| Autor: |
Lopansri, Bert, Miller, Russell, Burke, John, Levy, Mitchell, Opal, Steven, Rothman, Richard, D’Alessio, Franco, Venkataramana Sidhaye, Balk, Robert, Greenberg, Jared, Yoder, Mark, Gourang Patel, Gilbert, Emily, Afshar, Majid, Parada, Jorge, Martin, Greg, Esper, Annette, Kempker, Jordan, Mangala Narasimhan, Adey Tsegaye, Hahn, Stella, Mayo, Paul, McHugh, Leo, Rapisarda, Antony, Sampson, Dayle, Brandon, Roslyn, Seldon, Therese, Yager, Thomas, Brandon, Richard |
| Rok vydání: |
2019 |
| DOI: |
10.6084/m9.figshare.7750355.v1 |
| Popis: |
Analysis of Treated vs. Untreated SIRS Patients. Figure S7–1. Behavior of logistic regression models in ROC curve analysis. (A) Five variable model from Table S7–2, giving AUC = 0.72 (95% CI 0.63–0.81). (B) Four variable model from Table S7–3, giving AUC = 0.71 (95% CI 0.62–0.80). Figure S7–2. Machine learning attempts to identify combinations of clinical variables and demographic variables that discriminate between antibiotic treatment and no treatment, within the SIRS group. Recursive feature elimination was used, within a logistic regression (LR) or Random Forests (RF) model. Figure S7–3. Gini ranking of individual words in the “physician comments” field of the case report form, for SIRS patients. The ranking is based on contribution toward discriminating antibiotic-treated vs. untreated SIRS patients. Abbreviations: cxr, chest x-ray; dka, diabetic ketoacidosis; mri, magnetic resonance imaging. Figure S7–4. Gini ranking of word-pairs in the “physician comments” field of the case report form, for SIRS patients. The ranking is based on contribution toward discriminating antibiotic-treated vs. untreated SIRS patients. Abbreviation: chf, congestive heart failure. Figure S7–5. Machine learning attempt to identify combinations of clinical variables, demographic variables, words, and word-pairs that discriminate between antibiotic treatment and no treatment, within the SIRS group. Recursive feature elimination was used, within a logistic regression (LR) or Random Forests (RF) model. Table S7–1. Test for ability of clinical and demographic parameters to distinguish between SIRS patients who received (AB+) or did not receive (AB−) therapeutic antibiotics. Diagnosis was by consensus RPD. Parameters are listed in order of decreasing significance (2-tailed p-value) as evaluated either by Student’s t test, assuming equal variances in the two groups (for continuous variables), or by a 2-proportions z-test ( www.vassarstats.net ) for categorical variables. Table S7–2. Use of logistic regression, to discriminate between SIRS patients who were treated vs. not treated with antibiotics. Five variable model. Table S7–3. Use of logistic regression, to discriminate between SIRS patients who were treated vs. not treated with antibiotics. Four variable model. (PDF 550 kb) |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
