PooledCohort

The goal of PooledCohort is to give researchers who study risk prediction for cardiovascular disease a clean interface to implement the Pooled Cohort Risk prediction equations.

Why do we want to use Pooled Cohort Risk prediction equations?

The 2017 American College of Cardiology and American Heart Association blood pressure guideline recommends using 10-year predicted atherosclerotic cardiovascular disease risk to guide the decision to initiate or intensify antihypertensive medication. The guideline recommends using the Pooled Cohort Risk prediction equations to predict 10-year atherosclerotic cardiovascular disease risk. Therefore, a new method for predicting atherosclerotic cardiovascular disease risk should be evaluated based on how much it can improve risk prediction relative to the Pooled Cohort Risk prediction equations.

Installation

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("bcjaeger/PooledCohort")

Example

A basic example below computes 10-year atherosclerotic cardiovascular risk using the original Pooled Cohort Risk equations for a person who is black/white and male/female with

First we will use a dataset that requires no modification of any variable to be plugged into predict_10yr_ascvd_risk()


library(PooledCohort)
library(dplyr, warn.conflicts = FALSE)

example_data <- data.frame(
  sex = c('female', 'female', 'male', 'male'),
  race = c('black', 'white', 'black', 'white'),
  age_years = rep(55, times = 4),
  chol_total_mgdl = rep(213, times = 4),
  chol_hdl_mgdl = rep(50, times = 4),
  bp_sys_mmhg = rep(120, times = 4),
  bp_meds = rep('no', times = 4),
  smoke_current = rep('no', times = 4),
  diabetes = rep('no', times = 4),
  stringsAsFactors = FALSE
)

example_data
#>      sex  race age_years chol_total_mgdl chol_hdl_mgdl bp_sys_mmhg bp_meds
#> 1 female black        55             213            50         120      no
#> 2 female white        55             213            50         120      no
#> 3   male black        55             213            50         120      no
#> 4   male white        55             213            50         120      no
#>   smoke_current diabetes
#> 1            no       no
#> 2            no       no
#> 3            no       no
#> 4            no       no

A convenient way to use predict_10yr_ascvd_risk() is within dplyr::mutate():


example_risk <- example_data %>% 
  mutate(
    risk = predict_10yr_ascvd_risk(
      sex = sex,
      race = race,
      age_years = age_years,
      chol_total_mgdl = chol_total_mgdl,
      chol_hdl_mgdl = chol_hdl_mgdl,
      bp_sys_mmhg = bp_sys_mmhg,
      bp_meds = bp_meds,
      smoke_current = smoke_current,
      diabetes = diabetes
    )
  ) %>% 
  select(sex, race, risk)

example_risk
#>      sex  race       risk
#> 1 female black 0.02998196
#> 2 female white 0.02054450
#> 3   male black 0.06061116
#> 4   male white 0.05378606

Data formatting

Data usually need to be modified slightly before being plugged into the Pooled Cohort Risk equations. For example, instead of a race variable with values of black and white, the data may have a race variable with values of african_american, caucasian, and other.


example_data_granular <- data.frame(
  sex = c('female', 'female', 'male', 'male'),
  race = c('african_american', 'caucasian', 'african_american', 'other'),
  age_years = rep(55, times = 4),
  chol_total_mgdl = rep(213, times = 4),
  chol_hdl_mgdl = rep(50, times = 4),
  bp_sys_mmhg = rep(120, times = 4),
  bp_meds = rep('no', times = 4),
  smoke_current = rep('no', times = 4),
  diabetes = rep('no', times = 4),
  stringsAsFactors = FALSE
)

While you can always modify variables in your data so that they meet the same format as the variables in example_data above, you may prefer to let predict_10yr_ascvd_risk() modify those variables for you:


# a mapping from the current race categories to 
# the 'black' and 'white' categories used by the
# Pooled Cohort Risk equations.

race_levels <- list(
  black = 'african_american',
  white = c('caucasian', 'other')
)

example_risk_granular <- example_data_granular %>% 
  mutate(
    risk = predict_10yr_ascvd_risk(
      sex = sex,
      race = race,
      race_levels = race_levels,
      age_years = age_years,
      chol_total_mgdl = chol_total_mgdl,
      chol_hdl_mgdl = chol_hdl_mgdl,
      bp_sys_mmhg = bp_sys_mmhg,
      bp_meds = bp_meds,
      smoke_current = smoke_current,
      diabetes = diabetes
    )
  ) %>% 
  select(sex, race, risk)

example_risk_granular
#>      sex             race       risk
#> 1 female african_american 0.02998196
#> 2 female        caucasian 0.02054450
#> 3   male african_american 0.06061116
#> 4   male            other 0.05378606

A picky estimator

predict_10yr_ascvd_risk() is a picky estimator that will throw hard stops at you if it doesn’t like the data you give it. In particular, if the input data has continuous variables with values outside the range of recommended values for the Pooled Cohort equations, you will get an error message.


predict_10yr_ascvd_risk(
  sex = 'male',
  race = 'black',
  age_years = 35, # age recommendation: 40-79
  chol_total_mgdl = 213,
  chol_hdl_mgdl = 55,
  bp_sys_mmhg = 120,
  bp_meds = 'no',
  smoke_current = 'no',
  diabetes = 'no'
)
#> Error: min(age_years) is 35 but should be >= 40

This is meant to help you avoid mis-use of the Pooled Cohort Risk equations. However, if you must go outside the range of recommended values, you can set the argument override_boundary_errors to TRUE.


predict_10yr_ascvd_risk(
  sex = 'male',
  race = 'black',
  age_years = 35,
  chol_total_mgdl = 213,
  chol_hdl_mgdl = 55,
  bp_sys_mmhg = 120,
  bp_meds = 'no',
  smoke_current = 'no',
  diabetes = 'no',
  override_boundary_errors = TRUE
)
#> [1] 0.01969643