Dark Mode
Sepsis Patient Outcomes Dataset
Patient Health Records & Digital Health
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset is designed for sepsis survival prediction, a critical task in clinical medicine. It contains records of 110,204 patient admissions from 84,811 hospitalised individuals in Norway between 2011 and 2012. These patients were diagnosed with infections, systemic inflammatory response syndrome, sepsis due to causative microbes, or septic shock. The primary objective is to predict patient survival or death approximately 9 days after their medical record collection. Sepsis is a life-threatening condition associated with immediate death risk, making timely diagnosis and treatment challenging. The dataset supports the development of models to predict patient survival rapidly with minimal and easily retrievable medical features. It includes a primary cohort representing patients with potential sepsis preconditions (ante Sepsis-3 definition) and a study cohort reflecting admissions defined by the novel Sepsis-3 definition. A validation cohort from South Korea is also included for generalisability studies.
Columns
The dataset is structured into three distinct cohorts, each featuring 4 attributes or features. Specific column names and their detailed descriptions are not explicitly available in the provided materials. However, it is noted that the dataset contains sensitive information, including patient gender and age. All categorical variables within the dataset have undergone pre-encoding, eliminating the need for further preprocessing.
Distribution
The dataset is typically supplied in CSV file format and is organised into three distinct cohorts:
- Primary Cohort (Norway): Comprises 110,204 patient admissions with 4 features, stored in a file approximately 985.93 kB in size.
- Study Cohort (Norway): A subset of the primary cohort, containing 19,051 patient admissions with 4 features, found in a file approximately 171.27 kB in size.
- Validation Cohort (South Korea): Consists of 137 patients with 4 features, located in a file approximately 1.31 kB in size. The dataset's expected update frequency is never.
Usage
This dataset is ideal for developing and evaluating machine learning models for sepsis survival prediction. It supports various applications, including:
- Developing predictive tools for real-time patient outcome assessment in clinical settings.
- Informing and optimising timely diagnosis and treatment strategies for sepsis.
- Facilitating machine learning model selection through standard train-test or three-way holdout splits.
- Conducting external validation of predictive models to confirm their generalisability across diverse patient cohorts and geographical regions.
Coverage
The dataset covers patient admissions from Norway between 2011 and 2012 for both the primary and study cohorts. An additional validation cohort is included from South Korea. Demographically, the dataset contains information regarding patient gender and age. Specific notes on data availability for particular groups or years beyond the stated range are not provided.
License
Attribution 4.0 International (CC BY 4.0)
Who Can Use It
- Clinical Researchers: For investigating sepsis mortality, patient outcomes, and creating novel predictive algorithms.
- Data Scientists & Machine Learning Engineers: For building, training, and validating predictive models tailored for healthcare applications.
- Healthcare Professionals: To support informed decision-making processes concerning patient care and resource allocation.
- Academics: For educational purposes, research endeavours, and publications in the fields of medical prognostics and data science.
Dataset Name Suggestions
- Sepsis Patient Outcomes Dataset
- Norwegian and Korean Sepsis Survival
- Clinical Sepsis Prediction Cohorts
- Sepsis Mortality Prediction Data
- Healthcare Sepsis Patient Data
Attributes
Original Data Source:Sepsis Patient Outcomes Dataset
QUALITY