9.1 Phenotype outcome events

Load packages and reference.R which will be used in phenotyping diabetes (details in DM section below).

library(tidyverse)
library(data.table)
source("reference.R")

Import primary care data.

pc <- fread("generated_data/entire_gp_clinical_30March2021_formatted.txt")

Import outcome-specific dictionaries.

dm_dict <- readRDS("generated_data/dm_dict.RDS")
dr_dict <- readRDS("generated_data/dm_eye_disease_dict.RDS")
kidney_disease_case_dict <- readRDS("generated_data/kidney_disease_case_dict.RDS")

9.1.1 DM

Define T1D and T2D keywords.

type2_kw <- "type 2|type II|adult onset|non[ -]insulin[ -]dep|NIDD|maturity onset diabetes mellitus$"
type1_kw <- "type 1|type I[ $-]|juvenile|insulin-dependant|insulin dependent"

Add diabetes types to DM dictionary.

dm_dict <- dm_dict %>%
  mutate(dm_type = ifelse(grepl(type2_kw,term_description, ignore.case=T), "Type2", ifelse(grepl(type1_kw, term_description, ignore.case=T), "Type1", "Uncertain")))

Subset the primary care data to subjects with diabetes and identify which type of diabetes each event is associated with

diabetics_any <- pc %>%
  select(f.eid, code, event_dt) %>%
  filter(code %in% unique(dm_dict$code)) %>% distinct() %>%
  left_join(dm_dict %>% select(code,dm_type), by = "code")

Next, we select the terms in the DM-specific dictionary whose descriptions reflect more clear indications of DM. Defined as an object dm_descriptions in reference.R, the descriptions were selected with the help of an expert.

dm_codes_confident <-  dm_dict %>% filter(term_description %in% dm_descriptions) %>% .$code %>% unique

We define this list of terms with clear DM indications to be able to more accurately identify the first diagnosis date diabetes. Specifically, if a subject had terms predating the certain diagnosis which are less certain, we assume that the earlier diagnosis was the better start date. For example, consider a hypothetical subject with following diabetes events:

  • 2012: “Diabetic dietary review”
  • 2013: “Type II diabetes”

This individual will be considered type 2 diabetic patient with 2012 as the first date of DM diagnosis. Although the predating description “Diabetic dietary review” is a less certain of diagnosis, a subsequent diagnosis of “Type II diabetes” tells us this subject clearly has Type II diabetes. In contrast, the following subject would not be included as diabetes patient because the codes are not strong enough to indicate diabetes:

  • 2012: “Diabetic dietary review”
  • 2013: “Diabetic dietary review”

Based on this procedure, we will generate first occurrence event table for diabetes. First, subset the event table with any indications of diabetes to events with clear indications of diabetes.

diabetics_simple <- 
  diabetics_any %>% 
  group_by(f.eid) %>% 
  filter(any(code %in% dm_codes_confident)) %>%
  mutate(event_dt = as.Date(event_dt)) %>%
  arrange(f.eid,event_dt) %>%
  distinct()

Generate first occurrence diabetes events.

dm_firstoccur_pc <- diabetics_simple %>%
  group_by(f.eid) %>% 
  arrange(event_dt) %>% 
  slice(1) %>%
  select(f.eid, event_dt)
saveRDS(diabetics_simple,"generated_data/dm_pc.RDS")
saveRDS(dm_firstoccur_pc,"generated_data/dm_firstoccur_pc.RDS")

9.1.2 DR

Phenotype diabetic eye disease event table.

dr_pc <- pc %>% filter(code %in% dr_dict$code) %>% 
  select(f.eid,event_dt) %>% 
  mutate(event_dt=as.Date(event_dt)) %>% 
  distinct() %>% arrange(f.eid,event_dt)

Generate first occurrence diabetic eye disease event table.

dr_firstoccur_pc <- dr_pc %>% group_by(f.eid) %>% arrange(event_dt) %>% slice(1)
saveRDS(dr_pc,"generated_data/dr_pc.RDS")
saveRDS(dr_firstoccur_pc,"generated_data/dr_firstoccur_pc.RDS")

9.1.3 DKD

Phenotype diabetic kidney disease event table.

kidney_disease_case_pc <- 
  pc %>% filter(code %in% kidney_disease_case_dict$code) %>% 
  select(f.eid,event_dt) %>%
  mutate(event_dt = as.Date(event_dt)) %>%
  distinct() %>% arrange(f.eid,event_dt)

Generate first occurrence diabetic kidney disease event table.

kidney_disease_case_firstoccur_pc <- kidney_disease_case_pc %>% group_by(f.eid) %>% arrange(event_dt) %>% slice(1)
saveRDS(kidney_disease_case_pc,"generated_data/kidney_disease_case_pc.RDS")
saveRDS(kidney_disease_case_firstoccur_pc,"generated_data/kidney_disease_case_firstoccur_pc.RDS")