top of page

Systemic Lupus Erythematosus Phenotyping & Cohort Definition

Real-world evidence (RWE) observational informatics study implementing a reproducible, validated electronic health record (EHR) phenotype for incident Systemic Lupus Erythematosus (SLE). The study enforces a baseline 365-day clean lookback period, excludes drug-induced lupus cases, and applies high-specificity confirmatory criteria (subsequent diagnosis or initiation of first-line therapies such as Hydroxychloroquine or systemic corticosteroids) to evaluate cohort attrition and scale across standardized OMOP Common Data Model (v5.3/v5.4) networks.

Business question: How effectively does a multi-stage rule-based computable phenotype (incorporating clean observation windows, drug-induced disease exclusions, and confirmatory diagnosis/medication exposure logic) filter false-positive incident SLE cases, and does the observed cohort attrition scale consistently from local database validation to large-scale synthetic claims benchmarks (CMS SynPUF 5%)?

Technology stack: OHDSI ATLAS, OMOP Common Data Model (CDM v5.3/v5.4), SQL Server, CIRCE / OHDSI JSON,                 Git / GitHub.

Key Insights

​

  • Robust Phenotype Specificity: Enforcing a clean 365-day baseline observation period paired with secondary confirmation criteria successfully eliminated isolated, provisional coding, narrowing 2,944 initial suspected events down to 254 qualified incident patients (8.63% retention rate).

  • Controlled Exclusion of Confounding Etiologies: Dedicated exclusion criteria for Drug-Induced Lupus safely filtered out secondary, drug-mediated clinical presentations, isolating true idiopathic Systemic Lupus Erythematosus cases.

  • Dual Confirmatory Pathways: Requiring either a secondary diagnosis within 31–730 days or initiation of guideline-directed first-line therapy (such as Hydroxychloroquine or systemic corticosteroids) within 0–365 days ensured high diagnostic positive predictive value (PPV) across real-world longitudinal records.

  • Multi-Scale Validation: The phenotype logic proved robust across environments, validating expected clinical rule behavior on a localized SQL Server CDM unit test and scaling without schema degradation to population-level benchmarks.

1. Data Source

​

  • Standardized Real-World Data Model: All clinical entities are standardized to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) v5.3 architecture, ensuring database-agnostic cohort reproducibility across global research networks.

  • Longitudinal Claims Benchmark: Tested and benchmarked against the CMS 2008–2010 Synthetic Public Use File (SynPUF 5%), capturing complete inpatient, outpatient, and prescription histories across approximately 110,000 synthetic patient lives.

  • Core Clinical Domains Queried:

  • Condition Occurrence & Condition Era: Evaluated index diagnoses (SNOMED-CT: 257628 Systemic lupus erythematosus) and exclusion markers (SNOMED-CT: 4180454 Drug-induced systemic lupus erythematosus).

  • Drug Exposure & Drug Era: Monitored baseline and confirmatory dispensings of first-line disease-modifying therapies (RxNorm: 1777087 Hydroxychloroquine, 1550557 Prednisone).

  • Observation Period & Person: Enforced ≥\ge≥ 365 days of continuous pre-index enrollment and adult eligibility (≥\ge≥ 18 years of age at index).

  • Cross-Dialect Execution Pipeline: Authored and compiled via OHDSI ATLAS and CIRCE into parameterized OHDSI SQL, executable directly against relational backends (T-SQL / Microsoft SQL Server) and network analytics packages.

2. Data Exploration
 

  • Declarative Cohort Construction via OHDSI ATLAS: Cohort logic was defined and compiled directly from an OMOP CDM v5.3 architecture using CIRCE and OHDSI ATLAS, generating portable JSON specifications alongside dialect-optimized T-SQL execution scripts.

  • Standardized Concept Set Engineering: Curated distinct clinical concept sets across standard OMOP vocabularies:

  • Primary Target Condition: [SLE] Systemic Lupus Erythematosus (SNOMED-CT: 257628 and descendants).

  • Etiological Exclusion: [SLE] Drug-Induced Lupus (SNOMED-CT: 4180454 and descendants).

  • Guideline-Directed First-Line Therapies: [SLE] First-Line Therapies (RxNorm: 1777087 Hydroxychloroquine, 1550557 Prednisone and descendants).

  • Multi-Stage Longitudinal Filtering: Implemented a computable phenotype pipeline enforcing a 365-day clean baseline observation window to identify true incident cases, followed by sequential clinical inclusion criteria:

  • Age $\ge$ 18 at index date to focus strictly on adult-onset manifestations.

  • Complete exclusion of prior or concurrent drug-induced lupus to eliminate secondary confounding.

  • Confirmatory secondary SLE diagnosis within 31–730 days post-index OR initiation of first-line therapies within 0–365 days post-index to maximize positive predictive value.

  • Dual-Environment Verification: Tested and executed across two independent platforms:

  • Local Unit Testing: Validated rule logic against boundary patient cases in Microsoft SQL Server / SSMS.

  • Population Benchmark: Executed against the CMS SynPUF 5% dataset (~110,000 synthetic patient lives), processing 2,944 initial index events down to 254 qualified patients.

3. Data Visualization
 

  • ATLAS Analytical Visualizations: Cohort attrition, baseline characteristics, and diagnostic reports were generated and evaluated directly within the OHDSI ATLAS analytical reporting engine.

  • Visualizations Captured for Study Documentation:

  • Cohort Attrition Flowchart: Displays the sequential qualification and exclusion of patients from the initial 2,944 index events down to the final 254 incident SLE patients, contrasting remaining population bars against excluded volumes.

  • Inclusion Rule Statistics Table: Detailed breakdown of sample size retention ($N$) and criterion satisfaction percentages across adult age restrictions (100%), drug-induced lupus exclusion (100%), and confirmatory secondary diagnosis / treatment initiation (8.63%).

  • Demographic Distributions: Age-at-index histograms and gender breakdowns illustrating the clinical distribution of adult-onset SLE patients.

  • Baseline Comorbidity & Medication Profiles: Ranked prevalence plots of pre-index clinical conditions and initial pharmacological regimens within the cohort observation window.

  • Treatment Sequence Sunburst Architecture: Sunburst trajectory designs mapping clinical lines of therapy, from first-line antimalarial initiation (Hydroxychloroquine) to corticosteroid escalation and steroid-sparing immunosuppressive maintenance.

Other projects

bottom of page