July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Development and validation of decision trees for predicting systemic diseases associated with uveitis
Author Affiliations & Notes
  • Zhenyu Zhong
    The First Affiliated Hospital of Chongqing Medical University, Chongqing, Chongqing, China
    Chongqing Key Laboratory of Ophthalmology and Chongqing Eye Institute, China
  • Peizeng Yang
    The First Affiliated Hospital of Chongqing Medical University, Chongqing, Chongqing, China
    Chongqing Key Laboratory of Ophthalmology and Chongqing Eye Institute, China
  • Footnotes
    Commercial Relationships   Zhenyu Zhong, None; Peizeng Yang, None
  • Footnotes
    Support  This study was supported by grants from National Key R&D Program of China (2016YFC0904000), Natural Science Foundation Major International (Regional) Joint Research Project (81720108009), Chongqing Key Laboratory of Ophthalmology (CSTC, 2008CA5003), Chongqing Science & Technology Platform and Base Construction Program (cstc2014pt-sy10002) and the Natural Science Foundation Project of Chongqing (cstc2017shmsA130073).
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 6693. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Zhenyu Zhong, Peizeng Yang; Development and validation of decision trees for predicting systemic diseases associated with uveitis. Invest. Ophthalmol. Vis. Sci. 2019;60(9):6693.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : Uveitis is often causally associated with systemic diseases, which is not easily recognized by clinicians. The early detection of extraocular manifestations may be useful for an accurate diagnosis, but a systematic workup has never been proposed. We aim to develop and validate models to predict systemic diseases associated with uveitis orientated on extraocular manifestations using decision tree algorithms in a large cohort of uveitis cases.

Methods : We examined the records of 15,373 consecutive uveitis cases, who were referred to our uveitis center across mainland China between April 2008 and August 2018. We investigated demographic characteristics, ocular involvement, diagnosis and systemic manifestations for each patient. A total of 25 variables including sex, age at uveitis presentation, preceding events and general manifestations were analyzed for their potential roles as predictors to systemic diseases using multivariable logistic-regression. Decision tree algorithms were developed to predict systemic entities with predictors selected from medical history and extraocular presentations in uveitis patients. Two thirds of uveitis cases were randomly selected as the training group, and the remaining third was used as the testing group to validate the models with parameters, including sensitivity, specificity, classification accuracy and the area under the receiver operating characteristic curve (AUC).

Results : With the predictors identified, we developed decision trees for the four most common uveitis associated systemic diseases, Vogt-Koyanagi Harada (VKH), Behçet’s disease (BD), ankylosing spondylitis (AS) and juvenile idiopathic arthritis (JIA), using 10,173 samples randomly selected. When we tested the models with the remaining set of 5,200 patients, excellent specificity and accuracy (all ≥ 95.0%) were found, although sensitivity was limited, except for BD (sensitivity = 79.9%) . An AUC above 0.9 was found for BD (AUC, 95% CI: 0.951, 0.937-0.964) and JIA (0.986, 0.978-0.993). For VKH and AS, AUC values were 0.835 (0.815-0.855) and 0.789 (0.755-0.824), respectively.

Conclusions : Analysis of extraocular manifestations may be beneficial for an early diagnosis of systemic diseases associated with uveitis. Further development of decision tree models provides specific workup steps during the diagnosis process for a given patient with uveitis.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×