This repository includes PEDSnet resources that may be useful for planned/ongoing bone marrow studies. The repository is structure using the PEDSnet standard framework and set up to be run against an EHR in the PEDSnet data model on a Postgres database; code will likely need to be adapted to run in other settings.
Code for a modifidied KDIGO (mKDIGO) definition of AKI that was implemented for a prior feasibility query is in driver_aki.R. The cohort output at line 404 is the set of patients who meet the AKI definition below. Codesets that are used in the code can be found in the specs subdirectory.
Code for the KDIGO AKI definition (which uses CRRT data) has been added to the repo (as of 2/17/2026) as well and is in driver_kdigo_aki.R. Additional functions for running this code are defined in cohort_demographics.R and in (cohort_kdigo_aki.R)[https://github.com/PEDSnet/gluck_bone_marrow_resources/blob/main/code/cohort_kdigo_aki.R]. Documentation of this definition is included here.
This repository includes code for computing eGFR values; it is part of the CKD Phenotype code used for the Clinical Insights Report--code for all steps up through the step at which eGFRs are computed was included. Functions are defined in cohort_egfr.R and called in driver_egfr.R. Accompanying codesets can found in the specs subdirectory. Please see the function definitions for further documentation.
This repository includes the code, adapted for the Nephrology Data Insights dashboard, for computing the Chronic Kidney Disease phenotype.
PLEASE NOTE the PRESERVE definition requires a lower eGFR bound of 15, so this line may need to be changed based on study requirements
This repository includes code to compute the eGFR elements of this outcome. Functions are defined in cohort_renal_outcome_CER.R and executed in driver_renal_outcome_CER.R
Transplant and chronic dialysis phenotypes developed for PRESERVE can be found on PEDSpace (with links to all relevant code & codesets).
This repository includes code to classify & compute proteinuria. Some manual review of lab results may be required to address data quality issues & improve classification completeness.
Functions are called in cohort_proteinuria_CER.R and executed in driver_proteinuria_CER.R
The following codesets are needed to execute the analysis
- lab_urine_protein_qual -- qualitative urine protein labs (in PRESERVE Variables repository)
- lab_urine_protein_quant -- quantitative urine protein labs (in PRESERVE Variables repository)
- vs_result_unit -- PCORnet result_unit valueset
- vs_result_qual -- PCORnet result_qual valueset
- vs_result_modifier -- PCORnet result_modifier valueset
The PMCA algorithm is available to run in both the PCORnet and PEDSnet CDMs. The PCORnet implementation can be found in this repository, where the functions are in cohort_pmca_pcornet.R and the execution is in driver_pmca_pcornet.R.
The PEDSnet implementation can be found on PEDSpace
This code relies on the PMCA "cardiac" body system in order to identify patients in this subcohort. The functions can be found in cohort_cardiac_CER.R and the execution can be found in driver_cardiac_CER.R
The code included in this repository will compute ED visits and hospitalization days (based on length of stay) per person year for the study cohort. The function can be found cohort_visits_ppy_CER.R and the execution in driver_visits_ppy_CER.R
If you require any additional concept sets aside from what is listed here, please take a look in PEDSpace, our metadata repository where we archive concept sets used in prior studies.
A concept set to identify patients who had undergone hematopoietic stem cell transplants was developed for a prior PEDSnet study. The team took a lenient approach in reviewing this codeset, as the investigator wanted to cast a wider net for chart reviews.
The overall false positive rate was 16%, which included both cases where no transplant had occurred and cases where the transplant was but for a different disease. The team reviewed the codeset after chart reviews and found that none of the codes consistently yielded all the false positive cases. Among the 84% identified as true positives, 85% had transplant dates in the EHR that were within three days of the dates recorded by the chart reviewers. For all analyses, the team used the transplant dates provided by the chart reviewers.
A concept set from the PRESERVE Variables repository to identify patients with a diabetes diagnosis
A concept set from the PRESERVE Variables repository to identify patients with a hypertension diagnosis. Another concept set is also available that excludes pregnancy-related hypertension diagnoses.
Relevant drug concept sets are stored in the public PRESERVE Variables repository. For antihypertensives, includes concept sets for ACE inhibitors, ARBs, Beta Blockers, Calcium Channel Inhibitors, Loop Diuretics, & Thiazides, with files for full concept sets as well as concept sets without IV drugs.
It also includes two concept sets for nephrotoxic drugs -- these will need to be re-evaluated to confirm they align with the current NINJA classifications.