Publications

BACKGROUND: Addressing goals of care (GOC) serves a crucial role throughout the course of a patient's illness, but for it to be useful and enduring, documentation must provide enough detail to be actionable for the next clinician. Our aim is to describe GOC documentation content and quality across four clinical specialties (palliative care, hospitalists, internal medicine [IM] residents, and critical care) using structured assessment of high versus low quality documentation.

METHODS: Literature review suggests that five domains are most important for GOC notes: participants, prognosis, patient goals and values, treatment options, and advance directives. As part of an internal quality improvement initiative at our quaternary academic referral center in the Northeast United States, a novel note assessment tool was constructed and iteratively adjusted to achieve 80% interrater reliability. We then used this to review inpatient GOC notes from patients admitted under IM services with or without a palliative care consult (n = 100, 25 notes from each clinical specialty).

RESULTS: Only 69 out of 100 notes could be classified as GOC notes based on our tool and definitions. Except for palliative care specialists, high-quality documentation occurred infrequently in all specialties (palliative care 85.7%, critical care specialists 22.2%, hospitalists 20%, and IM residents 20%). Prognosis was both infrequently and poorly documented in all four groups, including palliative care. The remaining 31 notes were unrelated to GOC and commonly documented clinical updates, code status in isolation, or were left blank.

CONCLUSION: Using a novel assessment tool, we found that, with the general exception of palliative care specialists, GOC documentation was inconsistent and of low quality. This highlights the need to improve the consistency and quality of GOC documentation. We also submit that the definitions and examples outlined in our novel note assessment tool can double as a guide to teach providers how to improve the quality of their GOC documentation.

BACKGROUND: Cancer survivors often experience declining engagement in digital health interventions (DHIs). However, predictors of engagement with provider-guided DHIs remain unclear. Nurse WRITE (Nurse Written Representational Intervention To Ease Symptoms), an effective 8-week nurse-directed symptom management DHI, offers an opportunity to identify factors influencing engagement and enhance intervention efficacy evaluation.

OBJECTIVE: This study aims to (1) understand engagement phenomena (dimensions, influencing factors, and challenges), and (2) assess the relationship between engagement and patient symptom control in Nurse WRITE.

METHODS: In this secondary analysis of the Nurse WRITE arm of a 3-arm symptom management trial, we examined data from 68 women with recurrent ovarian cancer to assess socioaffective and cognitive engagement through message board activity, as well as behavioral engagement through website usage data. Regression analyses examined patient characteristics, engagement, and symptom control perceptions. Through content analysis, we explored participant challenges and activities before disengagement.

RESULTS: Education was significantly associated with selected cognitive, socioaffective, and behavioral engagement indicators, including cognitive activity count, total word count, completion of symptom care plans, and plan reviews after false discovery rate correction. The most common engagement challenges included worsening health and treatment, busy family life, and website difficulties. Moderate and low engagers also experienced confusion about the intervention timeline and process. Among low engagers, 63.2% (24/38) discontinued communication at specific intervention phases: introduction (8/24, 33.3%), symptom representational assessment (5/24, 20.8%), and goal setting and planning (5/24, 20.8%). Improved symptom control at the end of the intervention was significantly associated with overall engagement, cognitive and socioaffective activity count, question completion percentage, total word count, and completed symptom care plans after false discovery rate correction.

CONCLUSIONS: Education was associated with selected cognitive, socioaffective, and behavioral engagement indicators in Nurse WRITE. Future provider-guided DHIs should consider strategies to support participants with lower educational attainment, address common engagement barriers, and reengage participants during critical intervention phases. Meaningful engagement across cognitive, socioaffective, and behavioral dimensions may be important for improving outcomes while balancing protocol adherence with flexibility.

BACKGROUND: Up to 30% of women with breast cancer experience cancer-related cognitive impairment (CRCI) even before they receive treatment. The authors conducted a post hoc analysis to examine the relationship between neighborhood socioeconomic disadvantage and CRCI in treatment-naive postmenopausal women with early stage breast cancer.

METHODS: The Exercise Program in Cancer and Cognition was a single-blind, 6-month, randomized controlled trial evaluating moderate-intensity aerobic exercise versus usual activity on neurocognitive function. By using baseline data from treatment-naive participants (n = 100), neighborhood socioeconomic disadvantage was assessed using the Area Deprivation Index (national percentile). Cognitive function was measured using composite domain scores. Associations were examined using correlation and linear regression analyses, adjusting for potential covariates/confounders.

RESULTS: On average, participants were aged 63 years and had 16 years of education. Greater neighborhood disadvantage was associated with poorer cognitive function in domains of verbal memory (p = .032), working memory (p = .043), mental flexibility (p = .049), and processing speed (p = .026). In adjusted analyses, working memory remained associated with the Area Deprivation Index (p = .019) with adjustment for age; however, associations with the Area Deprivation Index were attenuated for verbal memory after adjustment for education (p = .067), for mental flexibility after adjustment for depressive symptoms (p = .175), and for processing speed after adjustment for age and body mass index (p = .065).

CONCLUSIONS: Neighborhood socioeconomic disadvantage may contribute to cognitive vulnerability before cancer treatment. However, these findings suggest that years of education and depressive symptoms should be factored into the consideration of neighborhood-level factors when assessing CRCI risk among women with breast cancer (Clinicaltrials.gov identifier NCT02793921).

BACKGROUND: Preterm birth affects approximately 10% of U.S. births, and children born preterm face twice the lifetime risk of chronic kidney disease (CKD). Despite this, kidney health surveillance after preterm birth is uncommon. Although clinical decision support (CDS) tools are widely used in pediatric practice, none address CKD risk stratification after preterm birth. This study assessed pediatric clinician perspectives on facilitators and barriers to CDS tool use, in general and for pediatric CKD risk stratification.

METHODS: We conducted a qualitative descriptive study using semistructured interviews with neonatologists, general pediatricians, and pediatric nephrologists in the United States (December 2023-April 2024). Interviews were conducted by video or teleconference, digitally recorded, and professionally transcribed. Thematic analysis followed COREQ guidelines, and sampling continued until thematic saturation was confirmed.

RESULTS: Twenty-five pediatric clinicians participated (44% neonatologists, 44% general pediatricians, and 12% nephrologists; median age 39 years, 76% female, 52% White, 88% non-Hispanic, and 80% academic practice). Clinicians reported strong preferences for CDS tools that efficiently support workflows, integrate with the electronic health record (EHR), and provide actionable recommendations with caregiver education. Key concerns included unintended consequences such as false reassurance, over-referral to nephrology, and care burden for families with limited subspecialty access. All participants endorsed the need for a pediatric CKD risk stratification tool.

CONCLUSIONS: Pediatric clinicians prefer EHR-integrated, evidence-based, family-centered CDS tools to guide CKD risk identification after preterm birth. These findings represent an important step toward developing a pediatric kidney disease risk stratification CDS tool.

OBJECTIVE: Recent growth in hospice has raised concerns about variability in access and quality. This study examines U.S. county-level presence of hospice providers.

METHODS: Quality was assessed using 2023 Centers for Medicare & Medicaid Services Consumer Assessment of Healthcare Providers and Systems Hospice Survey Star Ratings. County-level sociodemographic characteristics were obtained from the U.S. Census Bureau's American Community Survey 2023 five-year sample.

RESULTS: Of 7024 hospice providers, 29.2% had publicly available ratings, of which 55.3% were high quality (>3 stars). Of 3222 counties, at least one hospice provider was located in 41.4%, and a high-quality provider was present in 23.6%. Counties with high-quality providers tend to be urban and have higher income, larger populations, higher education, and lower poverty and uninsurance rates.

CONCLUSIONS: Many U.S. counties have no hospice providers. The presence of high-quality care is markedly lower in rural and poor counties. Missing star ratings limit the utility of public quality data for patient and family decision-making.

FUNDING: Funding (NIH #1OT2OD037972-01).

OBJECTIVE: Artificial intelligence (AI) may enhance the efficiency and personalization of breast cancer (BC) treatment care. It is imperative to include patient viewpoint into AI design for clinical care but there is limited research exploring how survivors perceive AI to inform BC care. To explore BC survivors' understanding of, trust in, and acceptance of AI applications to inform BC treatment, and to examine variations in perspective by age, race/ethnicity, income, and neighborhood deprivation index.

MATERIALS AND METHODS: This IRB-approved qualitative descriptive study used purposive sampling of adult BC survivors recruited from a large academic cancer center and affiliated community outreach programs. Inclusion included the ability to speak and understand English and BC diagnosis within the past 10 years, participants completed a brief demographic survey followed by a 15-30-minute semi-structured interview focusing on awareness of AI, perceived benefits, risks, equity implications, and any suggested strategies for AI to be reincorporated into BC care. Interviews were audio-recorded, transcribed verbatim, and analyzed using thematic and content analysis. A second reviewer verified coding. Demographic data were analyzed descriptively and dichotomized.

RESULTS: Participant (n = 20) ages range from 38 to 74 years old (mean 57); 80% identified as White, 20% as Black. Most participants had some college education, and neighborhood area deprivation index scores, mean 68.2, (standard deviation 19.2) spanned 8-95 (indicating broad socioeconomic representation). Time since diagnosis ranged from 1 to 10 years. Three themes are identified: 1) Awareness of AI ranging from no knowledge to very familiar. 2) Concerns included loss of human interaction, clinician over-reliance on algorithms, and security assurance (data privacy breaches and model safety). Acceptance of AI was highest when framed as augmenting rather than replacing clinician judgment. 3) Equity-some survivors viewed AI as a force that might reduce disparities, while others feared exclusion of under-represented groups due to non-diverse datasets or institutional access barriers. Notably, all Black participants emphasized the necessity of dataset representation to avoid perpetuating inequities. Variations in perspective by dichotomized age, income, years since diagnosis, and neighborhood deprivation index were otherwise not identified.

CONCLUSION: AI is welcomed when it enhances, but does not substitute for human-led care, and when transparency, safety, equitable inclusion, and governance are assured. These findings underscore the imperative and proactive attention needed to ensure patients with BC participate in AI system design, development, evaluation, and deployment, particularly when related to critical BC treatment decision-making.

BACKGROUND: Clinicians place comfort measures only (CMO) orders for hospitalized patients at the end-of-life when a decision has been made to focus on patient comfort and allow the natural dying process to occur.

OBJECTIVES: Our primary aim was to assess the associations of specialty palliative consults (SPC) or documented goals of care conversations (GOCC) with in-hospital mortality among patients with CMO orders.

DESIGN: We completed a retrospective cross-sectional study of data from the electronic medical record.

SETTING/PARTICIPANTS: We assessed all adult patients with CMO orders admitted to four hospitals in the United States between 2022 and 2024.

MEASUREMENTS: We used descriptive statistics and multivariable logistic regression and linear regression models to identify the association of SPC and documented GOCC with patient outcomes.

RESULTS: Of 6789 hospitalized patients with CMO orders, 48% were female, and these patients had median age 71 years. Seventy-three percent died in-hospital, and 22% were discharged with hospice. SPC placed anytime during hospital admission were associated with lower in-hospital mortality (aOR 0.4, 95% CI 0.3-0.6, p < 0.0001) and higher discharge with hospice (aOR 2.4, 95% CI 1.8-3.2, p < 0.0001). In contrast, documented GOCC anytime during admission were associated with higher in-hospital mortality (aOR 1.8, 95% CI 1.3-2.4, p = 0.0004) and lower discharge with hospice (aOR 0.5, 95% CI 0.4-0.7, p = 0.0003).

CONCLUSIONS: For patients with CMO orders, SPC, rather than documented GOCC, were associated with lower in-hospital mortality and higher receipt of hospice. Future research should explore reasons why only SPC, and not documented GOCC, were associated with these findings.

OBJECTIVES: To understand the unmet needs of bladder cancer survivors in the acute recovery after cystectomy and how remote monitoring may improve their care. These results are part of user-centered design studies on a remote monitoring recovery aid.

METHODS: We invited bladder cancer survivors who underwent a radical cystectomy and their support persons from bladder cancer survivorship groups. We conducted semi-structured interviews regarding participants' experiences during recovery from cystectomy, their perspectives regarding unmet needs, and the resources they used during the perioperative period, emphasizing the 2 weeks following hospital discharge. Transcripts were coded using a deductive/inductive approach. Coded excerpts were summarized to develop themes and subthemes. Findings were refined as needed by consensus to accurately capture participants' experiences.

RESULTS: A total of 20 bladder cancer survivors and 5 primary support persons were included. Patient participants were on average 3 years from their cystectomy. Barriers to acute recovery were categorized into system- and patient-level challenges, including difficulty connecting with knowledgeable providers and with information, as well as mental/emotional and physical hurdles. Three key facilitators to patients' acute recovery following cystectomy included identifying indicators of physical progress, facilitating access to provider recommendations, and implementing the survivor perspective into recovery support.

CONCLUSIONS: This study classified barriers and facilitators to cystectomy recovery. The majority of participants expressed interest in remote monitoring programs. Opportunities exist for stakeholders to improve the acute recovery experience for cystectomy patients through mobile interventions delivered via smartphones and wearable devices.