Evaluation of Machine-Learning Algorithms for Predicting Opioid Overdose Risk Among Medicare Beneficiaries With Opioid Prescriptions

Lo-Ciganic, Wei Hsuan; Huang, James; Zhang, Hao Helen; Weiss, Jeremy C.; Wu, Yonghui; Kwoh, C. Kent; Donohue, Julie M.; Cochran, Gerald; Gordon, Adam J.; Malone, Daniel C.; Kuza, Courtney C.; Gellad, Walid F.

doi:10.1001/jamanetworkopen.2019.0968

Cited by 168 publications

(212 citation statements)

References 61 publications

Supporting

Mentioning

191

Contrasting

Order By: Relevance

“…These ML algorithms have demonstrated their successful applications (either classification or prediction) in clinical practice. 17–19,27,28,35,36 The details of these ML algorithms and model training processes are attached in the eMethods in the Supplement.…”

Section: Methodsmentioning

confidence: 99%

“…Machine learning (ML) is a good analytical method for solving classification problems through identification of implicit data patterns from complex data. 26 The ML method outperforms traditional statistical methods because of its excellent ability to handle complex interactions between large amount of predictors and good performance in non-linear classification problems 27 ML has been successfully applied in several clinical fields. 27–36 Thus, the ML algorithm is especially appropriate for analyzing complex data such as MALDI-TOF spectra.…”

Section: Introductionmentioning

confidence: 99%

“…26 The ML method outperforms traditional statistical methods because of its excellent ability to handle complex interactions between large amount of predictors and good performance in non-linear classification problems 27 ML has been successfully applied in several clinical fields. 27–36 Thus, the ML algorithm is especially appropriate for analyzing complex data such as MALDI-TOF spectra. However, to our knowledge, few studies have used ML in the analysis of MALDI-TOF spectra for rapidly reporting VRE fm , and the case numbers in these studies were insufficient, and so, ML algorithm generalization has been limited.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Rapidly predicting vancomycin resistance ofEnterococcus faeciumthrough MALDI-TOF MS spectrum obtained in real-world clinical microbiology laboratory

Wang

Chung

et al. 2020

Preprint

View full text Add to dashboard Cite

50Enterococcus faecium is one of the leading pathogens in the world. In this study, we proposed 51 a strategy to rapidly and accurately distinguish vancomycin-resistant Enterococcus faecium 52 (VREfm) and vancomycin-susceptible E. faecium (VSEfm) to help doctors correctly determine 53 the use of vancomycin by a machine learning (ML)-based algorithm. A predictive model was 54 developed and validated to distinguish VREfm and VSEfm by analyzing MALDI-TOF MS 55 spectra of unique E. faecium isolates from different specimen types. Firstly, 5717 mass spectra, 56 including 2795 VREfm and 2922 VSEfm, were used to develop the algorithm. And 2280 mass 57 spectra of isolates, namely 1222 VREfm and 1058 VSEfm, were used to externally validate the 58 algorithm. The random forest-based algorithm demonstrated good classification performances 59 for overall specimens, whose mean AUROC in 5-fold cross validation, time-wise validation, 60 and external validation was all greater than 0.84. For the detection of VREfm in blood, sterile 61 body fluid, urinary tract, and wound, the AUROC in external validation was also greater than 62 0.84. The predictions with algorithms were significantly more accurate than empirical 63 antibiotic use. The accuracy of antibiotics administration could be improved by 30%. And the 64 algorithm could provide rapid antibiotic susceptibility results at least 24 hours ahead of routine 65 laboratory tests. The turn-around-time of antibiotic susceptibility could be reduced by 50%. In 66 conclusion, a ML algorithm using MALDI-TOF MS spectra obtained in routine workflow 67 accurately differentiated VREfm from VSEfm, especially in blood and sterile body fluid, which 68 can be applied to facilitate the clinical testing process due to its accuracy, generalizability, and 69 rapidness. 70 71 3 Introduction 72 Enterococcus spp. is one of the leading pathogens in healthcare-associated infection. 1 73 Enterococcal infection could cause urinary tract infection, blood stream infection, and even 74 mortality. 2 Until recently, vancomycin was virtually the only drug that could be consistently 75 relied on for treating multidrug-resistant enterococcal infections 3,4 . Vancomycin-resistant 76 Enterococcus (VRE) has led to heavy burden on healthcare worldwide since its first-time 77 isolation. 5,6 Enterococcus faecalis and E. faecium are the 2 most commonly isolated 78 Enterococcus spp. in clinical practice. 1 VRE faecium (VREfm) has received considerably more 79 attention than VRE faecalis (VREfs) because most of the clinically isolated VRE is E. faecium 80 in the recent decades 4,7 and VREfm causes more severe infection than VREfs 8,9 . Early detection 81 of vancomycin resistance is essential for successfully treating VREfm infection. 10 Vancomycin 82 could be discontinued, and antimicrobial agents could be replaced with other antibiotics (eg, 83 linezolid and daptomycin) based on the laboratory results of vancomycin resistance 11,12 . 84 Patients' prognosis could be improved and further drug resistance development ...

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rapidly predicting vancomycin resistance ofEnterococcus faeciumthrough MALDI-TOF MS spectrum obtained in real-world clinical microbiology laboratory

Wang

Chung

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…For low-prevalence events, high specificity can mask very low PPV. 19 Therefore, we began our analyses by characterizing the prevalence of our prediction targets. Figure 1 shows each participant's prevalence for each of the three dependent variables: heroin craving, cocaine craving, and stress.…”

Section: Introductionmentioning

confidence: 99%

Prediction of stress and drug craving ninety minutes in the future with passively collected GPS data

et al. 2020

View full text Add to dashboard Cite

Just-in-time adaptive interventions (JITAIs), typically smartphone apps, learn to deliver therapeutic content when users need it. The challenge is to "push" content at algorithmically chosen moments without making users trigger it with effortful input. We trained a randomForest algorithm to predict heroin craving, cocaine craving, or stress (reported via smartphone app 3x/day) 90 min into the future, using 16 weeks of field data from 189 outpatients being treated for opioid-use disorder. We used only one form of continuous input (along with person-level demographic data), collected passively: an indicator of environmental exposures along the past 5 h of movement, as assessed by GPS. Our models achieved excellent overall accuracy-as high as 0.93 by the end of 16 weeks of tailoring-but this was driven mostly by correct predictions of absence. For predictions of presence, "believability" (positive predictive value, PPV) usually peaked in the high 0.70s toward the end of the 16 weeks. When the prediction target was more rare, PPV was lower. Our findings complement those of other investigators who use machine learning with more broadly based "digital phenotyping" inputs to predict or detect mental and behavioral events. When target events are comparatively subtle, like stress or drug craving, accurate detection or prediction probably needs effortful input from users, not passive monitoring alone. We discuss ways in which accuracy is difficult to achieve or even assess, and warn that high overall accuracy (including high specificity) can mask the abundance of false alarms that low PPV reveals.npj Digital Medicine (2020) 3:26 ; https://doi.

show abstract

“…Furthermore, such measures of risk were often derived from studies that used traditional statistical methods to identify risk factors. Traditional statistical approaches are limited in their capacity to combine potentially predictive features [13]. The effective combination of ASD-prevalence in the family, prescription drugs, and socioeconomic variables may yield greater predictive power than one factor alone.…”

Section: Introductionmentioning

confidence: 99%

Identification of newborns at risk for autism using electronic medical records and machine learning

et al. 2020

View full text Add to dashboard Cite

Background. Current approaches for early identification of individuals at high risk for autism spectrum disorder (ASD) in the general population are limited, and most ASD patients are not identified until after the age of 4. This is despite substantial evidence suggesting that early diagnosis and intervention improves developmental course and outcome. The aim of the current study was to test the ability of machine learning (ML) models applied to electronic medical records (EMRs) to predict ASD early in life, in a general population sample. Methods. We used EMR data from a single Israeli Health Maintenance Organization, including EMR information for parents of 1,397 ASD children (ICD-9/10) and 94,741 non-ASD children born between January 1st, 1997 and December 31st, 2008. Routinely available parental sociodemographic information, parental medical histories, and prescribed medications data were used to generate features to train various ML algorithms, including multivariate logistic regression, artificial neural networks, and random forest. Prediction performance was evaluated with 10-fold cross-validation by computing the area under the receiver operating characteristic curve (AUC; C-statistic), sensitivity, specificity, accuracy, false positive rate, and precision (positive predictive value [PPV]). Results. All ML models tested had similar performance. The average performance across all models had C-statistic of 0.709, sensitivity of 29.93%, specificity of 98.18%, accuracy of 95.62%, false positive rate of 1.81%, and PPV of 43.35% for predicting ASD in this dataset. Conclusions. We conclude that ML algorithms combined with EMR capture early life ASD risk as well as reveal previously unknown features to be associated with ASD-risk. Such approaches may be able to enhance the ability for accurate and efficient early detection of ASD in large populations of children.

show abstract

Evaluation of Machine-Learning Algorithms for Predicting Opioid Overdose Risk Among Medicare Beneficiaries With Opioid Prescriptions

Cited by 168 publications

References 61 publications

Rapidly predicting vancomycin resistance ofEnterococcus faeciumthrough MALDI-TOF MS spectrum obtained in real-world clinical microbiology laboratory

Rapidly predicting vancomycin resistance ofEnterococcus faeciumthrough MALDI-TOF MS spectrum obtained in real-world clinical microbiology laboratory

Prediction of stress and drug craving ninety minutes in the future with passively collected GPS data

Identification of newborns at risk for autism using electronic medical records and machine learning

Contact Info

Product

Resources

About