creatifun / predictriskscores Goto Github PK

This is an R package featuring functions for generating PREDICT cardiovascular disease (CVD) risk scores.

• Estimate 5 year CVD risk in people with and without established CVD
• Estimate 5 year CVD risk in people who have experienced an acute coronary syndrome (ACS)
• Estimate 5 year CVD risk in people without established Type-II diabetes
• Estimate 5 year major bleeding risk in people without established CVD

1. NoPriorCVDRisk calculates the 5 year absolute risk of CVD for people without a history of atherosclerotic CVD. Published in the Lancet (Pylypchuk et al. 2018) and does not include BMI as a predictor. Full Article

2. NoPriorCVDRisk_BMI calculates the 5 year absolute risk of CVD for people without a history of atherosclerotic CVD. Published as a Ministry of Health's HISO standard and adds BMI as a predictor to the Lancet equation. HISO Standard

3. NoPriorCVDRisk_Policy calculates the 5 year absolute risk of CVD for people without a history of atherosclerotic CVD that is intended for use at policy or general population level. Publushed in the IJE (Mehta et al. 2018) and uses routinely-collected health information as predictors. Full Article

4. NoPriorCVDBleedRisk calculates the 5 year absolute risk of major bleeding for people without a history of atherosclerotic CVD. Published in AnnIntMed (Selak et al. 2019). Full Article

5. NoPriorT2DRisk creates a 5 year absolute risk of CVD for people without a history of type-II diabetes. Published as a Ministry of Health's HISO standard and includes BMI as a predictor. HISO Standard

6. PostCVDRisk creates a 5 year absolute risk of CVD for people with prior CVD. This equation is yet to be published. However, it is an update to the previously published equation in Heart (Poppe et al. 2017). Full Article

7. PostACSRisk creates a 5 year absolute risk for people for people who have experienced an ACS event. Published in heart (Poppe et al. 2019). Full Article

To install package:

remotes::install_github("billy-nz/PredictRiskScores")

To load the package:

library(PredictRiskScores)

Each function can be used as either a calculator or vectoriser. When used as a one-off calculator, a dataset is not required. This can be handy when checking the risk estimate for an individual person or for quickly seeing the effect of changing a single parameter value.

PostACSRisk(sex="F", age=65, eth=Indian, nzdep=5, smoker=0, diabetes=0,
            af=0, hf=1, acsdays=65, acstype="NSTEMI", bmi=NA, sbp=118,
            tchdl=3.3, hba1c=NA, scr=52, bpl=1, lld=1, athrombi=1)

[1] 0.4877

All parameter values can be numeric or encoded. Values for some parameters such as BMI and HbA1C can be NA if the value is unknown. See values in R documentation. Binary variables can take TRUE/FALSE boolean values.

PostACSRisk(sex=0, age=65, eth=43, nzdep=5, smoker=0, diabetes=0,
            af=0, hf=1, acsdays=65, acstype=1, bmi=NA, sbp=118,
            tchdl=3.3, hba1c=NA, scr=52, bpl=T, lld=F, athrombi=1)

[1] 0.4877

Binary parameters can take a variety of input values. For example, the value for TRUE can be T, Y, Yes, or 1, while the value for FALSE can be F, N, No, or 0. Quotations are not required and case sensitivity is ignored.

NoPriorCVDRisk(sex=F, age=30, eth=M, exsmoker=0, smoker=TRUE, nzdep=3, diabetes=Y,
               af=F, familyhx=1, lld=y, athrombi=yes, bpl=T, sbp=150, tchdl=5)

[1] 0.0549

When a dataset is supplied, a risk score is produced for each row of data. Each argument requires the variable name from the dataset. This can be handy when risk estimates are required for each row of data, or when datasets require vectorisation.

NoPriorCVDRisk(dat=DATA, sex=sex, age=age, eth=ethnicity, exsmoker=exsmoker, smoker=current_smk, nzdep=nzdep, 
              diabetes=diab_status, af=hx_af, familyhx=familyhx, lld=lld, athrombi=athrombotics, bpl=bpl, 
              sbp=sbp, tchdl=tchdl)

[1] 0.3327 0.0489 0.0620 0.5441 0.0688 0.1672 0.5054 0.0442 0.0150 0.0387

When a dataset is supplied, the function returns a numeric vector of risk scores. Each element of the vector is positioned as per row index. As such, the resulting numeric vector can be assigned back to the dataset as a new variable.

DATA$riskscores <- NoPriorCVDRisk(dat=DATA, sex=sex, age=age, eth=ethnicity, exsmoker=exsmoker, smoker=current_smk,
                                  nzdep=nzdep, diabetes=diab_status, af=hx_af, familyhx=familyhx, lld=lld, 
                                  athrombi=athrombotics, bpl=bpl, sbp=sbp, tchdl=tchdl)l)

The suite of functions in this package can be integrated into both data.table and dplyr. For example, when datasets are extremely large, consider using data.table along with the := notation. In the example below, a new column called riskscore is created. The data.table syntax might seem confusing at first but it offers fast and efficient performance.

library(data.table); setDT(DATA)

DATA[, riskscore := NoPriorCVDRisk(dat=DATA, sex=sex, age=age, eth=ethnicity, exsmoker=exsmoker, smoker=current_smk,
                                  nzdep=nzdep, diabetes=diab_status, af=hx_af, familyhx=familyhx, lld=lld,
                                  athrombi=athrombotics, bpl=bpl, sbp=sbp, tchdl=tchdl)]

Users who are more comfortable with dplyr will find that the functions works with mutate. In the example below, a new column called riskscore is created.

library(dplyr)

DATA %>%
  mutate(riskscore = NoPriorCVDRisk(dat=DATA, sex=sex, age=age, eth=ethnicity, exsmoker=exsmoker, smoker=current_smk,
                                    nzdep=nzdep, diabetes=diab_status, af=hx_af, familyhx=familyhx, lld=lld,
                                    athrombi=athrombotics, bpl=bpl, sbp=sbp, tchdl=tchdl))

Each function contains 3 dot parameters: dp, allow.age, and allow.na. A function will:

• by default return risk scores to 4 decimal places. The number of decimal places can be changed using the argument dp. For example, dp = 6 will produce risk scores containing 6 decimal places.
• by default return risk scores for people between 18 and 110 years of age. Ages 18 - 29 are calculated as 30; and 80 - 110 as 79. However, the risk prediction equations were developed from a cohort of people aged 30 to 74 years. Using the argument allow.age will determine whether a function is applied to ages outside of 30 - 74. If allow.age is set to FALSE, then age values between 30 and 74 must be provided, else NA is returned as the risk score.
• by default return risk scores where the missing values NA is treated as 0, No, or FALSE. This applies only to binary variables including smoking status. If allow.na is set to FALSE, then binary inputs including smoking status must have a value, else NA will be returned as the risk score.

library(PredictRiskScores)

?NoPriorCVDRisk
?NoPriorCVDRisk_BMI
?NoPriorCVDRisk_Policy
?NoPriorCVDBleedRisk
?NoPriorT2DRisk
?PostCVDRisk
?PostACSRisk

• Billy Wu / [email protected]

I built a web-based risk calculator using the bleeding risk equation developed by Selak et al (2019) and the general CVD risk equation developed by Pylypchuk et al (2018), to provides clinicians with an individualised estimate of the CVD benefit and bleeding harms of aspirin for their patients without established CVD. Full Article.

Link to the risk calculator here

• Estimate 5 year CVD risk in people with severe mental illness
• Estimate 5 year diabetes risk in people without established diabetes
• Estimate 5 year acute CVD risk in people with established ACS
• Estimate 5 year acute bleeding risk in people with established ACS