The RAPID project: Using electronic healthcare records to develop tools to help GPs determine who is at risk of getting rheumatoid arthritis

Status: Ongoing

Rheumatoid arthritis (RA) is a long-term disease that involves inflammation of the joints causing pain and stiffness. The disease has impacts well beyond the joint, causing often quite profound fatigue as well as impacts on the heart, lungs and bones. If joint inflammation is not well controlled, damage to the joints can occur rapidly and can be permanent. This can lead to significant problems in the long-term, causing pain and difficulty with function, including with daily activities and work. Currently, there is no cure for RA, and most people need to take medicines to control inflammation, often for the rest of their lives. These medicines can have serious side effects, and do not work for everyone with RA. In addition, many of them are very expensive.

Early treatment is better for patients. If a patient who starts to develop RA is treated within three months of the onset of their symptoms, they are more likely to go into remission and less likely to have long-term joint damage. Unfortunately, for many patients, early treatment does not happen. There are many reasons for this including patients delaying seeking help from GPs, delays in GPs identifying patients as being at risk of RA or referring them to rheumatologists promptly, and long hospital waiting lists once a GP has made a referral.

RA can be hard for GPs to diagnose as a large proportion of GP consultations are about joint problems, but most are not related to inflammatory arthritis such as RA. At the same time, the early symptoms of RA can be difficult to recognise. Supporting GPs to recognise patients at risk would help them to refer these patients to a rheumatologist rapidly. This would increase patients’ chances of being diagnosed and treated early.

So far there is no reliable method to accurately identify people who are at risk of RA in general practice. Lots of information about a patient’s risk of RA is held in their electronic healthcare record. This includes their demographic features such as age, sex and previous medical history including past conditions, symptoms and blood tests. This information could help a GP identify patients with RA, but GPs do not always have the time to combine this information to predict if a patient has RA. We want to find out whether a computer program could aid GPs in their clinical decision making by drawing together this information to make predictions on the patient’s risk of RA. We are investigating this in the RAPID project, which is funded by the National Institute for Health and Care Research.

Project aims

A patient’s electronic healthcare record contains valuable information that can help determine their risk of RA. Researchers can access certain parts of these records in an anonymised form and this can be used to conduct research. Previous research has shown that these anonymised healthcare records can be used to build programs that can predict if a patient is at risk of a certain health condition. This has been found to be effective for a variety of conditions, but this has not yet been done for RA. As a result, using these electronic healthcare records, we are designing, developing, and testing a computer program called RAPID, which will identify patients at high risk of early RA and support GPs in making clinical decisions for their care.

To make this programme, we looked at existing research and spoke to doctors and patients to find out what makes a patient more likely to have RA. These included symptoms of RA such as hand pain and stiffness, pre-existing health conditions, and patient characteristics such as age, sex, and smoking history. Using analysis within the computer program, we are working out how much each risk factor is likely to increase the patient’s risk of having RA. These will then be combined by the computer program to work out the likelihood that each patient has RA given their mix of symptoms and other risk factors. We are also checking that our prediction programme does not discriminate against people from different socioeconomic or ethnic groups, who might be more or less likely to see their doctor about some symptoms. We have already published some of our analyses to check this in a scientific journal.

We are also looking at the performance of the computer program. We are doing this by looking at how well the program can tell apart who does and does not have RA (this is called discrimination) and whether the number of RA diagnoses the program predicts is similar to the actual number of RA diagnoses in the patient records (this is called calibration). To do these checks, we will first use the anonymised records used to build the program. However, this does not tell us how well the program would do in patients who are different to those whose records we used to build the program. To solve this, we will do the same tests but using patient records from a different dataset of electronic healthcare records and another research team will independently repeat the tests in patient records from Wales.

Patient and public involvement

Many of our patient research partners have been involved in our previous studies that have investigated why treatment is often delayed for people with RA. They have told us that it can take a long time for GPs to recognise that they should be referred to a rheumatologist and that it is important to find ways to reduce referral time.

Patient research partners have been involved in RAPID from the beginning. They helped us to apply for funding to do the research, and to decide what symptoms should be included in our prediction model. One of them has also taken part in monthly project update meetings, to make sure that patients’ views are represented and help us make sense of our findings. They have also helped us to write reports of our findings, including this summary. R2P2 patient research partners were also invited to an event where we presented our findings so far to patients, GPs, rheumatologists, patient organizations and policy makers, and asked for their advice about what further research is needed and how best to use our findings in ways that will be useful to GPs and beneficial for patients.

This article was written by Ben Hammond (Medical PhD student), Elspeth Insch (Patient Research Partner) and Marie Falahee (Researcher).