When I tried to join and create days a difference between two dates (don’t have any missing value here) using difftime, there are some rows that have NA and others are just fine. There are two things that I notice:

1. Those NA also seem to be clustered by patient ID, so I am not sure what is going on,
2. If I run difftime separately from data table, it produced all difference just fine, no NA.

I expect the difftime does not produce NA since two date variables don’t have NA when running in data table.

## generate data 1
pat_characteristics <- function(n_group){
  patid <- seq(1:n_group)
  
  birth_date <- rdate(n_group, min = as.Date('01/01/1940',"%d/%m/%Y"), 
                      max = as.Date('31/12/2000',"%d/%m/%Y"))
  
  sex <- sample(c("F","M"), n_group, replace = TRUE)
  
  edu <- sample(c("AA","B","M","PHD","MD"), n_group, replace = TRUE)
  
  f_born <- sample(c("US-born","f-born", "Others"), n_group, replace = TRUE)
  
  state <- sample(state.abb, n_group, replace = TRUE)
  
  pat1 <- data.table(patid, birth_date, sex, edu, f_born, state)
  
  return(pat1)
}


## generate data 2
encounter <- function(n_group, n_each_group){
  patid <- rep(seq(1:n_group), each=n_each_group)
  enc_date <- rdate(n_group*n_each_group, min = as.Date('01/01/2014',"%d/%m/%Y"), 
                    max = as.Date('31/12/2020',"%d/%m/%Y"))
  ICD_code <- sample(1:100, n_group*n_each_group, replace=TRUE)
  
  pat2 <- data.table(patid, enc_date, ICD_code)
  
  return(pat2)
}


set.seed(1)
pat1 <- pat_characteristics(10)
enc1 <- encounter(10,5)


d <- enc1[pat1[birth_date > as.Date("01/01/1970", "%d/%m/%Y")], on="patid"
          ][sex=="F", age := enc_date- birth_date]
difftime(d$enc_date, d$birth_date)