Can Cluster-Boosted Regression Improve Prediction of Death and Length of Stay in the ICU?

Rouzbahman, Mahsa; Jovičić, Aleksandra; Chignell, Mark

doi:10.1109/jbhi.2016.2525731

Cited by 29 publications

(18 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…A list of potential predictors, and how the predictors were coded can be found in Appendix 1. For continuous variables (WBC count, hemoglobin, and age), we used k means clustering to choose cut-points (see Appendix 1) in order to improve accuracy and decrease over-fitting [24,25]. The outcomes were coded as binary outcomes.…”

Section: Methodsmentioning

confidence: 99%

Derivation of decision rules to predict clinically important outcomes in acute flank pain patients

Wang

Rodriguez

Fahimi

et al. 2017

The American Journal of Emergency Medicine

View full text Add to dashboard Cite

Objective Routine CT for patients with acute flank pain has not been shown to improve patient outcomes, and it may unnecessarily expose patients to radiation and increased costs. As preliminary steps toward the development of a guideline for selective CT, we sought to determine the prevalence of clinically important outcomes in patients with acute flank pain and derive preliminary decision rules. Methods We analyzed data from a randomized trial of CT vs. ultrasonography for patients with acute flank pain from 15 EDs between October 2011 and February 2013. Clinically important outcomes were defined as inpatient admission for ureteral stones and alternative diagnoses. Clinically important stones were defined as stones requiring urologic intervention. We sought to derive highly sensitive decision rules for both outcomes. Results Of 2759 participants, 236 (8.6%) had a clinically important outcome and 143 (5.2%) had a clinically important stone. A CDR including anemia (hemoglobin <13.2 g/dl), WBC count >11 000/μl, age > 42 years, and the absence of CVAT had a sensitivity of 97.9% (95% CI 94.8–99.2%) and specificity of 18.7% (95% 17.2–20.2%) for clinically important outcome. A CDR including hydronephrosis, prior history of stone, and WBC count <8300/μl had a sensitivity of 98.6% (95% CI 94.5–99.7%) and specificity of 26.0% (95% 24.2–27.7%) for clinically important stone. Conclusions We determined the prevalence of clinically important outcomes in patients with acute flank pain, and derived preliminary high sensitivity CDRs that predict them. Validation of CDRs with similar test characteristics would require prospective enrollment of 2100 patients.

show abstract

Section: Methodsmentioning

confidence: 99%

Derivation of decision rules to predict clinically important outcomes in acute flank pain patients

Wang

Rodriguez

Fahimi

et al. 2017

The American Journal of Emergency Medicine

View full text Add to dashboard Cite

show abstract

“…After each patient in the test set was assigned to a cluster, the associated regression functions obtained from the train set were used to predict the death status for a particular patient. We conducted the experiments with four different datasets and compared our results with the baseline results from previous work [4], which only considered numerical features. The first configuration was the existing data combined with the selected data of DESCRIPTION.…”

Section: F Evaluation Setting and Statistical Analysismentioning

confidence: 99%

“…Therefore, it is feasible to share summarized personal health information across different health data repositories of many organizations without privacy violations. Rouzbahman et al [4] performed cluster analysis to partition patients into similar groups. They then predicted patient mortality and length of stay in the ICU using regression prediction for each cluster.…”

Section: Introductionmentioning

confidence: 99%

“…In addition, textual data cannot be directly used for statistical analyses, unless converted into numerical form. Although domain experts can evaluate and interpret textual data, the reliability of the data may be questionable [4].…”

Section: Introductionmentioning

confidence: 99%

“…Data presentations of similar patients is a promising complementary method, which can support clinical decision making with the current patient [13], and can be achieved without additional effort from domain experts. Rouzbahman et al [4] performed clustering and predicted patient mortality and length of stay in the ICU using regression analysis for each cluster. Grouping similar patients into cohorts and using temporal projection of that cohorts, Ebadollahi et al [14] can predict the trajectory planning of a patient's physiological data.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Enhancing Predictive Power of Cluster-Boosted Regression With Text-Based Indexing

et al. 2019

Self Cite

View full text Add to dashboard Cite

Clustering prior to regression analysis improves the accuracy of prediction in clinical decision making. However, most previously described methods focused on numerical data only. This paper investigated how well textual features can improve the accuracy of regression predictions. Preliminary diagnosis, diagnosis summary, and drug names used in prescriptions as provided in the MIMIC II dataset were used to derive textual features. We proposed the bag-of-entities indexing method, which relies on named entity recognition, a machine learning technique used for locating and identifying words into predefined classes. The proposed technique captured meaningful phrases from texts in health records and represented them in numerical vector format. Dimensionality of the data space was reduced using principal component analysis. The additional well-tuned textual features were then combined with existing numerical features in using cluster-boosted regression to predict patient mortality in ICU. The experimental results showed prediction improvement obtained from textual features over the use of numerical features only. We found that using the proposed indexing method outperformed traditional word-vector representation approaches (bag-of-words and bag-of-bigrams) as well as a state-of-the-art approach (Doc2vec) in terms of resulting accuracy in predicting death status. Moreover, instead of directly interpreting, the identifiable individual features were grouped into types and summarized. The summarized de-identified data of textual features handled by the proposed framework can support predictive classification while also reducing privacy concerns. Grouping of similar patients based on their electronic health records also benefits physicians through the improved differential diagnosis and effective treatment planning. INDEX TERMS Bag-of-entities, clustering, electronic health record, named entity recognition, regression analysis, summarized data.

show abstract

A Survival Certification Model Based on Active Learning over Medical Insurance Data

2019

View full text Add to dashboard Cite

Can Cluster-Boosted Regression Improve Prediction of Death and Length of Stay in the ICU?

Cited by 29 publications

References 17 publications

Derivation of decision rules to predict clinically important outcomes in acute flank pain patients

Derivation of decision rules to predict clinically important outcomes in acute flank pain patients

Enhancing Predictive Power of Cluster-Boosted Regression With Text-Based Indexing

A Survival Certification Model Based on Active Learning over Medical Insurance Data

Contact Info

Product

Resources

About