Provider Perceptions of a Novel Inpatient Co-Rounding Model Integrating Medical Oncology, Neuro-Oncology, and Radiation Oncology for the Care of Patients with Advanced Cancer

PURPOSE/OBJECTIVE(S)

Patients (pts) with advanced cancer require interdisciplinary care. Although tumor boards are well-established in the outpatient setting, few studies have evaluated interventions for improving consultative care coordination for hospitalized pts with cancer. The purpose of this study was to evaluate a novel inpatient co-rounding model of care in which medical-, neuro-, and radiation-oncology consult teams rounded jointly, with the hypothesis that primary referring team perception of the alignment of the recommendations from these consult services would improve post-implementation.

MATERIALS/METHODS

An inpatient co-rounding model was implemented in September 2021 for hospitalized pts with solid malignancies at a tertiary medical center. Attending physicians, nurse practitioners, fellows, and residents from oncologic consulting services met virtually twice weekly to discuss pt care. Providers from the two most common primary services for pts with cancer at this hospital (hospital medicine and neurosurgery) were surveyed via institutional email listservs. The survey included Likert-type questions about the quality of inpatient consultation and the alignment of recommendations across three consulting oncological specialty services. The pre-intervention survey was distributed prior to model implementation, and the post-intervention survey was distributed 9 months later. Wilcoxon rank-sum tests were used to compare responses from the pre-and post-intervention surveys.

RESULTS

At each session, a median of 6 providers attended (range, 4-8 providers), and a median of 6 pts were discussed (range, 4-8 pts). Among 331 providers surveyed, 119 completed the pre-intervention survey (36% response rate), and 34 completed the post-intervention survey (10% response rate). Respondents were 81 (53%) internal medicine attending physicians/hospitalists, 55 (36%) internal medicine resident physicians, 6 (4%) neurosurgery advanced practice providers, 6 (4%) neurosurgery attending physicians, and 5 (3%) neurosurgery resident physicians. When asked to rate agreement with the statement that consultant recommendations from medical-, neuro-, and radiation-oncology were aligned, respondents were significantly more likely to perceive alignment 9 months post-implementation (67% strongly agree) compared to pre-implementation (23% strongly agree, p = 0.0001). There was high satisfaction with the quality of medical-, neuro-, and radiation-oncology consultations at both time points, with no statistical difference pre- vs. post-implementation of the co-rounding model.

CONCLUSION

A novel inpatient co-rounding model of care was successfully launched between medical-, neuro-, and radiation-oncology. Primary teams perceived greater alignment in recommendations between these consulting services after project implementation. Future directions include evaluating the impact of this co-rounding model on patient outcomes.

Identifying Common Topics in Patient Portal Messages with Unsupervised Natural Language Processing

PURPOSE/OBJECTIVE(S)

Patient portal messaging is an increasingly important form of communication between patients and medical providers. This has become particularly relevant in oncology, where patients undergo intense longitudinal treatments that require frequent communication regarding symptoms, appointments, and diagnostic results. The rise in the volume of these messages has significantly increased the workload of medical providers and consequent physician burn-out. Natural language processing (NLP), particularly transformer-based models, may offer an automated approach to characterize the content of patient messages and improve message triage and routing. In this study, we employed a state-of-the-art language model (Bidirectional Encoder Representations from Transformers; BERT) to identify data-derived categories of representative topics from real-world data thereby providing basic information to build an appropriate routing system.

MATERIALS/METHODS

Patient-generated portal messages sent to a messaging pool for a single institution radiation oncology department from 2014 to 2023 were extracted. BERTopic, an NLP-based topic modeling technique based on BERT was optimized for topic modeling of patient messages. Uniform Manifold Approximation and Projection (UMAP) was used to reduce dimensionality and visualize topic relationships across messages. The BERTopic-identified topic categories were subsequently labeled manually by one of the physician investigators. Differences of number of messages over time were assessed using t-tests.

RESULTS

A total of 47,492 messages were retrieved. The average number of messages per month from a single patient ranged from 1 to 18 (median 1.67, interquartile range 1.0-2.4). The total volume of patient messages showed a ten-fold increase over the study period, with 101 messages per month sent in 2014 and 999 messages per month in 2022 (p<0.001). BERTopic initially identified 35 topics whose relationships and degrees of overlap were visualized by UMAP. Due to physician-identified similarities, these topics were reduced into 13 categories. The most frequent topic category was messages about laboratory tests or imaging studies: 24.3%, followed by messages expressing appreciation: 18.9%, scheduling discussions: 15.6%, symptom-related messages: 11%, and treatment-related messages: 10.7%.

CONCLUSION

Patient portal messages sent to a single institution radiation oncology department have increased dramatically in volume since implementation, corresponding to a broader national trend. NLP successfully identified common subject themes across patient messages, many of which are related to scheduling. This presents potential opportunities to apply NLP to automate message routing in the future.

Natural Language Processing Identification of Symptoms in Emergency Department Visits in Patients Receiving Radiation

PURPOSE/OBJECTIVE(S)

Patients undergoing radiotherapy (RT) for cancer often require emergency department (ED) attention with possible hospitalization. Designing strategies to mitigate hospital admissions requires understanding the causal symptoms to tailor interventional strategies. Natural language processing (NLP) has previously been shown to accurately identify documented symptoms and may help characterize factors contributing to admission. The objective of this study was to use NLP to identify documented symptoms during ED visits and their association with subsequent hospital admission of patients undergoing RT.

MATERIALS/METHODS

A de-identified, single tertiary-care institution cohort of patients who received radiation between 2013 and 2022 was identified from the electronic health record using International Classification of Disease (ICD) and Current Procedural Terminology (CPT) codes. We applied a previously validated clinical Text Analysis and Knowledge Extraction System (cTAKES)-based NLP pipeline to extract Common Terminology Criteria for Adverse Events (CTCAE) encoded symptoms from ED encounter clinical notes. Chi-squared testing was used to compare demographics, and logistic regression was used to identify symptoms associated with subsequent admission from ED visits.

RESULTS

We identified 14,007 patients who received radiation, of whom 270 (1.9%) experienced 302 ED visits during their radiation course. 141 (46.7%) of ED visits resulted in an admission. Among patients with an ED encounter, there were no differences in admission rates based on primary language (p = 0.771), sex (p = 0.824), marital status (p = 0.753), race (p = 0.222), or age (p = 0.123). In admitted patients, the top 5 symptoms were pain (94.3%), nausea (92.1%), vomiting (73.7%), constipation (70.9%), and weakness (63.8%). In patients who did not require admission, the most common symptoms were pain (84.5%), nausea (67.1%), vomiting (47.2%), headache (36.6%), and weakness (35.4%). The 10 symptoms most associated with admission from the ED based on logistic regression were malaise (OR 21.7, [95% CI 10.1 - 51.0]), lethargy (19.1, [8.5 - 51.3]), flushing (15.7, [8.6 - 30.4]), agitation (12.4, [3.5 - 78.7]), somnolence (10.3, [4.7 - 25.9]), fall (8.5, [3.7 - 23.2]), fatigue (7.8, [4.6 - 13.4]), constipation (6.9, [4.2 - 11.6]), nausea (5.8, [3.0 - 12.2]), and swelling (5.4, [3.3 - 9.1]).

CONCLUSION

Admitted and non-admitted ED patients with cancer seen in the ED during radiotherapy are documented to experience similar symptoms, but certain symptoms are associated with a higher risk of hospital admission. NLP may offer a mechanism for early, automated identification to facilitate supportive interventions for patients at high risk for admission during radiotherapy.

MRI of hippocampal volume loss in early Alzheimer's disease in relation to ApoE genotype and biomarkers

Hippocampal volume change over time, measured with MRI, has huge potential as a marker for Alzheimer's disease. The objectives of this study were: (i) to test if constant and accelerated hippocampal loss can be detected in Alzheimer's disease, mild cognitive impairment and normal ageing over short periods, e.g. 6-12 months, with MRI in the large multicentre setting of the Alzheimer's Disease Neuroimaging Initiative (ADNI); (ii) to determine the extent to which the polymorphism of the apolipoprotein E (ApoE) gene modulates hippocampal change; and (iii) to determine if rates of hippocampal loss correlate with cerebrospinal fluid (CSF) biomarkers of Alzheimer's disease, such as the beta-amyloid (Abeta(1-42)) and tau proteins (tau). The MRI multicentre study included 112 cognitive normal elderly individuals, 226 mild cognitive impairment and 96 Alzheimer's disease patients who all had at least three successive MRI scans, involving 47 different imaging centres. The mild cognitive impairment and Alzheimer's disease groups showed hippocampal volume loss over 6 months and accelerated loss over 1 year. Moreover, increased rates of hippocampal loss were associated with presence of the ApoE allele epsilon4 gene in Alzheimer's disease and lower CSF Abeta(1-42) in mild cognitive impairment, irrespective of ApoE genotype, whereas relations with tau were only trends. The power to measure hippocampal change was improved by exploiting correlations statistically between successive MRI observations. The demonstration of considerable hippocampal loss in mild cognitive impairment and Alzheimer's disease patients over only 6 months and accelerated loss over 12 months illustrates the power of MRI to track morphological brain changes over time in a large multisite setting. Furthermore, the relations between faster hippocampal loss in the presence of ApoE allele epsilon4 and decreased CSF Abeta(1-42) supports the concept that increased hippocampal loss is an indicator of Alzheimer's disease pathology and a potential marker for the efficacy of therapeutic interventions in Alzheimer's disease.