Table 1. Recent Applications of Machine Learning in Digital Healthcare Epidemiology Referencea
No. of Data Type Participants Study Aim Savin et al., 201830 Surveillance data 2,286 individuals
Allen et al, 201638 Surveillance data and social media (Twitter)
Beeler et al, 201828 EMR
Escobar et al, 201722
Ehrentraut et al, 201839
EMR EMR Kuo et al, 201840 EMR and surveillance data Oh et al, 201820
Parreco et al, 201829
Sanger et al, 201642
Gómez-Vallejo et al, 201643
Lu et al, 201844
Santillana et al, 201545
EMR EMR
Identify risk factors for healthcare–associated ventriculitis and meningitis
Not reported Influenza surveillance 70,218 individuals 11,251 individuals
Prediction of Clostridioides difficile recurrence
120 individuals Prediction of healthcare- associated infections
1,836 individuals 256,732 admissions 22,201
individuals with a central line
Prediction of daily risk for C. difficile infection
Prediction of CLABSI and mortality
Surveillance data 851 individuals Prediction of SSI EMR
Multiple data sourcesb
Multiple data sourcesc
5,385 cases
Detection and classification of healthcare-associated infections
Not reported Influenza surveillance Prediction of SSI Prediction of CLABSI
Main Analytical Methods
– LASSO – Random forest – XG boost
– SVM
– Random forest – Logistic regression
– Random forest – Logistic regression
– SVM – Gradient tree boosting
– ANN – Logistic regression
– L2 regularized logistic regression
– Deep learning – Logistic regression – Gradient tree boosting
– Naïve Bayes classifier
– Logistic regression
– Naïve Bayes classifier
– PART algorithm – LASSO
Not reported Real-time influenza surveillance – SVM – Decision tree regression
– LASSO Sohn et al, 201746 Surveillance data 751 individuals Detection of SSI Pak et al, 201747 EMR
171,938 visits Estimating costs and changes in length of hospital stay for C. difficile infection
– NLP – Bayesian network – Logistic regression
– Logistic regression with elastic net regularization
S/P O
Combination of NLP and Bayesian network provided the best accuracy to detect SSI, and Bayesian network was more accurate than ridge estimator logistic regression.
Machine learning was used for propensity score development. Note. ANN, artificial neural network; CLABSI, central line-associated bloodstream infection; EMR, electronic medical records; I, identification of risk factors; LASSO, least absolute shrinkage and selection operator; NLP, natural language processing; P,
prediction; O, other; S, surveillance; SSI, surgical site infection; SVM, support vector machine; XG boost, extreme gradient booster aArticles published between October 2015 and June 2018 were included based on a Medline search and a bibliographic screening of the selected articles. bCombines data from different sources (ie, Boston Public Health Commission, Google Trends, Twitter, FluNearYou and electronic medical records). cCombines data from different sources (ie, Centers for Disease Control and Prevention, electronic medical records, Google Trends, Twitter, FluNearYou, and Google Flu Trends).
S/P S/P S/P
Machine learning successfully classified healthcare-associated infection: an automated rule was more accurate than a predefined definition.
Combining information from multiple models resulted in the best predictive performance.
Decision-tree regression resulted in the most robust and accurate predictions.
Application Conclusion/Lessons Learned I
P P P/(I) P P P/I P
Tree-based machine learning algorithms performed better than multivariable logistic regression and allowed detection of non-linear time-dependent variables.
SVM accurately predicted influenza-like illnesses at a local level.
Random forest had higher accuracy for prediction of CLABSI than logistic regression.
Both methods had poor performance for predicting C. difficile recurrence.
Gradient Tree Boosting performed best for predicting healthcare-associated infections. Simple preprocessing of data increased predictive accuracy.
ANN using post-operative data performed best for prediction of SSI.
Machine learning can accurately predict C. difficile infection, but data-driven predictions must be tailored locally.
Inconclusive41: Use of crude imbalanced data was not better than best guess for prediction of CLABSI. Other machine-learning algorithms were marginally more predictive than logistic regression.
P
Naïve Bayes classifier predicted SSI more accurately, with marginal gain, than did logistic regression.
1460
Jan A. Roth et al
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124
