Drug-adherence-trajectories

Analysis code for the manuscript "Longitudinal patterns and determinants of statin adherence in over one million individuals from Finland and Italy" - Andrea Corbetta, Katherine M. Logan, Matteo Ferro, Luisa Zuccolo, Markus Perola, Andrea Ganna, Emanuele Di Angelantonio, Francesca Ieva (medRxiv 2026.01.26.26344722; doi: https://doi.org/10.64898/2026.01.26.26344722)

Pipelines (Finregistry / Finland and Regione Lombardia / Italy) to construct long-term medication purchase histories, compute adherence time-series (e.g., MPR), smooth trajectories, and run functional trajectory modelling with downstream clustering and association analyses.

Important: This repository contains code only. The underlying registry data are not included. Many scripts use environment-specific/hard-coded paths that you will need to adapt.

---

Data inputs (high level)

You will need individual-level longitudinal purchase/dispensing data with variables of this type (names differ by source):

• person identifier (e.g., ID)
• dispensing/purchase date (e.g., DATE)
• medication identifier (e.g., ATC code)
• quantity / pack information to derive pills and days supplied
• optional daily dose assumptions/derivations (some scripts filter to 1 tablet/day)

The Lombardia pipeline includes logic to merge nearby purchases (e.g., within 7 days), compute pills per purchase, and derive both point and cumulative MPR (including capped versions at 1).

---

Workflow (analysis scripts)

Scripts are numbered to reflect the intended order.

1. Process purchase histories

   • Finregistry/Analysis/1_process_trajectories.R
   • RegioneLombardia/Analysis/1_process_purchases.R

2. Select cohort & summarise

   • Finregistry/Analysis/2_select&summarise.R
   • RegioneLombardia/Analysis/2_select_summarise.R (and _4covid.R variant)

3. Smooth adherence trajectories

   • Finregistry/Analysis/3_smooth_adherence.R
   • RegioneLombardia/Analysis/3_smooth_traj.R

4. Functional data analysis

   • */Analysis/4_fda.R with helper functions in */Analysis/fda_funs.R

5. Create covariates / cluster trajectories

   • 5.1_create_endpoints.R, 5.2_create_socioeco.R, 5.3_create_medication.R
   • Lombardia additionally: 5.4_create_visits.R
   • Clustering scripts:
     • Finregistry/Analysis/5_cluster_curves.R
     • RegioneLombardia/Analysis/5_cluster.R

6. Assemble covariate table

   • */Analysis/6_covariates.R

7. Fit models

   • */Analysis/7_models.R
   • Lombardia also: models_PC.R and summary_stats.R

---