Skip to contents

Derive ICU Binary Flag per Patient

Source: R/burden.R, R/yll.R
dot-derive_icu_binary.Rd

Collapses unit-type data (one row per drug test per patient) to a patient-level binary ICU indicator using the "ever in ICU" rule: if a patient has any row with an ICU-type location during that admission, ICU = 1; otherwise ICU = 0.

Collapses unit-type data to a patient-level binary ICU indicator using the "ever in ICU" rule. Returns integer 0/1 or a 3-level factor when missing data is high.

Usage

.derive_icu_binary(
  data,
  patient_id_col = "PatientInformation_id",
  unit_type_col = "unit_type",
  icu_values = c("ICU", "Intensive Care", "Critical Care", "PICU", "NICU"),
  missing_threshold = 0.1
)

.derive_icu_binary(
  data,
  patient_id_col = "PatientInformation_id",
  unit_type_col = "unit_type",
  icu_values = c("ICU", "Intensive Care", "Critical Care", "PICU", "NICU"),
  missing_threshold = 0.1
)

Arguments

data

Data frame at drug-test level.

patient_id_col

Character. Default "PatientInformation_id".

unit_type_col

Character. Default "unit_type".

icu_values

Character vector. ICU location labels.

missing_threshold

Numeric. Proportion above which a 3-level factor is returned. Default 0.10.

Value

Patient-level data frame: patient_id_col and ICU (integer 0/1, or factor "ICU"/"Ward"/"Unknown" when missing is high).

Patient-level data frame with patient_id_col and ICU.

Details

Missing unit-type values are coded as NA. When the proportion of patients with all unit-type rows missing exceeds missing_threshold, the function returns a three-level ordered factor ("ICU" / "Ward" / "Unknown") and emits a message recommending that the "Unknown" level be included in the model.