Investigate care home cis markers (#989)

In the care home methodology update #945 the cis markers were working
but not for episodes nested within a stay, it would wrongly assign the
'next' episode counter. This was due to the lag function assigning NA to
the first row.

This update fixes the nested cases and i think it is working properly
now for all scenarios. Where the first row is NA for
`previous_discharge_date` this will use the `ch_discharge_date` which
will be the first episode.

We also use the following code, adapted from Social care scripts to
create the cis markers. This simplifies the code and seems to work for
all scenarios. e.g. when the first episode linked to a chi is NA and the
next two episodes are continuous, this will return 1,2,2 correctly.
Previously this would return 1,2,3.
```
mutate(continuous_marker = accumulate(new_episode[-1], .init = 1, ~ if (.y == "Y") {.x + 1} else {.x}))
```

R/process_sc_all_care_home.R

@@ -43,7 +43,7 @@ process_sc_all_care_home <- function(
         .data[["ch_discharge_date"]]
       )
     ) %>%
-    dplyr::full_join(sc_demog_lookup, # this is the correct join.
+    dplyr::right_join(sc_demog_lookup, # this is the correct join.
       by = c("sending_location", "social_care_id")
     ) %>%
     replace_sc_id_with_latest() %>%
@@ -201,7 +201,6 @@ process_sc_all_care_home <- function(
 
   # Compare to Deaths Data
   # match ch_episode data with deaths data
-  # TO DO should this be boxi nrs death dates instead of IT extract deaths?
   matched_deaths_data <- ch_episode %>%
     dplyr::left_join(refined_death,
       by = "chi"
@@ -237,56 +236,97 @@ process_sc_all_care_home <- function(
   # creates a CIS  flag for CHI across all of scotland
   # and a CIS for social care ID and sending location for just that LA
   ch_chi_markers <- matched_deaths_data %>%
-    # uses the chi to flag continuous stays. Will flag cases even if in another LA
+    # Group the data by chi
     dplyr::group_by(.data[["chi"]]) %>%
-    # create variable for previous discharge date + 1 day
-    dplyr::mutate(previous_discharge_date_chi = dplyr::lag(.data[["ch_discharge_date"]]) + lubridate::days(1L)) %>%
-    # TRUE/FALSE flag for if admission date is before or equal to previous discharge date + 1 day
-    dplyr::mutate(continuous_stay_flag_chi = tidyr::replace_na(.data[["ch_admission_date"]] <= previous_discharge_date_chi, FALSE)) %>%
-    # different to code in above sections.
-    # we want to uniquely identify all cases where the flag is FALSE. and only the first case where the flag is TRUE
-    # to do this create a variable of the flag in the previous row
-    dplyr::mutate(previous_continuous_stay_flag_chi = tidyr::replace_na(dplyr::lag(.data[["continuous_stay_flag_chi"]]), FALSE)) %>%
-    dplyr::mutate(continuous_stay_chi = ifelse(continuous_stay_flag_chi == FALSE |
-      (continuous_stay_flag_chi == TRUE & previous_continuous_stay_flag_chi == FALSE), FALSE, TRUE)) %>%
-    dplyr::group_by(
-      .data[["chi"]],
-      .data[["continuous_stay_chi"]]
+    # Set up previous_discharge_date
+    # The lag function will set the first row to NA.
+    dplyr::mutate(
+      # We want to flag the first episode per chi with row_number
+      row_number = dplyr::row_number(),
+      # create variable for previous discharge date + 1 day
+      previous_discharge_date_chi = dplyr::lag(.data[["ch_discharge_date"]]) +
+        lubridate::days(1L),
+      # if the first row is NA, set this to the ch_discharge_date
+      previous_discharge_date_chi = dplyr::if_else(row_number == 1, .data[["ch_discharge_date"]],
+        .data[["previous_discharge_date_chi"]]
+      )
     ) %>%
-    # gives cases their unique CIS identifier
-    dplyr::mutate(ch_chi_cis = ifelse(continuous_stay_chi == FALSE, dplyr::row_number(), NA)) %>%
-    dplyr::group_by(
-      .data[["social_care_id"]],
-      .data[["sending_location"]]
+    # flag continuous stays and create marker
+    # calculate number of days between start_date and end_date on the previous episode
+    dplyr::mutate(
+      days_to_next_rec = floor(
+        lubridate::time_length(lubridate::interval(
+          .data[["previous_discharge_date_chi"]],
+          .data[["ch_admission_date"]]
+        ), "days")
+      ),
+      # if there is more than 1 day between (or the last ep for the individual) flag as new ep (Y)
+      # if there is < 1 day (i.e. a pause of up to 1 day or stays overlap flag as same ep (N))
+      new_episode = dplyr::if_else(is.na(days_to_next_rec) | days_to_next_rec > 1, "Y", "N")
+    ) %>%
+    # create continuous marker using flag for new stay
+    dplyr::mutate(
+      ch_chi_cis = purrr::accumulate(new_episode[-1],
+        .init = 1,
+        ~ if (.y == "Y") {
+          .x + 1
+        } else {
+          .x
+        }
+      )
     ) %>%
-    # fills in CIS identifier for all cases
-    tidyr::fill(ch_chi_cis, .direction = c("down"))
+    dplyr::ungroup()
 
 
   # This is the same but uses the social care id and sending location so can be used for
   # episodes that are not attached to a CHI number
   # This will restrict continuous stays to each Local Authority
   sc_ch_id_markers <- ch_chi_markers %>%
+    # uses social_care_id and sending_location to flag continuous stays.
+    # Will flag cases even if in another LA
     dplyr::group_by(.data[["social_care_id"]], .data[["sending_location"]]) %>%
-    # create variable for previous discharge date + 1 day
-    dplyr::mutate(previous_discharge_date_sc = dplyr::lag(.data[["ch_discharge_date"]]) + lubridate::days(1L)) %>%
-    # TRUE/FALSE flag for if admission date is before or equal to previous discharge date + 1 day
-    dplyr::mutate(continuous_stay_flag_sc = tidyr::replace_na(.data[["ch_admission_date"]] <= previous_discharge_date_sc, FALSE)) %>%
-    # we want to uniquely identify all cases where the flag is FALSE. and only the first case where the flag is TRUE
-    # to do this create a variable of the flag in the previous row
-    dplyr::mutate(previous_continuous_stay_flag_sc = tidyr::replace_na(dplyr::lag(.data[["continuous_stay_flag_sc"]]), FALSE)) %>%
-    dplyr::mutate(continuous_stay_sc = ifelse(continuous_stay_flag_sc == FALSE |
-      (continuous_stay_flag_sc == TRUE & previous_continuous_stay_flag_sc == FALSE), FALSE, TRUE)) %>%
-    dplyr::group_by(.data[["social_care_id"]], .data[["sending_location"]], .data[["continuous_stay_sc"]]) %>%
-    # gives cases their unique CIS identifier
-    dplyr::mutate(ch_sc_id_cis = ifelse(continuous_stay_sc == FALSE, dplyr::row_number(), NA)) %>%
-    dplyr::group_by(.data[["social_care_id"]], .data[["sending_location"]]) %>%
-    # fills in CIS identifier for all cases
-    tidyr::fill(ch_sc_id_cis, .direction = c("down")) %>%
+    # Set up previous_discharge_date
+    # The lag function will set the first row to NA.
+    dplyr::mutate(
+      # We want to flag the first episode per sc id and sending_location with row_number
+      row_number = dplyr::row_number(),
+      # create variable for previous discharge date + 1 day
+      previous_discharge_date_sc = dplyr::lag(.data[["ch_discharge_date"]]) +
+        lubridate::days(1L),
+      # if the first row is NA, set this to the ch_discharge_date
+      previous_discharge_date_sc = dplyr::if_else(row_number == 1, .data[["ch_discharge_date"]],
+        .data[["previous_discharge_date_sc"]]
+      )
+    ) %>%
+    # flag continuous stays and create marker
+    # calculate number of days between start_date and end_date on the previous episode
+    dplyr::mutate(
+      days_to_next_rec = floor(
+        lubridate::time_length(lubridate::interval(
+          .data[["previous_discharge_date_sc"]],
+          .data[["ch_admission_date"]]
+        ), "days")
+      ),
+      # if there is more than 1 day between (or the last ep for the individual) flag as new ep (Y)
+      # if there is < 1 day (i.e. a pause of up to 1 day or stays overlap flag as same ep (N))
+      new_episode = dplyr::if_else(is.na(days_to_next_rec) | days_to_next_rec > 1, "Y", "N")
+    ) %>%
+    # create continuous marker using flag for new stay
+    dplyr::mutate(
+      ch_sc_id_cis = purrr::accumulate(new_episode[-1],
+        .init = 1,
+        ~ if (.y == "Y") {
+          .x + 1
+        } else {
+          .x
+        }
+      )
+    ) %>%
+    dplyr::ungroup() %>%
+    # remove variables no longer needed
     dplyr::select(
-      -previous_discharge_date_chi, -continuous_stay_flag_chi, -previous_continuous_stay_flag_chi, -continuous_stay_chi,
-      -previous_discharge_date_sc, -continuous_stay_flag_sc, -previous_continuous_stay_flag_sc, -continuous_stay_sc,
-      -dis_after_death
+      -previous_discharge_date_chi, -previous_discharge_date_sc, -row_number,
+      -days_to_next_rec, -new_episode
     )
 
 
@@ -361,6 +401,7 @@ process_sc_all_care_home <- function(
       "ch_provider_description",
       "ch_nursing",
       "ch_adm_reason",
+      "type_of_admission",
       "sc_latest_submission"
     ) %>%
     slfhelper::get_anon_chi()