Jawad Najar
Pronouns: He/Him
Research Mentor(s): Flora Rajaei
Research Mentor School/College/Department: DCMB / Medicine
Program: UROPF
Session: Session 3 (11:00am – 11:50am)
Authors:
Abstract
This research seeks to target patients that are infected with sepsis, more specifically we focus on those infected in the United States. The individuals in our research are provided to us by a widely known critical care database called Mimic-III. This dataset has 38,597 unique adult patients all of which were patients in the Beth Israel Deaconess Medical Center in Boston, Massachusetts1, 5000 of which (will insert exact number when we are finished with querying) were identified as potential sepsis patients based on the Sepsis-3 criteria2. Once found, we created a table using the 29 variables that were used in the paper3. These variables can be grouped into 9 distinct categories which are demographic characteristics, organ dysfunction, inflammation, pulmonary, cardiovascular/hemodynamic, renal, hepatic, hematological, and other. They were measured using the CareVue and MetaVision Electronic Health Record systems. To preprocess the data, we used the programming language Python, utilizing the Pandas, NumPy, and SkLearn libraries. In order to handle the missing data in continuous variables we used mean imputation, however with categorical data, this methodology would not work. Due to the fact that we could not use categorical data in our analysis since the K-Means algorithm does not support it, we opted into using One-Hot Encoding. This implicitly fixed our issue of missing categorical data since One-Hot Encoding is able to represent those with missing data in a binary structure without losing data. After the preprocessing, we can finally apply K-Means Clustering which was hypothesized to have four phenotypes in the paper3. These phenotypes were Mechanical Ventilation, Administration of Vasopressors, Admitted to the Intensive Care Unit, and In-Hospital Mortality. 1. Johnson AEW, Pollard TJ, Shen L, et al. Mimic-III, a freely accessible Critical Care Database. Nature News. https://www.nature.com/articles/sdata201635. Published May 24, 2016. Accessed February 23, 2023. 2. Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4968574/. Published February 23, 2016. Accessed February 23, 2023. 3. Seymour CW, Kennedy JN, Wang S, et al. Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis. JAMA. 2019;321(20):2003. doi:10.1001/jama.2019.5791
