Project Summary

What Is the Research About?

Many diseases have a relapsing remitting disease course, whereby disease symptoms tend to wax and wane over time, with or without treatment. The ups and downs of disease activity (i.e., disease trajectory) can make it difficult to determine if a therapy is working. It is also especially challenging for patients as they go through a roller coaster of ill health.

In this study, the research team will develop a machine learning (ML) method to predict disease trajectories in patients with relapsing remitting diseases, with a focus on juvenile idiopathic arthritis. The team wants to learn if the method can help predict how likely it is for a patient to achieve an inactive disease state and identify patients who can achieve inactive disease without use of medications that can potentially have serious side effects. The team will also apply the method to study how different factors interact over time to affect disease trajectories.

The ability to predict disease trajectories can help clinicians decide which care management strategy will work best for a patient given the patient’s past disease history and identify patients who are likely to do well without use of medications.

How Can This Project Help Improve Research Methods?

Currently, most studies evaluate outcomes in relapsing remitting diseases at a prespecified time interval, which may not accurately reflect how a therapy affects disease trajectories over time. Predicting disease trajectories, in the context of treatment patterns and other factors affecting disease outcome that also change over time, is a challenging task that cannot be adequately addressed by most conventional statistical approaches. Furthermore, accurate prediction of disease trajectories requires frequent data collection, and researchers may not have this data.

This project will help improve research methods for predicting disease trajectories. The research will adopt an approach that engages clinicians and patient partners in the model development process so that the models reflect both domain knowledge and the lived experience of patients. Lessons learned from this process will inform other studies on how best to incorporate knowledge from stakeholders into ML models. The research team will also evaluate the utility of patient-reported outcomes (PROs) in facilitating more accurate prediction of disease trajectories.

What Is the Research Team Doing?

The team will develop and apply a ML approach, known as Dynamic Bayesian Network, to model and predict disease trajectories. The research has two parts. In the first part, clinical data such as disease symptoms, treatment patterns, laboratory results taken during clinic visits and adverse events will be used to develop the models. These data are captured in electronic health data. In the second part, the team will evaluate if use of PROs can help predict disease trajectories. For this analysis, PROs collected during and between clinic visits will be incorporated into the models. PRO data between clinic visits can provide more frequent assessment of disease activity, which may improve prediction of disease trajectories. The team will use between-visit PRO data collected via a mobile app platform to test if increasing the frequency of PRO data collection can enhance prediction of disease trajectories.

*Methods to Support Innovative Research on AI and Large Language Models Supplement
This study received supplemental funding to build on existing PCORI-funded comparative clinical effectiveness research (CER) methods studies to improve understanding of emerging innovations in large language models (LLMs).

Project Information

Mei-Sing Ong, PhD
Harvard Pilgrim Health Care ,Inc.

Key Dates

36 months
March 2023


State State The state where the project originates, or where the primary institution or organization is located. View Glossary
Last updated: April 12, 2024