OBJECTIVES: A patient's view on and subsequent compliance with treatment may vary depending on the pathology, treatment constraints and patient profile. Thus, the selection of a treatment that better fits the profile and therefore the preferences of a patient could encourage better compliance. The objective of this study was to identify profiles of patients sharing similar expectations through a Multiple Correspondence Analysis (MCA) associated with unsupervised clustering methods.

METHODS: An online questionnaire was presented, between August 2018 and April 2019, to French patients registered on the Lupus community of Carenity.com, a patient platform generating real-life data. MCA and unsupervised clustering methods (hierarchical, k-modes, partitioning around medoids) have been conducted using the medico-social profile and two preference variables ("dosage form" and "most important criterion for treatment, efficacy aside"). The robustness of these results was assessed by internal validation with comparisons between clustering methods.

RESULTS: 268 lupus patients responded to the survey, 96% were women with an average age of 44.3 years. MCA and unsupervised clustering methods have highlighted three clusters, similar between clustering methods. Cluster 1 (59%) included patients with few comorbidities, who do not have the capacity to identify approaching flare-ups, already mostly under oral treatments and favoring an oral treatment with limited side effects. Cluster 2 (13%), included younger patients, who had already participated in clinical trials, favoring implants and the compatibility of treatments with pregnancy. Cluster 3 (28%) included patients who had little to no control over their lupus and many comorbidities, a third being already on injectable treatments, desiring mainly injectable options and with the main target of decreasing corticosteroids.

CONCLUSIONS: This study demonstrates the value of real-life data directly generated through online patient communities to identify patient profiles with similar treatment preferences. Personalization of care based on patient profiles could help to improve compliance.