Infection trajectories and dementia risk in type 2 diabetes

Dementia poses a significant global health burden, with rising prevalence and substantial economic and social costs. Type 2 diabetes mellitus (T2DM) is a well-established risk factor for dementia, yet the impact of recurrent infections—a common complication in T2DM—on dementia risk remains unclear. This study investigates the role of longitudinal infection trajectories in dementia development. This retrospective cohort study utilized Taiwan’s National Health Insurance Research Database to identify 33,080 patients with T2DM aged ≥ 40 years who were hospitalized with an infection in 2011. Group-based trajectory modeling characterized infection patterns over the subsequent 5 years (2011–2016). Four trajectory groups were identified in the full cohort: infrequent (N = 29,891), decreasing (N = 2,081), increasing (N = 1,013), and frequent infections (N = 95). We then implemented a landmark design: among patients who were alive and dementia-free on January 1, 2016 (N = 31,667), incident dementia was ascertained during the 5-year outcome period (2016–2020). Cox proportional hazards models estimated the associations between infection trajectories and incident dementia, adjusting for age, sex, frailty, diabetes severity, Charlson Comorbidity Index, and other confounders; death was treated as censoring in the primary models and as a competing event in Fine–Gray sensitivity analyses. During the 2016–2020 landmark outcome period, 1,069 incident dementia cases occurred among the 31,667 patients at risk (798, 218, 47, and 6 cases in the infrequent, decreasing, increasing, and frequent infection groups, respectively). Compared with the infrequent infection group, patients with increasing and frequent infections had significantly higher dementia risk (adjusted hazard ratio [HR] 1.48, 95% CI 1.29–1.69, and 1.92, 95% CI 1.22–3.02, respectively). Five-year cumulative dementia incidence increased stepwise from 3.1% in the infrequent group to 6.3% in the frequent group. Infection trajectories were also strongly associated with all-cause mortality, with adjusted HRs of 2.69 (95% CI 2.51–2.89) and 4.78 (95% CI 4.09–5.58) for the increasing and frequent infection groups, respectively. Recurrent infection trajectories significantly elevate dementia risk in T2DM patients. These findings extend beyond prior single-episode infection studies by introducing a trajectory-based, methodologically robust approach that minimizes survival bias and incorporates frailty and comorbidity indices. They underscore the need for infection prevention and sustained management—such as vaccination, timely antimicrobial therapy, and glycemic control—to mitigate long-term cognitive decline in this high-risk population.