Research Papers:

Development of a novel score for the prediction of hospital mortality in patients with severe sepsis: the use of electronic healthcare records with LASSO regression

Zhongheng Zhang _ and Yucai Hong

PDF  |  HTML  |  Supplementary Files  |  How to cite

Oncotarget. 2017; 8:49637-49645. https://doi.org/10.18632/oncotarget.17870

Metrics: PDF 2765 views  |   HTML 3417 views  |   ?  


Zhongheng Zhang1 and Yucai Hong1

1Department of Emergency Medicine, Sir Run-Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou, 310016, China

Correspondence to:

Zhongheng Zhang, email: [email protected]

Keywords: sepsis, LASSO, prediction

Received: March 28, 2017     Accepted: May 01, 2017     Published: May 15, 2017


Background and objective: There are several disease severity scores being used for the prediction of mortality in critically ill patients. However, none of them was developed and validated specifically for patients with severe sepsis. The present study aimed to develop a novel prediction score for severe sepsis.

Results: A total of 3206 patients with severe sepsis were enrolled, including 1054 non-survivors and 2152 survivors. The LASSO score showed the best discrimination (area under curve: 0.772; 95% confidence interval: 0.735-0.810) in the validation cohort as compared with other scores such as simplified acute physiology score II, acute physiological score III, Logistic organ dysfunction system, sequential organ failure assessment score, and Oxford Acute Severity of Illness Score. The calibration slope was 0.889 and Brier value was 0.173.

Materials and Methods: The study employed a single center database called Medical Information Mart for Intensive Care-III) MIMIC-III for analysis. Severe sepsis was defined as infection and acute organ dysfunction. Clinical and laboratory variables used in clinical routines were included for screening. Subjects without missing values were included, and the whole dataset was split into training and validation cohorts. The score was coined LASSO score because variable selection was performed using the least absolute shrinkage and selection operator (LASSO) technique. Finally, the LASSO score was evaluated for its discrimination and calibration in the validation cohort.

Conclusions: The study developed the LASSO score for mortality prediction in patients with severe sepsis. Although the score had good discrimination and calibration in a randomly selected subsample, external validations are still required.

Creative Commons License All site content, except where otherwise noted, is licensed under a Creative Commons Attribution 4.0 License.
PII: 17870