Skip to content

Latest commit

 

History

History
97 lines (56 loc) · 7.58 KB

README.md

File metadata and controls

97 lines (56 loc) · 7.58 KB
Raw

𝗟𝗮𝗯𝗥𝗜 𝗦𝗵𝗶𝗻𝘆 𝗜𝗻𝘀𝘁𝗮𝗹𝗹𝗲𝗿

The LabRI System is a Shiny-based application designed for the estimation and verification of reference intervals in clinical laboratories. It combines RMarkdown scripts, Shiny applications, and automation scripts (.bat) to streamline its use without requiring direct interaction with RStudio. This repository includes three key components:

  • install_packages.bat: Triggered by the shortcut LabRI Package Installer, this file executes the install_packages.R script, automating the installation and updating of the required R packages to properly configure the environment.
  • app.bat: Triggered by the shortcut LabRI, this file executes the launch_app.R script, initializing the Shiny application and opening the interactive graphical interface to run the LabRI_script.Rmd, which implements the LabRI method.
  • LabRI_script.Rmd: The primary script that implements the LabRI method, responsible for estimating and verifying reference intervals, and producing comprehensive HTML reports.

The image above provides an example of the initial interface of the LabRI Shiny Application, demonstrating how users configure essential parameters for data analysis. The "Name of the Responsible Specialist" section captures the analyst's name, while the "Define the Dataset" section allows users to upload a .csv, .xls, or .xlsx file and select the relevant data column. A status bar indicates the system's progress during processing. This streamlined interface ensures intuitive navigation and efficient setup for reference interval estimation and verification.

The LabRI System is distributed primarily as an executable installer, which organizes files, configures directories, and creates desktop shortcuts for easy access. As the installer relies on .bat files for automation, it is only compatible with the Windows operating system. The executable can be downloaded using the link below:


The LabRI Method is a core component of the LabRI System, serving as the analytical backbone for the estimation and verification of reference intervals in laboratory data. It comprises a set of algorithms, sub-algorithms, and mathematical procedures, implemented primarily in the LabRI_script.Rmd file.

The LabRI Method is structured into two main modules:

  • Estimation Module: Focuses on adaptive, multi-criteria estimation of reference intervals through data cleaning, transformation, and clustering techniques, utilizing algorithms such as refineR and reflimR, available in the R packages refineR and reflimR, respectively, along with the Expectation-Maximization (EM) algorithm, supported by packages like mclust and mixR.

  • Verification Module: Ensures the validity of estimated reference intervals through a three-level analysis, which evaluates statistical uncertainty, equivalence, and concordance, making the intervals reliable for clinical application.

𝗔. 𝗘𝘀𝘁𝗶𝗺𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝘂𝗹𝗲

The LabRI method provides an adaptive and multi-criteria approach for the indirect estimation of reference intervals. This module integrates data cleaning, transformation, and clustering techniques, utilizing the refineR, reflimR, and EM algorithms. By combining parametric and non-parametric percentile approaches, the method estimates population reference intervals based on the number of clusters in the truncated distribution.

Characteristics of the LabRI Method

  • Adaptive:

    • Adjusts dynamically based on data structure and characteristics, applying appropriate cleaning and transformation techniques.

    • For multi-cluster distributions, the Centroid of Windsorized Reference Limits is applied to the reference limits estimated by refineR and reflimR. This involves a two-stage process: first, the Two-stage Winsorization sub-algorithm estimates the winsorized reference limits, adding robustness against extreme values. Next, the Hartigan-Wong Centroid Reference Limits sub-algorithm calculates the centroid, with the x and y coordinates representing the lower and upper reference limits, respectively, yielding a centralized and stable estimate. When clusters are sufficiently distant from each other, the EM algorithm is also incorporated to further refine the reference interval estimate.

    • For single-cluster distributions, the EM algorithm applies parametric and non-parametric methods to derive the best reference interval estimate.

  • Multi-criteria:

    • Incorporates multiple criteria and methods for robust and comprehensive estimation and verification of reference intervals.

𝗕. 𝗩𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝘂𝗹𝗲

To ensure reliability in clinical practice, it is crucial for laboratories to verify their reference intervals (RIs) before routine application. This verification is especially important for RIs derived through indirect methods.

Structure of the Verification Module

The Verification Module performs a three-level analysis to assess whether the compared reference limits are equivalent:

  1. First-Level Analysis ~ Statistical Uncertainty: Assesses the magnitude of statistical uncertainty in the reference limits.

  2. Second-Level Analysis ~ Distance Criterion Based on Equivalence Testing: Compares the LabRI-estimated reference limit with a comparative limit to evaluate practical significance.

  3. Third-Level Analysis ~ Concordance Evaluation: Evaluates concordance using tests like Fleiss’ Kappa, Lin’s Concordance Correlation Coefficient, and Flagging Rates.

Details of the Three-Level Analysis

  • First-Level Analysis:

    • Evaluates statistical uncertainty associated with reference limits. If uncertainty is within acceptable bounds, the analysis proceeds to the second level.

  • Second-Level Analysis:

    • Compares the LabRI reference limit with a comparative reference limit using equivalence testing to assess practical significance.

  • Third-Level Analysis:

    • Conducted if the second-level analysis suggests "Possible Equivalence" or "Probable Equivalence". This level incorporates confidence intervals and uses Fleiss' Kappa, Lin's Concordance Correlation Coefficient, and Flagging Rates to ensure robust verification.

A simple usage tutorial, covering the installation of R and RStudio and instructions for using the Shiny tool, can be found on the Grupo Lab R website:

𝗖𝗼𝗻𝘁𝗮𝗰𝘁

You are welcome to:

Submit suggestions and Bugs at: https://github.com/labrgrupo/LabRI_Tool/issues

Write an Email with any questions and problems to: alancdias@hotmail.com or labrgrupo@gmail.com

Link to the publication: