icuscoreR

icuscoreR is intended to collate R functions which apply clinical scoring systems to ICU patient data.

Currently a set of function calculate the Acute Physiology and Chronic Health Evaluation (APACHE) III score (Knaus et al. (1991)) for patient data pertaining to the first 24 hours of admission to an intensive care unit.

The R Data Frame must contain the field names described in the Australia New Zealand Intensive Care Society (ANZICS) Adult Patient Data (APD) Data Dictionary (Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation (2022)). All field names are case insensitive, including a mandatory IcuEpisodeId field. HI and LO suffixes on field names such as HRHI (the highest heart rate in 24 hours from ICU admission) are optional, so HRHI can be HR if only 1 extreme is available, or if there are multiple rows of data for each IcuEpisodeId.

For greater detail, after installing the package as below, read vignette(apache3).

Extensive and clearly written resources regarding variable collection, nomeclature and units of measure used by this package are available, hosted by ANZICS at ANZICS Data Collection Tools.

Installation

You can install the development version of icuscoreR on GitHub with:

# 'remotes' package will be installed if devtools is installed
# install.packages("remotes")  
remotes::install_github("mcshaz/icuscoreR", build_vignettes = TRUE)

Alternatively, you may also wish to use pak::pkg_install("mcshaz/icuscoreR"), Although you will not have access to the vignette("apache3").

Example

See the vignette(apache3) for an in depth explanation of the following example:

library(icuscorer)
data("ap3_appendix_example")
ap3_patient_data <- add_miss_ap3_cols(ap3_appendix_example)
#> The column names conform to required nomenclature
ap3.scores <- apache3(ap3_appendix_example)
ap3.comorb.scores <- apache3_comorb(ap3_appendix_example)
data("ap3_appendix_gas")
ap3.abg.scores <- apache3_resp(ap3_appendix_gas)

ap3.scores <- merge(ap3.scores, ap3.abg.scores, by = "icuepisodeid")
ap3.scores <- merge(ap3.scores, ap3.comorb.scores, by = "icuepisodeid")
# Calculating the total Apache III score using
ap3.scores$totalap3 <- rowSums(ap3.scores[,endsWith(colnames(ap3.scores), 'ap3score')], na.rm = TRUE)
ap3.scores
#>   icuepisodeid age.ap3score hr.ap3score map.ap3score temp.ap3score rr.ap3score
#> 1            1            5           7            6             0           9
#> 2            2           17           5            0             0           0
#>   hct.ap3score wcc.ap3score creat.ap3score urineop.ap3score urea.ap3score
#> 1            3            5              7                5            12
#> 2            3            0              4                4             7
#>   Na.ap3score album.ap3score bili.ap3score gluc.ap3score gcs.ap3score
#> 1           0              6             6             3            3
#> 2           0              0             5             0            0
#>   o2.ap3score ph.ap3score comorb.ap3score totalap3
#> 1          11           9              10      107
#> 2           0           0               0       45

Note

• At present the use cases for this package seem too niche to publish it to CRAN , but please feel free to create an issue if you would prefer this to simplify your workflow.
• If you believe other (adult or paediatric) scores should be added (APACHE2, ANZROD, PIM2, PIM3, Trauma Severity Score, SOFA …) please feel free to create an issue (± pull request).

Contributing

Contributions are welcome. However before submitting a pull request please:

• Create a Github issue first.
• Try and minimise dependencies beyond those existing (e.g. DPLYR) unless there is a reasonable case to do so.
• Please include documentation(roxygen2) and tests (testthat) with the pull request.
• Code style follows the Tidyverse style guide.
• Read the design principles of the library below:

Design Principles

• Consider performance and while maintaing readability, try to keep functions vectorised (e.g. avoid dplyr::rowwise). It is conceivable these functions could be executed on a multinational ICU database with 100 000 rows of patient data (and many fold more Arterial Blood Gas results).
• Consider different use cases and database structures.
• Reference clear and publicly available data dictionaries.
• Generate composite scores where feasible - the researcher using the score may wish to include or exclude certain components because that is the exact component under investigation.
• Differentiate NA from a score of 0 even if 0 is the final score contributed.

References

Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation. 2022. Adult Patient Database Data Dictionary. 6.1 ed. Prahran, Victoria, Australia. https://www.anzics.org/wp-content/uploads/2021/03/ANZICS-APD-Dictionary-Version-6.1.pdf.

Knaus, William A, Douglas P Wagner, Elizabeth A Draper, Jack E Zimmerman, Marilyn Bergner, Paulo G Bastos, Carl A Sirio, et al. 1991. “The APACHE III Prognostic System.” Chest 100 (6): 1619–36.