Predicting the diagnosis of inflammatory and autoimmune arthritis using a data driven approach.

Inflammatory arthritis is easy to diagnose once the characteristic changes in the joints is present, for example fusion of the sacroiliac joints for those with ankylosing spondylitis. However, the characteristic changes are normally present after damage has been done, the person has already experienced a great deal of pain and damage and the window of opportunity to stop the disease in the early stages has been lost. An early diagnosis means that disease modifying drugs can be used to slow the progression and possibly greatly change the course of the disease or switch it off. This is especially important as the loss of function occurs in the first 10 years of disease onset and loss of function is the main factor leading to high costs of disease in terms of loss of work and need for assistance and a carer. However, in the early stages many symptoms of inflammatory arthritis are common to aging, injury or wear and tear, for example fatigue, joint pain, swelling. This means they can easily be dismissed by the patient and the doctor and simply treated with pain killers and anti-inflammatory and long delays can occur when symptoms do not resolve. Thus, there is a need to identify predictors that can be used by a GP to rapidly direct and refer patients to specialist treatment. However, considering the waiting lists to see a specialist, it is also important that the predictors have a high positive predictive value so that extra burden is not placed on the hospital system causing greater delays in diagnosis of early disease. New investments in bringing together routinely collected data and new developments in pattern detection methods using large complex datasets means that it is now possible to take advantage of a data driven method approach to tackle this problem. Until now the approach has been taking clinical knowledge from secondary care physicians and applying this knowledge to what they think should be present in primary care records. This study aims to take a data driven approach and to examine early predictors of very early autoimmune inflammatory arthritis. It is possible that the early predictors are all common symptoms such as an infection, antibiotic use, multiple prescriptions for anti-inflammatory drugs and pain killers, time off work as recommended by GP, fatigue and fever. However, the pattern of these common symptoms and timing of when they occur relative to each other, may be the main predictive factor.

Inflammatory and autoimmune arthritis (e.g. Rheumatoid arthritis, Spondyloarthritis, Psoriatic arthritis, Lupus) affects 1 in 12 women and 1 in 20 men. Aggressive treatment for inflammatory arthritis in the first few months has been shown to reduce the rate of disease progression. There is increasing evidence that the first few months after symptom onset represent a pathologically distinct phase of disease. This very early phase may translate into a therapeutic window of opportunity during which it may be possible to permanently switch off the disease process. However, catching the disease in the very early stages is difficult and for many people it can take years to be given a diagnosis and then this window of opportunity is lost. This fellowship proposes to use a data driven approach to examine if early predictors (such as symptoms, laboratory tests, co-morbidities, co-medication history, administrative codes) can be used to identify those who are likely to be given a diagnosis of inflammatory arthritis. The work examines the frequency of predictive codes before and after the diagnosis in those with the disease compared to controls. However, the temporal relationship is very important and so this requires the development and testing of novel analytical approaches. For example, a severe infection 7 years before a diagnosis may be predictive but an infection at any time in the patients history may not predictive. A pilot aspect of this work examining ankylosing spondylitis and separately psoriatic spondyloarthritis has been funded by the pharmacological company UCB and so this has direct relevance to be commercialised by pharma and technology industries. This Fellowship will build on this parallel work to expand new models developed to other conditions such as Lupus, Rheumatoid arthritis and enteropathic spondyloarthropathy.