Patient Centered Adaptive Treatment Strategies Using Bayesian Causal Inference

During routine clinical care, patients and physicians are often confronted with the following
questions: "Given my (my child's) responses to the previous treatments, what is the best
treatment option for me (my child)?" (by a patient/parent) and "What treatment should we
recommend to patients who fail to respond to the first (or second) line of treatment?" (by a
physician). Both questions are at the heart of patient centered outcomes research and
clinical care, yet answers to these questions are seriously hindered by the lack of adequate
analytic methods that appropriately take into account the fact that treatments, as well as
the determinants of a treatment decision, vary over time during the course of the disease.
Case in point: despite many medication options, polyarticular Juvenile Idiopathic Arthritis
(pJIA) is often refractory, and requires better adaptive treatment strategies (ATS). Three
ATS were recommended by a panel of experts for pJIA patients, but they need adequate analysis
methods to evaluate and identify better ATS using observational data. Motivated by our
patient-centered questions, and rigorously designed to evaluate the clinical effectiveness of
patient centered adaptive treatment strategies (PCATS), the proposed method development will
directly address: "development and dissemination of methods for adequate analysis of data in
cases where the treatment/exposure varies over time", an area of interest identified by
PCORI.

Accomplishing the proposed study will provide much needed double robust Bayesian causal
inference methods that take the challenges of analyzing large registry and electronic health
records including model uncertainty, large dimensional covariates and the unmeasured
confounders, into account. A web-based userfriendly analytic computational package will be
developed to allow easy application of the proposed methods. These developments will: 1)
immediately offer methods and computational tools for evaluating clinical effectiveness and
informing optimal ATS, 2) in the near future, enable shared-decision making tools for
identifying optimal PCATS at the point-of-care, and 3) eventually enable a rapid learning
system that will facilitate optimal PCATS. This study will have an immediate impact to
children and stakeholders of JIA and a long-term broad impact to many chronically ill
patients. Successful completion of this project will significantly move PCORI closer to its
mission of helping "people make better informed healthcare decisions and improve healthcare
delivery and outcomes".

Participants should meet the following inclusion criteria:

1. Age ≤19 years at baseline

2. Diagnosed with ployarticular course of juvenile idiopathic arthritis (pcJIA) follow
the operational case definition of pcJIA presented in Table 2 of the journal article
published by Ringold et al. (2014).

3. Meet new patient definition, i.e. diagnosed with pcJIA no more than 6 month at the
first clinical encounter captured in the database

4. Taken DMARDs no more than 9 months after diagnosed with pcJIA

Participants will be excluded if they meet the following exclusion criteria:

1. Systematic JIA patients according to the ILAR code

2. Patients with comorbidities of inflammatory bowel disease (IBD), celiac disease, and
trisomy 21.