Machine Learning May Diagnose Psoriatic Arthritis Earlier Save

A machine learning approach was applied to two different population cohorts and was shown to be effective at the early identification of undiagnosed psoriatic arthritis (PsA).

It is estimated that PsA affects approximately 0.27% of the adult population, and 20% of those with with psoriasis (PsO). There is great debate and efforts in the dermatology community on how to best identify potential PsA patients and refer them to expert care. 

A proprietary machine learning tool (PredictAI™) was developed for identification of undiagnosed PsA patients.

This retrospective study two adult population cohorts from Maccabi Healthcare Service (2008 to 2020): a general adult population ("GP Cohort") including patients with and without psoriasis and the Psoriasis cohort ("PsO Cohort") including psoriasis patients only. Each cohort was divided into two non-overlapping train and test sets. The PredictAI™ model was trained and evaluated with 3 years of data predating the reference event (the diagnosis of PsA) by at least one year.

From a total of 2096 ultimately diagnosed PsA patients, an earlier diagnosis (from the undiagnosed PsO cohorts) was identified with a specificity of 90% one and four years before the reference event, with a sensitivity of 51% (1yr) and 38% (4 yrs), and a PPV of 36.1% (1 yr) and 29.6% (4 yrs), respectively.

In the GP cohort an earlier diagnosis was made with a specificity of 99%, also 1 and 4 years before the reference event.

Overall, the model achieved a sensitivity of 43% and 32% and a PPV of 10.6% and 8.1%, respectively.

While the specificity of a PsA diagnosis was very high 1-4 years before the reference PsA diagnosis, the sensitivity was better at 1 year (compared to 4 years) before the reference event. It is possible that such machine learning tools may lead to early identification of undiagnosed PsA patients, allowing for earlier intervention and better patient outcomes.