Abstract
Introduction
Diabetes affects 537 million people worldwide, with type 2 diabetes (T2D) estimated to account for most cases. Type 1 diabetes (T1D), latent autoimmune diabetes in adults (LADA) and other specific types due to other causes remain under-recognised, especially LADA, given the absence of a standardised definition. In the province of Quebec (Canada), no population-based prevalence and incidence estimates are available for all diabetes phenotypes. Current medico-administrative algorithms fail to distinguish among diabetes subtypes, limiting accurate surveillance and effective prevention and clinical strategies.
Methods and analysis
This retrospective, longitudinal study (1997-2027) will be conducted in Quebec.
Phase 1 will assess classification performance of the Corsenac
et al.
(2022) medico-administrative algorithms. Diagnostic performance metrics will be calculated per phenotype (T1D, T2D, LADA and others), using three independent subsamples as references in a first cohort (A1-A2-A3; n≈5200). It composed respectively of: (A1) self-reported diagnoses of T1D and LADA; (A2) diagnoses of T2D and other specific types due to others causes, confirmed by a physician; and (A3) general population respondents reporting diabetes status and phenotypes (if applicable). Subsamples are structured to capture the diversity and relative proportions of diabetes phenotypes, ensuring sufficient statistical power for population-based and subgroup analyses. All records (A1, A2, A3) will be probabilistically linked with medico-administrative and pharmaceutical claims from the
Régie de l’assurance maladie du Québec
(RAMQ) by the
Institut de la statistique du Québec
(ISQ). Machine learning methods will be then applied to refine algorithmic definitions for the four phenotypes (T1D, T2D, LADA, others).
In phase 2, refined algorithms will be applied to a second medico-administrative population-based cohort, named C (n=50 000) to produce the first simultaneous prevalence and incidence estimates for the four phenotypes.
Analyses will be restricted to individuals continuously covered by RAMQ’s public drug insurance, which provides pharmaceutical data and covers 46% of the Quebec population. Calibration on population margins (phase 1) and standardised inverse probability weighting (phase 2) will reweight estimates (performance metrics in phase 1 and frequencies in phase 2) to represent the general Quebec population.
Ethics and dissemination
The different ethics boards of partner institutions approved the feasibility of the study. The study was registered on ClinicalTrials.gov,
NCT06573905
. Results will be disseminated through scientific and public health channels.