ACR Meeting Abstracts

ACR Meeting Abstracts

  • Home
  • Meetings Archive
    • ACR Convergence 2020
    • 2020 ACR/ARP PRSYM
    • 2019 ACR/ARP Annual Meeting
    • 2018 ACR/ARHP Annual Meeting
    • 2017 ACR/ARHP Annual Meeting
    • 2017 ACR/ARHP PRSYM
    • 2016-2009 Meetings
    • Download Abstracts
  • Keyword Index
  • Advanced Search
  • Your Favorites
    • Favorites
    • Login
    • Register
    • View and print all favorites
    • Clear all your favorites
  • Meeting Resource Center

Abstract Number: 1371

Create an Algorithm of Outcome Forecasting and Decision Making for RA Treatment: Data Mining and Machine Learning via the Smart System of Disease Management (SSDM)

Rong Mu1, Chun Li 2, Jing Yang 3, Xiaohan Wang 4, Bin Wu 5, Fengxiao Zhang 6, Yong Wang 7, Li Qin 8, Cundong Mi 9, Huifang Guo 10, Wei Wei 11, Wei Liu 12, Qingchun Huang 13, Jing Lu 14, Yang Liu 15, Huiling Li 16, Bing Wu 17, Hui Xiao 18, Yuhua Jia 19 and Fei Xiao 20, 1People's Hospital, Beijing University Medical School, Beijing, China (People's Republic), 2People's Hospital, Beijing University Medical School, Beijing, Beijing, China (People's Republic), 3Central Hospital of MianYang, Mianyang, Sichuan, China (People's Republic), 4Anyang district hospital, Fuyang, Hainan, China (People's Republic), 5Chongqing Hospital of Traditional Chinese Medicine, Chongqing, Chongqing, China (People's Republic), 6Hebei General Hospital, Shijiazhuang, Hebei, China (People's Republic), 7The first Hospital Affiliated to AMU (Southernwest Hospital), Chongqing, China (People's Republic), 8The No.3 People’s Hospital of HuZhou City, Huzhou, Zhejiang, China (People's Republic), 9The Second Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi, China (People's Republic), 10The Second Hospital of Hebei Medical University, Shijiazhuang, Hebei, China (People's Republic), 11Tianjin Medical University General Hospital, Tianjin, Tianjin, China (People's Republic), 12First Teaching Hospital of Tianjin University of Traditional Chinese Medicine, Tianjin, China (People's Republic), 13The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou, China (People's Republic), 14First Affiliated Hospital of China Medical University, Shenyang, Liaoning, China (People's Republic), 15Affiliated Hospital of Inner Mongolia Medical University, Huhehaote, Nei Mongol, China (People's Republic), 16Hubei Hospital of Traditional Chinese Medicine, Wuhan, Hubei, China (People's Republic), 17Shanghai Gothic Internet Technology Co., Ltd, Shanghai, China (People's Republic), 18Shanghai Gothic Internet Technology Co., Ltd, shanghai, Shanghai, China (People's Republic), 19Shanghai Gothic Internet Technology Co., Ltd., Shanghai, Shanghai, China (People's Republic), 20Shanghai Gothic Internet Technology Co., Ltd., shanghai, China (People's Republic)

Meeting: 2019 ACR/ARP Annual Meeting

Keywords: Big data, decision analysis and self-management, Electronic Health Record, Rheumatoid arthritis (RA)

  • Tweet
  • Email
  • Print
Save to PDF
Session Information

Date: Monday, November 11, 2019

Session Title: RA – Diagnosis, Manifestations, & Outcomes Poster II: Treatments, Outcomes, & Measures

Session Type: Poster Session (Monday)

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

Background/Purpose: Daily health care deals with a wide spectrum of RA patients from statuses of remission (R) to disease activities at low (L), moderate (M) and high (S) based on DAS28. Each patient has 4 potential outcomes of either R, or L, or M, or S at certain time layer. But predicting efficacy and selecting regimens are in trial-and-error style. The objective of this study is to develop an algorithm for efficacy prediction and regimen selection in RA via SSDM.

Methods: SSDM includes 2 APPs for both physicians and patients. The patients input medication and laboratory test results, and perform self-evaluation (DAS28) after training. The data synchronizes to the physician’s APP and the advices could be delivered.

In order to develop a prediction model, top 2 regimens, Leflunomide (LEF, 5-10 mg/d) and LEF+MTX (7.5-10mg/w) were selected as samples from SSDM database. Repeat evaluation of DAS28 scores were extracted at 3 months interval.

The study was divided into 2 phases: phase 1. To develop Markov model and generate the probability (P) of prior believe, P(B|A), based on Bayes Theorem, and phase 2. Calculate P of posterior believe, P(A|B), and validate the reliability comparing with true evidence of outcome via Intra-group correlation coefficient (ICC).

Results: From June 2014 to May 2019, 54,149 RA patients from 587 centers registered in SSDM. Among those with 3 months data available, 416 patients with LEF and 307 with LEF+MTX who registered before June 1 of 2018 were enrolled in phase 1, and 180 patients with LEF and 120 with LEF+MTX after June 1 of 2018 were included as phase II. There were no significant difference considering gender, age, race, disease history and DAS28 distribution between 2 phases. Phase I: Four baseline status (P(A)) and 16 outcomes at month 3 (P(B)) for LEF or LEF+MTX are presented in graph 1a and 1b, respectively, based on Markov Chain. The algorithm of probabilistic inference for next month 3 outcome was created as Σ p{Xt+1|Xt (Rt, Lt, Mt, Ht)}=1 0< t< ∞. The beta distribution and P(B|A) were generated. Phase II: The probability of outcome prediction (P(B)) for patients was calculated using the DAS 28 at baseline via the Bayes algorithm: P(A|B)=P(B|A)*P(A)/P(B) and were shown in graph 2a & 2b. The true outcomes of DAS 28 were obtained through repeat self-valuation at month 3 served as evidence (graph 3a & 3b). Comparing the distribution of evidence with those in P(A|B), ICC for LEF was 0.90, and ICC for LEF+MTX was 0.96, which indicated high inter-observer reliability. Based on the algorithm, the inferences of efficacy for both LEF and LEF+MTX were obtained. The advices were that LEF at low dose was good for RA with R and L, but for RA with H, adding MTX was preferred. Both were good for M.

Conclusion: Through patterns extraction, data mining and modeling, a master algorithm of efficacy prediction and decision making for RA were developed. Following data inputting and machine leaning, an artificial intelligent system in assisting clinical practice may be achieved with SSDM.


Disclosure: R. Mu, None; C. Li, None; J. Yang, None; X. Wang, None; B. Wu, None; F. Zhang, None; Y. Wang, None; L. Qin, None; C. Mi, None; H. Guo, None; W. Wei, None; W. Liu, None; Q. Huang, None; J. Lu, None; Y. Liu, None; H. Li, None; B. Wu, None; H. Xiao, None; Y. Jia, None; F. Xiao, None.

To cite this abstract in AMA style:

Mu R, Li C, Yang J, Wang X, Wu B, Zhang F, Wang Y, Qin L, Mi C, Guo H, Wei W, Liu W, Huang Q, Lu J, Liu Y, Li H, Wu B, Xiao H, Jia Y, Xiao F. Create an Algorithm of Outcome Forecasting and Decision Making for RA Treatment: Data Mining and Machine Learning via the Smart System of Disease Management (SSDM) [abstract]. Arthritis Rheumatol. 2019; 71 (suppl 10). https://acrabstracts.org/abstract/create-an-algorithm-of-outcome-forecasting-and-decision-making-for-ra-treatment-data-mining-and-machine-learning-via-the-smart-system-of-disease-management-ssdm/. Accessed April 17, 2021.
  • Tweet
  • Email
  • Print
Save to PDF

« Back to 2019 ACR/ARP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/create-an-algorithm-of-outcome-forecasting-and-decision-making-for-ra-treatment-data-mining-and-machine-learning-via-the-smart-system-of-disease-management-ssdm/

Advanced Search

Your Favorites

You can save and print a list of your favorite abstracts by clicking the “Favorite” button at the bottom of any abstract. View your favorites »

ACR Convergence: Where Rheumatology Meets. All Virtual. November 5-9.

ACR Pediatric Rheumatology Symposium 2020

© COPYRIGHT 2021 AMERICAN COLLEGE OF RHEUMATOLOGY

Wiley

  • Home
  • Meetings Archive
  • Advanced Search
  • Meeting Resource Center
  • Online Journal
  • Privacy Policy
  • Permissions Policies
loading Cancel
Post was not sent - check your email addresses!
Email check failed, please try again
Sorry, your blog cannot share posts by email.
This site uses cookies: Find out more.