Article Text

Download PDFPDF

  1. J. Sparks1,
  2. W. Huang1,
  3. B. Lu1,
  4. S. Huang1,
  5. A. Cagan2,
  6. V. Gainer2,
  7. S. Finan3,
  8. G. Savova3,
  9. D. Solomon1,
  10. E. Karlson1,
  11. K. Liao1
  1. 1Brigham and Women’s Hospital, Boston, United States of America
  2. 2Partners HealthCare, Boston, United States of America
  3. 3Boston Children’s Hospital, Boston, United States of America


Background: Patients with rheumatoid arthritis (RA) are at increased risk of serious infections, with considerable excess morbidity and mortality after pneumonia. RA-related autoantibodies such as anti-cyclic citrullinated peptide (CCP) and rheumatoid factor (RF) may be generated at inflamed pulmonary mucosa prior to clinical RA onset. Therefore, patients with seropositive RA may be at increased risk for pneumonia after RA diagnosis due to subclinical pulmonary injury.

Objectives: We investigated whether seropositive RA was associated with increased pneumonia risk compared to seronegative RA.

Methods: We performed a retrospective cohort study among RA patients seen at a health care system in Boston, MA. RA patients were identified using a previously validated electronic health record (EHR) algorithm incorporating billing codes, natural language processing (NLP) of notes, medications, and laboratory results at 97% specificity1. We constructed an incident RA cohort using NLP for the index date of initial mention of RA. All patients were required to have both CCP and RF data from clinical care to determine serologic RA phenotype. We used semi-supervised machine learning approaches to identify pneumonia using billing codes and terms extracted using NLP, with the Centers for Disease Control definition of pneumonia from medical record review as a gold standard. The area under the receiver operating curve (AUROC) for this billing code+NLP pneumonia algorithm was 0.94 compared to the standard rule-based pneumonia algorithm (billing code on inpatient discharge) AUROC of 0.86 (p<0.001). Smoking status was extracted using NLP methods. Other covariates, including a previous validated weighted RA multimorbidity score2, were determined using structured EHR data. We used Cox regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for pneumonia adjusting for potential confounders.

Results: We analyzed a total of 4,110 patients with incident RA and both CCP/RF data available. Mean age at index date was 53.0 years (SD 14.8), 77.2% were female, and 79.8% were CCP+ or RF+. During 32,248 patient-years of follow-up (mean 7.8 years/patient), we identified 240 pneumonia cases. Patients with seropositive RA had a HR of 1.99 (95%CI 1.30-3.01, Table) for pneumonia compared to patients with seronegative RA, adjusted for age, sex, smoking, index year, ESR level, glucocorticoid use, DMARD use, and weighted RA multimorbidity score. While CCP+ RA (HR 1.91, 95%CI 1.23-2.97) and RF+ RA (HR 2.07, 95%CI 1.35-3.16) had increased pneumonia risk compared to seronegative RA, the CCP+RF- RA subgroup had no association with pneumonia (HR 0.67, 95%CI 0.23-1.93).

Conclusion: Patients with incident seropositive RA, particularly RF+ RA, had increased risk for pneumonia throughout the RA disease course that was not explained by measured confounders including smoking status, multimorbidity, medications, and ESR level. Further studies should investigate how RF+ may predispose RA patients to later develop pneumonia after clinical RA diagnosis.

References: [1]Liao KP, Cai T, Gainer V, et al. Electronic medical records for discovery research in rheumatoid arthritis. Arthritis Care Res. 2010;62(8):1120–1127.

[2]Radner H, Yoshida K, Mjaavatten MD, et al. Development of a multimorbidity index: Impact on quality of life using a rheumatoid arthritis cohort. Semin Arthritis Rheum. 2015;45(2):167–173.

Disclosure of Interests: Jeffrey Sparks Consultant of: Bristol-Myers Squibb, Optum, Janssen, Gilead, Weixing Huang: None declared, Bing Lu: None declared, Sicong Huang: None declared, Andrew Cagan: None declared, Vivian Gainer: None declared, Sean Finan: None declared, Guergana Savova: None declared, Daniel Solomon Grant/research support from: Funding from Abbvie and Amgen unrelated to this work, Elizabeth Karlson: None declared, Katherine Liao: None declared

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.