Chao Yan1, Cheng Gao2, Ziqi Zhang1, Wencong Chen3,4, Bradley A Malin1,2,3, E Wesley Ely4,5, Mayur B Patel4,5,6,7,8,9,10, You Chen1,2. 1. Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, Tennessee, USA. 2. Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 3. Department of Biostatistics, School of Medicine, Vanderbilt University, Nashville, Tennessee, USA. 4. Critical Illness, Brain Dysfunction, and Survivorship Center, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 5. Geriatric Research and Education Clinical Center, Tennessee Valley Healthcare System, U.S. Department of Veteran Affairs, Nashville, Tennessee, USA. 6. Section of Surgical Sciences, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 7. Department of Surgery, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 8. Department of Neurosurgery, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 9. Department of Hearing & Speech Sciences, Vanderbilt University Medical Center, Nashville, Tennessee, USA. 10. Division of Trauma, Surgical Critical Care, and Emergency General Surgery, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Abstract
OBJECTIVE: In intensive care units (ICUs), a patient's brain function status can shift from a state of acute brain dysfunction (ABD) to one that is ABD-free and vice versa, which is challenging to forecast and, in turn, hampers the allocation of hospital resources. We aim to develop a machine learning model to predict next-day brain function status changes. MATERIALS AND METHODS: Using multicenter prospective adult cohorts involving medical and surgical ICU patients from 2 civilian and 3 Veteran Affairs hospitals, we trained and externally validated a light gradient boosting machine to predict brain function status changes. We compared the performances of the boosting model against state-of-the-art models-an ABD predictive model and its variants. We applied Shapley additive explanations to identify influential factors to develop a compact model. RESULTS: There were 1026 critically ill patients without evidence of prior major dementia, or structural brain diseases, from whom 12 295 daily transitions (ABD: 5847 days; ABD-free: 6448 days) were observed. The boosting model achieved an area under the receiver-operating characteristic curve (AUROC) of 0.824 (95% confidence interval [CI], 0.821-0.827), compared with the state-of-the-art models of 0.697 (95% CI, 0.693-0.701) with P < .001. Using 13 identified top influential factors, the compact model achieved 99.4% of the boosting model on AUROC. The boosting and the compact models demonstrated high generalizability in external validation by achieving an AUROC of 0.812 (95% CI, 0.812-0.813). CONCLUSION: The inputs of the compact model are based on several simple questions that clinicians can quickly answer in practice, which demonstrates the model has direct prospective deployment potential into clinical practice, aiding in critical hospital resource allocation.
OBJECTIVE: In intensive care units (ICUs), a patient's brain function status can shift from a state of acute brain dysfunction (ABD) to one that is ABD-free and vice versa, which is challenging to forecast and, in turn, hampers the allocation of hospital resources. We aim to develop a machine learning model to predict next-day brain function status changes. MATERIALS AND METHODS: Using multicenter prospective adult cohorts involving medical and surgical ICU patients from 2 civilian and 3 Veteran Affairs hospitals, we trained and externally validated a light gradient boosting machine to predict brain function status changes. We compared the performances of the boosting model against state-of-the-art models-an ABD predictive model and its variants. We applied Shapley additive explanations to identify influential factors to develop a compact model. RESULTS: There were 1026 critically ill patients without evidence of prior major dementia, or structural brain diseases, from whom 12 295 daily transitions (ABD: 5847 days; ABD-free: 6448 days) were observed. The boosting model achieved an area under the receiver-operating characteristic curve (AUROC) of 0.824 (95% confidence interval [CI], 0.821-0.827), compared with the state-of-the-art models of 0.697 (95% CI, 0.693-0.701) with P < .001. Using 13 identified top influential factors, the compact model achieved 99.4% of the boosting model on AUROC. The boosting and the compact models demonstrated high generalizability in external validation by achieving an AUROC of 0.812 (95% CI, 0.812-0.813). CONCLUSION: The inputs of the compact model are based on several simple questions that clinicians can quickly answer in practice, which demonstrates the model has direct prospective deployment potential into clinical practice, aiding in critical hospital resource allocation.
Authors: M van den Boogaard; L Schoonhoven; E Maseda; C Plowright; C Jones; A Luetz; P V Sackey; P G Jorens; L M Aitken; F M P van Haren; R Donders; J G van der Hoeven; P Pickkers Journal: Intensive Care Med Date: 2014-01-18 Impact factor: 17.440
Authors: Nita Khandelwal; Erin K Kross; Ruth A Engelberg; Norma B Coe; Ann C Long; J Randall Curtis Journal: Crit Care Med Date: 2015-05 Impact factor: 7.598
Authors: William D Schweickert; Mark C Pohlman; Anne S Pohlman; Celerina Nigos; Amy J Pawlik; Cheryl L Esbrook; Linda Spears; Megan Miller; Mietka Franczyk; Deanna Deprizio; Gregory A Schmidt; Amy Bowman; Rhonda Barr; Kathryn E McCallister; Jesse B Hall; John P Kress Journal: Lancet Date: 2009-05-14 Impact factor: 79.321
Authors: Heidi Lindroth; Lisa Bratzke; Suzanne Purvis; Roger Brown; Mark Coburn; Marko Mrkobrada; Matthew T V Chan; Daniel H J Davis; Pratik Pandharipande; Cynthia M Carlsson; Robert D Sanders Journal: BMJ Open Date: 2018-04-28 Impact factor: 2.692