EUVIMEDEUVIMED entdecken
CrossrefInfokarte

Predictive Modeling of Surgical Site Infections Using Sparse Laboratory Data

Prabhu RV Shankar, Anupama Kesari, Priya Shalini, N. Kamalashree, Charan Bharadwaj, Nitika Raj, Sowrabha Srinivas, Manu Shivakumar, Anand Raj Ulle, Nagabhushana N. Tagadur

International Journal of Big Data and Analytics in Healthcare · 2018

DOI: 10.4018/ijbdah.2018010102 ↗
AbstractBibliografische Angaben

Abstract

As part of a data mining competition, a training and test set of laboratory test data about patients with and without surgical site infection (SSI) were provided. The task was to develop predictive models with training set and identify patients with SSI in the no label test set. Lab test results are vital resources that guide healthcare providers make decisions about all aspects of surgical patient management. Many machine learning models were developed after pre-processing and imputing the lab tests data and only the top performing methods are discussed. Overall, RANDOM FOREST algorithms performed better than Support Vector Machine and Logistic Regression. Using a set of 74 lab tests, with RF, there were only 4 false positives in the training set and predicted 35 out of 50 SSI patients in the test set (Accuracy 0.86, Sensitivity 0.68, and Specificity 0.91). Optimal ways to address healthcare data quality concerns and imputation methods as well as newer generalizable algorithms need to be explored further to decipher new associations and knowledge among laboratory biomarkers and SSI.

Wissen, das zu Ihrer Arbeit passt.

Vom kostenlosen Einstieg über die persönliche Recherche bis zur vollständigen Research Suite.

Tarife ansehen →

Sind alle Verlagsvolltexte enthalten?

Nein. Bibliografische Nachweise, Abstracts und Volltexte sind unterschiedliche Inhalte. Ob ein Volltext verfügbar ist, hängt von der Quelle, Open Access und Ihren vorhandenen Lizenzen ab. Ein EUVIMED-Tarif ersetzt keine zusätzlichen Verlagsrechte.

Tarife ansehen →