Background/Purpose: The identification of patients with recent-onset rheumatoid arthritis (RA) is often desirable to create inception cohorts of patients. We evaluated an approach to identify the timing of RA onset using administrative claims data from public (Medicare) and commercial health plans.
Methods: The study sample consisted of RA patients participating in Corrona, a large North American RA registry, linked to administrative medical and pharmacy claims data from Medicare (2006 to 2011) or a U.S. commercial health plan (2005-2012). We estimated year of RA onset in the claims data using several algorithms that were based on the following factors: 1) different ICD-9 diagnosis code for RA (e.g. 714.0, 714.2, and 714.81. vs. 714.X, code from physician visit claim vs. any claim); and 2) length of observable time in the health plan (>1 vs. > 2 years) preceding the first diagnosis code for RA, with exclusions for use of any disease modifying anti-rheumatic drugs. We compared the estimated year of RA onset using the claims-based algorithms to that recorded by rheumatologists in the Corrona registry (gold standard). We reported accuracy as a positive predictive value (PPV), calculated if the year of RA onset from the claims data agreed (+- 1 year) with that documented in Corrona. We conducted a subgroup analysis limited to patients whose disease duration was 2 years or less at their first Corrona rheumatologist visit to improve the reliability of disease onset ascertainment by reducing recall bias and misclassification of the gold standard.
Results: In the main analysis, using ICD-9 codes 714.0, 714.2, 714.81 from a physician visit, the PPVs for accurately classifying year of RA onset ranged from 62% to 68%. When ICD-9 codes 714.x from any type of claim were used, PPVs were higher, ranging from 67% to 100%. In subgroup analysis of patients with more recently diagnosed RA, PPVs were much higher, ranging from 91-100%.
Conclusion: Claims-based algorithms can be used with high validity to identify patients with recent onset RA. Additional research will focus on reasons and opportunities to reduce misclassification of disease onset.
Table: PPVs of claims-based algorithms to identify recent onset RA compared to gold standard of rheumatologist report
|
|
|
Medicare |
Commercial Health Plan |
||
|
Claims-Based Algorithm |
Look Back Period |
Total N |
PPV |
Total N |
PPV |
|
All Patients |
|||||
|
714.0, 714.2, 714.21 on physician claim |
>=365 days |
144 |
62% |
21 |
67% |
|
>=730 days |
84 |
68% |
12 |
67% |
|
|
714.xx from any claim |
>=365 days |
182 |
67% |
34 |
71% |
|
>=730 days |
93 |
77% |
12 |
100% |
|
|
|
|||||
|
Patients with year of RA diagnosis within 2 years from time of physician ascertainment* |
|||||
|
714.0, 714.2, 714.21 on physician claim |
>=365 days |
120 |
91% |
24 |
100% |
|
>=730 days |
80 |
91% |
15 |
100% |
|
|
714.xx from any claim |
>=365 days |
139 |
94% |
22 |
91% |
|
>=730 days |
73 |
92% |
15 |
93% |
|
*The purpose of the subgroup analysis is to identify patients whose year of onset is less likely to be biased due to recalling events occurred a long time ago
Disclosure:
J. Zhang,
None;
F. Xie,
None;
L. Chen,
None;
J. D. Greenberg,
Corrona, LLC.,
1,
Corrona, LLC.,
3,
AstraZeneca, Celgene, Novartis and Pfizer,
5;
J. R. Curtis,
Roche, Genentech, UCB Pharma, Janssen, CORRONA, Amgen, Pfizer, BMS, Crescendo, AbbVie,
2,
Roche, Genentech, UCB Pharma, Janssen, CORRONA, Amgen, Pfizer, BMS, Crescendo, AbbVie,
5.
« Back to 2014 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/evaluation-of-a-methodological-approach-to-determine-timing-of-rheumatoid-arthritis-disease-onset-using-administrative-claims-data/