Utility of Network Science and Intensive Longitudinal Data in the Study of Symptom Associations in Rheumatoid Arthritis

Student thesis: Doctoral ThesisDoctor of Philosophy


Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune condition primarily affecting the small joints of the extremities. RA patients suffer physical symptoms and has a high comorbidity with depression, with a prevalence rate of around 17%. These symptoms fluctuate wildly; sometimes throughout a day, and cross-sectional data analysis will only allow for one time point to be evaluated. Thus, it is important to carry out intensive follow up to allow tracking through fluctuations. A scoping review carried out showed that there a lack of high frequency follow up studies to investigate the associations between multiple physical and/or psychological symptoms in the field of musculoskeletal disorders. These showed important gaps in literature in the field of RA that needs to be addressed.

The overarching aim of this thesis is to explore physical and psychological symptom associations using advanced quantitative methods on a longitudinal dataset in RA patients; and to explore the feasibility of network science in this field of research. This will be addressed via three main objectives: 1) to investigate the feasibility of collecting intensive longitudinal data with the help of a wearable device in the field of RA; 2) to explore associations between physical symptoms, psychological symptoms and other important variables in RA patients; and 3) to test the usability of the network approach in evaluating multiple symptoms. These objectives will be addressed through data collected from two studies involving intensively collected (multiple times per-day) symptoms ratings: IA-COVID and APPro. In addition, cross-sectional data from the TITRATE-US study and routinely collected data from the KCL Rheumatology IMPARTS Patient Reported Outcome system are used to evaluate the usefulness of symptoms network approaches prior to applying these methods to longitudinal data.

To address the feasibility of collecting intensive longitudinal data, the APPro study showed that there was a success rate of 33.8% when recruiting patients to the study and the APPro study showed that the average compliance rate was 88.75%. This is higher than the proposed recommended compliance rate of 80% by several studies, and higher than the 73% that was shown in a review on EMA studies that utilises wearables as well on youth. Both longitudinal studies used mixed effects regression and provided novel insights into the bidirectional association between physical and psychological symptoms in RA patients. In the IA-COVID study, it was discovered that there were significantly less social contact and higher loneliness level during period of lockdown during the COVID-19 pandemic, and during this state, increased social contact was significantly associated with lower physical symptoms in the next time period. It also showed that positive affect was the only symptom that influenced physical activity in the next time period, suggesting that high positive affect would increase physical activity in the next time period. The APPro study demonstrated that there were significantly lower physical symptoms and, surprisingly, lower positive affect after the initiation of a new biologic treatment. After a new treatment, psychological symptoms have a significant impact on physical symptoms in the next period.

When examining the relationship between multiple symptoms, it becomes complicated to separately fit and interpret many different models (i.e. at least one model per-outcome assessed). This limited the analyses described above to focusing on broad constructs of psychological well-being rather than individual symptoms. Network science approaches were utilised in every empirical chapter to provide an insight into the associations between specific physical and psychological symptoms. Distinct clusters of physical symptoms, psychological symptoms, and inflammatory markers in RA were identified using a network approach. Individual network plots also showed that fluctuant symptom plot differs from a stable symptom plot where fatigue is highly connected to psychological symptoms rather than physical symptoms. The influence of each symptom was also looked at in APPro using centrality values. The influence of joint stiffness dropped dramatically after a new treatment, showing the effects of the treatment on inflammation. Both before and after symptom plots also showed psychological symptoms to be the most influential nodes, further showing the importance of positive and negative affect in RA patients.

In conclusion, this thesis displayed a framework for the recruitment and assessment of intensive longitudinal data in the field of RA. It also discovered several novel associations between symptoms and quality of life variables, and also reinforced the importance of psychological symptoms in a RA patient, especially after a new treatment. Fatigue was also discovered to be the symptom that has the most influence on the activation of psychological symptoms, and were shown to be affected by psychological symptoms tremendously in different situations. Network science also proved to be a methodology that could reveal new information, but more work is required to discover its full potential.
Date of Award1 Sept 2022
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorSam Norton (Supervisor), James Galloway (Supervisor) & Faith Matcham (Supervisor)

Cite this