Predicting in-hospital mortality in pneumonia-associated septic shock patients using Classification and Regression Tree (CART) methodology..
CCCF ePoster library. Karvellas C. Oct 4, 2017; 198148; 90
Dr. Constantine Karvellas
Dr. Constantine Karvellas
Login now to access Regular content available to all registered users.

You may also access this content "anytime, anywhere" with the Free MULTILEARNING App for iOS and Android
Abstract
Rate & Comment (0)
#90



Predicting in-hospital mortality in pneumonia-associated septic shock patients using Classification and Regression Tree (CART) methodology..

Karvellas, Constantine1,2; Speiser, Jaime Lynn3; Sligl, Wendy1,4; Kumar, Anand5,6.

1. Department of Critical Care Medicine, University of Alberta, Edmonton, Canada.

2. Division of Gastroenterology, University of Alberta, Edmonton, Canada.

3. Department of Biostatistics, Wake Forest University, Winston Salem, United States.

4. Division of Infectious Diseases, University of Alberta, Edmonton, Canada.

5. Department of Critical Care, University of Manitoba, Winnipeg, Canada.

6 Division of Infectious Diseases, University of Manitoba, Winnipeg, Canada.


Background/Aim: Bacterial pneumonia and septic shock are associated with substantial morbidity and mortality. Classification and Regression Tree (CART) methodology allows the development of predictive models using binary splits and offers an intuitive method for predicting outcome using processes familiar to clinicians. We aimed to improve determinations of prognosis at the time of admission for pneumonia and septic shock using CART model analysis.
Methods: CART models were applied to all pneumonia-associated septic shock patients between 1996 and 2015 (n=4222) from the international, multicenter CATSS database. The association between patient and practice-related factors (time delay to appropriate antimicrobial therapy, severity of illness) and in-hospital mortality were evaluated. The accuracy in prediction of outcome (AC), sensitivity (SN), specificity (SP), and area under receiver-operating curve (AUROC) of the final model were evaluated in training (n=2111) and testing (n=2111) sets.
Results: In the overall cohort (n=4222, mean age 62 years, 61% male), overall mortality at hospital discharge was 51%. Sixty-three percent (n=2652) were culture positive (tracheal aspirate/sputum or blood), 21% (n=876) had co-existent bacteremia and 35% (n=1075) had nosocomial infections.  Of culture positive patients, the most common pathogens were staphylococcus sp. (n=702/2652, 27%), streptococcus sp. (n=658, 25%), pseudomonas sp. (n=267, 10%), Escherichia coli (n=225, 8.5%), Klebsiella sp. (n=183, 6.9%) and Haemophilus influenzae (n=118, 4.4%). On ICU admission, mean (SD) APACHEII was 26(8) and lactate 4.1 (3.9) mmol/L. While in ICU, 89% (n=3760) required mechanical ventilation and 11% (n=464) required new renal replacement therapy. Of 3048 patients who received appropriate antimicrobial therapy after the development of hypotension (shock), the mean delay to therapy was 10.9 hours. In the training set (n=2111) a new CART model (see Figure 1) using APACHEII ≥ 28, lactate ≥ 6.3 mmol/L, age > 65 and delay to appropriate antimicrobial therapy ≥ 6.6 hours yielded predictive AC 73%, SP 75%, SN  71% and AUROC 0.75. In the testing set (n=2111), the CART model offered predictive AC 69%, SP 72%, SN 65%, AUROC 0.72.

























Model Accuracy Specificity Sensitivity AUROC
Training
(n=2111)
0.73 0.75 0.71 0.75
Testing
(n=2111)
0.69 0.72 0.65 0.72


 
 
Conclusion: Overall mortality in patients with pneumonia and septic shock is high (~51%). Delay to appropriate antimicrobial therapy, admission severity of illness (APACHEII), serum lactate and advanced age discriminated those patients who survived to hospital discharge and those who did not. CART offer simple prognostic models with good performance.

    This eLearning portal is powered by:
    This eLearning portal is powered by MULTIEPORTAL
Anonymous User Privacy Preferences

Strictly Necessary Cookies (Always Active)

MULTILEARNING platforms and tools hereinafter referred as “MLG SOFTWARE” are provided to you as pure educational platforms/services requiring cookies to operate. In the case of the MLG SOFTWARE, cookies are essential for the Platform to function properly for the provision of education. If these cookies are disabled, a large subset of the functionality provided by the Platform will either be unavailable or cease to work as expected. The MLG SOFTWARE do not capture non-essential activities such as menu items and listings you click on or pages viewed.


Performance Cookies

Performance cookies are used to analyse how visitors use a website in order to provide a better user experience.


Save Settings