Skip to contents

swereg

swereg is an R package for manipulating and analyzing healthcare registry data in epidemiological research. It uses a three-stage framework for creating longitudinal data skeletons and integrating multiple health registries.

Registry integration

swereg is designed for healthcare registries with realistic synthetic data included:

  • Hospital registers: Diagnosis codes (ICD-10) and surgical procedures
  • Prescription registers: Medication data with ATC codes and treatment duration
  • Death registers: Underlying and multiple causes with proper variable names
  • Administrative data: Demographics and socioeconomic data
  • Synthetic datasets: fake_demographics, fake_diagnoses, fake_prescriptions, fake_cod, etc.

The skeleton approach

swereg uses a skeleton concept: you build strong ‘bones’ (time structure) then attach ‘muscles’ (data) systematically through three stages:

  1. skeleton1_create - Raw data integration
  2. skeleton2_clean - Data cleaning and derived variables (analysis-ready for most studies)
  3. skeleton3_analyze - Memory-efficient batching (only needed for huge datasets with limited RAM)

Features

  • Three-stage workflow: Systematic skeleton1→skeleton2→skeleton3 progression
  • Time structure: ISO year-weeks with automatic Sunday date calculation and person-time calculations
  • Flexible date parsing: Handles Swedish registry dates with varying precision (4/6/8 characters)
  • Healthcare integration: Hospital diagnosis and surgical procedures with automatic pattern matching
  • Prescription analysis: Prescription data with treatment duration and ATC code patterns
  • Mortality data: Death registry with underlying and multiple causes
  • Synthetic data: Realistic synthetic registry data for development and testing
  • Scalable processing: Memory-efficient batching for large populations (when needed)
  • Reproducible workflow: Standardized methodology for registry-based epidemiological research

Core functions

Data structure

  • create_skeleton() - Create longitudinal data skeleton with individual IDs, time periods, and person-time calculations
  • make_lowercase_names() - Standardize column names and clean date columns across datasets
  • parse_swedish_date() - Parse Swedish registry dates with varying precision (4/6/8 characters)

Data integration

Specialized functions

Installation

You can install the development version of swereg from GitHub:

# Install devtools if not already installed
install.packages("devtools")

# Install swereg
devtools::install_github("papadopoulos-lab/swereg")

Getting started

Follow the vignettes in order to learn the skeleton approach:

1. Understand the concept 

vignette("skeleton-concept")

Learn the foundational skeleton approach and why it works for registry data analysis.

2. Build your skeleton (skeleton1_create) 

vignette("skeleton1-create")

Create the time structure and integrate raw registry data:

library(data.table)

# Load synthetic registry data (included with package)
data("fake_person_ids", package = "swereg")
data("fake_demographics", package = "swereg")
data("fake_diagnoses", package = "swereg")

# Step 1: Create the skeleton (good bones)
skeleton <- swereg::create_skeleton(
  ids = fake_person_ids[1:100],
  date_min = "2020-01-01",
  date_max = "2022-12-31"
)

# Step 2: Apply standardization (required for all registry data)
swereg::make_lowercase_names(fake_demographics, date_column = "fodelseman")
swereg::make_lowercase_names(fake_diagnoses, date_column = "indatum")

# Step 3: Attach demographic data (muscles)
demographics_subset <- fake_demographics[lopnr %in% fake_person_ids[1:100]]
swereg::add_onetime(skeleton, demographics_subset, id_name = "lopnr")

# Step 4: Add diagnosis patterns (^ prefix automatically added)
diagnosis_patterns <- list(
  "diabetes" = c("E10", "E11"),
  "depression" = c("F32", "F33")
)
diagnoses_subset <- fake_diagnoses[lopnr %in% fake_person_ids[1:100]]
swereg::add_diagnoses(skeleton, diagnoses_subset, "lopnr", diags = diagnosis_patterns)

# Result: skeleton1_create ready for skeleton2_clean stage
head(skeleton)

3. Clean your data (skeleton2_clean) - Analysis ready for most studies 

vignette("skeleton2-clean")

Clean variables and create derived clinical indicators using only data within the skeleton. For most analyses, this is your final step - you can proceed directly with statistical modeling using the skeleton2_clean output.

4. Batching for huge datasets (skeleton3_analyze) - Optional 

vignette("skeleton3-analyze")

Only needed if you have huge datasets (>100,000 individuals) and limited RAM. Learn memory-efficient batching techniques. Most users can skip this and use skeleton2_clean output directly for analysis.

When do you need each stage?

  • skeleton1_create: Required for all users - integrates raw data into time structure
  • skeleton2_clean: Required for all users - creates analysis-ready variables
  • skeleton3_analyze: Optional - only needed for memory-constrained processing of very large datasets

Key principles

  1. Sequential stages: Each stage builds on the previous one
  2. Time structure: ISO year-weeks provide precise temporal alignment
  3. Pattern matching: Automatic prefix handling for medical codes
  4. Analysis flexibility: skeleton2_clean output works for most statistical models
  5. Memory efficiency: skeleton3_analyze provides batching when RAM is limited

Typical workflow

Most users follow this pattern:

  1. Read vignette(“skeleton-concept”) to understand the approach
  2. Create skeleton1 with your data using the skeleton1-create vignette
  3. Clean to skeleton2 using the skeleton2-clean vignette
  4. Analyze skeleton2 directly with your preferred statistical methods (glm, survival, etc.)

Only large-scale studies with memory constraints need skeleton3_analyze.

Documentation

Complete documentation and tutorials available at: https://papadopoulos-lab.github.io/swereg/

Citation

If you use swereg in your research, please cite:

[Citation information to be added]