Faculty Bibliography

Early identification of Alzheimer's disease (AD) and AD-related dementias (ADRD) has high clinical significance, both because of the potential to slow decline through initiating FDA-approved therapies and managing modifiable risk factors, and to help persons living with dementia and their families to plan before cognitive loss makes doing so challenging. However, substantial racial and ethnic disparities in early diagnosis currently lead to additional inequities in care, urging accurate and inclusive risk assessment programs. In this study, we trained an artificial intelligence foundation model to represent the electronic health records (EHR) data with a vast cohort of 1.2 million patients within a large health system. Building upon this foundation EHR model, we developed a predictive Transformer model, named TRADE, capable of identifying risks for AD/ADRD and mild cognitive impairment (MCI), by analyzing the past sequential visit records. Amongst individuals 65 and older, our model was able to generate risk predictions for various future timeframes. On the held-out validation set, our model achieved an area under the receiver operating characteristic (AUROC) of 0.772 (95% CI: 0.770, 0.773) for identifying the AD/ADRD/MCI risks in 1 year, and AUROC of 0.735 (95% CI: 0.734, 0.736) in 5 years. The positive predictive values (PPV) in 5 years among individuals with top 1% and 5% highest estimated risks were 39.2% and 27.8%, respectively. These results demonstrate significant improvements upon the current EHR-based AD/ADRD/MCI risk assessment models, paving the way for better prognosis and management of AD/ADRD/MCI at scale.

BACKGROUND:Medicare-funded home healthcare (HHC) delivers skilled nursing, therapy, and related services through visits to the patient's home. Nearly one-third (31%) of HHC patients have diagnosed dementia, but little is currently known regarding how HHC utilization and care delivery differ for persons living with dementia (PLwD). METHODS:We drew on linked 2012-2018 Health and Retirement Study and Medicare claims for a national cohort of 1 940 community-living older adults. We described differences in HHC admission, length of stay, and referral source by patient dementia status and used weighted, multivariable logistic and negative binomial models to estimate the relationship between dementia and HHC visit type and intensity while adjusting for sociodemographic characteristics, health and functional status, and geographic/community factors. RESULTS:PLwD had twice the odds of using HHC during a 2-year observation period, compared to those without dementia (odds ratio [OR]: 2.03; p < .001). They were more likely to be referred to HHC without a preceding hospitalization (49.4% vs 32.1%; p < .001) and incurred a greater number of HHC episodes (1.4 vs 1.0; p < .001) and a longer median HHC length of stay (55.8 days vs 40.0 days; p < .001). Among post-acute HHC patients, PLwD had twice the odds of receiving social work services (unadjusted odds ratio [aOR]: 2.15; p = .008) and 3 times the odds of receiving speech-language pathology services (aOR: 2.92; p = .002). CONCLUSIONS:Findings highlight HHC's importance as a care setting for community-living PLwD and indicate the need to identify care delivery patterns associated with positive outcomes for PLwD and design tailored HHC clinical pathways for this patient subpopulation.

INTRODUCTION/BACKGROUND:This study aimed to assess emergency nurses' perceived barriers toward engaging patients in serious illness conversations. METHODS:Using a mixed-method (quant + QUAL) convergent design, we pooled data on the emergency nurses who underwent the End-of-Life Nursing Education Consortium training across 33 emergency departments. Data were extracted from the End-of-Life Nursing Education Consortium post-training questionnaire, comprising a 5-item survey and 1 open-ended question. Our quantitative analysis employed a cross-sectional design to assess the proportion of emergency nurses who report that they will encounter barriers in engaging seriously ill patients in serious illness conversations in the emergency department. Our qualitative analysis used conceptual content analysis to generate themes and meaning units of the perceived barriers and possible solutions toward having serious illness conversations in the emergency department. RESULTS:A total of 2176 emergency nurses responded to the survey. Results from the quantitative analysis showed that 1473 (67.7%) emergency nurses reported that they will encounter barriers while engaging in serious illness conversations. Three thematic barriers-human factors, time constraints, and challenges in the emergency department work environment-emerged from the content analysis. Some of the subthemes included the perceived difficulty of serious illness conversations, delay in daily throughput, and lack of privacy in the emergency department. The potential solutions extracted included the need for continued training, the provision of dedicated emergency nurses to handle serious illness conversations, and the creation of dedicated spaces for serious illness conversations. DISCUSSION/CONCLUSIONS:Emergency nurses may encounter barriers while engaging in serious illness conversations. Institutional-level policies may be required in creating a palliative care-friendly emergency department work environment.

IMPORTANCE/UNASSIGNED:Large language models (LLMs) are crucial for medical tasks. Ensuring their reliability is vital to avoid false results. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. OBJECTIVE/UNASSIGNED:Evaluate ChatGPT and LlaMA-2 performance in extracting MMSE and CDR scores, including their associated dates. METHODS/UNASSIGNED:Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. RESULTS/UNASSIGNED:For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. CONCLUSIONS/UNASSIGNED:In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.

BACKGROUND/UNASSIGNED:Patients with head and neck cancer (HNC) often experience high symptom burden leading to lower quality of life (QoL). OBJECTIVE/UNASSIGNED:This study aims to conceptually model optimal cutpoint by examining where total number of patient-reported symptoms exceeds patients' coping capacity, leading to a decline in QoL in patients with HNC. METHODS/UNASSIGNED:Secondary data analysis of 105 individuals with HNC enrolled in a clinical usefulness study of the NYU Electronic Patient Visit Assessment (ePVA)©, a digital patient-reported symptom measure. Patients completed ePVA and European Organization for Research and Treatment of Cancer (EORTC©) QLQ-C30 v3.0. The total number of patient-reported symptoms was the sum of symptoms as identified by the ePVA questionnaire. Analysis of variance (ANOVA) was used to define optimal cutpoint. RESULTS/UNASSIGNED:<.0001). CONCLUSIONS/UNASSIGNED:In HNC, defining optimal cutpoints in the total number of patient-reported symptoms is feasible. IMPLICATIONS FOR PRACTICE/UNASSIGNED:Cutpoints in the total number of patient-reported symptoms may identify patients experiencing a high symptom burden from HNC. FOUNDATIONAL/UNASSIGNED:Using optimal cutpoints of the total number of patient-reported symptoms may help effectively align clinical resources with patients' symptom burden.

PURPOSE/UNASSIGNED:Utilizing a participatory approach, we sought to co-design a 12-week Green Activity Program (GAP) with Hispanic/Latino individuals living with memory challenges and their care partners, local outdoor professionals, and healthcare providers. METHODS/UNASSIGNED:Participants were recruited via convenience and snowball sampling in the Bronx, New York with Hispanic/Latino persons living with memory challenges and care partners, outdoor activity professionals, and interdisciplinary healthcare providers/dementia experts. Co-design occurred iteratively with 5 focus groups and 4 individual interviews lasting 30-90 min and focused on program and research design. Sessions were recorded and transcribed. Utilizing directed content analysis data was coded using a priori codes program design and research design. RESULTS/UNASSIGNED:were identified including community-based organizations and primary care. CONCLUSION/UNASSIGNED:Co-design was a successful form of engagement for people living with memory challenges that enabled participants to help design key elements of the GAP and research design. Our processes, findings, and recommendations for tailoring co-design to engage Hispanic/Latino people living with memory challenges can inform the development of other programs for this population.

BACKGROUND:We sought to describe neuropsychiatric symptoms (NPS) among people living with dementia (PLWD) from diverse racial and ethnic groups receiving home health services while accounting for dementia severity, individual symptom prevalence, and neighborhood disadvantage. METHODS:A prospective study using cross-sectional data from n = 192 PLWD receiving skilled home healthcare in New Jersey enrolled in the Dementia Symptom Management at Home Program trial. We prospectively measured symptom prevalence with the Neuropsychiatric Inventory Questionnaire and dementia severity using the Quick Dementia Rating System. A one-way ANOVA determined NPS prevalence by dementia severity (mild, moderate, severe). Fisher's exact tests were used to assess the association of individual symptom prevalence with race and ethnicity and cross tabs to descriptively stratify individual symptom prevalence by dementia severity among groups. A Pearson correlation was performed to determine if a correlation existed among neighborhood disadvantages measured by the Area Deprivation Index (ADI) state decile scores and NPS prevalence and severity. RESULTS:Participants identified as non-Hispanic White (50%), non-Hispanic Black (30%), or Hispanic (13%). NPS were prevalent in 97% of participants who experienced 5.4 ± 2.6 symptoms with increased severity (10.8 ± 6.6) and care partner distress (13.8 ± 10.8). NPS increased with dementia severity (p = 0.004) with the greatest difference seen between individuals with mild dementia (4.3 ± 2.3) versus severe dementia (5.9 ± 2.3; p = 0.002). Few differences were found in symptom prevalence by racial and ethnic sub-groups. Nighttime behaviors were higher in non-Hispanic Black (78%), compared with non-Hispanic Whites (46%) with moderate dementia, p = 0.042. State ADI scores were not correlated with the number of NPS reported, or severity. CONCLUSIONS:NPS were prevalent and increased with dementia severity with commonalities among racial and ethnic groups with varying levels of neighborhood disadvantage. There is a need for effective methods for improving NPS identification, assessment, and management broadly for homebound PLWD.

Understanding the complex care needs of seriously ill adults with multiple chronic conditions with and without cancer is critical for the delivery of high-quality serious illness and palliative care at the end of life. The objective of this secondary data analysis of a multisite randomized clinical trial in palliative care was to elucidate the clinical profile and complex care needs of seriously ill adults with multiple chronic conditions and to highlight key differences among those with and without cancer at the end of life. Of the 213 (74.2%) older adults who met criteria for multiple chronic conditions (eg, 2 or more chronic conditions requiring regular care with limitations of daily living), 49% had a diagnosis of cancer. Hospice enrollment was operationalized as an indicator for severity of illness and allowed for the capture of complex care needs of those deemed to be nearing the end of life. Individuals with cancer had complex symptomatology with a higher prevalence of nausea, drowsiness, and poor appetite and end of life and lower hospice enrollment. Individuals with multiple chronic conditions without cancer had lower functional status, greater number of medications, and higher hospice enrollment. The care of seriously ill older adults with multiple chronic conditions requires tailored approaches to improve outcomes and quality of care across health care settings, particularly at the end of life.

OBJECTIVE:To investigate mental health needs and barriers to seeking mental health support in Medicaid-funded Assisted Living Facility (M-ALF). DESIGN:A multimethod, qualitative-dominant descriptive design using questionnaires and semistructured interviews. SETTING AND PARTICIPANTS:The study occurred at a M-ALF in the Bronx, New York. A researcher in residence recruited 13 residents (11 Black or African American, 2 Asian) using purposive sampling. METHODS:Demographic data and mental health indicators (depression, anxiety, stress, hopelessness) were measured with questionnaires (Center for Epidemiological Studies Depression Scale, Edmonton Symptom Assessment System, Perceived Stress Scale, Beck Hopelessness Survey) and analyzed with descriptive statistics. Interviews were conducted between June and November 2021, transcribed, and analyzed using conventional content analysis. RESULTS:Thirteen residents (mean age: 73.4 years, mean length of stay: 3.5 years; range: 1.0-7.5) completed data collection. Quantitatively indicators of unmet mental health were common. Qualitatively, residents reported barriers to mental health access to address depression, anxiety, and substance use. This was accompanied by concerns surrounding loss of autonomy, mistrust for M-ALF organizational support, isolation and uncertainty about how to receive mental health support. Perspectives were shaped by past experiences with institutional living, serious illness, and being unhoused. Themes and subthemes were (1) mental health need (unmet mental health need, depression, and anxiety and seeking support through non-mental health resources) and (2) barriers to mental health support (dissatisfaction with M-ALF care, perceived threats to autonomy, desire for autonomy that leads to diminished care seeking). CONCLUSION AND IMPLICATIONS:Residents of M-ALF have mental health needs for which care is stymied by loss of autonomy, lack of resources, and the M-ALF environment. Residents use unconventional resources to address needs that may be neither efficient nor effective. Novel mental health interventions and processes are needed to improve mental health access and should prioritize residents' desire for autonomy and the unique circumstances of living in M-ALF.