User Guide#

Welcome to the MedModels User Guide!

MedModels is a powerful and versatile open-source Python package designed to streamline the analysis of real-world evidence data within the healthcare domain. This user guide is your one-stop shop for everything you need to know about leveraging MedModels’ functionalities to unlock valuable insights from your medical data.

Whether you’re a seasoned data scientist or just starting your foray into medical data analysis, MedModels empowers you with a comprehensive toolkit to tackle various challenges. Here’s a glimpse into what MedModels can offer:

MedRecord

The MedRecord class provides a comprehensive interface for managing medical records using node and edge data structures. This class offers various methods to create, modify, and query these records. Here’s a brief summary of its capabilities:

  1. Initialization and Schema Management:

    • The class can be initialized with or without a schema.

    • Supports setting and getting schema information.

  2. Data Import and Export:

    • Can create instances from tuples, Pandas, or Polars DataFrames.

    • Supports importing data from example datasets and RON files.

    • Can export the record to RON files.

  3. Node and Edge Operations:

    • Add, remove, and query nodes and edges.

    • Supports operations like adding nodes from different data formats (tuples, DataFrames).

    • Provides methods to get node and edge attributes, and manage their connections.

  4. Grouping and Indexing:

    • Allows creating and managing groups of nodes and edges.

    • Can add or remove nodes and edges to/from groups.

    • Provides methods to query nodes and edges in specific groups.

  5. Querying and Selection:

    • Supports complex querying using node and edge operations.

    • Can select nodes and edges based on various conditions and operations.

    • Allows direct selection using indexing with operations.

  6. Connectivity and Relationships:

    • Methods to find incoming and outgoing edges for nodes.

    • Can determine edge endpoints and find edges connecting specific nodes.

    • Supports both directed and undirected neighbor queries.

  7. Clearing and Counting:

    • Can clear all data in the record.

    • Provides methods to count nodes, edges, and groups.

  8. Properties and Accessors:

    • Properties to get node, edge, and group lists.

    • Provides indexers for node and edge attributes, allowing flexible querying and manipulation.

Overall, the MedRecord class is a robust tool for handling complex medical data with a flexible and efficient approach to manage nodes (patients, conditions, etc.) and edges (relationships, treatments, etc.) in a graph-based structure.

TreatmentEffect

The TreatmentEffect class is designed for analyzing treatment effects within medical records. The class has the following capabilities:

  1. Initialization and Configuration:

    • Initializes with groups of treatment and outcome nodes.

    • Configurable parameters include patient groups, time attributes, washout periods, grace periods, follow-up periods, and criteria for filtering control groups.

  2. Group Identification and Filtering:

    • Identifies patients who underwent treatment and experienced outcomes.

    • Finds control groups with similar criteria but without undergoing the treatment.

    • Applies customizable criteria filtering and time constraints between treatment and outcome.

  3. Matching and Comparison:

    • Supports optional matching of control groups to treatment groups using specified matching methods and covariates.

    • Parameters for matching include essential covariates, one-hot covariates, matching models, number of neighbors, and hyperparameters.

  4. Outcome Analysis:

    • Determines patients who had the outcome after the treatment, and optionally before the treatment if configured.

    • Applies washout periods to exclude patients who experienced outcomes during specified times.

  5. Control Group Handling:

    • Identifies control patients and applies filters to ensure appropriate comparison groups.

    • Distinguishes between control patients who experienced the outcome and those who did not.

  6. Estimation and Reporting:

    • Provides properties to create Estimate and Report objects for detailed analysis and reporting of treatment effects.

  7. Support for Temporal Analysis:

    • Utilizes functions like find_node_in_time_window for temporal analysis of treatment and outcome relationships.

Overall, the TreatmentEffect class offers a robust framework for analyzing the impact of treatments in medical records, with flexible configurations for group identification, filtering, matching, and outcome analysis.