CKD-Bone: Frailty Fractures in Patients with CKD

Feasibility analysis for studying frailty fractures in patients with chronic kidney disease (CKD), using OMOP CDM and OHDSI R packages.

Study Design

Base cohort: Patients aged ≥50 with a frailty fracture (hip, thoracic/lumbar vertebra, pelvis, wrist, proximal forearm) in the study period.

Attrition waterfall:

1. N1: Frailty fracture (age ≥50, study period)
2. N2: eGFR measurement within ±4 weeks of fracture
3. N3: eGFR < 44 → Analysis cohort
4. N4: ≥1 creatinine in 5-year pre-fracture lookback
5. N5: eGFR < 30

Requirements

• R ≥ 4.3
• Access to an OMOP CDM (SQL Server) with write permissions
• renv for package management

Setup

# 1. Clone the repo and restore packages
renv::restore()

# 2. Copy the config template and fill in your site-specific values
cp config.yml.example config.yml

Edit config.yml with your database connection details:

default:
  database:
    driver: "ODBC Driver 17 for SQL Server"
    server: "YOUR_SERVER_HERE"
    database: "YOUR_DATABASE_HERE"
    cdm_schema: "YOUR_CDM_SCHEMA"
    write_schema: "YOUR_WRITE_SCHEMA"
    write_prefix: "feas2_"
    db_name: "YOUR_DB_DISPLAY_NAME"

Note: config.yml is in .gitignore and will not be committed. Each site maintains their own copy.

Then in R:

# 3. (Optional) Adjust study parameters in feasibility_run.R section 0
# 4. Set generate_codelists = TRUE for first run
# 5. Source or run interactively
source("feasibility_run.R")

Key Files

File	Description
feasibility_run.R	Main analysis script (run interactively, section by section)
CKD-Bone.Rproj	RStudio project file (sets working directory, editor preferences)
.Rprofile	Bootstraps renv on session start
rural_urban.csv	ONS Rural-Urban Classification 2021 lookup (LSOA → Rural/Urban)
config.yml	Site-specific database configuration (not committed — copy from config.yml.example)
config.yml.example	Template config with placeholder values for partner sites
renv/	Package management (renv lockfile + activate script)
Results/Feasibility2/	Output directory (HTML report, DOCX tables, codelist CSV)

Configurable Parameters (Section 0)

study_start            <- as.Date("2023-01-01")
study_end              <- as.Date("2026-06-25")
min_age                <- 50
egfr_threshold         <- 44       # eGFR cut-off
egfr_window_days       <- 28       # ±4 weeks around fracture
lookback_creatinine_days <- 1825   # 5 years
generate_codelists     <- TRUE     # Set FALSE after first run (uses cached CSV)

Outputs

• Feasibility_Report.html — Self-contained HTML report with all tables and findings
• Table1_Full_Characteristics.docx — Full Table 1 (demographics, death, kidney function, comorbidities, eGFR timeliness)
• Attrition_Waterfall.docx — N1→N5 waterfall
• Missingness_Analysis_Cohort.docx — Variable availability/missingness
• all_codelists_for_partners.csv — Exportable codelists for partner OMOP sites
• CodeUse_*.docx — Database-wide code usage reports

Important Notes on Feasibility Code

• LTHT-specific patches: The R script includes server-side type patches specific to the LTHT OMOP instance (integer64 handling, date casting via TRY_CAST, measurement_time coercion). Partner sites may need to adjust or remove these patches (see Section 1, "Server-side type patches for LTHT") depending on their CDM implementation.
• IMD (Index of Multiple Deprivation): The feasibility code uses IMD as a concept from the OMOP CDM. Sites must have IMD mapped as a concept in their CDM for deprivation analysis to work, or should be able to adjust the code to join IMD from LSOA (see location note below).
• Rural/Urban Classification: The code uses LSOA (Lower Layer Super Output Area) from the location_source_value field to map patients to rural/urban categories via the rural_urban.csv lookup file.
• Frailty Score: Uses LTHT-specific concept ID 40483383 plus keyword-generated concepts for "frailty" and "rockwood frailty score". Partner sites should verify which frailty concept IDs are available in their local vocabulary.

Partner Site Instructions

1. Clone this repo
2. Run renv::restore() to install packages
3. Copy config.yml.example to config.yml and fill in your site-specific database details:
   • server — your SQL Server hostname
   • database — your OMOP database name
   • cdm_schema — schema containing OMOP CDM tables
   • write_schema — schema where you have write permissions (for temp cohort tables)
   • write_prefix — prefix for tables written by the script (default "feas2_")
   • db_name — a display name for your database (used in reports)
4. Set generate_codelists = TRUE (first run) or use the shared all_codelists_for_partners.csv
5. Run interactively — pre-flight checks will validate your config and permissions immediately

OMOP Packages Used

• CDMConnector
• CodelistGenerator
• CohortConstructor
• PatientProfiles
• CohortCharacteristics
• PhenotypeR
• OmopSketch