In johnmackintosh/patientcounter: Count Hospital Patients Quickly

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

patientcounter

How many patients were in the hospital at 10 AM yesterday?
How many were in during each 15 minute spell between 2pm and 6pm?
How many were in during the last week, by hour?

This package aims to make answering these questions easier and quicker.

No SQL? No problem!

If you have time in, time out, a unique patient identifier, and optionally, a grouping variable to track moves between departments, this package will tell you how many patients were 'IN' at any time, at whatever granularity you need.

Installation

patientcounter is not on CRAN yet, so install the development version from GitHub with:

# install.packages("remotes") # if not already installed
remotes::install_github("johnmackintosh/patientcounter")

Example

Obtain data for each individual patient, by hour, for each hour of their stay.
Note we are restricting the outputs to keep this readable

library(patientcounter)
patient_count <- interval_census(beds, 
identifier = 'patient',
admit = 'start_time', 
discharge = 'end_time', 
group_var = 'bed', 
time_unit = '1 hour', 
results = "patient", 
uniques = TRUE)

head(patient_count)

To obtain summary data for every hour, for all combined patient stays:

library(patientcounter)
patient_count_hour <- interval_census(beds, 
identifier = 'patient',
admit = 'start_time', 
discharge = 'end_time', 
group_var = 'bed', 
time_unit = '1 hour', 
results = "total", 
uniques = TRUE)

head(patient_count_hour)

Note that you also receive the base date and base hour for each interval to enable easier filtering of the results.

Grouped values

This example shows grouping results by bed and hour.
The first ten rows of the resulting grouped values are shown

library(patientcounter)
grouped <- interval_census(beds, 
identifier = 'patient',
admit = 'start_time', 
discharge = 'end_time', 
group_var = 'bed', 
time_unit = '1 hour',
results = "group", 
uniques = FALSE)

head(grouped[bed %chin% c('A', 'B')],10)

General Help

You must 'quote' your variables, for the time being at least..

Results

Set results to 'patient' for 1 row per patient per interval for each interval in the patient stay.
Set results to 'group' to get a count per group per interval.
Remember this will also be influenced by the 'uniques' argument - set it to FALSE to ensure each move in each location is counted.
Set results to 'total' for a summary of the data set - interval, base_hour and count.

Tracking moves within the same interval with 'uniques'

To count individual patients ONLY, leave 'uniques' at the default value of 'TRUE'.
To count patient moves between locations during intervals, set uniques to 'FALSE'. Patients who occupy beds in different locations during each interval are accounted for in each location. They will be counted at least twice during an interval - both in their initial location and their new location following a move.

Timezones

Everything is easier if you use "UTC" by default. You can attempt to coerce the final results yourself using lubridate::force_tz()

To find your system timezone:

Sys.timezone()

Time Unit

See ? seq.POSIXt for valid values

E.G. '1 hour', '15 mins', '30 mins'

Time Adjust

Want to count those in between 10:01 to 11:00? You can do that using 'time_adjust_period' - set it to 'start_min' and then set 'time_adjust_interval' to 1.

10:00 to 10:59?
Yes, that's possible as well - set 'time_adjust_period' to 'end_min' and set 'time_adjust_interval' as before. You can set these periods to any value, as long as it makes sense in relation to your chosen time_unit.

Here we adjust the start_time by 5 minutes

library(patientcounter)
patient_count_time_adjust <- interval_census(beds, 
identifier = 'patient',
admit = 'start_time', 
discharge = 'end_time', 
group_var = 'bed', 
time_unit = '1 hour', 
time_adjust_period = 'start_min',
time_adjust_value = 5,
results = "total", 
uniques = TRUE)

head(patient_count_time_adjust)