Related Experiment Video
Updated: Jan 1, 2026
02:28
Author Spotlight: Self-Assessment Protocol for Predicting Psoriatic Arthritis in Psoriasis Patients
Published on: March 1, 2024
713
Validation of claims-based algorithms for psoriatic arthritis
Hemin Lee1, Julia A Ford2, Yinzhu Jin1
1Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts.
Pharmacoepidemiology and Drug Safety
|December 19, 2019
Summary
Developing accurate claims-based algorithms is crucial for identifying psoriatic arthritis (PsA) patients. Algorithms combining diagnosis codes and medication data show high positive predictive values (PPVs), aiding real-world evidence generation for PsA treatments.
Area of Science:
- Rheumatology
- Health Informatics
- Real-World Evidence
Background:
- Psoriatic arthritis (PsA) management is evolving with new therapies.
- Generating real-world evidence (RWE) on drug safety and effectiveness is essential.
- Accurate identification of PsA patients in claims data is needed for RWE studies.
Purpose of the Study:
- To develop and validate claims-based algorithms for identifying patients with psoriatic arthritis (PsA).
- To assess the accuracy of these algorithms using linked electronic medical records.
- To support RWE generation for comparative drug studies in PsA.
Main Methods:
- Seven claims-based algorithms were developed using diagnosis codes and medication dispensing data (Medicare Parts A/B/D, 2012-2014).
- Algorithms were validated against a gold standard of treating physician diagnoses via independent physician chart review.
- Positive predictive values (PPVs) and 95% confidence intervals were calculated for each algorithm.
Main Results:
- The PPVs of the seven algorithms ranged from 75.2% to 88.6%.
- Algorithm 7 (≥2 PsA diagnosis codes, ≥1 rheumatologist code, ≥1 PsA medication dispensing) achieved the highest PPV (88.6%).
- Algorithm 6 (≥2 PsA diagnosis codes, ≥1 PsA medication dispensing) demonstrated a PPV of 82.4%.
Conclusions:
- All seven developed algorithms showed moderately high PPVs (75%-89%) for identifying PsA.
- Algorithms combining diagnosis codes and prescription claims are valid and efficient for PsA patient identification in claims data.
- Broader diagnostic algorithms without prescription data also offer reasonably good PPVs.
