Rheumatoid arthritis (RA) is a painful, long-term disease that involves inflammation of the joints and other parts of the body. If the inflammation is not controlled, it can cause irreversible joint damage and disability. If a patient is treated within three months, it is more likely that symptoms will ease and the patient is less likely to have long-term joint damage (1, 2). Previous research shows that patients often experience difficulties in getting referral to a rheumatologist (3). Supporting GPs to recognise patients at risk would help them to refer these patients rapidly. This would increase patients’ chances of being diagnosed and treated early. However, there is no reliable method to accurately identify people who are at risk of RA in general practice.
We aim to develop and validate a tool to predict who is at risk of developing RA, based on patients’ symptoms and other information recorded in their medical records. We will develop it as a web-based risk calculator and explore how to integrate it with electronic health records as a GP alert tool. We will use existing medical databases to identify signs and symptoms that are associated with a higher risk of RA. These will include patient characteristics such as age, sex, and ethnicity, and clinical characteristics such as symptoms and laboratory test results. We will develop a prediction tool that can be used to calculate risk for someone based on these characteristics. We will check this works well by testing it in different databases.
We will undertake the development and validation of a risk prediction model for rheumatoid arthritis (RA) termed RAPID in patients presenting with signs and symptoms indicative of RA. This involves identifying candidate predictors, developing and internally validating the model, and externally validating RAPID. We will use CPRD Aurum to identify the candidate predictors, develop RAPID, and for internal validation. CPRD Gold will be used for independent external validation of RAPID.
Adult patients 18 years or over and registered with participating general practices between 01 January 2000 and 31 December 2020 will be included in the study. The cohort will be patients presenting with signs and symptoms indicative of RA (identified by relevant SNOMED CT codes). Patients with a previous history of RA will be excluded. Each patient can contribute to the cohort after a minimum registration period of 12 months. Outcomes (at 5 years, and also yearly 1-5 years) will be defined as clinical codes for RA and evidence of initiation of a disease-modifying anti-rheumatic drugs (DMARDs). Patients will be followed from index date until the earliest of outcome date, patient transfer date, last date of data collection, death date, or study end date.
Predictors that are strongly associated with the outcome or explain observed variation in the outcome become candidates for inclusion in RAPID. Potential candidate predictors for RAPID will be chosen based on review of the literature, the outcomes of a consensus meeting for clinical relevance, and assessment of the data quality of the potential candidate predictors recorded in the databases. Using all the candidate predictors, we will develop RAPID using a Cox proportional hazards model. Internal validation will be carried out using bootstrapping. External validation includes assessment of discrimination (C-statistics), calibration (calibration plots, calibration-in-the large, calibration slope), and clinical utility (net benefit and decision curves).
• Adjusted hazard ratio (aHR) for the risk of developing RA in patients with signs and symptoms.
• Incidence rate of RA in the cohort.
Karim Raza - Chief Investigator - University of Birmingham
Dawit Zemedikun - Corresponding Applicant - University of Birmingham
Ben Hammond - Collaborator - University of Birmingham
Joht Singh Chandan - Collaborator - University of Birmingham
Krishnarajah Nirantharakumar - Collaborator - University of Birmingham
Nicola Adderley - Collaborator - University of Birmingham
Ben Hammond - Collaborator - University of Birmingham
Joht Singh Chandan - Collaborator - University of Birmingham
Patient Level Index of Multiple Deprivation