반응형
0. Reference
1. Sample dataset
2. ARR
3. Interpretation
0. Reference
- https://en.wikipedia.org/wiki/Risk_difference
Risk difference - Wikipedia
From Wikipedia, the free encyclopedia The adverse outcome (black) risk difference between the group exposed to the treatment (left) and the group unexposed to the treatment (right) is −0.25 (RD = −0.25, ARR = 0.25). The risk difference (RD), excess ris
en.wikipedia.org
Confidence Interval for Two Independent Samples, Dichotomous Outcome
Confidence Interval for Two Independent Samples, Dichotomous Outcome It is common to compare two independent groups with respect to the presence or absence of a dichotomous characteristic or attribute, (e.g., prevalent cardiovascular disease or diabetes, c
sphweb.bumc.bu.edu
1. Sample dataset
library(dplyr)
library(data.table)
set.seed(230706)
# toy data
dt<-data.table(
pid = sample(1:5000, 2000, replace = F),
mace = sample(0:1, 2000, replace = T, prob = c(0.8,0.2)),
mace_sub = sample(c("acute_mi","is","cr","death","cvd_death"), 2000, replace = T, prob = rep(0.2,5)),
dm = sample(0:1, 2000, replace = T, prob = c(0.8,0.2)),
dementia = sample(0:1, 2000, replace = T, prob = c(0.8,0.2)),
myalgia = sample(0:1, 2000, replace = T, prob = c(0.8,0.2)),
age = sample(0:99, 2000, replace = T),
group = sample(c("Atorvastatin","Pravastatin"), 2000, replace = T, prob = c(0.8,0.2)),
from = sample(c("0","1","2","3","4","5","6","7","8"), 2000, replace = T),
index_year = sample(2006:2022, 2000, replace = T),
event_tars_year = sample(0:15, 2000, replace = T)
)
glimpse(dt)
2. ARR
dt2 <- dt %>%
mutate(
is = ifelse(mace_sub=="is",1,0),
acute_mi = ifelse(mace_sub=="acute_mi",1,0),
cr = ifelse(mace_sub=="cr",1,0),
cvd_death = ifelse(mace_sub=="cvd_death",1,0),
death = ifelse(mace_sub=="death",1,0)
)
# risk difference custom query
risk_difference <- function(dt, outcome){
# 0: atorvastatin, 1: pravastatin
exposed <- dt$group
outcome <- dt[,get(outcome)]
# 위험 차이 계산
risk_exposed <- sum(exposed == "Pravastatin" & outcome == 1) / sum(exposed == "Pravastatin")
risk_unexposed <- sum(exposed == "Atorvastatin" & outcome == 1) / sum(exposed == "Atorvastatin")
risk_difference <- risk_exposed - risk_unexposed
# 95% 신뢰구간 계산
n_exposed <- sum(exposed == "Pravastatin")
n_unexposed <- sum(exposed == "Atorvastatin")
se_difference <- sqrt(risk_exposed * (1 - risk_exposed) / n_exposed + risk_unexposed * (1 - risk_unexposed) / n_unexposed)
z_value <- qnorm(0.975) # 95% 신뢰수준에서의 Z값
ci <- risk_difference + c(-1, 1) * z_value * se_difference
return(paste0(round(risk_difference,3)," [",round(ci[1],3),";",round(ci[2],3),"]"))
}
paste0("Test:",risk_difference(dt,'mace'))
dt_rd_all = data.table(sub_name = "MACE",Risk_Diff = risk_difference(dt,"mace"))
for (i in 1:length(unique(dt2$mace_sub))) {
sub_name = unique(dt2$mace_sub)[i]
Risk_Diff = risk_difference(dt2, sub_name)
dt_rd_sub = data.table(sub_name, Risk_Diff)
dt_rd_all = rbind(dt_rd_all, dt_rd_sub)
}
(dt_rd_all)
3. Interpretation
- Risk Diff 값 (%) 만큼의 절대 발생률 차이가 있다.
- CI가 0.0을 포함한다면 통계적으로 유의하지 않다.
