extract_medical_history: Extract medical history of subjects per healthcare provider

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

Description

This function spread medical histories into different columns.

Usage

extract_medical_history(population, n_batch = 50, cl = 1)

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.
n_batch	Number of batch, a non-negative integer of number of batch in which a number of subjects per provider are grouped. This is intended to ensure the computation fits to the memory.
cl	Parallel cluster, a non-negative integer of number of CPU cluster used for computation in parallel. Set to 1 if no parallelism is expected.

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

Target population data, a data frame with rows for visits and standardized columns (please see Details above), except code. Each code are spread as a column consisting the number of days from the latest admission_date on which this code have ever been encountered to admission_date of each visit. If no encounter is found, then the value returns NA to denote censored data. This is because the code may be encountered before db_start_date. If the latest date and that of each visit are the same, then the value returns 0.

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 and sample some of them
outcome=
  extract_outcome(population,'O1[4-5]',min,-1,'Z3[3-7]',max,0) %>%
  group_by(outcome) %>%
  slice(sample(seq(n()),ceiling(n()*0.0125),F)) %>%
  ungroup()

## Filter medical history before the date of either event or non-event
input=
  outcome %>%
  right_join(population,by='subject_id') %>%
  select(visit_id, everything()) %>%
  filter(admission_date<latest_date) %>%
  select(-outcome,-latest_date)

## Extract medical history of subjects per healthcare provider
mh_table=extract_medical_history(input)

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