Record linkage and deduplication of individual-level data, such as repeated spells in hospital, or recurrent cases of infection is a common task in epidemiological analysis and other fields of research.
The diyar package aims to provide a simple and flexible implementation of deterministic record linkage and episode grouping for the application of case definitions in epidemiological analysis.
# Install the latest CRAN release
install.packages("diyar")
# Or, install the development version from GitHub
install.packages("devtools")
devtools::install_github("OlisaNsonwu/diyar")
There are two main aspects of the diyar package; multistage record grouping (record_group()) and episode grouping (fixed_episodes(), rolling_episodes() and episode_group()) for applying case definitions in epidemiological analysis. number_line objects are used for both.
number_line objects - series of real numbers on a number line. These can be manipulated and merged.library(diyar)
l <- as.Date("01/04/2019", "%d/%m/%Y"); r <- as.Date("30/04/2019", "%d/%m/%Y")
nl <- number_line(l, r)
nl
#> [1] "2019-04-01 -> 2019-04-30"
reverse_number_line(nl)
#> [1] "2019-04-30 <- 2019-04-01"
shift_number_line(nl, -2)
#> [1] "2019-03-30 -> 2019-04-28"
expand_number_line(nl, 2)
#> [1] "2019-03-30 -> 2019-05-02"
number_line_sequence(nl, by =3)
#> [[1]]
#> [1] "2019-04-01" "2019-04-04" "2019-04-07" "2019-04-10" "2019-04-13"
#> [6] "2019-04-16" "2019-04-19" "2019-04-22" "2019-04-25" "2019-04-28"fixed_episodes(), rolling_episodes() and episode_group() - Group records into chronological episodes. epid objects are now the default output. Use to_s4 or to_df() to change this to data.framesdata(infections);
db <- infections[c("date")]
# Dates
db$date
#> [1] "2018-04-01" "2018-04-07" "2018-04-13" "2018-04-19" "2018-04-25"
#> [6] "2018-05-01" "2018-05-07" "2018-05-13" "2018-05-19" "2018-05-25"
#> [11] "2018-05-31"
# Fixed episodes
db$f_epid <- fixed_episodes(date = db$date, case_length = 15, display = F, group_stats = T)
#> Episode grouping complete: 0 record(s) with a unique ID.
# Rolling episodes
db$r_epid <- rolling_episodes(date = db$date, case_length = 15, recurrence_length = 40, display = F,
group_stats = T)
#> Episode grouping complete: 0 record(s) with a unique ID.
db[c("f_epid","r_epid")]
#> f_epid r_epid
#> 1 E.01 2018-04-01 -> 2018-04-13 (C) E.1 2018-04-01 -> 2018-05-31 (C)
#> 2 E.01 2018-04-01 -> 2018-04-13 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 3 E.01 2018-04-01 -> 2018-04-13 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 4 E.04 2018-04-19 -> 2018-05-01 (C) E.1 2018-04-01 -> 2018-05-31 (R)
#> 5 E.04 2018-04-19 -> 2018-05-01 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 6 E.04 2018-04-19 -> 2018-05-01 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 7 E.07 2018-05-07 -> 2018-05-19 (C) E.1 2018-04-01 -> 2018-05-31 (D)
#> 8 E.07 2018-05-07 -> 2018-05-19 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 9 E.07 2018-05-07 -> 2018-05-19 (D) E.1 2018-04-01 -> 2018-05-31 (D)
#> 10 E.10 2018-05-25 -> 2018-05-31 (C) E.1 2018-04-01 -> 2018-05-31 (D)
#> 11 E.10 2018-05-25 -> 2018-05-31 (D) E.1 2018-04-01 -> 2018-05-31 (D)record_group() - Perform multistage deterministic linkages while addressing missing data using a specified list of alternative matching criteria or matching range of values. pid objects are now the default output. Use to_s4 or to_df() to change this to a data.frames# Two stages of record grouping
data(staff_records);
staff_records$pids_a <- record_group(staff_records, sn = r_id, criteria = c(forename, surname),
data_source = sex, display = FALSE)
#> Record grouping complete: 1 record(s) with a unique ID.
staff_records
#> r_id forename surname sex dataset pids_a
#> 1 1 James Green M Staff list P.1 (CRI 02)
#> 2 2 <NA> Anderson M Staff list P.2 (CRI 02)
#> 3 3 Jamey Green M Pay slips P.1 (CRI 02)
#> 4 4 <NA> F Pay slips P.4 (No Hit)
#> 5 5 Derrick Anderson M Staff list P.2 (CRI 02)
#> 6 6 Darrack Anderson M Pay slips P.2 (CRI 02)
#> 7 7 Christie Green F Staff list P.1 (CRI 02)Find out more here!
Please report any bug or issues with using this package here.