Session Information
Date: Sunday, November 8, 2015
Title: Vasculitis Poster I
Session Type: ACR Poster Session A
Session Time: 9:00AM-11:00AM
Background/Purpose:
To facilitate clinical care and research, validated algorithms are needed to
accurately identify patients with granulomatosis with
polyangiitis (GPA; Wegener’s). This study, sought to
evaluate and validate case-finding algorithms for GPA in 2 large health
administrative databases.
Methods: 250
patients were randomly selected from 2 large healthcare systems (125 patients
per system) using the International Classification of Diseases version 9 (ICD9)
code for GPA: 446.4. Because eosinophilic granulomatosis with polyangiitis
(Churg-Strauss) shares the same ICD9 code with GPA, patients
who have ever had ICD9 codes 288.3 (eosinophilia) or 493.*
(asthma) were excluded. 30 case-finding algorithms were constructed using a
combination of ICD9 code, encounter type (1 inpatient
ICD9 code on 3 consecutive days or 2 outpatient ICD9 codes 3 months apart), physician
specialty (Rheumatology, Nephrology, Pulmonology, or Otorhinolaryngology), use
of immunosuppressive medications, and whether an ANCA test was ordered. 2 time
periods for ICD9 code submission were examined: 1 year and 2 years. The
diagnosis was confirmed by chart review using the modified ACR classification
criteria or the Chapel Hill Consensus Conference definitions for GPA. Patients
who did not have a clear diagnosis were excluded from the analysis.
Results: 97
patients from the first healthcare system (site 1) and 98 patients from the second
system (site 2) were included in the analysis after excluding patients with
unclear diagnosis. 38/97 (39%) patients had a confirmed diagnosis of GPA at site
1 and 66/98 (67%) patients at site 2. Table 1 shows the positive and negative
predictive values of the studied algorithms in each healthcare system. An
algorithm excluding patients with eosinophilia or asthma and including the encounter
type and the physician specialty had the highest average positive predictive value
(PPV: 87.1%). An algorithm excluding patients with eosinophilia or asthma and including
the physician specialty had the highest average negative predictive value (NPV:
83.1%). There were no differences between the 2 time periods with regards to
the algorithms’ PPV or NPV.
Conclusion:
Case-finding algorithms can accurately identify patients with GPA using large administrative
databases. A simple algorithm excluding patients with eosinophilia or asthma
and including the encounter type and the physician specialty has the highest
positive predictive value. Similarly, an algorithm excluding patients with
eosinophilia or asthma and including the physician specialty has the highest
negative predictive value. These algorithms can be used to assemble population-based
cohorts of patients with GPA and facilitate future research in healthcare use, outcomes,
and comparative effectiveness.
Table 1. Test characteristics
of the GPA algorithms in both healthcare systems
|
GPA Algorithms
|
Site 1 (n=97)
|
Site 2 (n=98)
|
||
|
|
PPV |
NPV |
PPV |
NPV |
|
ICD9+
|
46.2 |
94.1 |
72.9 |
69.2 |
|
Encounter
|
58.3 |
79.6 |
90.8 |
78.8 |
|
Specialty**
|
77.1 |
82.3 |
91.0 |
83.9 |
|
Medications
|
64.4 |
82.7 |
78.5 |
78.9 |
|
ANCA
|
68.0 |
70.8 |
82.5 |
66.7 |
|
ICD9+Encounter+
|
||||
|
Specialty*
|
82.1 |
78.3 |
92.1 |
77.1 |
|
Medications
|
71.0 |
75.8 |
90.8 |
78.8 |
|
ANCA
|
78.8 |
70.5 |
91.5 |
69.2 |
|
Specialty+Medications
|
80.0 |
75.0 |
92.1 |
77.1 |
|
Specialty+ANCA
|
82.4 |
70.0 |
91.5 |
69.2 |
|
Medications+ANCA
|
75.0 |
67.9 |
91.5 |
69.2 |
|
Specialty+Medications+ANCA
|
80.0 |
68.3 |
81.8 |
79.7 |
|
ICD9+Specialty+
|
||||
|
Medications
|
77.4 |
78.8 |
91.0 |
83.9 |
|
ANCA
|
78.9 |
70.5 |
90.3 |
72.2 |
|
Medications+ANCA
|
81.2 |
69.1 |
90.3 |
72.2 |
|
ICD9+Medications+
|
||||
|
ANCA
|
70.0 |
68.8 |
83.6 |
67.7 |
|
ICD9+Encounter+
|
||||
|
(1 year)
|
52.4 |
69.6 |
89.8 |
66.7 |
|
(2 years)
|
52.4 |
70.9 |
89.8 |
66.7 |
|
Specialty (1 year)
|
78.3 |
73.0 |
91.4 |
67.5 |
|
Specialty (2 years)
|
79.2 |
74.0 |
91.4 |
67.5 |
|
Medications (1 year)
|
67.9 |
72.5 |
89.8 |
66.7 |
|
Medications (2 years)
|
67.9 |
72.5 |
89.8 |
66.7 |
|
ANCA (1 year)
|
75.0 |
67.9 |
90.7 |
61.4 |
|
ANCA (2 years)
|
76.5 |
68.8 |
90.7 |
61.4 |
|
Specialty+Medications (1 year)
|
77.3 |
72.0 |
91.4 |
67.5 |
|
Specialty+Medications (2 years)
|
77.3 |
72.0 |
91.4 |
67.5 |
|
Specialty+ANCA (1 year)
|
80.0 |
68.3 |
90.7 |
61.4 |
|
Specialty+ANCA (2 years)
|
81.2 |
69.1 |
90.7 |
61.4 |
|
Specialty+Medications+ANCA (1 year)
|
80.0 |
68.3 |
90.7 |
61.4 |
|
Specialty+Medications+ANCA (2 years)
|
80.0 |
68.3 |
90.7 |
61.4 |
|
* Algorithm with the highest average PPV ** Algorithm with the highest average NPV. PPV: positive predictive value. NPV: negative predictive value. ICD9: ICD9 code 446.4 excluding ICD9 code 288.3 (eosinophilia) or ICD9 code 493.* (asthma). ENCOUNTER: 1 inpatient ICD9 code on 3 consecutive days or 2 outpatient ICD9 codes 3 months apart. SPECIALTY: a Rheumatologist, Pulmonologist, Otorhinolaryngologist, or a Nephrologist involved in the care of the patient. MEDICATIONS: an immunosuppressive medication used. ANCA: ANCA test ordered. 1 year: encounter ICD9 codes submitted within 12 months. 2 years: encounter ICD9 codes submitted within 24 months. |
||||
To cite this abstract in AMA style:
Sreih AG, Annapureddy N, Byram K, Casey G, Frangiosa V, George M, Sangani S, Sharim R, Merkel PA. Evaluation and Validation of Case-Finding Algorithms for the Identification of Patients with Granulomatosis with Polyangiitis in Large Healthcare Administrative Databases [abstract]. Arthritis Rheumatol. 2015; 67 (suppl 10). https://acrabstracts.org/abstract/evaluation-and-validation-of-case-finding-algorithms-for-the-identification-of-patients-with-granulomatosis-with-polyangiitis-in-large-healthcare-administrative-databases/. Accessed .« Back to 2015 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/evaluation-and-validation-of-case-finding-algorithms-for-the-identification-of-patients-with-granulomatosis-with-polyangiitis-in-large-healthcare-administrative-databases/