Identifying Goals of Care Conversations in the Electronic Health Record Using Natural Language Processing and Machine Learning

ContextGoals-of-care discussions are an important quality metric in palliative care. However, goals-of-care discussions are often documented as free text in diverse locations. It is difficult to identify these discussions in the electronic health record (EHR) efficiently.ObjectivesTo develop, train, and test an automated approach to identifying goals-of-care discussions in the EHR, using natural language processing (NLP) and machine learning (ML).MethodsFrom the electronic health records of an academic health system, we collected a purposive sample of 3183 EHR notes (1435 inpatient notes and 1748 outpatient notes) from 1426 patients with serious illness over 2008–2016, and manually reviewed each note for documentation of goals-of-care discussions. Separately, we developed a program to identify notes containing documentation of goals-of-care discussions using NLP and supervised ML. We estimated the performance characteristics of the NLP/ML program across 100 pairs of randomly partitioned training and test sets. We repeated these methods for inpatient-only and outpatient-only subsets.ResultsOf 3183 notes, 689 contained documentation of goals-of-care discussions. The mean sensitivity of the NLP/ML program was 82.3% (SD 3.2%), and the mean specificity was 97.4% (SD 0.7%). NLP/ML results had a median positive likelihood ratio of 32.2 (IQR 27.5–39.2) and a median negative likelihood ratio of 0.18 (IQR 0.16–0.20). Performance was better in inpatient-only samples than outpatient-only samples.ConclusionUsing NLP and ML techniques, we developed a novel approach to identifying goals-of-care discussions in the EHR. NLP and ML represent a potential approach toward measuring goals-of-care discussions as a research outcome and quality metric.     

Functional neuroanatomy of language without speech: An ALE meta‐analysis of sign language

Sign language (SL) conveys linguistic information using gestures instead of sounds. Here, we apply a meta‐analytic estimation approach to neuroimaging studies (N = 23; subjects = 316) and ask whether SL comprehension in deaf signers relies on the same primarily left‐hemispheric cortical network implicated in spoken and written language (SWL) comprehension in hearing speakers. We show that: (a) SL recruits bilateral fronto‐temporo‐occipital regions with strong left‐lateralization in the posterior inferior frontal gyrus known as Broca''s area, mirroring functional asymmetries observed for SWL. (b) Within this SL network, Broca''s area constitutes a hub which attributes abstract linguistic information to gestures. (c) SL‐specific voxels in Broca''s area are also crucially involved in SWL, as confirmed by meta‐analytic connectivity modeling using an independent large‐scale neuroimaging database. This strongly suggests that the human brain evolved a lateralized language network with a supramodal hub in Broca''s area which computes linguistic information independent of speech.     

Using Natural Language Processing on Electronic Health Records to Enhance Detection and Prediction of Psychosis Risk

BackgroundUsing novel data mining methods such as natural language processing (NLP) on electronic health records (EHRs) for screening and detecting individuals at risk for psychosis. MethodThe study included all patients receiving a first index diagnosis of nonorganic and nonpsychotic mental disorder within the South London and Maudsley (SLaM) NHS Foundation Trust between January 1, 2008, and July 28, 2018. Least Absolute Shrinkage and Selection Operator (LASSO)-regularized Cox regression was used to refine and externally validate a refined version of a five-item individualized, transdiagnostic, clinically based risk calculator previously developed (Harrell’s C = 0.79) and piloted for implementation. The refined version included 14 additional NLP-predictors: tearfulness, poor appetite, weight loss, insomnia, cannabis, cocaine, guilt, irritability, delusions, hopelessness, disturbed sleep, poor insight, agitation, and paranoia. ResultsA total of 92 151 patients with a first index diagnosis of nonorganic and nonpsychotic mental disorder within the SLaM Trust were included in the derivation (n = 28 297) or external validation (n = 63 854) data sets. Mean age was 33.6 years, 50.7% were women, and 67.0% were of white race/ethnicity. Mean follow-up was 1590 days. The overall 6-year risk of psychosis in secondary mental health care was 3.4 (95% CI, 3.3–3.6). External validation indicated strong performance on unseen data (Harrell’s C 0.85, 95% CI 0.84–0.86), an increase of 0.06 from the original model. ConclusionsUsing NLP on EHRs can considerably enhance the prognostic accuracy of psychosis risk calculators. This can help identify patients at risk of psychosis who require assessment and specialized care, facilitating earlier detection and potentially improving patient outcomes.     

QRICH1 variants in Ververi‐Brady syndrome—delineation of the genotypic and phenotypic spectrum

Ververi‐Brady syndrome (VBS, # 617982) is a rare developmental disorder, and loss‐of‐function variants in QRICH1 were implicated in its etiology. Furthermore, a recognizable phenotype was proposed comprising delayed speech, learning difficulties and dysmorphic signs. Here, we present four unrelated individuals with one known nonsense variant (c.1954C > T; p.[Arg652*]) and three novel de novo QRICH1 variants, respectively. These included two frameshift mutations (c.832_833del; p.(Ser278Leufs*25), c.1812_1813delTG; p.(Glu605Glyfs*25)) and interestingly one missense mutation (c.2207G > A; p.[Ser736Asn]), expanding the mutational spectrum. Enlargement of the cohort by these four individuals contributes to the delineation of the VBS phenotype and suggests expressive speech delay, moderate motor delay, learning difficulties/mild ID, mild microcephaly, short stature and notable social behavior deficits as clinical hallmarks. In addition, one patient presented with nephroblastoma. The possible involvement of QRICH1 in pediatric cancer assumes careful surveillance a key priority for outcome of these patients. Further research and enlargement of cohorts are warranted to learn about the genetic architecture and the phenotypic spectrum in more detail.     

Situating language in a minimal social context: how seeing a picture of the speaker’s face affects language comprehension

Natural use of language involves at least two individuals. Some studies have focused on the interaction between senders in communicative situations and how the knowledge about the speaker can bias language comprehension. However, the mere effect of a face as a social context on language processing remains unknown. In the present study, we used event-related potentials to investigate the semantic and morphosyntactic processing of speech in the presence of a photographic portrait of the speaker. In Experiment 1, we show that the N400, a component related to semantic comprehension, increased its amplitude when processed within this minimal social context compared to a scrambled face control condition. Hence, the semantic neural processing of speech is sensitive to the concomitant perception of a picture of the speaker’s face, even if irrelevant to the content of the sentences. Moreover, a late posterior negativity effect was found to the presentation of the speaker’s face compared to control stimuli. In contrast, in Experiment 2, we found that morphosyntactic processing, as reflected in left anterior negativity and P600 effects, is not notably affected by the presence of the speaker’s portrait. Overall, the present findings suggest that the mere presence of the speaker’s image seems to trigger a minimal communicative context, increasing processing resources for language comprehension at the semantic level.     

From Gesture to Sign Language: Conventionalization of Classifier Constructions by Adult Hearing Learners of British Sign Language

There has long been interest in why languages are shaped the way they are, and in the relationship between sign language and gesture. In sign languages, entity classifiers are handshapes that encode how objects move, how they are located relative to one another, and how multiple objects of the same type are distributed in space. Previous studies have shown that hearing adults who are asked to use only manual gestures to describe how objects move in space will use gestures that bear some similarities to classifiers. We investigated how accurately hearing adults, who had been learning British Sign Language (BSL) for 1–3 years, produce and comprehend classifiers in (static) locative and distributive constructions. In a production task, learners of BSL knew that they could use their hands to represent objects, but they had difficulty choosing the same, conventionalized, handshapes as native signers. They were, however, highly accurate at encoding location and orientation information. Learners therefore show the same pattern found in sign-naïve gesturers. In contrast, handshape, orientation, and location were comprehended with equal (high) accuracy, and testing a group of sign-naïve adults showed that they too were able to understand classifiers with higher than chance accuracy. We conclude that adult learners of BSL bring their visuo-spatial knowledge and gestural abilities to the tasks of understanding and producing constructions that contain entity classifiers. We speculate that investigating the time course of adult sign language acquisition might shed light on how gesture became (and, indeed, becomes) conventionalized during the genesis of sign languages.     

Tools for Language: Patterned Iconicity in Sign Language Nouns and Verbs

When naming certain hand-held, man-made tools, American Sign Language (ASL) signers exhibit either of two iconic strategies: a handling strategy, where the hands show holding or grasping an imagined object in action, or an instrument strategy, where the hands represent the shape or a dimension of the object in a typical action. The same strategies are also observed in the gestures of hearing nonsigners identifying pictures of the same set of tools. In this paper, we compare spontaneously created gestures from hearing nonsigning participants to commonly used lexical signs in ASL. Signers and gesturers were asked to respond to pictures of tools and to video vignettes of actions involving the same tools. Nonsigning gesturers overwhelmingly prefer the handling strategy for both the Picture and Video conditions. Nevertheless, they use more instrument forms when identifying tools in pictures, and more handling forms when identifying actions with tools. We found that ASL signers generally favor the instrument strategy when naming tools, but when describing tools being used by an actor, they are significantly more likely to use more handling forms. The finding that both gesturers and signers are more likely to alternate strategies when the stimuli are pictures or video suggests a common cognitive basis for differentiating objects from actions. Furthermore, the presence of a systematic handling/instrument iconic pattern in a sign language demonstrates that a conventionalized sign language exploits the distinction for grammatical purpose, to distinguish nouns and verbs related to tool use.