A Pragmatic Machine Learning Model to Predict Carbapenem Resistance

Infection caused by carbapenem resistant (CR) organisms is a rising problem in the United States. While the risk factors for antibiotic resistance are well known, there remains a large need for the early identification of antibiotic resistant infections. Using machine learning (ML), we sought to develop a prediction model for carbapenem resistance. All patients >18 years of age admitted to a tertiary-care academic medical center between Jan 1, 2012 and Oct 10, 2017 with ≥1 bacterial culture were eligible for inclusion. All demographic, medication, vital sign, procedure, laboratory, and culture/sensitivity data was extracted from the electronic health record. Organisms were considered CR if a single isolate was reported as intermediate or resistant. CR and non-CR patients were temporally matched to maintain positive/negative case ratio. Extreme gradient boosting was used for model development. In total, 68,472 patients met inclusion criteria with 1,088 CR patients identified. Sixty-seven features were used for predictive modeling. The most important features were number of prior antibiotic days, recent central venous catheter placement, and inpatient surgery. After model training, the area under the receiver operating characteristic curve was 0.846. The sensitivity of the model was 30%, with a positive predictive value (PPV) of 30% and a negative predictive value of 99%. Using readily available clinical data, we were able to create a ML model capable of predicting CR infections at the time of culture collection with a high PPV.

Download Full-text

Patient-Specific Predictive Antibiogram in Decision Support for Empiric Antibiotic Treatment

Infection Control and Hospital Epidemiology ◽

10.1017/ice.2020.1205 ◽

2020 ◽

Vol 41 (S1) ◽

pp. s521-s522

Author(s):

Debarka Sengupta ◽

Vaibhav Singh ◽

Seema Singh ◽

Dinesh Tewari ◽

Mudit Kapoor ◽

...

Keyword(s):

Machine Learning ◽

Antimicrobial Resistance ◽

Model Building ◽

Medical Center ◽

Bacterial Species ◽

Model Performance ◽

The United States ◽

Patient Specific ◽

Gradient Boosting ◽

Comparative Performance

Background: The rising trend of antibiotic resistance imposes a heavy burden on healthcare both clinically and economically (US$55 billion), with 23,000 estimated annual deaths in the United States as well as increased length of stay and morbidity. Machine-learning–based methods have, of late, been used for leveraging patient’s clinical history and demographic information to predict antimicrobial resistance. We developed a machine-learning model ensemble that maximizes the accuracy of such a drug-sensitivity versus resistivity classification system compared to the existing best-practice methods. Methods: We first performed a comprehensive analysis of the association between infecting bacterial species and patient factors, including patient demographics, comorbidities, and certain healthcare-specific features. We leveraged the predictable nature of these complex associations to infer patient-specific antibiotic sensitivities. Various base-learners, including k-NN (k-nearest neighbors) and gradient boosting machine (GBM), were used to train an ensemble model for confident prediction of antimicrobial susceptibilities. Base learner selection and model performance evaluation was performed carefully using a variety of standard metrics, namely accuracy, precision, recall, F1 score, and Cohen κ. Results: For validating the performance on MIMIC-III database harboring deidentified clinical data of 53,423 distinct patient admissions between 2001 and 2012, in the intensive care units (ICUs) of the Beth Israel Deaconess Medical Center in Boston, Massachusetts. From ~11,000 positive cultures, we used 4 major specimen types namely urine, sputum, blood, and pus swab for evaluation of the model performance. Figure 1 shows the receiver operating characteristic (ROC) curves obtained for bloodstream infection cases upon model building and prediction on 70:30 split of the data. We received area under the curve (AUC) values of 0.88, 0.92, 0.92, and 0.94 for urine, sputum, blood, and pus swab samples, respectively. Figure 2 shows the comparative performance of our proposed method as well as some off-the-shelf classification algorithms. Conclusions: Highly accurate, patient-specific predictive antibiogram (PSPA) data can aid clinicians significantly in antibiotic recommendation in ICU, thereby accelerating patient recovery and curbing antimicrobial resistance.Funding: This study was supported by Circle of Life Healthcare Pvt. Ltd.Disclosures: None

Download Full-text

Communications and Screening for 2019 Novel Coronavirus at a Tertiary-Care Medical Center

Infection Control and Hospital Epidemiology ◽

10.1017/ice.2020.578 ◽

2020 ◽

Vol 41 (S1) ◽

pp. s84-s84

Author(s):

Lorinda Sheeler ◽

Mary Kukla ◽

Oluchi Abosi ◽

Holly Meacham ◽

Stephanie Holley ◽

...

Keyword(s):

Screening Tool ◽

Medical Center ◽

Academic Medical Center ◽

Tertiary Care ◽

The United States ◽

World Health ◽

Communication Campaign ◽

Electronic Screening ◽

Novel Coronavirus ◽

The Impact

Background: In December of 2019, the World Health Organization reported a novel coronavirus (severe acute respiratory coronavirus virus 2 [SARS-CoV-2)]) causing severe respiratory illness originating in Wuhan, China. Since then, an increasing number of cases and the confirmation of human-to-human transmission has led to the need to develop a communication campaign at our institution. We describe the impact of the communication campaign on the number of calls received and describe patterns of calls during the early stages of our response to this emerging infection. Methods: The University of Iowa Hospitals & Clinics is an 811-bed academic medical center with >200 outpatient clinics. In response to the coronavirus disease 2019 (COVID-19) outbreak, we launched a communications campaign on January 17, 2020. Initial communications included email updates to staff and a dedicated COVID-19 webpage with up-to-date information. Subsequently, we developed an electronic screening tool to guide a risk assessment during patient check in. The screening tool identifies travel to China in the past 14 days and the presence of symptoms defined as fever >37.7°C plus cough or difficulty breathing. The screening tool was activated on January 24, 2020. In addition, university staff contacted each student whose primary residence record included Hubei Province, China. Students were provided with medical contact information, signs and symptoms to monitor for, and a thermometer. Results: During the first 5 days of the campaign, 3 calls were related to COVID-19. The number of calls increased to 18 in the 5 days following the implementation of the electronic screening tool. Of the 21 calls received to date, 8 calls (38%) were generated due to the electronic travel screen, 4 calls (19%) were due to a positive coronavirus result in a multiplex respiratory panel, 4 calls (19%) were related to provider assessment only (without an electronic screening trigger), and 2 calls (10%) sought additional information following the viewing of the web-based communication campaign. Moreover, 3 calls (14%) were for people without travel history but with respiratory symptoms and contact with a person with recent travel to China. Among those reporting symptoms after travel to China, mean time since arrival to the United States was 2.7 days (range, 0–11 days). Conclusion: The COVID-19 outbreak is evolving, and providing up to date information is challenging. Implementing an electronic screening tool helped providers assess patients and direct questions to infection prevention professionals. Analyzing the types of calls received helped tailor messaging to frontline staff.Funding: NoneDisclosures: None

Download Full-text

An innovative resident-driven mortality case review curriculum to teach and drive system-based practice improvements in the United States

Journal of Educational Evaluation for Health Professions ◽

10.3352/jeehp.2018.15.31 ◽

2018 ◽

Vol 15 ◽

pp. 31

Author(s):

Nila S. Radhakrishnan ◽

Margaret C. Lo ◽

Rohit Bishnoi ◽

Subhankar Samal ◽

Robert Leverence ◽

...

Keyword(s):

Quality Improvement ◽

Systems Of Care ◽

Medical Center ◽

Academic Medical Center ◽

Tertiary Care ◽

The United States ◽

Case Review ◽

Failure To Rescue ◽

Internal Medicine Residents ◽

Mortality Case

Purpose: Traditionally, the morbidity and mortality conference (M&MC) is a forum where possible medical errors are discussed. Although M&MCs can facilitate identification of opportunities for systemwide improvements, few studies have described their use for this purpose, particularly in residency training programs. This paper describes the use of M&MC case review as a quality improvement activity that teaches system-based practice and can engage residents in improving systems of care. Methods: Internal medicine residents at a tertiary care academic medical center reviewed 347 consecutive mortalities from March 2014 to September 2017. The residents used case review worksheets to categorize and track causes of mortality, and then debriefed with a faculty member. Selected cases were then presented at a larger interdepartmental meeting and action items were implemented. Descriptive statistics and thematic analysis were used to analyze the results. Results: The residents identified a possible diagnostic mismatch at some point from admission to death in 54.5% of cases (n= 189) and a possible need for improved management in 48.0% of cases. Three possible management failure themes were identified, including failure to plan, failure to communicate, and failure to rescue, which accounted for 21.9%, 10.7 %, and 10.1% of cases, respectively. Following these reviews, quality improvement initiatives proposed by residents led to system-based changes. Conclusion: A resident-driven mortality review curriculum can lead to improvements in systems of care. This novel type of curriculum can be used to teach system-based practice. The recruitment of teaching faculty with expertise in quality improvement and mortality case analyses is essential for such a project.

Download Full-text

An Interpretable Early Dynamic Sequential Predictor for Sepsis-Induced Coagulopathy Progression in the Real-World Using Machine Learning

Frontiers in Medicine ◽

10.3389/fmed.2021.775047 ◽

2021 ◽

Vol 8 ◽

Author(s):

Ruixia Cui ◽

Wenbo Hua ◽

Kai Qu ◽

Heran Yang ◽

Yingmu Tong ◽

...

Keyword(s):

Machine Learning ◽

Real World ◽

Time Series Data ◽

Time Window ◽

Medical Center ◽

Characteristic Curve ◽

Series Data ◽

Gradient Boosting ◽

Early Management ◽

Extreme Gradient Boosting

Sepsis-associated coagulation dysfunction greatly increases the mortality of sepsis. Irregular clinical time-series data remains a major challenge for AI medical applications. To early detect and manage sepsis-induced coagulopathy (SIC) and sepsis-associated disseminated intravascular coagulation (DIC), we developed an interpretable real-time sequential warning model toward real-world irregular data. Eight machine learning models including novel algorithms were devised to detect SIC and sepsis-associated DIC 8n (1 ≤ n ≤ 6) hours prior to its onset. Models were developed on Xi'an Jiaotong University Medical College (XJTUMC) and verified on Beth Israel Deaconess Medical Center (BIDMC). A total of 12,154 SIC and 7,878 International Society on Thrombosis and Haemostasis (ISTH) overt-DIC labels were annotated according to the SIC and ISTH overt-DIC scoring systems in train set. The area under the receiver operating characteristic curve (AUROC) were used as model evaluation metrics. The eXtreme Gradient Boosting (XGBoost) model can predict SIC and sepsis-associated DIC events up to 48 h earlier with an AUROC of 0.929 and 0.910, respectively, and even reached 0.973 and 0.955 at 8 h earlier, achieving the highest performance to date. The novel ODE-RNN model achieved continuous prediction at arbitrary time points, and with an AUROC of 0.962 and 0.936 for SIC and DIC predicted 8 h earlier, respectively. In conclusion, our model can predict the sepsis-associated SIC and DIC onset up to 48 h in advance, which helps maximize the time window for early management by physicians.

Download Full-text

Interactive iBook-Based Patient Education in a NeuroTrauma Clinic

Neurosurgery ◽

10.1093/neuros/nyx095 ◽

2017 ◽

Vol 81 (5) ◽

pp. 787-794 ◽

Cited By ~ 9

Author(s):

Ronald Sahyouni ◽

Amin Mahmoodi ◽

Amir Mahmoodi ◽

Ramin R Rajaii ◽

Bima J Hasjim ◽

...

Keyword(s):

Medical Center ◽

Academic Medical Center ◽

Tertiary Care ◽

Educational Interventions ◽

The United States ◽

Outpatient Setting ◽

Educational Material ◽

Patient Knowledge ◽

Academic Medical ◽

Baseline Knowledge

Abstract BACKGROUND Traumatic brain injury (TBI) is a leading cause of death and disability in the United States. Educational interventions may alleviate the burden of TBI for patients and their families. Interactive modalities that involve engagement with the educational material may enhance patient knowledge acquisition when compared to static text-based educational material. OBJECTIVE To determine the effects of educational interventions in the outpatient setting on self-reported patient knowledge, with a focus on iPad-based (Apple, Cupertino, California) interactive modules. METHODS Patients and family members presenting to a NeuroTrauma clinic at a tertiary care academic medical center completed a presurvey assessing baseline knowledge of TBI or concussion, depending on the diagnosis. Subjects then received either an interactive iBook (Apple) on TBI or concussion, or an informative pamphlet with identical information in text format. Subjects then completed a postsurvey prior to seeing the neurosurgeon. RESULTS All subjects (n = 152) significantly improved on self-reported knowledge measures following administration of either an iBook (Apple) or pamphlet (P < .01, 95% confidence interval [CI]). Subjects receiving the iBook (n = 122) performed significantly better on the postsurvey (P < .01, 95% CI), despite equivalent presurvey scores, when compared to those receiving pamphlets (n = 30). Lastly, patients preferred the iBook to pamphlets (P < .01, 95% CI). CONCLUSION Educational interventions in the outpatient NeuroTrauma setting led to significant improvement in self-reported measures of patient and family knowledge. This improved understanding may increase compliance with the neurosurgeon's recommendations and may help reduce the potential anxiety and complications that arise following a TBI.

Download Full-text

CT Radiomics and Machine-Learning Models for Predicting Tumor-Stroma Ratio in Patients With Pancreatic Ductal Adenocarcinoma

Frontiers in Oncology ◽

10.3389/fonc.2021.707288 ◽

2021 ◽

Vol 11 ◽

Author(s):

Yinghao Meng ◽

Hao Zhang ◽

Qi Li ◽

Fang Liu ◽

Xu Fang ◽

...

Keyword(s):

Machine Learning ◽

Pancreatic Ductal Adenocarcinoma ◽

Predictive Value ◽

Tumor Stroma ◽

Ductal Adenocarcinoma ◽

Gradient Boosting ◽

Rank Test ◽

Training Set ◽

Extreme Gradient Boosting ◽

Validation Set

PurposeTo develop and validate a machine learning classifier based on multidetector computed tomography (MDCT), for the preoperative prediction of tumor–stroma ratio (TSR) expression in patients with pancreatic ductal adenocarcinoma (PDAC).Materials and MethodsIn this retrospective study, 227 patients with PDAC underwent an MDCT scan and surgical resection. We quantified the TSR by using hematoxylin and eosin staining and extracted 1409 arterial and portal venous phase radiomics features for each patient, respectively. Moreover, we used the least absolute shrinkage and selection operator logistic regression algorithm to reduce the features. The extreme gradient boosting (XGBoost) was developed using a training set consisting of 167 consecutive patients, admitted between December 2016 and December 2017. The model was validated in 60 consecutive patients, admitted between January 2018 and April 2018. We determined the XGBoost classifier performance based on its discriminative ability, calibration, and clinical utility.ResultsWe observed low and high TSR in 91 (40.09%) and 136 (59.91%) patients, respectively. A log-rank test revealed significantly longer survival for patients in the TSR-low group than those in the TSR-high group. The prediction model revealed good discrimination in the training (area under the curve [AUC]= 0.93) and moderate discrimination in the validation set (AUC= 0.63). While the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value for the training set were 94.06%, 81.82%, 0.89, 0.89, and 0.90, respectively, those for the validation set were 85.71%, 48.00%, 0.70, 0.70, and 0.71, respectively.ConclusionsThe CT radiomics-based XGBoost classifier provides a potentially valuable noninvasive tool to predict TSR in patients with PDAC and optimize risk stratification.

Download Full-text

A Machine Learning Approach to Identify Predictors of Potentially Inappropriate Non-Steroidal Anti-Inflammatory Drugs (NSAIDs) Use in Older Adults with Osteoarthritis

International Journal of Environmental Research and Public Health ◽

10.3390/ijerph18010155 ◽

2020 ◽

Vol 18 (1) ◽

pp. 155

Author(s):

Jayeshkumar Patel ◽

Amit Ladani ◽

Nethra Sambamoorthi ◽

Traci LeMasters ◽

Nilanjana Dwibedi ◽

...

Keyword(s):

Machine Learning ◽

Older Adults ◽

The United States ◽

Gradient Boosting ◽

Study Cohort ◽

Real World Data ◽

Extreme Gradient Boosting ◽

Inflammatory Drugs ◽

Anti Inflammatory ◽

Using Data

Evidence from some studies suggest that osteoarthritis (OA) patients are often prescribed non-steroidal anti-inflammatory drugs (NSAIDs) that are not in accordance with their cardiovascular (CV) or gastrointestinal (GI) risk profiles. However, no such study has been carried out in the United States. Therefore, we sought to examine the prevalence and predictors of potentially inappropriate NSAIDs use in older adults (age > 65) with OA using machine learning with real-world data from Optum De-identified Clinformatics® Data Mart. We identified a retrospective cohort of eligible individuals using data from 2015 (baseline) and 2016 (follow-up). Potentially inappropriate NSAIDs use was identified using the type (COX-2 selective vs. non-selective) and length of NSAIDs use and an individual’s CV and GI risk. Predictors of potentially inappropriate NSAIDs use were identified using eXtreme Gradient Boosting. Our study cohort comprised of 44,990 individuals (mean age 75.9 years). We found that 12.8% individuals had potentially inappropriate NSAIDs use, but the rate was disproportionately higher (44.5%) in individuals at low CV/high GI risk. Longer duration of NSAIDs use during baseline (AOR 1.02; 95% CI:1.02–1.02 for both non-selective and selective NSAIDs) was associated with a higher risk of potentially inappropriate NSAIDs use. Additionally, individuals with low CV/high GI (AOR 1.34; 95% CI:1.20–1.50) and high CV/low GI risk (AOR 1.61; 95% CI:1.34–1.93) were also more likely to have potentially inappropriate NSAIDs use. Heightened surveillance of older adults with OA requiring NSAIDs is warranted.

Download Full-text

Machine Learning Algorithms for Predicting Fatty Liver Disease

Annals of Nutrition and Metabolism ◽

10.1159/000513654 ◽

2021 ◽

pp. 1-8

Author(s):

Xieyi Pei ◽

Qingqing Deng ◽

Zhuo Liu ◽

Xiang Yan ◽

Weiping Sun

Keyword(s):

Machine Learning ◽

Liver Disease ◽

Fatty Liver ◽

Predictive Value ◽

Fatty Liver Disease ◽

Machine Learning Algorithms ◽

High Rate ◽

Gradient Boosting ◽

Linear Discriminant ◽

Extreme Gradient Boosting

Background: Fatty liver disease (FLD) has become a rampant condition. It is associated with a high rate of morbidity and mortality in a population. The condition is commonly referred as FLD. Early prediction of FLD would allow patients to take necessary preventive, diagnosis, and treatment. The main objective of this research is to develop a machine learning (ML) model to predict FLD that can help medics to classify individuals at high risk of FLD, make novel diagnosis, management, and prevention for FLD. Methods: Total of 3,419 subjects were recruited with 845 having been screened for FLD. Classification models were used in the detection of the disease. These models include logistic regression (LR), random forest (RF), artificial neural networks (ANNs), k-nearest neighbors (KNNs), extreme gradient boosting (XGBoost), and linear discriminant analysis (LDA). Predictive accuracy was assessed by area under curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value. Results: We demonstrated that ML models give more accurate predictions, the best accuracy reached to 0.9415 in the XGBoost model. Feature importance analysis not only confirmed some well-known FLD risk factors, but also demonstrated several novel features for predicting the risk of FLD, such as hemoglobin. Conclusion: By implementing the XGBoost model, physicians can efficiently identify FLD in general patients; this would help in prevention, early treatment, and management of FLD.

Download Full-text

Utility of Dual Skin Tests to Evaluate Tuberculin Skin Test Reactions of 10 to 14 mm in Healthcare Workers

Infection Control and Hospital Epidemiology ◽

10.1086/502143 ◽

2003 ◽

Vol 24 (11) ◽

pp. 821-824 ◽

Cited By ~ 7

Author(s):

Bryan J. Marsh ◽

Joshua San Vicente ◽

C. Fordham von Reyn

Keyword(s):

Occupational Health ◽

Skin Test ◽

Healthcare Workers ◽

Medical Center ◽

Skin Testing ◽

Academic Medical Center ◽

Tertiary Care ◽

Health Department ◽

The United States ◽

Skin Tests

AbstractObjective:To define the utility of 10- to 14-mm reactions to a Mycobacterium tuberculosis purified protein derivative (PPD) skin test for healthcare workers (HCWs).Design:Blinded dual skin testing, using PPD and M. avium sensitin, of HCWs at a single medical center who had a 10-to 14-mm reaction to PPD when tested by personnel from the Occupational Health Department as part of routine annual screening.Setting:A single tertiary-care academic medical center.Participants:Employees of the medical center who underwent routine annual PPD screening and were identified by the Occupational Health Department as having a reaction of 10 to 14 mm to PPD.Results:Nineteen employees were identified as candidates and 11 underwent dual skin testing. Only 4 (36%) had repeat results for PPD in the 10- to 14-mm range, whether read by Occupational Health Department personnel or study investigators. For only 5 (45%) of the subjects did the Occupational Health Department personnel and study investigators concur (± 3 mm) on the size of the PPD reaction. Two of the 4 subjects with reactions of 10 to 14 mm as measured by the study investigators were M. avium sensitin dominant, 1 was PPD dominant, and 1 was nondominant.Conclusion:A reaction of 10 to 14 mm to PPD should not be used as an indication for the treatment of latent tuberculosis (TB) infection in healthy HCWs born in the United States with no known exposure to TB.

Download Full-text

Machine learning identifies girls with central precocious puberty based on multisource data

JAMIA Open ◽

10.1093/jamiaopen/ooaa063 ◽

2020 ◽

Author(s):

Liyan Pan ◽

Guangjian Liu ◽

Xiaojian Mao ◽

Huiying Liang

Keyword(s):

Machine Learning ◽

Precocious Puberty ◽

Laboratory Tests ◽

Medical Center ◽

Central Precocious Puberty ◽

Stimulation Test ◽

Youden Index ◽

Gradient Boosting ◽

Single Source ◽

Extreme Gradient Boosting

Abstract Objective The study aimed to develop simplified diagnostic models for identifying girls with central precocious puberty (CPP), without the expensive and cumbersome gonadotropin-releasing hormone (GnRH) stimulation test, which is the gold standard for CPP diagnosis. Materials and methods Female patients who had secondary sexual characteristics before 8 years old and had taken a GnRH analog (GnRHa) stimulation test at a medical center in Guangzhou, China were enrolled. Data from clinical visiting, laboratory tests, and medical image examinations were collected. We first extracted features from unstructured data such as clinical reports and medical images. Then, models based on each single-source data or multisource data were developed with Extreme Gradient Boosting (XGBoost) classifier to classify patients as CPP or non-CPP. Results The best performance achieved an area under the curve (AUC) of 0.88 and Youden index of 0.64 in the model based on multisource data. The performance of single-source models based on data from basal laboratory tests and the feature importance of each variable showed that the basal hormone test had the highest diagnostic value for a CPP diagnosis. Conclusion We developed three simplified models that use easily accessed clinical data before the GnRH stimulation test to identify girls who are at high risk of CPP. These models are tailored to the needs of patients in different clinical settings. Machine learning technologies and multisource data fusion can help to make a better diagnosis than traditional methods.

Download Full-text