Validity of Ankylosing Spondylitis Diagnoses in the Health Improvement Network

Maureen Dubreuil^1,2, Christine Peloquin³, Yuqing Zhang⁴, Hyon Choi⁵, Robert D Inman⁶ and Tuhina Neogi⁷, ¹Rheumatology, VA Boston Healthcare System, Boston, MA, ²Rheumatology, Boston University School of Medicine, Boston, MA, ³Clinical Epidemiology Unit, Boston University School of Medicine, Boston, MA, ⁴BUSM, Boston, MA, ⁵Division of Rheumatology, Allergy and Immunology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, ⁶Immunlogy and Institute of Medical Science, Toronto Western Hospital, University of Toronto, Toronto, ON, Canada, ⁷Clinical Epidemiology, BUSM, Boston, MA

Meeting: 2015 ACR/ARHP Annual Meeting

Date of first publication: September 29, 2015

Keywords: Validity and ankylosing spondylitis (AS)

Session Information

Date: Sunday, November 8, 2015

Title: Spondylarthropathies and Psoriatic Arthritis - Clinical Aspects and Treatment Poster I: Clinical Aspects and Assessments

Session Type: ACR Poster Session A

Session Time: 9:00AM-11:00AM

Background/Purpose: Because ankylosing spondylitis (AS) is relatively uncommon, large electronic medical record (EMR) databases with longitudinal follow up offer an important opportunity for epidemiologic research in AS. However, the validity of AS diagnoses recorded by a general practitioner in such databases is unknown. We sought to assess the validity of several algorithms for identifying AS patients in The Health Improvement Network (THIN).

Methods: THIN is an EMR database of over 10 million persons in the UK, with data entered by general practitioners (GPs). In 2014, we administered a questionnaire to the GPs of 100 patients aged 18-59 years for whom at least one AS diagnostic code was recorded during years 2000-2013. As high positive predictive value (PPV) is of critical importance in epidemiologic studies of AS (i.e.- accurate identification of subjects who truly have AS), our questionnaire was designed to determine the PPV of an AS diagnostic code, using the GP’s clinical impression as the “gold standard”. We also determined characteristics for other AS case identification algorithms including: more than one AS diagnostic code, absence of osteoarthritis (OA) or rheumatoid arthritis (RA) codes, prescription of a nonsteroidal anti-inflammatory drug (NSAID), or presence of a disease modifying anti-rheumatic drug (DMARD) or biologic.

Results: Questionnaires were returned for 85 out of 100 patients with an AS code, and in 61 of those patients the GP’s clinical impression confirmed the AS diagnosis, resulting in an overall positive predictive value (PPV) of 72% (Table). The highest PPV (89%) was with an algorithm requiring two AS codes at least 7 days apart, however PPV was also high for an algorithm requiring at least one AS diagnostic code plus a DMARD or biologic drug prescription (86%). Sensitivity was reduced with algorithms requiring 2 AS codes (64%) and a DMARD/biologic prescription (30%). Algorithms also requiring prescription of an NSAID, or the absence of an OA or RA code had lower PPV (71-75%) and higher sensitivity (95-98%).

Conclusion: AS case identification algorithms of: (A) two AS diagnostic codes separated by at least 7 days or (B) one AS diagnosis plus a DMARD or biologic prescription provided the highest PPV in THIN. One or both of these algorithms should be used for AS case identification in epidemiologic studies in THIN.

Table. Ankylosing spondylitis case identification algorithms and characteristics in The Health Improvement Network
Algorithm	Total (N)	Confirmed AS (N)	PPV (95% CI)	Sensitivity* (95% CI)
One or more AS codes	85	61	71.8 % (60.5-83.1)	N/A
Two AS codes, > 7 days apart	44	39	88.6 % (78.7-98.6)	63.9 % (48.9-79.0)
AS + absence of OA code	77	58	75.3 % (76.0-97.3)	95.1 % (89.5-100)
AS + absence of RA code	80	58	72.5 % (61.0-84.0)	95.1 % (89.5-100)
AS + DMARD or Biologic	21	18	85.7 % (69.5-100)	29.5 % (8.4-50.6)
AS + NSAID	84	60	71.4 % (60.0-82.9)	98.4 % (95.1-100)
AS = ankylosing spondylitis, DMARD = disease modifying anti-rheumatic drug, NSAID =nonsteroidal anti-inflammatory drug, OA = osteoarthritis, RA = rheumatoid arthritis, N/A= not assessable. *Sensitivity with the additional feature (e.g. absence of an OA code) for a verified diagnosis of AS by GP report among patients with at least one code for AS, not the overall sensitivity and specificity of the algorithm as false negatives were unavailable.

Disclosure: M. Dubreuil, None; C. Peloquin, None; Y. Zhang, None; H. Choi, None; R. D. Inman, None; T. Neogi, None.

To cite this abstract in AMA style:

Dubreuil M, Peloquin C, Zhang Y, Choi H, Inman RD, Neogi T. Validity of Ankylosing Spondylitis Diagnoses in the Health Improvement Network [abstract]. Arthritis Rheumatol. 2015; 67 (suppl 10). https://acrabstracts.org/abstract/validity-of-ankylosing-spondylitis-diagnoses-in-the-health-improvement-network/. Accessed .

« Back to 2015 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/validity-of-ankylosing-spondylitis-diagnoses-in-the-health-improvement-network/