Duration Hospitalisation of a patient

[michaelwib]

Hello,

I was wondering if I could get some help on this: In the transfers table, the physical location of patient is given, while the admissions table denotes when a patient has been admitted to the hospital. But why do the times not line up, that is why is the admittime not the same as the intime in transfers for the first hospital ward?

For example, hadm_id = 20020072 has an admittime at 07:06:00, but in transfers the first ward after the ED is at 08:08:00.
That is, if I am interested in the duration of the hospitalisation, what would be the appropriate time interval to consider?

[Bright381]

Hi,
The reason is that the transfers table only tracks changes in patient location (see the documentation). In this case, the patient likely remained in the ED until 08:08:00. If you look in the services table, you'll see that the transfer time for this patient aligns with their admission time.

[michaelwib]

Thank you so much for your reply. I also thought about using the services table, but it does to always seem to be the case that the admission time aligns with the transfer time in services. For example, hadm_id = 29904280 we have an admission time at 07:15:00, and a transfer time in services at 18:46:00. This aligns with the earliest entry in the transfers table (Medicine/Cardiology ward).

That is, I don't really understand the 11 hour gap between the admission time and the service/first ward time in this case.

[Bright381]

I initially thought this might be linked to the admission_type, since in this case it appears to be related to a planned surgery.

However, when analyzing patients across different gap durations between admittime and their first transfertime — whether short (≤ 1 hour) or longer — it seems that admission_type alone doesn't fully explain the variation. Other factors may be influencing the timing.

Here’s the R code I used:

mimiciv$hosp$admissions %>%
  left_join(
    mimiciv$hosp$services %>%
      group_by(hadm_id) %>%
      summarise(transfertime=min(transfertime)) %>%
      ungroup,
    by='hadm_id'
  ) %>% 
  select(admittime, admission_type, transfertime) %>%
  collect %>%
  count(
    (difftime(admittime, transfertime, units='hours')<1 & difftime(admittime, transfertime, units='hours')>-1),
    admission_type
  )

which returns:

Edit: Also I am extremely skeptical regarding the reliability of admission_type and admission_location, since many patients recorded in ed.edstays do not have ER as an admission_location and ED-related admission_type, and vice-versa.