Skip to contents

preventr

Why preventr?

The main goal of preventr is to implement the Predicting Risk of cardiovascular disease EVENTs (PREVENT) equations, released in 2023 by the American Heart Association (AHA). It also permits comparison of the PREVENT equations with their de facto predecessor, the Pooled Cohort Equations (PCEs). The PCEs were originally released in 2013 as part of the American College of Cardiology (ACC)/AHA Guideline on the Assessment of Cardiovascular Risk, and they were revised by Yadlowsky and colleagues in 2018, though these revisions were never endorsed by the ACC/AHA.

The PREVENT equations are a large collection of predictive/prognostic models for predicting the risk of cardiovascular disease events over the next 10 and 30 years. Specifically, estimation includes both 10- and 30-year risk of 5 events:

  • Total cardiovascular disease (CVD), which includes atherosclerotic CVD (ASCVD) and heart failure as defined below
  • ASCVD, which includes coronary heart disease (CHD) and stroke as defined below
  • Heart failure (often abbreviated HF, but not herein)
  • CHD, which includes nonfatal myocardial infarction (MI) and fatal CHD
  • Stroke

The predicted risk of each of these 5 outcomes is defined by separate beta coefficients and intercepts. Additionally, the beta coefficients and intercepts are sex-specific, yielding 10 sets of beta coefficients and intercepts for the 5 outcomes predicted by the PREVENT equations.

But wait, there’s more. In addition, there are 5 model “types”:

  • the base model
  • the base model adding HbA1c
  • the base model adding urine albumin-to-creatinine ratio (UACR)
  • the base model adding social deprivation index (SDI)
  • the base model adding HbA1c, UACR, and SDI (also called the “full model”)

We are now up to 50 sets of beta coefficients and intercepts.

Finally, there are separate models for 10- and 30-year risks, bringing the grand total to 100 sets of beta coefficients and intercepts for the entirety of the PREVENT equations.

preventr takes care of all of that for you. This includes selecting among the 5 model “types”, but there is an option to specify this yourself if you want. The example below gives a quick overview, but the function documentation has much more detail, including many more examples.

preventr also permits estimation of 10-year ASCVD risk from the PCEs, but of note, the PREVENT equations have better calibration than the PCEs, as demonstrated in the original article describing the derivation and validation of the PREVENT equations.

Installation

Install the released version of preventr from CRAN with:

install.packages("preventr")

You can install the development version of preventr from GitHub with:

# install.packages("devtools")
devtools::install_github("martingmayer/preventr")

Using preventr

Despite the work it does, I designed preventr with the goal of having a simple, intuitive API. The “workhorses” are “behind the scenes”, culminating in one function (and synonym) being exposed to the user for estimation of risk:

# Very basic example; see function documentation for much more detail and
# many more examples
# 
# If the package is loaded (e.g., `library(preventr)`) or the function is made 
# available  via some other means (e.g., importing as part of development 
# of another package), use of `preventr::` before the function name is not
# strictly necessary; however, this approach to calling functions can often be
# helpful for clarity of code, for avoiding potential namespace conflicts, etc. 
# However, a full discussion of pros and cons and when one approach might be
# favored over another is beyond the scope of this package or this comment.
preventr::estimate_risk(
  age = 50,
  sex = "female",
  sbp = 160,
  bp_tx = TRUE,
  total_c = 200,
  hdl_c = 45,
  statin = FALSE,
  dm = TRUE,
  smoking = FALSE,
  egfr = 90,
  bmi = 35
)
#> PREVENT estimates are from: Base model.
#> $risk_est_10yr
#> # A tibble: 1 × 8
#>   total_cvd ascvd heart_failure   chd stroke model over_years input_problems
#>       <dbl> <dbl>         <dbl> <dbl>  <dbl> <chr>      <int> <chr>         
#> 1     0.147 0.092         0.081 0.044  0.054 base          10 <NA>          
#> 
#> $risk_est_30yr
#> # A tibble: 1 × 8
#>   total_cvd ascvd heart_failure   chd stroke model over_years input_problems
#>       <dbl> <dbl>         <dbl> <dbl>  <dbl> <chr>      <int> <chr>         
#> 1      0.53 0.354          0.39 0.198  0.221 base          30 <NA>

There’s an app for that

In addition to the R package, I created a Shiny app that is driven by the preventr package and also includes things like risk visualization and several options for customization of the output. The app is available at:

https://martingmayer.shinyapps.io/prevent-equations

Easier-to-remember URLs:

Calling preventr::app() will also open the user’s default browser and navigate to the Shiny app.

Why not 1.0.0?

First and most importantly: I have tested the package’s functionality extensively. That is not the reason for the < 1.0.0 release.

Rather, I developed the API with an eye toward maximizing simplicity and intuitiveness. Thus, while I do not anticipate any major/breaking changes to the API, I remain open to the idea there may be improvements to the API that may surface, either from my own development ideas or after more people use this package. This is the only reason preventr is < 1.0.0. In fact, this happened with previous updates when I had additional development ideas to improve the utility of the package (all while maintaining backward compatibility), including:

  • adding the ability to call calc_egfr() and calc_bmi() (or synonyms) for the corresponding arguments in estimate_risk() (or its synonym est_risk()).

  • adding several more capabilities to estimate_risk() and est_risk(), such as the ability to estimate with the PCEs, the ability to collapse the return where applicable, and the ability to pass a data frame with the option to have the results added back to that source data frame.

Users should rest assured if any changes come to the API, I will avoid breaking changes unless they are necessary or there is a compelling argument for making such changes; likewise, I will aim to communicate any such changes clearly and in advance if possible. Again, I consider all of this very unlikely, but in the off chance something like that does arise, I also want others to rest assured I will make efforts to ensure surprises and/or headaches are minimized or avoided if possible, essentially following the best practices outlined at the excellent reference R Packages.

How to cite

If you use preventr in your work, please cite the package as follows:

citation("preventr")

Acknowledgments

This package would, of course, not be possible without the efforts from the authors of the PREVENT equations. Additionally, the Social Deprivation Index (SDI) is a key element informing the PREVENT equations. Citations for both appear below, as does acknowledgment of the other packages and software I used in creating preventr.

The PREVENT equations

Khan SS, Matsushita K, Sang Y, Ballew SH, Grams ME, Surapaneni A, Blaha MJ, Carson AP, Chang AR, Ciemins E, Go AS, Gutierrez OM, Hwang SJ, Jassal SK, Kovesdy CP, Lloyd-Jones DM, Shlipak MG, Palaniappan LP, Sperling L, Virani SS, Tuttle K, Neeland IJ, Chow SL, Rangaswami J, Pencina MJ, Ndumele CE, Coresh J; Chronic Kidney Disease Prognosis Consortium and the American Heart Association Cardiovascular-Kidney-Metabolic Science Advisory Group. Development and Validation of the American Heart Association’s PREVENT Equations. Circulation. 2024 Feb 6;149(6):430-449. Epub 2023 Nov 10.

Social Deprivation Index (SDI)

Social deprivation index (SDI). Robert Graham Center - Policy Studies in Family Medicine & Primary Care. (2018, November 5). Retrieved December 13, 2023, from https://www.graham-center.org/maps-data-tools/social-deprivation-index.html.

Other literature, packages, and software

Estimation via the original and revised PCEs would not be possible without the efforts from the respective groups behind this work. Thank you to the authors of this work:

preventr also uses the dplyr and zipcodeR packages, the latter of which is used entirely “behind the scenes” as a data set to help validate zip codes (and thus zipcodeR is not imported as part of using preventr). Thank you to the authors and maintainers of these packages.

The authors and maintainers of packages dedicated to developing packages also deserve recognition. These include but are not necessarily limited to: devtools, roxygen2, usethis, testthat, and pkgdown. The rmarkdown package, though not dedicated to package development, also deserves explicit recognition, as it directly and significantly contributes to the feasibility and functionality of packages like pkgdown and roxygen2. Thank you to the authors and maintainers of these packages.

A huge thank you is also in order for the R Core Team and others who have contributed to maintaining R, as well as the CRAN Team. Your tireless work is, of course, foundational to all the above packages.

Lastly, I also used DALL·E 3 within Microsoft Copilot to help generate the logo. Thank you to the authors and maintainers of DALL·E 3 and Microsoft Copilot.