FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2896540029> ?p ?o ?g. }

• W2896540029 abstract "The increasing prevalence of digitized clinical datacreates new opportunities to use machine learning to unlockclinical insights, and ultimately improve healthcare delivery.However, while data from Electronic Health Records (EHRs) havebecome common, they present unique challenges. Clinical data arenoisy, sparse, irregularly sampled, and often biased in theirrecording of health state and care patterns. Further, much of themost important information used by care staff is recorded inunstructured text notes that are not easily deciphered bynon-experts. In this work, we present machine learning methods thatdistill large amounts of text-based clinical data into latentrepresentations. These representations are then used to predict avariety of important outcomes. In particular, we focus onprediction tasks that can provide evidence-based risk assessmentand forecasting in settings with guidelines that have nottraditionally been data-driven. We consider several abstractionsfor clinical narrative text, and evaluate their utility on commonpredictive tasks, such as mortality and readmission. We argue thata good representation will improve performance on these tasks andthat multiple representations may be necessary, as different modelsexcel on differing tasks. We present three case studies in which weuse representations of clinical text to improve performance ofclinical prediction tasks. First, we augment predictive models thatused baseline clinical features by including features from clinicalprogress notes [31].These features are derived using LatentDirichlet Allocation (LDA) and incorporated as features usingper-patient topic membership. Notably, this representation has thebenefit of interpretable topics over which each patient can berepresented as a distribution. Second, we explore the expressivepower of clinical prose by evaluating the performance of severalcommon models on both downstream clinical tasks and their abilityto identify information contained in patients' notes [7]. Thisstands in contrast to much prior work that positions the utility ofa given model solely with respect to its ability to improvedownstream clinical performance. Such extrinsic evaluations areblind to much of the insight contained in the notes, thusmotivating the need for intrinsic evaluations. Finally, we use thetext-based metadata associated with EHR encodings to allow thetransfer of predictive models from one database to another [35].Existing machine learning methods typically assume consistency inhow semantically equivalent information is encoded. However, theway information is recorded differs across institutions and overtime, often rendering potentially useful data obsolescent. Toaddress this problem, we map database-specific representations ofthe information to a shared set of semantic concepts, thus allowingmodels to be built from or transition across differentdatabases." @default.
• W2896540029 created "2018-10-26" @default.
• W2896540029 creator A5023123863 @default.
• W2896540029 date "2018-06-01" @default.
• W2896540029 modified "2023-09-23" @default.
• W2896540029 title "Leveraging text representations for clinical predictive tasks" @default.
• W2896540029 cites W131564882 @default.
• W2896540029 cites W1657774248 @default.
• W2896540029 cites W1880262756 @default.
• W2896540029 cites W1898928487 @default.
• W2896540029 cites W1919152067 @default.
• W2896540029 cites W1964986019 @default.
• W2896540029 cites W1979573023 @default.
• W2896540029 cites W1996796871 @default.
• W2896540029 cites W200246157 @default.
• W2896540029 cites W2004166947 @default.
• W2896540029 cites W2008441672 @default.
• W2896540029 cites W2011861268 @default.
• W2896540029 cites W2035038290 @default.
• W2896540029 cites W2057780602 @default.
• W2896540029 cites W2072644219 @default.
• W2896540029 cites W2076652359 @default.
• W2896540029 cites W2083634842 @default.
• W2896540029 cites W2095425517 @default.
• W2896540029 cites W2101234009 @default.
• W2896540029 cites W2104381725 @default.
• W2896540029 cites W2107182304 @default.
• W2896540029 cites W2132755184 @default.
• W2896540029 cites W2139646386 @default.
• W2896540029 cites W2146089916 @default.
• W2896540029 cites W2153635508 @default.
• W2896540029 cites W2154048976 @default.
• W2896540029 cites W2157963512 @default.
• W2896540029 cites W2167032822 @default.
• W2896540029 cites W2169800130 @default.
• W2896540029 cites W2208683342 @default.
• W2896540029 cites W236448318 @default.
• W2896540029 cites W2405994726 @default.
• W2896540029 cites W2774292910 @default.
• W2896540029 cites W2788104422 @default.
• W2896540029 cites W2949547296 @default.
• W2896540029 cites W2963956191 @default.
• W2896540029 cites W2120156715 @default.
• W2896540029 hasPublicationYear "2018" @default.
• W2896540029 type Work @default.
• W2896540029 sameAs 2896540029 @default.
• W2896540029 citedByCount "0" @default.
• W2896540029 crossrefType "dissertation" @default.
• W2896540029 hasAuthorship W2896540029A5023123863 @default.
• W2896540029 hasConcept C119857082 @default.
• W2896540029 hasConcept C136197465 @default.
• W2896540029 hasConcept C154945302 @default.
• W2896540029 hasConcept C160735492 @default.
• W2896540029 hasConcept C162324750 @default.
• W2896540029 hasConcept C171686336 @default.
• W2896540029 hasConcept C17744445 @default.
• W2896540029 hasConcept C199539241 @default.
• W2896540029 hasConcept C204321447 @default.
• W2896540029 hasConcept C23123220 @default.
• W2896540029 hasConcept C2522767166 @default.
• W2896540029 hasConcept C2776359362 @default.
• W2896540029 hasConcept C41008148 @default.
• W2896540029 hasConcept C500882744 @default.
• W2896540029 hasConcept C50522688 @default.
• W2896540029 hasConcept C71472368 @default.
• W2896540029 hasConcept C94625758 @default.
• W2896540029 hasConceptScore W2896540029C119857082 @default.
• W2896540029 hasConceptScore W2896540029C136197465 @default.
• W2896540029 hasConceptScore W2896540029C154945302 @default.
• W2896540029 hasConceptScore W2896540029C160735492 @default.
• W2896540029 hasConceptScore W2896540029C162324750 @default.
• W2896540029 hasConceptScore W2896540029C171686336 @default.
• W2896540029 hasConceptScore W2896540029C17744445 @default.
• W2896540029 hasConceptScore W2896540029C199539241 @default.
• W2896540029 hasConceptScore W2896540029C204321447 @default.
• W2896540029 hasConceptScore W2896540029C23123220 @default.
• W2896540029 hasConceptScore W2896540029C2522767166 @default.
• W2896540029 hasConceptScore W2896540029C2776359362 @default.
• W2896540029 hasConceptScore W2896540029C41008148 @default.
• W2896540029 hasConceptScore W2896540029C500882744 @default.
• W2896540029 hasConceptScore W2896540029C50522688 @default.
• W2896540029 hasConceptScore W2896540029C71472368 @default.
• W2896540029 hasConceptScore W2896540029C94625758 @default.
• W2896540029 hasLocation W28965400291 @default.
• W2896540029 hasOpenAccess W2896540029 @default.
• W2896540029 hasPrimaryLocation W28965400291 @default.
• W2896540029 hasRelatedWork W1486762633 @default.
• W2896540029 hasRelatedWork W2149028520 @default.
• W2896540029 hasRelatedWork W2170817709 @default.
• W2896540029 hasRelatedWork W2284250673 @default.
• W2896540029 hasRelatedWork W2306340896 @default.
• W2896540029 hasRelatedWork W2549442941 @default.
• W2896540029 hasRelatedWork W2586051142 @default.
• W2896540029 hasRelatedWork W2743778505 @default.
• W2896540029 hasRelatedWork W2770545708 @default.
• W2896540029 hasRelatedWork W2955712006 @default.
• W2896540029 hasRelatedWork W2963795072 @default.
• W2896540029 hasRelatedWork W2973767881 @default.
• W2896540029 hasRelatedWork W2979949198 @default.
• W2896540029 hasRelatedWork W3030501723 @default.

• next