Figure 1. Summary of filtering chronic and polypharmacy patients where two years of prescription data are considered.

First, all data is partitioned into time periods of three consecutive months. Each three-month interval is used to: first, define the number of concurrent use of medications, and second, check for T2D or CVD medications. Next, the year before prediction time point (PTP) is used to remove all patients with previous polypharmacy (i.e., number of concurrent medications is higher than 4) or previous CVD or T2D chronic condition (i.e., at least one chronic medication is taken every 3 months). Finally, the year following PTP is used to select only the patients with a chronic condition, while polypharmacy in this year is used to define positive and negative patients.