Viv3ckj/move validation data

analysis/config.py

@@ -12,4 +12,8 @@
 
 # measures_definition_pf_medications.py
 start_date_measure_medications = "2023-11-01"
-monthly_intervals_measure_medications = 9
+monthly_intervals_measure_medications = 9
+
+# measures_definition_pf_consultation_pf_counts.py
+start_date_measure_med_counts = "2024-02-01"
+monthly_intervals_measure_med_counts = 6

analysis/measures_definition_pf_consultation_med_counts.py

@@ -0,0 +1,72 @@
+from ehrql import INTERVAL, create_measures, months
+from ehrql.tables.tpp import (
+    patients,
+    clinical_events,
+    practice_registrations,
+)
+from ehrql.tables.raw.tpp import medications
+
+from config import start_date_measure_med_counts, monthly_intervals_measure_med_counts
+from codelists import (
+    pharmacy_first_consultation_codelist,
+    pharmacy_first_med_codelist,
+)
+from pf_variables_library import select_events
+
+# Script taken from Pharmacy First Data Development (for top 10 PF meds table)
+
+measures = create_measures()
+measures.configure_dummy_data(population_size=1000)
+
+start_date = start_date_measure_med_counts
+monthly_intervals = monthly_intervals_measure_med_counts
+
+registration = practice_registrations.for_patient_on(INTERVAL.end_date)
+
+# Select Pharmacy First events during interval date range
+pharmacy_first_events = select_events(
+    clinical_events, 
+    start_date=INTERVAL.start_date, 
+    end_date=INTERVAL.end_date).where(
+        clinical_events.snomedct_code.is_in(
+            pharmacy_first_consultation_codelist
+        )
+    )
+
+pharmacy_first_ids = pharmacy_first_events.consultation_id
+has_pharmacy_first_consultation = pharmacy_first_events.exists_for_patient()
+
+# Select Pharmacy First consultations during interval date range
+selected_medications = select_events(
+    medications,
+    start_date=INTERVAL.start_date, end_date=INTERVAL.end_date
+).where(medications.consultation_id.is_in(pharmacy_first_ids))
+
+# First medication for each patient
+first_selected_medication = (
+    selected_medications.sort_by(selected_medications.date).first_for_patient().dmd_code
+)
+# Boolean variable that selected medication is part of pharmacy first med codelists
+has_pharmacy_first_medication = first_selected_medication.is_in(pharmacy_first_med_codelist)
+
+# Numerator, patients with a PF medication
+# This allows me to count all (first) medications linked to a PF consultation
+numerator = first_selected_medication.is_not_null()
+
+# Denominator, registered patients (f/m) with a PF consultation
+denominator = (
+    registration.exists_for_patient()
+    & patients.sex.is_in(["male", "female"])
+    & has_pharmacy_first_consultation
+)
+
+measures.define_measure(
+    name="pf_medication_count",
+    numerator = first_selected_medication.is_not_null(),
+    denominator=denominator,
+    group_by={
+        "dmd_code": first_selected_medication,
+        "pharmacy_first_med": has_pharmacy_first_medication,
+    },
+    intervals=months(monthly_intervals).starting_on(start_date),
+)

analysis/measures_definition_pf_descriptive_stats.py

@@ -2,7 +2,7 @@
 from ehrql.tables.raw.tpp import medications
 from ehrql.tables.tpp import practice_registrations, patients, clinical_events
 
-from pf_variables_library import select_events_from_codelist, select_events_by_consultation_id
+from pf_variables_library import select_events
 from codelists import (
     pharmacy_first_med_codelist,
     pharmacy_first_consultation_codelist,
@@ -20,24 +20,42 @@
 registration = practice_registrations.for_patient_on(INTERVAL.end_date)
 
 # Function to retrieve consultation ids from clinical events that are PF consultations
-pharmacy_first_ids = select_events_from_codelist(
-    clinical_events, pharmacy_first_consultation_codelist
+pharmacy_first_ids = select_events(
+    clinical_events, codelist=pharmacy_first_consultation_codelist
 ).consultation_id
 
 # Function to retrieve selected events using pharmacy first ids
-selected_clinical_events = select_events_by_consultation_id(
-    clinical_events, pharmacy_first_ids
+selected_clinical_events = select_events(
+    clinical_events, consultation_ids=pharmacy_first_ids
 ).where(clinical_events.date.is_on_or_between(INTERVAL.start_date, INTERVAL.end_date))
 
-selected_med_events = select_events_by_consultation_id(medications, pharmacy_first_ids).where(
+selected_med_events = select_events(
+    medications, consultation_ids=pharmacy_first_ids).where(
     medications.date.is_on_or_between(INTERVAL.start_date, INTERVAL.end_date)
 )
 
+selected_clinical_pathways = select_events(
+    clinical_events, codelist=pharmacy_first_conditions_codelist
+).where(clinical_events.date.is_on_or_between(INTERVAL.start_date, INTERVAL.end_date))
+
 # Create variable which contains boolean values of whether pharmacy first event exists for patient
-has_pf_consultation = select_events_from_codelist(selected_clinical_events, pharmacy_first_consultation_codelist).exists_for_patient()
+has_pf_consultation = select_events(
+    selected_clinical_events, codelist=pharmacy_first_consultation_codelist).exists_for_patient()
 
 # PF consultations with PF clinical condition
-has_pf_condition = select_events_from_codelist(selected_clinical_events, pharmacy_first_conditions_codelist).exists_for_patient()
+has_pf_condition = select_events(
+    selected_clinical_events, codelist=pharmacy_first_conditions_codelist).exists_for_patient()
+
+# Dates of pharmacy first consultations
+pharmacy_first_dates = select_events(clinical_events, codelist=pharmacy_first_consultation_codelist).date
+
+# Specify whether a patient has been prescribed a PF medication on the same day as a PF consultation code
+has_pfmed_on_pfdate = selected_med_events.where(medications.date.is_in(pharmacy_first_dates)).exists_for_patient()
+
+# Specify whether a patient has a PF condition
+has_pfpathway_on_pfdate = selected_clinical_pathways.where(
+    selected_clinical_pathways.date.is_in(pharmacy_first_dates)
+).exists_for_patient()
 
 # PF consultations with prescribed PF medication
 has_pf_medication = selected_med_events.where(
@@ -52,6 +70,12 @@
 )
 measures.define_defaults(denominator=denominator)
 
+# Denominator without the pf_consultation constraint
+study_population = (
+    registration.exists_for_patient()
+    & patients.sex.is_in(["male", "female"])
+)
+
 # Measures for PF consultations with PF medication
 measures.define_measure(
     name="pf_with_pfmed",
@@ -71,3 +95,27 @@
     numerator=has_pf_condition & has_pf_medication,
     intervals=months(monthly_intervals).starting_on(start_date),
 )
+
+# Measures for PF medications prescribed on the same day as PF consultation 
+measures.define_measure(
+    name="pfmed_on_pfdate",
+    numerator=has_pfmed_on_pfdate,
+    denominator=study_population,
+    intervals=months(monthly_intervals).starting_on(start_date),
+)
+
+# Measures for PF conditions diagnosed on the same day as PF consultation 
+measures.define_measure(
+    name="pfpathway_on_pfdate",
+    numerator=has_pfpathway_on_pfdate,
+    denominator=study_population,
+    intervals=months(monthly_intervals).starting_on(start_date),
+)
+
+# Measures for PF conditions diagnosed and PF med prescribed on the same day as PF consultation 
+measures.define_measure(
+    name="pfmed_and_pfpathway_on_pfdate",
+    numerator=has_pfmed_on_pfdate & has_pfpathway_on_pfdate,
+    denominator=study_population,
+    intervals=months(monthly_intervals).starting_on(start_date),
+)

lib/functions/create_tables.R

@@ -129,3 +129,42 @@ create_clinical_conditions_codes_table <- function(title) {
       heading.subtitle.font.size = "small"
     )
 }
+
+#  Create top 5 table grouped by pharmacy_first_med status
+#  Data needs to have the following columns:
+#  pharmacy_first_med
+#  term
+#  count
+#  ratio_by_group
+gt_top_meds <- function(data) {
+  data |>
+    gt(
+      groupname_col = "pharmacy_first_med",
+      rowname_col = "term"
+    ) %>%
+    tab_header(
+      title = "Top 5 medications linked to Pharmacy First consultations",
+      subtitle = "Timeframe: 1st Feb 2024 to 31st July 2024"
+    ) %>%
+    cols_label(
+      term = md("**Medication**"),
+      count = md("**Count**"),
+      ratio_by_group = md("**%**")
+    ) %>%
+    fmt_number(
+      columns = count,
+      decimals = 0
+    ) %>%
+    fmt_percent(
+      columns = ratio_by_group,
+      decimals = 1
+    ) %>%
+    tab_style(
+      style = cell_text(weight = "bold"),
+      locations = cells_row_groups(groups = everything())
+    ) %>%
+    tab_stub_indent(
+      rows = everything(),
+      indent = 3
+    )
+}

lib/functions/eps_erd_prescribing_data.r

@@ -0,0 +1,36 @@
+library(tidyverse)
+library(janitor)
+library(here)
+library(httr)
+
+# Function to download and read the xlsx files
+read_xlsx_from_url <- function(url_list, sheet = NULL, skip = NULL, ...) {
+  temp_file <- tempfile(fileext = ".xlsx")
+  GET(
+    url_list,
+    write_disk(temp_file, overwrite = TRUE)
+  )
+  readxl::read_xlsx(
+    temp_file,
+    col_names = TRUE,
+    .name_repair = janitor::make_clean_names,
+    sheet = sheet,
+    skip = skip,
+    ...
+  )
+}
+
+df <- read_xlsx_from_url(
+  "https://github.com/user-attachments/files/17774058/EPS.and.eRD.Prescribing.Dashboard.July.2024.xlsx",
+  skip = 2,
+  sheet = "Historical Data"
+)
+
+df_filtered <- df %>%
+  select(month, region_code, practice_code, eps_items, erd_items) %>%
+  filter(month %in% c(202402, 202403, 202404, 202405, 202406, 202407)) %>%
+  mutate(month = ym(month))
+
+df_filtered |> write_csv(
+  here("lib", "validation", "data", "eps_erd_prescribing_2024-02-01_to_2024-07-01.csv")
+)

project.yaml

@@ -47,4 +47,13 @@ actions:
        --output output/population/pf_population.csv.gz
     outputs:
       highly_sensitive:
-        cohort: output/population/pf_population.csv.gz
+        cohort: output/population/pf_population.csv.gz
+
+  generate_pf_med_counts_measures:
+    run: >
+      ehrql:v1 generate-measures analysis/measures_definition_pf_consultation_med_counts.py
+      --dummy-tables dummy_tables
+      --output output/measures/consultation_med_counts_measures.csv
+    outputs:
+      moderately_sensitive:
+        measure: output/measures/consultation_med_counts_measures.csv