README.md

In StatGent/RiskStandard: Package for benchmarking clinical care centers

RiskStandard

This R-package implements statistical methods for benchmarking clinical care centers based on a binary quality indicator such as 30-day mortality. For each center we provide directly or indirectly standardized risks based on fixed center effects outcome models that incorporate patient-specific baseline covariates to adjust for differential case-mix. The user can choose to apply the Firth correction in the outcome model to maintain convergence in the presence of very small centers.

Input data must contain for each patient:

1. patient-specific covariates to adjust for in the analysis, e.g., age and baseline severity,
2. a hospital code where the patient was treated and
3. a binary quality outcome, e.g., 30-day mortality.

Three example datasets are included in the package.

Contents

This repository contains both the extracted R-package that can be installed by using the R-package devtools as well as the compressed RiskStandard_0.0.6.tar.gz that can be installed as follows:

install.packages("./RiskStandard_0.0.6.tar.gz", 
                 repos = NULL, type = "source")

Included in this package are the following functions:

1. standardizeRisks(): estimates the (directly or indirectly) standardized risks. When some patients have missing values for a categorical patient covariate, two ways to handle the missingness are offered: missing='completeCase' performs a complete case analysis, missing='dummyCategory' adds a separate category to the fitted outcome model, allowing for a missing value effect. When some patients have missing values for a continuous patient covariate, the function will perform a complete case analysis.
2. labelCenters(): uses the output from the standardizeRisks() function to classify the centers as having 'low', 'accepted' or 'high' mortality risk. The clinically relevant cut-off boundaries can be adapted by the user.
3. plotRisks(): for indirectly standardized outcomes, this function generates a descriptive scatterplot of the observed against the expected risk under the average care level for patients of that center. For directly standardized outcomes, this function generates a descriptive scatterplot of the observed against the estimated directly standardized risk for each center.
4. plotCenterLabels(): center performance classification can be visualised using the estimated standardized risk and variance per center from the output of standardizeRisks() and classification labels from labelCenters().
5. funnelPlot(): the funnelplot will compare institutional performance. The estimated standardized risks are plotted against a measure of precision. Care centers with an estimated standardized risk lying outside the 95% control limits are flagged as outlying centers.

as well as three illustrative datasets on which the functions can be run:

• smallCaseMix with small differences in patient mix across centers,
• largeCaseMix with large differences in patient mix across centers, and
• largeCaseMix_missing which is based on largeCaseMix but where the consciousness level is missing for some patients.

Documentation

A detailed description of all functions and datasets can be found in the vignette.

Reference to the package

Please use the following reference when referencing to the RiskStandard package:

Varewyck, M. and Van Messem, A. (2015). RiskStandard. https://github.com/StatGent/RiskStandard

DOI:

Authors & Licence

The package was developed by Stat-Gent CRESCENDO, Ghent University. The code made available in this repository is licenced under the GPL-3 licence (see LICENCE for details).

StatGent/RiskStandard documentation built on May 9, 2019, 1:57 p.m.