Adherence to mesalazine and identification of patients with ulcerative colitis in healthcare administrative databases of Tuscany (Italy)

Background: Drug supply can be used to identify patients with a specific disease in healthcare administrative databases (HAD), particularly when they have a univocal indication. For drugs with multiple indications, other elements must be considered, such as dosing or scheduled regimen. This is the case of mesalazine (MZ), a drug indicated for ulcerative colitis (UC), but also for Crohn's disease and other not classifiable inflammatory bowel diseases. Objectives: To explore adherence trajectories to MZ, in the attempt of identifying variables useful to detect patients with UC in HAD. Methods: In this retrospective cohort study, data was retrieved from administrative healthcare databases in Tuscany. Patients who had at least a first MZ dispensation between 6/1/2011 and 6/30/2016 (index date, ID) were included. We performed four sub-group analyses: 1) Patients with at least 1 MZ dispensation in the follow-up (FUP); 2) Patients with at least 2 MZ dispensations in the first 3 months of FUP; 3) Patients with MZ dosage >1.5 grams between the 1st and the 2nd dispensation; 4) Patients with MZ dosage >2.5 grams between the 1st and the 2nd dispensation. Each sub-group was stratified by age: ≤65 and >65, separately. We estimated adherence to MZ monthly, through the Medication Possession Ratio and computed adherence trajectories with a 3 step procedure: 1) computation of 24 statistical measures; 2) factor analysis; 3) cluster analysis. We performed two sensitivity analyses: 1) Patients >65 years old with UC ICD-9 code, 2) Patients ≤65 years old with the UC exemption code. Results: The sub-groups 1, 2, 3 and 4 included 47.750, 34.037, 29.730 and 26.165 MZ users, respectively. In all the sub-groups and stratification by age, 3 clusters of adherence were identified: high, medium low and low. The high adherence trajectory showed a mean adherence over 40% in the FUP, while patients in the low and medium low trajectories discontinued MZ within 5 and 15 months, respectively. The first and second sensitivity analysis included 409 and 1514 patients, respectively. Two clusters were identified in both analyses: high, around 80%, and moderate, up to 30% at the end of FUP. No MZ discontinuation was observed in the sensitivity analyses on ascertained UC patients. Conclusions: In the main analyses, two out of three trajectories were characterized by MZ discontinuation, while in the trajectories calculated for the patients with UC codes, no MZ use interruption was detected. This finding could help the identification of true UC patients in the overall population of MZ users.