extract_outcome: Extract outcome of subjects

In herdiantrisufriyana/medhist: A preprocessor to construct medical history table from data source

Description

This function makes a list of subjects with either event or non-event outcome, that visit healthcare providers up to the predefined latest date.

Usage

extract_outcome(
  population,
  icd10_event,
  latest_event = min,
  day_to_event = -1,
  icd10_nonevent,
  latest_nonevent = max,
  day_to_nonevent = 0,
  verbose = T
)

Arguments

population	Target population data, a data frame with rows for visits and standardized columns (please see Details below). If only a single healthcare provider, then make healthcare_id consisting only a single identification number.
icd10_event	Event code, a character of ICD-10 code for either diagnosis or procedure, that identify subjects with event outcome. This parameter should be a regular expression if detecting >1 ICD-10 codes. The detection is not exact but anywhere within the character sequence; thus, the code with more than k-mer will be extracted.
latest_event	Latest date of event, a function to choose admission_date with the icd10_event among many for each subject. This date will be adjusted with day_to_event to limit visits accounted for medical histories, since these are considerably before the outcome. The function is conceivably min() and max() for the earliest (preferred) and latest date, respectively.
day_to_event	Day to event, an integer to adjust latest_event for limiting visits accounted for medical histories. These are considerably before the outcome, i.e. up to -1 day.
icd10_nonevent	Non-event code, a character of ICD-10 code for either diagnosis or procedure, that identify subjects with non-event outcome. This parameter should be a regular expression if detecting >1 ICD-10 codes. Make this parameter as '' (character of none) if all except event codes are used. The detection is not exact but anywhere within the character sequence; thus, the code with more than k-mer will be extracted.
latest_nonevent	Latest date of non-event, a function to choose admission_date with the icd10_nonevent among many for each subject. This date will be adjusted with day_to_nonevent to limit visits accounted for medical histories, since these are considerably up to the day of the outcome. The function is conceivably min() and max() for the earliest and latest date (preferred), respectively.
day_to_nonevent	Day to non-event, an integer to adjust latest_nonevent for limiting visits accounted for medical histories. These are considerably up to the day of the outcome, i.e. up to 0 day.
verbose	Verbosity, a logical indicating whether progress should be shown.

Details

Target population data consisted visit_id,subject_id, healthcare_id, admission_date, code, and db_start_date. Columns of visit_id,subject_id, and healthcare_id are characters of identification numbers identifying respectively unique visits, subjects, and healthcare providers. Columns of admission_date and db_start_date are dates of subject admission in a visit and database start of recording period, respectively. Column of code is a character of ICD-10 code for either diagnosis or procedure regardless the number of digit or k-mer.

Value

Subject list, a data frame with rows for unique subjects and columns of subject_id, latest_date, and outcome. The last column is a factor of which non-event is the first class between non-event and event.

Examples

## Create input example
data(visit_cap)
data(visit_ffs)
data(visit_drg)
data(diagnosis)

population=
  list(visit_cap,visit_ffs,visit_drg) %>%
  lapply(select,visit_id,subject_id,healthcare_id,admission_date) %>%
  do.call(rbind,.) %>%
  left_join(diagnosis,by='visit_id') %>%
  filter(!code_type%in%c('Admission diagnosis')) %>%
  select(-code_type) %>%
  mutate(db_start_date=as.Date('2015-01-01')) %>%
  .[!duplicated(.),]

## Extract outcome of subjects
outcome=extract_outcome(population,'O1[4-5]',min,-1,'Z3[3-7]',max,0)

herdiantrisufriyana/medhist documentation built on June 24, 2021, 3:41 a.m.