Determine first (weighted) isolates

Determine first (weighted) isolates of all microorganisms of every patient per episode and (if needed) per specimen type.

first_isolate(
  x,
  col_date = NULL,
  col_patient_id = NULL,
  col_mo = NULL,
  col_testcode = NULL,
  col_specimen = NULL,
  col_icu = NULL,
  col_keyantibiotics = NULL,
  episode_days = 365,
  testcodes_exclude = NULL,
  icu_exclude = FALSE,
  specimen_group = NULL,
  type = "keyantibiotics",
  ignore_I = TRUE,
  points_threshold = 2,
  info = interactive(),
  include_unknown = FALSE,
  ...
)

filter_first_isolate(
  x,
  col_date = NULL,
  col_patient_id = NULL,
  col_mo = NULL,
  ...
)

filter_first_weighted_isolate(
  x,
  col_date = NULL,
  col_patient_id = NULL,
  col_mo = NULL,
  col_keyantibiotics = NULL,
  ...
)

Arguments

x	a `data.frame` containing isolates.
col_date	column name of the result date (or date that is was received on the lab), defaults to the first column of with a date class
col_patient_id	column name of the unique IDs of the patients, defaults to the first column that starts with 'patient' or 'patid' (case insensitive)
col_mo	column name of the IDs of the microorganisms (see `as.mo()`), defaults to the first column of class `mo`. Values will be coerced using `as.mo()`.
col_testcode	column name of the test codes. Use `col_testcode = NULL` to not exclude certain test codes (like test codes for screening). In that case `testcodes_exclude` will be ignored.
col_specimen	column name of the specimen type or group
col_icu	column name of the logicals (`TRUE`/`FALSE`) whether a ward or department is an Intensive Care Unit (ICU)
col_keyantibiotics	column name of the key antibiotics to determine first weighted isolates, see `key_antibiotics()`. Defaults to the first column that starts with 'key' followed by 'ab' or 'antibiotics' (case insensitive). Use `col_keyantibiotics = FALSE` to prevent this.
episode_days	episode in days after which a genus/species combination will be determined as 'first isolate' again. The default of 365 days is based on the guideline by CLSI, see Source.
testcodes_exclude	character vector with test codes that should be excluded (case-insensitive)
icu_exclude	logical whether ICU isolates should be excluded (rows with value `TRUE` in column `col_icu`)
specimen_group	value in column `col_specimen` to filter on
type	type to determine weighed isolates; can be `"keyantibiotics"` or `"points"`, see Details
ignore_I	logical to determine whether antibiotic interpretations with `"I"` will be ignored when `type = "keyantibiotics"`, see Details
points_threshold	points until the comparison of key antibiotics will lead to inclusion of an isolate when `type = "points"`, see Details
info	print progress
include_unknown	logical to determine whether 'unknown' microorganisms should be included too, i.e. microbial code `"UNKNOWN"`, which defaults to `FALSE`. For WHONET users, this means that all records with organism code `"con"` (contamination) will be excluded at default. Isolates with a microbial ID of `NA` will always be excluded as first isolate.
...	parameters passed on to the `first_isolate()` function

Source

Methodology of this function is strictly based on:

M39 Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, 4th Edition, 2014, Clinical and Laboratory Standards Institute (CLSI). https://clsi.org/standards/products/microbiology/documents/m39/.

Value

A logical vector

Details

WHY THIS IS SO IMPORTANT
To conduct an analysis of antimicrobial resistance, you should only include the first isolate of every patient per episode (ref). If you would not do this, you could easily get an overestimate or underestimate of the resistance of an antibiotic. Imagine that a patient was admitted with an MRSA and that it was found in 5 different blood cultures the following week. The resistance percentage of oxacillin of all S. aureus isolates would be overestimated, because you included this MRSA more than once. It would be selection bias.

All isolates with a microbial ID of NA will be excluded as first isolate.

The functions filter_first_isolate() and filter_first_weighted_isolate() are helper functions to quickly filter on first isolates. The function filter_first_isolate() is essentially equal to:

 x %&gt;%
   mutate(only_firsts = first_isolate(x, ...)) %&gt;%
   filter(only_firsts == TRUE) %&gt;%
   select(-only_firsts)

The function filter_first_weighted_isolate() is essentially equal to:

 x %&gt;%
   mutate(keyab = key_antibiotics(.)) %&gt;%
   mutate(only_weighted_firsts = first_isolate(x,
                                               col_keyantibiotics = "keyab", ...)) %&gt;%
   filter(only_weighted_firsts == TRUE) %&gt;%
   select(-only_weighted_firsts)

Key antibiotics

There are two ways to determine whether isolates can be included as first weighted isolates which will give generally the same results:

Using type = "keyantibiotics" and parameter ignore_I

Any difference from S to R (or vice versa) will (re)select an isolate as a first weighted isolate. With ignore_I = FALSE, also differences from I to S|R (or vice versa) will lead to this. This is a reliable method and 30-35 times faster than method 2. Read more about this in the key_antibiotics() function.
Using type = "points" and parameter points_threshold

A difference from I to S|R (or vice versa) means 0.5 points, a difference from S to R (or vice versa) means 1 point. When the sum of points exceeds points_threshold, which default to 2, an isolate will be (re)selected as a first weighted isolate.

Stable lifecycle

The lifecycle of this function is stable. In a stable function, we are largely happy with the unlying code, and major changes are unlikely. This means that the unlying code will generally evolve by adding new arguments; we will avoid removing arguments or changing the meaning of existing arguments.

If the unlying code needs breaking changes, they will occur gradually. To begin with, the function or argument will be deprecated; it will continue to work but will emit an message informing you of the change. Next, typically after at least one newly released version on CRAN, the message will be transformed to an error.

Examples

# `example_isolates` is a dataset available in the AMR package.
# See ?example_isolates.

if (FALSE) {
library(dplyr)
# Filter on first isolates:
example_isolates %>%
  mutate(first_isolate = first_isolate(.)) %>%
  filter(first_isolate == TRUE)

# Now let's see if first isolates matter:
A <- example_isolates %>%
  group_by(hospital_id) %>%
  summarise(count = n_rsi(GEN),            # gentamicin availability
            resistance = resistance(GEN))  # gentamicin resistance

B <- example_isolates %>%
  filter_first_weighted_isolate() %>%      # the 1st isolate filter
  group_by(hospital_id) %>%
  summarise(count = n_rsi(GEN),            # gentamicin availability
            resistance = resistance(GEN))  # gentamicin resistance

# Have a look at A and B.
# B is more reliable because every isolate is counted only once.
# Gentamicin resistance in hospital D appears to be 3.7% higher than
# when you (erroneously) would have used all isolates for analysis.


## OTHER EXAMPLES:

# Short-hand versions:
example_isolates %>%
  filter_first_isolate()

example_isolates %>%
  filter_first_weighted_isolate()


# set key antibiotics to a new variable
x$keyab <- key_antibiotics(x)

x$first_isolate <- first_isolate(x)

x$first_isolate_weighed <- first_isolate(x, col_keyantibiotics = 'keyab')

x$first_blood_isolate <- first_isolate(x, specimen_group = "Blood")
}