Question:
I am trying to calculate the number of assays a given patient has within pre-specified time periods. E.g., between 14 and 45 days after a patient receives a third dose of vaccine, how many assays were taken? However, I don’t want to include any assays taken after the patient receives a fourth dose of vaccine.My dataset is in long format and contains a variable indicating each date that an assay was completed, and variables for the date of each vaccination. Below is a contrived example of my data frame.
My first thought was to use
case_when to make a flag for the cases in which the assay_date is between 14 and 45 days after the dose_3_date, but not after the dose_4_date, and then sum the flags somehow. Below is what I’ve written so far:df %>% mutate(post = case_when(assay_date >= dose_3_date+14 & assay_date <= dose_3_date+45 & assay_date <= dose_4_date & !is.na(dose_4_date) ~ 1), post3 = case_when(assay_date >= dose_3_date+60 & assay_date <= dose_3_date+120 & assay_date <= dose_4_date & !is.na(dose_4_date) ~ 1), post6 = case_when(assay_date >= dose_3_date+135 & assay_date <= dose_3_date+210 & assay_date <= dose_4_date & !is.na(dose_4_date) ~ 1)) [/code]
The above code works well for patients with a dose_4_date, but results in NA values for those with a “missing” dose_4_date. I’m unsure how I can ignore the NAs for patients with a missing dose_4_date.I’m also unsure how to sum the flags afterward.
Any advice would be greatly appreciated!
Answer:
library(data.table)
# dummy data
df <- data.table(id = rep(c(1,2), times=c(4,3))
, assay_date = c('20mar2021', '06jun2021', '24sep2021', '19nov2021', '29apr2021', '23may2021', '15jun2021')
, dose_3_date = rep(c('22feb2021', '02apr2021'), times=c(4,3))
, dose_4_date = c(rep(c('17aug2021', NA), times=c(4,3)))
); df
# set as data.table if yours isn't one already
setDT(df)
# as.Date
x <- c("assay_date", "dose_3_date", "dose_4_date")
df[, (x) := lapply(.SD, \(i) as.Date(i, format="%d%b%Y")), .SDcols=x
][, date_diff := assay_date - dose_3_date # calculate date diff
]
# flag rows which fit criteria
df[date_diff %between% c(14, 45)
& (assay_date <= dose_4_date
| is.na(dose_4_date)
)
, fits_criteria := 1
]
# count per patient
df[, .(assays_in_period = sum(fits_criteria, na.rm=T)), id]
id assays_in_period
1: 1 1
2: 2 1
[/code]
If you have better answer, please add a comment about this, thank you!