airway_18 {nemsqar} | R Documentation |
Airway-18 Calculation
Description
This function processes and analyzes the dataset to calculate the "Airway-18" NEMSQA metric. It includes cleaning and transforming several columns related to patient data, airway procedures, and vital signs, and it returns a cleaned dataset with the relevant calculations. The final calculation is an assessment of the successful last invasive airway procedures performed during an EMS response originating from a 911 request in which waveform capnography is used for tube placement confirmation.
Usage
airway_18(
df = NULL,
patient_scene_table = NULL,
procedures_table = NULL,
vitals_table = NULL,
airway_table = NULL,
response_table = NULL,
erecord_01_col,
incident_date_col = NULL,
patient_DOB_col = NULL,
epatient_15_col,
epatient_16_col,
eresponse_05_col,
eprocedures_01_col,
eprocedures_02_col,
eprocedures_03_col,
eprocedures_06_col,
eairway_02_col = NULL,
eairway_04_col = NULL,
evitals_01_col,
evitals_16_col,
confidence_interval = FALSE,
method = c("wilson", "clopper-pearson"),
conf.level = 0.95,
correct = TRUE,
...
)
Arguments
Value
A data.frame summarizing results for two population groups (Adults and Peds) with the following columns:
-
pop
: Population type (Adults and Peds). -
numerator
: Count of incidents meeting the measure. -
denominator
: Total count of included incidents. -
prop
: Proportion of incidents meeting the measure. -
prop_label
: Proportion formatted as a percentage with a specified number of decimal places. -
lower_ci
: Lower bound of the confidence interval forprop
(ifconfidence_interval = TRUE
). -
upper_ci
: Upper bound of the confidence interval forprop
(ifconfidence_interval = TRUE
).
Author(s)
Nicolas Foss, Ed.D., MS, Samuel Kordik, BBA, BS
Examples
# If you are sourcing your data from a SQL database connection
# or if you have your data in several different tables,
# you can pass table inputs versus a single data.frame or tibble
# create tables to test correct functioning
# patient table
patient_table <- tibble::tibble(
erecord_01 = rep(c("R1", "R2", "R3", "R4", "R5"), 2),
incident_date = rep(as.Date(c("2025-01-01", "2025-01-05", "2025-02-01",
"2025-01-01", "2025-06-01")), 2),
patient_dob = rep(as.Date(c("2000-01-01", "2020-01-01", "2023-02-01",
"2023-01-01", "1970-06-01")), 2),
epatient_15 = rep(c(25, 5, 2, 2, 55), 2), # Ages
epatient_16 = rep(c("Years", "Years", "Years", "Years", "Years"), 2)
)
# response table
response_table <- tibble::tibble(
erecord_01 = rep(c("R1", "R2", "R3", "R4", "R5"), 2),
eresponse_05 = rep(2205001, 10)
)
# vitals table
vitals_table <- tibble::tibble(
erecord_01 = rep(c("R1", "R2", "R3", "R4", "R5"), 2),
evitals_01 = lubridate::as_datetime(c("2025-01-01 23:02:00",
"2025-01-05 12:03:00", "2025-02-01 19:04:00", "2025-01-01 05:05:00",
"2025-06-01 13:01:00", "2025-01-01 23:02:00",
"2025-01-05 12:03:00", "2025-02-01 19:04:00", "2025-01-01 05:05:00",
"2025-06-01 13:06:00")),
evitals_16 = rep(c(5, 6, 7, 8, 9), 2)
)
# airway table
airway_table <- tibble::tibble(
erecord_01 = rep(c("R1", "R2", "R3", "R4", "R5"), 2),
eairway_02 = rep(lubridate::as_datetime(c("2025-01-01 23:05:00",
"2025-01-05 12:02:00", "2025-02-01 19:03:00", "2025-01-01 05:04:00",
"2025-06-01 13:06:00")), 2),
eairway_04 = rep(4004019, 10)
)
# procedures table
procedures_table <- tibble::tibble(
erecord_01 = rep(c("R1", "R2", "R3", "R4", "R5"), 2),
eprocedures_01 = rep(lubridate::as_datetime(c("2025-01-01 23:00:00",
"2025-01-05 12:00:00", "2025-02-01 19:00:00", "2025-01-01 05:00:00",
"2025-06-01 13:00:00")), 2),
eprocedures_02 = rep("No", 10),
eprocedures_03 = rep(c(16883004, 112798008, 78121007, 49077009,
673005), 2),
eprocedures_06 = rep(9923003, 10)
)
# Run the function
# Return 95% confidence intervals using the Wilson method
airway_18(df = NULL,
patient_scene_table = patient_table,
procedures_table = procedures_table,
vitals_table = vitals_table,
response_table = response_table,
airway_table = airway_table,
erecord_01_col = erecord_01,
incident_date_col = incident_date,
patient_DOB_col = patient_dob,
epatient_15_col = epatient_15,
epatient_16_col = epatient_16,
eresponse_05_col = eresponse_05,
eprocedures_01_col = eprocedures_01,
eprocedures_02_col = eprocedures_02,
eprocedures_03_col = eprocedures_03,
eprocedures_06_col = eprocedures_06,
evitals_01_col = evitals_01,
evitals_16_col = evitals_16,
eairway_02_col = eairway_02,
eairway_04_col = eairway_04,
confidence_interval = TRUE
)