Psoriatic arthritis (PsA) is a type of arthritis characterised by its association with the disfiguring skin disease psoriasis. Around 20% of people with psoriasis will go on to develop PsA, which can cause permanent joint damage and disability. Common joints affected include the hands, feet, and spine, making daily tasks painful and resulting in a lower quality of life. Diagnosis of PsA can be difficult due to its similarities with other diseases, and it often remains undiagnosed. This has negative effects on health outcomes, as a delay in treatment increases the chances that permanent joint damage has already occurred. The primary aim of this study is to analyse anonymized primary care records of patients with psoriasis to characterise patterns in the clinical symptoms, blood tests and prescriptions of patients who go on to develop PsA. Furthermore, statistical models will be produced to quantify psoriasis patients’ risk of developing PsA, which could be used to improve disease screening selection. A further complication of PsA diagnosis is that 10-15% of patients may not be clinically diagnosed with psoriasis before they develop PsA. A secondary study objective will be to characterise patterns in the clinical symptoms, blood tests and prescriptions in the 5 years preceding any PsA diagnosis, regardless of psoriasis status. This study will improve patient care by aiding primary care physicians in identifying PsA symptoms, allowing patients to receive further diagnosis and treatment by rheumatologists in secondary care earlier, improving disease outcomes.
Psoriatic arthritis (PsA) is a systemic, inflammatory disease associated with the skin disease psoriasis. PsA can be difficult to diagnose in primary care due to its heterogeneous nature and similarity to other forms of arthritis. This study’s objective is to build predictive models to assist with the earlier detection of PsA in primary care. The primary aim is to develop a novel temporal Bayesian network to model patterns of clinical symptoms, blood tests and prescriptions associated with an increased probability of developing PsA in a cohort of patients with incident psoriasis. The model’s performance will be compared to two common approaches for similar tasks in the literature: a Cox proportional hazards model and a neural network. This will be a cohort study consisting of cases of incident psoriasis so that the disease trajectory can be followed prospectively over time in the data.
However, not all PsA cases in the CPRD will have an incident psoriasis diagnosis. This could be due to psoriasis diagnosis concurrently or after their PsA diagnosis, or before the patient’s data entered the CPRD. To include these scenarios in the study, a secondary aim will be to incorporate all incident cases of PsA in a case-control study. Cases (those with incident PsA) and controls (those with no PsA diagnosis) will be required to have at least 5 years of data collection before the matched PsA index date to capture the progressive nature of the prodromal phase of PsA. This methodology will be used to verify that the direction of associations found in the primary study extends to the wider PsA population. Findings from this study could highlight characteristics of the pre-clinical phase of PsA, and predictive models could aid screening procedures by selecting patients at an elevated risk for referral to secondary care rheumatologists.
PsA diagnosis; clinical symptoms, blood tests and prescriptions that could be predictors of PsA.
Theresa Smith - Chief Investigator - University of Bath
Alex Rudge - Corresponding Applicant - University of Bath
Anita McGrogan - Collaborator - University of Bath
Julia Snowball - Collaborator - University of Bath
Neil McHugh - Collaborator - University of Bath
Rachel Charlton - Collaborator - University of Bath
William Tillett - Collaborator - University of Bath