Sepsis Screening Tool

A Streamlit application for automatically screening research records for diagnostics in antimicrobial resistance in sepsis using a keyword-based approach.

Features

• Upload CSV files containing search results with columns such as "Title", "Abstract", and "Authors"
• Automatically screen records using keywords from AMR, diagnostic, and sepsis categories
• Calculate scores based on keyword matches in both title and abstract fields
• Classify records as "Include", "Exclude", or "Maybe" based on configurable thresholds
• Manually review and override classifications
• Export final screening results to CSV

Installation

1. Clone this repository:

   git clone <repository-url>
   cd sepsis_screening
   

2. Install required dependencies:

   pip install -r requirements.txt
   

Usage

1. Run the Streamlit application:

   streamlit run app.py
   

2. Open your web browser and navigate to the URL shown in the terminal (typically http://localhost:8501)

3. Upload your CSV file containing research records (you can use the provided sample_data.csv for testing)

4. Adjust the threshold settings in the sidebar as needed

5. Review and manually classify records if necessary

6. Download the results using the provided link

Sample Data

A sample dataset (sample_data.csv) is included in this repository for testing purposes. It contains 12 mock research records related to sepsis, antimicrobial resistance, and diagnostic methods. Each record includes title, abstract, authors, publication year, and journal information.

Input Data Format

The application expects a CSV file with the following columns:

• Title: The title of the research paper/record
• Abstract: The abstract text of the research paper/record
• Optional columns like Authors, Year, and Journal will also be displayed if present

Keyword Categories

The screening is based on three categories of keywords:

1. AMR Keywords: Terms related to antimicrobial resistance
2. Diagnostic Keywords: Terms related to diagnostic methods and tests
3. Sepsis Keywords: Terms related to sepsis and bloodstream infections

You can view the specific keywords in each category by expanding the respective sections in the sidebar.

Scoring System

Each record receives a score based on the number of keyword matches found in both the title and abstract. Records are then classified as:

• Include: Records with scores at or above the high threshold
• Exclude: Records with scores below the low threshold
• Maybe: Records with scores between the two thresholds

The thresholds can be adjusted in the sidebar to fine-tune the screening process.

License

MIT License