Data Model and Basics

synthetic data will be added later and full example will be provided.

Description

This vignette describes the data structure required by the package and demonstrates selected classification functions.
The healthpopR package uses a simple common data model (CDM) with predefined variable names.

The expected datasets, variables, and their definitions are presented.

Data model showing relationships between tables

population

Population data should include only one row per ID and few DATE variables.

Variable Name	Description
ID	Patient identifier
DATE_BIRTH	Date of patient birth
DATE_DEATH	Date of when patient died or NA
DATE_MIGRATION	Date of when patient migrated to abroad or NA

diagnoses

Diagnoses data is follow-up of the patients diagnoses over the timeline.

Variable Name	Description
ID	Patient identifier
DGREG	Diagnose Code Source (ICD10, ICD9 or ICD8)
SRC	Registry source (ex. hilmo, avohilmo etc.)
DATE	Date of the diagnose
DG	Full Diagnose Code (ex. E11, H903, H0410)
ICD10_3LETTERS	Only with ICD10 code. Diagnose Code in first 3 letters
ICD10_CLASS	Only with ICD10 code. ICD-10 Diagnose Class (Ex. E00–E90, H60–H95, H00–H59)
DATE_BIRTH	Day of patient birth
AGE	Only with murtumat Patient age on diagnose date

other datasets

These will be modified in future. Now these are used only in cox modelling phases

data_sosioeconomic

Data Sosioeconomic includes few extra variables for the population data.

Variable Name	Description
ID	Patient identifier
edu	education in factor leveling
bmi	Body Mass Index
bmi_cat1	Body Mass Index in factor leveling

data_dates

Data dates should include more date variables for the population.

Variable Name	Description
ID	Patient identifier
vpvmbl	Baseline answering date
spvm to mpvmfps3y	Multiple follow-up timepoints (exact definitions to be documented)

Basic Data Handling

library(dplyr)
library(healthpopR)

Load data

# synthetic raw datasets
population <- arrow::read_parquet("../data/synth_population.parquet")
diagnoses <- arrow::read_parquet("../data/synth_diagnoses.parquet")
population_variables <- arrow::read_parquet("../data/synth_population_variables.parquet")

population data should have these variables:

Variable Name	Description
ID	Patient identifier
DATE_BIRTH	Date of patient birth
DATE_DEATH	Date of when patient died or NA
DATE_MIGRATION	Date of when patient migrated to abroad or NA

diagnoses data should have these variables:

Variable Name	Description
ID	Patient identifier
DGREG	Diagnose Code Source (ICD10, ICD9 or ICD8)
SRC	Registry source (ex. hilmo, avohilmo etc.)
DATE	Date of the diagnose
DG	Full Diagnose Code (ex. E11, H903, H0410)

To diagnoses data we will create also these variables:

ICD10_3LETTERS | Only with ICD10 code. Diagnose Code in first 3 letters |
ICD10_CLASS | Only with ICD10 code. ICD-10 Diagnose Class (Ex. E00–E90, H60–H95, H00–H59) |
DATE_BIRTH | Day of patient birth |
AGE | Only with murtumat Patient age on diagnose date |

population_variables should have these variables:

Variable Name	Description
ID	Patient identifier
edu	Education, in our example it factor with categories Low, Med and High
bmi	Body Mass Index as double

For bmi we will categorized variables.

Categorize ICD-10 Diagnoses

ICD-10 Codes are useful to categorize to 3 Letters and to upper category:

diagnoses <- diagnoses %>% 
  mutate(
    ICD10_3LETTERS = healthpopR::recode_icd10_3letters(DG, DGREG = DGREG),
    ICD10_CLASS = healthpopR::recode_icd10_class(ICD10_3LETTERS, DGREG = DGREG)
  ) %>% 
  left_join(population %>% select(ID, DATE_BIRTH), by = "ID") %>% 
  mutate(AGE = floor(lubridate::time_length(difftime(DATE, DATE_BIRTH), "years")))

Categorize population_variables BMI

BMI can be categorized to 4 or 6 category classes. We can use these variables in Cox model in future.

population_variables <- population_variables %>% 
  mutate(
    bmi_cat1 = factor(healthpopR::categorize_bmi(bmi, levels = 6)),
    bmi_cat2 = factor(healthpopR::categorize_bmi(bmi, levels = 4)),
    edu = factor(edu)
  )