Identifying change in variable by time in R - r

Question

I am looking at how extubation rates in an intensive care unit have changed over the course of the pandemic.
I have a data set which has hourly timestamps next to a category of airway types which simplified looks like this:
Time
AirwayStatus
2020/01/01 00:00
ETT/LMA
2020/01/01 01:00
ETT/LMA
2020/01/01 02:00
Own Airway
2020/01/01 03:00
Own Airway
2020/01/01 04:00
ETT/LMA
What I am effectively looking to do is find the times when the patient is extubated (ETT/LMA turns to Own Airway) and also when intubated (own airway to ETT/LMA). Eventually I want to be able to see how often an extubated patient has to be re-intubated.
Within 48 hours this is known as a failed extubation and we are expecting to see vastly different data during the pandemic compared to before.
The ideas I have so far are creating a seperate column with the airwayStatus of the prior hour and then if these are not the same then counting this. This seems unsophisticated though and I was hoping some of you clever people may have a nicer option.
Thank you in advance

Answer 1

Using dplyr from tidyverse:
Supposing you have a dataframe (or tibble) df and patient(?) id ID:
library(dplyr)
df <- tibble(
ID = c(1,1,1,1,1),
Time = c("2020/01/01 00:00", "2020/01/01 01:00", "2020/01/01 02:00", "2020/01/01 03:00", "2020/01/01 04:00"),
AirwayStatus = c("ETT/LMA", "ETT/LMA", "Own Airway", "Own Airway", "ETT/LMA"))
df <- df %>%
group_by(ID) %>%
arrange(Time) %>%
mutate(
Extubated = ifelse(AirwayStatus == "Own Airway" & lag(AirwayStatus) == "ETT/LMA", TRUE, FALSE),
Intubated = ifelse(AirwayStatus == "ETT/LMA" & lag(AirwayStatus) == "Own Airway", TRUE, FALSE))
result <- df %>%
summarise_at(c("Extubated", "Intubated"), sum, na.rm = TRUE)
result
Result:
# A tibble: 1 x 3
ID Extubated Intubated
<dbl> <int> <int>
1 1 1 1
This allows grouping by patient id which you will most likely do.
It's a bit longer than Oliver's answer though.

Answer 2

Your idea is the right way to go. You can skip storing intermediary results but they have to be estimated anyway. Lets assume your data is called df, then we could do something similar to
# Read table: (Could get read.table to work)
library(data.table)
df <- fread("Time AirwayStatus
2020/01/01 00:00 ETT/LMA
2020/01/01 01:00 ETT/LMA
2020/01/01 02:00 Own Airway
2020/01/01 03:00 Own Airway
2020/01/01 04:00 ETT/LMA")
setDF(df)
# Convert time to a date format
df$Time <- as.POSIXct(df$Time)
n <- nrow(df)
# Find changes
df$change <- with(df, c(FALSE, AirwayStatus[seq(n - 1)] != AirwayStatus[seq(2, n)]))
# estimate the length of time since last change
df$hours_between_change[df$change] <- with(df, diff(c(NA, Time[change])) / 3600)
df
Time AirwayStatus change hours_between_change
1 2020-01-01 00:00:00 ETT/LMA FALSE NA
2 2020-01-01 01:00:00 ETT/LMA FALSE NA
3 2020-01-01 02:00:00 Own Airway TRUE NA
4 2020-01-01 03:00:00 Own Airway FALSE NA
5 2020-01-01 04:00:00 ETT/LMA TRUE 2
Note I store the intermediate results here. We likely could make it a bit more readable using dplyr but this does the job.

Answer 3

Here is an approach using dplyr.
First, you might want to consider a separate column to indicate an intubation or extubation "event." If someone is "Own Airway" and then the previous row has "ETT/LMA", we assume the person has been extubated. The opposite can also be determined for intubation.
Then, you can filter and only focus on these events.
For each event, you may want to capture when the event is "Extubation", and then following event is "Intubation", and the time difference is < 48 hrs. If this is true, then the extubation is actually a "failed extubation."
This may handle situations where someone has data that begins with "Own Airway" and gets intubated (if no extubation event, then cannot be failed extubation). It will also keep extubation events where the time difference is > 48 hrs as well.
library(tidyverse)
df %>%
mutate(Event = case_when(
AirwayStatus == "Own Airway" & lag(AirwayStatus) == "ETT/LMA" ~ "Extubation",
AirwayStatus == "ETT/LMA" & lag(AirwayStatus) == "Own Airway" ~ "Intubation",
TRUE ~ NA_character_)
) %>%
filter(!is.na(Event)) %>%
mutate(Event = ifelse(
Event == "Extubation" & lead(Event) == "Intubation" & (lead(Time) - Time < 48),
"Failed Extubation",
Event
))
Output
Time AirwayStatus Event
1 2020-01-01 02:00:00 Own Airway Failed Extubation
2 2020-01-01 04:00:00 ETT/LMA Intubation
Data
df <- structure(list(Time = structure(c(1577858400, 1577862000, 1577865600,
1577869200, 1577872800), class = c("POSIXct", "POSIXt"), tzone = ""),
AirwayStatus = c("ETT/LMA", "ETT/LMA", "Own Airway", "Own Airway",
"ETT/LMA"), Event = c(NA, NA, "Extubated", NA, "Intubated"
)), row.names = c(NA, -5L), class = "data.frame")