Measuring Implicit Bias in ICU Notes Using Word-Embedding Neural Network Models.

BACKGROUND: Language in nonmedical data sets is known to transmit human-like biases when used in natural language processing (NLP) algorithms that can reinforce disparities. It is unclear if NLP algorithms of medical notes could lead to similar transmissions of biases. RESEARCH QUESTION: Can we identify implicit bias in clinical notes, and are biases stable across time and geography? STUDY DESIGN AND METHODS: To determine whether different racial and ethnic descriptors are similar contextually to stigmatizing language in ICU notes and whether these relationships are stable across time and geography, we identified notes on critically ill adults admitted to the University of California, San Francisco (UCSF), from 2012 through 2022 and to Beth Israel Deaconess Hospital (BIDMC) from 2001 through 2012. Because word meaning is derived largely from context, we trained unsupervised word-embedding algorithms to measure the similarity (cosine similarity) quantitatively of the context between a racial or ethnic descriptor (eg, African-American) and a stigmatizing target word (eg, nonco-operative) or group of words (violence, passivity, noncompliance, nonadherence). RESULTS: In UCSF notes, Black descriptors were less likely to be similar contextually to violent words compared with White descriptors. Contrastingly, in BIDMC notes, Black descriptors were more likely to be similar contextually to violent words compared with White descriptors. The UCSF data set also showed that Black descriptors were more similar contextually to passivity and noncompliance words compared with Latinx descriptors. INTERPRETATION: Implicit bias is identifiable in ICU notes. Racial and ethnic group descriptors carry different contextual relationships to stigmatizing words, depending on when and where notes were written. Because NLP models seem able to transmit implicit bias from training data, use of NLP algorithms in clinical prediction could reinforce disparities. Active debiasing strategies may be necessary to achieve algorithmic fairness when using language models in clinical research.

Time is Short: Tools to Integrate Palliative Care and Communication Skills Education into Your Surgical Residency.

The need to integrate palliative care (PC) training into surgical education has been increasingly recognized. Our aim is to describe a set of PC educational strategies, with a range of requisite resources, time, and prior expertise, to provide options that surgical educators can tailor for different programs. Each of these strategies has been successfully employed individually or in some combination at our institutions, and components can be generalized to other training programs. Asynchronous and individually paced PC training can be provided using existing resources published by the American College of Surgeons and upcoming SCORE curriculum modules. A multiyear PC curriculum, with didactic components of increasing complexity for more advanced residents, can be applied based on available time in the didactic schedule and local expertise. Simulation-based training in PC skills can be developed to provide objective competency-based training. Finally, a dedicated rotation on a surgical palliative care service can provide the most immersive experience with steps toward clinical entrustment of PC skills for trainees.

Negativity and Positivity in the ICU: Exploratory Development of Automated Sentiment Capture in the Electronic Health Record.

OBJECTIVES: To develop proof-of-concept algorithms using alternative approaches to capture provider sentiment in ICU notes. DESIGN: Retrospective observational cohort study. SETTING: The Multiparameter Intelligent Monitoring of Intensive Care III (MIMIC-III) and the University of California, San Francisco (UCSF) deidentified notes databases. PATIENTS: Adult (≥18 yr old) patients admitted to the ICU. MEASUREMENTS AND MAIN RESULTS: We developed two sentiment models: 1) a keywords-based approach using a consensus-based clinical sentiment lexicon comprised of 72 positive and 103 negative phrases, including negations and 2) a Decoding-enhanced Bidirectional Encoder Representations from Transformers with disentangled attention-v3-based deep learning model (keywords-independent) trained on clinical sentiment labels. We applied the models to 198,944 notes across 52,997 ICU admissions in the MIMIC-III database. Analyses were replicated on an external sample of patients admitted to a UCSF ICU from 2018 to 2019. We also labeled sentiment in 1,493 note fragments and compared the predictive accuracy of our tools to three popular sentiment classifiers. Clinical sentiment terms were found in 99% of patient visits across 88% of notes. Our two sentiment tools were substantially more predictive (Spearman correlations of 0.62-0.84, p values < 0.00001) of labeled sentiment compared with general language algorithms (0.28-0.46). CONCLUSION: Our exploratory healthcare-specific sentiment models can more accurately detect positivity and negativity in clinical notes compared with general sentiment tools not designed for clinical usage.

Utilization and Delivery of Specialty Palliative Care in the ICU: Insights from the Palliative Care Quality Network.

CONTEXT: Palliative care (PC) benefits critically ill patients but remains underutilized. Important to developing interventions to overcome barriers to PC in the ICU and address PC needs of ICU patients is to understand how, when, and for which patients PC is provided in the ICU. OBJECTIVES: Compare characteristics of specialty PC consultations in the ICU to those on medical-surgical wards. METHODS: Retrospective analysis of national Palliative Care Quality Network data for hospitalized patients receiving specialty PC consultation January 1, 2013 to December 31, 2019 in ICU or medical-surgical setting. 98 inpatient PC teams in 16 states contributed data. Measures and outcomes included patient characteristics, consultation features, process metrics and patient outcomes. Mixed effects multivariable logistic regression was used to compare ICU and medical-surgical units. RESULTS: Of 102,597 patients 63,082 were in medical-surgical units and 39,515 ICU. ICU patients were younger and more likely to have non-cancer diagnoses (all P < 0.001). While fewer ICU patients were able to report symptoms, most patients in both groups reported improved symptoms. ICU patients were more likely to have consultation requests for GOC, comfort care, and withdrawal of interventions and less likely for pain and/or symptoms (OR-all P < 0.001). ICU patients were less often discharged alive. CONCLUSION: ICU patients receiving PC consultation are more likely to have non-cancer diagnoses and less likely able to communicate. Although symptom management and GOC are standard parts of ICU care, specialty PC in the ICU is often engaged for these issues and results in improved symptoms, suggesting routine interventions and consultation targeting these needs could improve care.