. The intervention launched in Quito in the Highlands, then expanded to Tena in the Amazon and Manta on the Coast. In each region, four schools participated — two public and two private — divided into intervention and control groups.

A total of 378 children were enrolled. Three visits were conducted at each study site. We had approximately 1% lost to follow-up during the study period. During the first visit, the research team introduced the project, obtained informed consent, collected demographic information, and administered the adapted CKAQ-Short and WIIST-II-R instruments to assess CSA prevention knowledge and self-protection skills in both intervention and control groups. Participants in intervention schools also received the Kit Mi Escudo. Follow-up visits were conducted one month and five months after baseline assessment, during which participants from both groups were re-evaluated using the same instruments. Between visits, families in the intervention group were contacted by phone to support implementation of the kit and address questions regarding its use. Four digital tablets were raffled among intervention group students as part of participant compensation and engagement activities. All participating children — from both control and intervention groups — also received a $15 gift card in recognition of their contribution to the study.

Preliminary findings suggest that the effect of the Kit Mi Escudo on CSA prevention knowledge and self-protection skills varies across regions. Overall, the intervention appears to have a modest positive effect; however, some inconsistent findings were observed, potentially related to variability in instrument administration across study sites.

During the final study visit, focus groups were conducted with randomly selected caregivers and children from the intervention group to identify barriers and facilitators related to Kit Mi Escudo use and gather feedback on the intervention. While the target sample size for this qualitative component was 60 participants, only 24 were included due to logistical barriers in the Coast and Amazon regions. Focus group data were subsequently transcribed, coded and qualitatively analyzed to identify strengths and areas for improvement across the five games included in the kit. Overall, caregivers and children reported a positive perception of the intervention. Caregivers described the kit as a useful tool for addressing topics that are often difficult to discuss at home and for fostering deeper communication with their children. Children generally responded positively to the games, though preferences varied by activity and appeared to differ by age group.

Preliminary findings were shared with participating communities at the final study visit and presented at a work-in-progress violence-focused group at the University of Washington. A manuscript is currently in preparation alongside the finalization of analyses for Aims 1 and 2. Aim 3 results informed a successful grant proposal to support quality improvement of Kit Mi Escudo.

Investigators
Nicole Errett, Department of Environmental & Occupational Health Sciences
Tania Busch Isaksen, Department of Environmental & Occupational Health Sciences
Ann Bostrom, Evans School of Public Policy & Governance
Mary Hannah Smith, Department of Environmental & Occupational Health Sciences
Bradley Kramer, Public Health – Seattle & King County
Jillian Edge, Public Health – Seattle & King County
Reid Wolcott, National Weather Service Seattle
Jacob DeFlitch, National Weather Service Seattle
Suzanna Lindeman, National Weather Service Seattle
Jamie Vickery, NOAA Office of Oceanic and Atmospheric Research, Global Systems Laboratory
Jebb Q. Stewart, NOAA Office of Oceanic and Atmospheric Research, Global Systems Laboratory

Project summary
The goal of this project was to outline a framework of extreme cold response in King County, integrating weather information, health impacts and relevant response strategies. We worked with and heard from extreme cold response partners across the county, and used a variety of different methods to accomplish this goal. Our research scope has expanded to include the entire state of Washington through a two-year, $500,000 award from the National Oceanic and Atmospheric Administration (NOAA).

Our first aim was to assess the impact of extreme cold events on the use of emergency medical services (EMS) in King County, WA, and describe indicators that modify extreme cold event (ECE) risk. We collected and analyzed King County EMS data from 2019 through 2024 to identify if and how calls increased during extreme cold days. Our analysis was limited by a small sample size and lack of data of interest like ZIP code of residence, or race.

Our second aim was to examine ECE warning and response thresholds and planned interventions in large cities that have undertaken extreme cold planning and preparedness, including approaches to risk communication, which we accomplished by collecting and analyzing 17 extreme cold response plans. We summarized the proportion of plans that included various response strategies, and found significant variation across the plans with respect to format, hazard scope and types of strategies included.

Our third aim was to convene response partners to discuss their perception of exposures and vulnerabilities that influence ECE risk, identify health impacts of ECEs not captured through EMS data, and identify locally relevant ECE intervention points. To accomplish this aim, we hosted a workshop in March 2025 with 25 representatives of King County organizations currently involved in extreme cold preparedness, response or coordination, as well as state and regional response partners. Participants often considered and planned for ECEs alongside other winter weather hazards, rather than in isolation. They described two broad categories of people at risk of health impacts from ECEs, the first being those more likely to be exposed due to factors like occupation or recreational choices, and those more likely to be vulnerable to health impacts due to factors like age or underlying conditions. Participants discussed many existing ECE response interventions, and specifically acknowledged the utility of risk communication activities based on their relatively low cost and their potential to connect vulnerable groups with community resources. Agencies beyond those typically involved in emergency response may need to be engaged to fully address upstream factors that drive exposure and vulnerability to extreme cold.

Our fourth aim was to determine information needed for response partner decision-making to inform the type, timing, and mode of information provided by NWS meteorologists before, during and after ECEs. We conducted 19 interviews with individuals representing a wide variety of agencies in King County and neighboring jurisdictions. We are using a mental models approach to analyze the data to provide an enhanced understanding of how response partners conceptualize ECEs.

Investigators
Jingyi Li, UW Tacoma Nursing & Healthcare Leadership
Serena Jinchen Xie, Department of Biomedical Informatics and Medical Education
Weichao Yuwen, UW Tacoma Nursing & Healthcare Leadership
Trevor Cohen, Department of Biomedical Informatics and Medical Education
Michael Woo, Kin On Health Care Center
Boliver Choi, Chinese Information and Service Center
Paul Tan, Open Doors for Multicultural Families WA

Project summary
Our project aimed to develop and evaluate culturally responsive AI driven mental health conversational agents to support Chinese American family caregivers of older adults, a population that experiences high caregiving burden and persistent barriers to culturally and linguistically appropriate mental health support. While large language model (LLM)–based chatbots o er scalable mental health assistance, many existing systems often reflect Western-centric assumptions and fail to adequately incorporate culturally specific values, caregiving norms and communication styles.

Using a community-engaged, mixed-methods approach, we conducted 13 semi-structured interviews with Chinese American family caregivers of older adults and four focus groups with 21 staff members from community-based organizations serving Chinese American families. Data collection occurred in participants’ preferred languages (English, Mandarin, or Cantonese). We identified culturally salient caregiving challenges and self-care barriers, including filial obligation, caregiver guilt related to older adult social isolation, intergenerational communication challenges, perceived self-care as indulgence and difficulties navigating healthcare and social service systems. These findings informed the development of a structured cultural context database capturing caregiving challenges, underlying cultural interpretations, and exemplar culturally responsive responses grounded in caregivers lived experiences.

Building on this foundation, we designed and evaluated two cultural adaptation strategies for LLM-based mental health conversational agents: (1) a prompt-based cultural adaptation approach and (2) a dynamic context-engineering approach using retrieval augmented generation (RAG) to integrate relevant cultural context in real time. The prompt based approach embeds explicit cultural instructions directly into the LLM’s system prompt, while the context-engineering approach retrieves culturally grounded caregiving challenges, underlying cultural factors and example responses from a curated cultural context database developed through community-engaged qualitative research. In a blinded evaluation study with 36 Chinese-speaking participants, culturally adapted responses were rated significantly higher on cultural competence, cultural relevance, and perceived empathy. In a subsequent interactive, within-subject comparison study with 36 caregivers and community organization staff, the context-engineered agent consistently outperformed both a non-adapted baseline and a prompt-adapted agent across cultural responsiveness, empathy, therapeutic alliance and overall satisfaction, and was the only version to show statistically significant improvements over baseline across all major outcomes.

Finally, we conducted a randomized user study with 58 Chinese American family caregivers of older adults, comparing the dynamically adapted agent with a non-adapted baseline in a PHI Final Report realistic, single-session use scenario. Caregivers in both groups showed immediate reductions in negative emotional states (including guilt, fatigue and sadness). Notably, caregivers who interacted with the culturally adapted agent showed additional improvements in positive emotional states and were significantly more likely to recommend the agent within their cultural community, indicating greater cultural acceptability and trust. Together, these findings demonstrate that while prompt-based adaptation offers measurable benefits, dynamic context engineering provides more consistent and robust improvements in culturally responsive, empathic and therapeutically aligned AI-based mental health support for Chinese American caregivers.