The Experts below are selected from a list of 1743 Experts worldwide ranked by ideXlab platform
Marie-helene Metzger - One of the best experts on this subject based on the ideXlab platform.
-
Deep learning versus conventional machine learning for detection of healthcare-associated Infections in French clinical narratives
Methods of Information in Medicine, 2019Co-Authors: Sara Rabhi, Jérémie Jakubowicz, Marie-helene MetzgerAbstract:Objective: The objective of this article was to compare the performances of health care-associated Infection (HAI) detection between deep learning and conventional machine learning (ML) methods in French medical reports. Methods: The corpus consisted in different types of medical reports (discharge summaries, surgery reports, consultation reports, etc.). A total of 1,531 medical text documents were extracted and deidentified in three French university hospitals. Each of them was labeled as presence (1) or absence (0) of HAI. We started by normalizing the records using a list of preprocessing techniques. We calculated an overall performance metric, the F1 Score, to compare a deep learning method (convolutional neural network [CNN]) with the most popular conventional ML models (Bernoulli and multi-naïve Bayes, k-nearest neighbors, logistic regression, random forests, extra-trees, gradient boosting, support vector machines). We applied the hyperparameter Bayesian optimization for each model based on its HAI identification performances. We included the set of text representation as an additional hyperparameter for each model, using four different text representations (bag of words, term frequency–inverse document frequency, word2vec, and Glove). Results: CNN outperforms all other conventional ML algorithms for HAI classification. The best F1 Score of 97.7% ± 3.6% and best area under the curve score of 99.8% ± 0.41% were achieved when CNN was directly applied to the processed clinical notes without a pretrained word2vec embedding. Through receiver operating characteristic curve analysis, we could achieve a good balance between false notifications (with a specificity equal to 0.937) and system detection capability (with a sensitivity equal to 0.962) using the Youden's index reference. Conclusions: The main drawback of CNNs is their opacity. To address this issue, we investigated CNN inner layers' activation values to visualize the most meaningful phrases in a document. This method could be used to build a phrase-based medical assistant algorithm to help the Infection Control Practitioner to select relevant medical records. Our study demonstrated that deep learning approach outperforms other classification learning algorithms for automatically identifying HAIs in medical reports.
-
Deep Learning versus Conventional Machine Learning for Detection of Healthcare-Associated Infections in French Clinical Narratives.
Methods of information in medicine, 2019Co-Authors: Sara Rabhi, Jérémie Jakubowicz, Marie-helene MetzgerAbstract:The objective of this article was to compare the performances of health care-associated Infection (HAI) detection between deep learning and conventional machine learning (ML) methods in French medical reports. The corpus consisted in different types of medical reports (discharge summaries, surgery reports, consultation reports, etc.). A total of 1,531 medical text documents were extracted and deidentified in three French university hospitals. Each of them was labeled as presence (1) or absence (0) of HAI. We started by normalizing the records using a list of preprocessing techniques. We calculated an overall performance metric, the F1 Score, to compare a deep learning method (convolutional neural network [CNN]) with the most popular conventional ML models (Bernoulli and multi-naïve Bayes, k-nearest neighbors, logistic regression, random forests, extra-trees, gradient boosting, support vector machines). We applied the hyperparameter Bayesian optimization for each model based on its HAI identification performances. We included the set of text representation as an additional hyperparameter for each model, using four different text representations (bag of words, term frequency-inverse document frequency, word2vec, and Glove). CNN outperforms all other conventional ML algorithms for HAI classification. The best F1 Score of 97.7% ± 3.6% and best area under the curve score of 99.8% ± 0.41% were achieved when CNN was directly applied to the processed clinical notes without a pretrained word2vec embedding. Through receiver operating characteristic curve analysis, we could achieve a good balance between false notifications (with a specificity equal to 0.937) and system detection capability (with a sensitivity equal to 0.962) using the Youden's index reference. The main drawback of CNNs is their opacity. To address this issue, we investigated CNN inner layers' activation values to visualize the most meaningful phrases in a document. This method could be used to build a phrase-based medical assistant algorithm to help the Infection Control Practitioner to select relevant medical records. Our study demonstrated that deep learning approach outperforms other classification learning algorithms for automatically identifying HAIs in medical reports. Georg Thieme Verlag KG Stuttgart · New York.
Sara Rabhi - One of the best experts on this subject based on the ideXlab platform.
-
Deep learning versus conventional machine learning for detection of healthcare-associated Infections in French clinical narratives
Methods of Information in Medicine, 2019Co-Authors: Sara Rabhi, Jérémie Jakubowicz, Marie-helene MetzgerAbstract:Objective: The objective of this article was to compare the performances of health care-associated Infection (HAI) detection between deep learning and conventional machine learning (ML) methods in French medical reports. Methods: The corpus consisted in different types of medical reports (discharge summaries, surgery reports, consultation reports, etc.). A total of 1,531 medical text documents were extracted and deidentified in three French university hospitals. Each of them was labeled as presence (1) or absence (0) of HAI. We started by normalizing the records using a list of preprocessing techniques. We calculated an overall performance metric, the F1 Score, to compare a deep learning method (convolutional neural network [CNN]) with the most popular conventional ML models (Bernoulli and multi-naïve Bayes, k-nearest neighbors, logistic regression, random forests, extra-trees, gradient boosting, support vector machines). We applied the hyperparameter Bayesian optimization for each model based on its HAI identification performances. We included the set of text representation as an additional hyperparameter for each model, using four different text representations (bag of words, term frequency–inverse document frequency, word2vec, and Glove). Results: CNN outperforms all other conventional ML algorithms for HAI classification. The best F1 Score of 97.7% ± 3.6% and best area under the curve score of 99.8% ± 0.41% were achieved when CNN was directly applied to the processed clinical notes without a pretrained word2vec embedding. Through receiver operating characteristic curve analysis, we could achieve a good balance between false notifications (with a specificity equal to 0.937) and system detection capability (with a sensitivity equal to 0.962) using the Youden's index reference. Conclusions: The main drawback of CNNs is their opacity. To address this issue, we investigated CNN inner layers' activation values to visualize the most meaningful phrases in a document. This method could be used to build a phrase-based medical assistant algorithm to help the Infection Control Practitioner to select relevant medical records. Our study demonstrated that deep learning approach outperforms other classification learning algorithms for automatically identifying HAIs in medical reports.
-
Deep Learning versus Conventional Machine Learning for Detection of Healthcare-Associated Infections in French Clinical Narratives.
Methods of information in medicine, 2019Co-Authors: Sara Rabhi, Jérémie Jakubowicz, Marie-helene MetzgerAbstract:The objective of this article was to compare the performances of health care-associated Infection (HAI) detection between deep learning and conventional machine learning (ML) methods in French medical reports. The corpus consisted in different types of medical reports (discharge summaries, surgery reports, consultation reports, etc.). A total of 1,531 medical text documents were extracted and deidentified in three French university hospitals. Each of them was labeled as presence (1) or absence (0) of HAI. We started by normalizing the records using a list of preprocessing techniques. We calculated an overall performance metric, the F1 Score, to compare a deep learning method (convolutional neural network [CNN]) with the most popular conventional ML models (Bernoulli and multi-naïve Bayes, k-nearest neighbors, logistic regression, random forests, extra-trees, gradient boosting, support vector machines). We applied the hyperparameter Bayesian optimization for each model based on its HAI identification performances. We included the set of text representation as an additional hyperparameter for each model, using four different text representations (bag of words, term frequency-inverse document frequency, word2vec, and Glove). CNN outperforms all other conventional ML algorithms for HAI classification. The best F1 Score of 97.7% ± 3.6% and best area under the curve score of 99.8% ± 0.41% were achieved when CNN was directly applied to the processed clinical notes without a pretrained word2vec embedding. Through receiver operating characteristic curve analysis, we could achieve a good balance between false notifications (with a specificity equal to 0.937) and system detection capability (with a sensitivity equal to 0.962) using the Youden's index reference. The main drawback of CNNs is their opacity. To address this issue, we investigated CNN inner layers' activation values to visualize the most meaningful phrases in a document. This method could be used to build a phrase-based medical assistant algorithm to help the Infection Control Practitioner to select relevant medical records. Our study demonstrated that deep learning approach outperforms other classification learning algorithms for automatically identifying HAIs in medical reports. Georg Thieme Verlag KG Stuttgart · New York.
Mary K. Hayden - One of the best experts on this subject based on the ideXlab platform.
-
Daily skin cleansing with chlorhexidine did not reduce the rate of central-line associated bloodstream Infection in a surgical intensive care unit
Intensive Care Medicine, 2010Co-Authors: Kyle J. Popovich, Bala Hota, Robert Hayes, Robert A. Weinstein, Mary K. HaydenAbstract:Purpose Cleansing the skin of intensive care unit (ICU) patients daily with chlorhexidine gluconate (CHG) has been associated with beneficial effects, including a reduction in central-line-associated bacteremias (CLABSIs). Most studies have been done in medical ICUs. In this study, we evaluated the effectiveness of daily chlorhexidine skin cleansing on CLABSI rates in a surgical ICU. Methods In Fall 2005, the 30-bed surgical ICU at Rush University Medical Center discontinued daily soap-and-water bathing of patients and substituted skin cleansing with no-rinse, 2% CHG-impregnated cloths. This change was made without research investigator input or oversight. Using administrative, microbiological and Infection Control Practitioner databases, we compared rates of CLABSIs and blood culture contamination during soap-and-water bathing (September 2004–October 2005) and CHG cleansing (November 2005–October 2006) periods. Rates of other nosocomial Infections that were not expected to be affected by CHG bathing (secondary bacteremia, Clostridium difficile -associated diarrhea, ventilator-associated pneumonia, urinary tract Infection) were included as Control variables. Results There was no significant difference in the CLABSI rate between soap-and-water and CHG bathing periods (3.81/1,000 central line days vs. 4.6/1,000 central line days; p = 0.57). Blood culture contamination declined during CHG bathing (5.97/1,000 to 2.41/1,000 patient days; p = 0.003). Rates of other nosocomial Infections did not change significantly. Conclusions In this real-world effectiveness trial, daily cleansing of surgical ICU patients’ skin with CHG had no effect on CLABSI rates, but was associated with half the rate of blood culture contamination. Controlled trials in surgical ICUs are needed to determine whether CHG bathing can prevent Infections in this setting.
F. Lopes Cardoso - One of the best experts on this subject based on the ideXlab platform.
-
Ten key points for the appropriate use of antibiotics in hospitalised patients: a consensus from the Antimicrobial Stewardship and Resistance Working Groups of the International Society of Chemotherapy
International Journal of Antimicrobial Agents, 2016Co-Authors: G. Levy Hara, S.s. Kanj, L. Pagani, L. Abbo, A. Endimiani, H.f.l. Wertheim, C. Amábile-cuevas, P. Tattevin, S. Mehtar, F. Lopes CardosoAbstract:The Antibiotic Stewardship and Resistance Working Groups of the International Society for Chemotherapy propose ten key points for the appropriate use of antibiotics in hospital settings. (i) Get appropriate microbiological samples before antibiotic administration and carefully interpret the results: in the absence of clinical signs of Infection, colonisation rarely requires antimicrobial treatment. (ii) Avoid the use of antibiotics to ‘treat’ fever: use them to treat Infections, and investigate the root cause of fever prior to starting treatment. (iii) Start empirical antibiotic treatment after taking cultures, tailoring it to the site of Infection, risk factors for multidrug-resistant bacteria, and the local microbiology and susceptibility patterns. (iv) Prescribe drugs at their optimal dosing and for an appropriate duration, adapted to each clinical situation and patient characteristics. (v) Use antibiotic combinations only where the current evidence suggests some benefit. (vi) When possible, avoid antibiotics with a higher likelihood of promoting drug resistance or hospital-acquired Infections, or use them only as a last resort. (vii) Drain the infected foci quickly and remove all potentially or proven infected devices: Control the Infection source. (viii) Always try to de-escalate/streamline antibiotic treatment according to the clinical situation and the microbiological results. (ix) Stop unnecessarily prescribed antibiotics once the absence of Infection is likely. And (x) Do not work alone: set up local teams with an infectious diseases specialist, clinical microbiologist, hospital pharmacist, Infection Control Practitioner or hospital epidemiologist, and comply with hospital antibiotic policies and guidelines. © 2016
Trish M Perl - One of the best experts on this subject based on the ideXlab platform.
-
development of a comprehensive surgical site Infection communication plan to enhance Infection Control interventions
American Journal of Infection Control, 2005Co-Authors: Margaret Pass, Ann Richards, Sara E Cosgrove, Trish M PerlAbstract:ISSUE: While implementing interventions to lower rates of surgical site Infection (SSI) for coronary artery bypass graft procedures (CABGs), we realized we needed a more timely way of delivering data to those that could impact these outcome measures. PROJECT: A strategy was designed to enhance communication with the cardiac surgery division. A four-pronged approach was developed to provide real time communication regarding SSIs: 1) Weekly surveillance reports to the cardiac surgery division chief (CSDC) stating the number of Infections noted for the week and which SSIs will be in their calculated CABG surveillance rates. (SSIs in all types of cardiac surgery procedures are reported to the CSDC, even those that will not be used for calculating rates for CABGs.) The dean and administrative leaders of the hospital also receive weekly Infection data. 2) Quarterly retrospective data of CABGs is presented along with 2 years of quarterly past data at the hospital epidemiology and Infection Control meeting. This information is also distributed to the nurse manager of cardiac surgery to post on the bulletin board outside the cardiac surgical rooms. Data is also sent to each surgeon and presented at performance improvement meetings. 3) Regular communication by e-mail and during regular weekly rounds between the Infection Control Practitioner (ICP) assigned to cardiac surgery and the nurse manager(s) of units that “support” cardiac surgery patients occurs regarding SSIs. 4) Weekly communication between the ICP and the nurse Practitioner(s) in the cardiac clinic following up all cardiac surgery patients occurs regarding SSIs. RESULTS: Over the past 18 months, CABG SSI rates have fallen from a high of 12.16 SSIs/100 CABGs in quarter 1 of 2003, to a low of 5.15 SSIs/100 CABGs in quarter 3 of 2004 (58% decrease). LESSONS LEARNED: Communication is a vital element of any Infection Control intervention. SSI rates will not change if the appropriate personnel are not provided timely data in a consistent way. This plan utilizes surveillance data in a retrospective way for trending and in a prospective way for more timely intervention. The interchange of information between surgical division chief and ICP is encouraged and produces an ongoing relationship that contributes to the success of the intervention.