Type 2 diabetes is common in older adults and contributes substantially to morbidity and healthcare utilization. While intensive lifestyle interventions in randomized trials reduce diabetes incidence among high‑risk individuals, it is uncertain whether lower‑intensity, population‑based prevention programs embedded in routine care can affect diabetes incidence in the general older population. This study evaluated whether participation in a comprehensive health assessment and counseling program was associated with reduced incident diabetes among 70‑year‑olds in a defined Swedish region.
The Healthy Ageing Initiative (HAI) is a population‑wide preventive program offered to all 70‑year‑old residents in Umeå municipality. The pragmatic, low‑intensity intervention comprised two structured health assessment visits within routine care that evaluated lifestyle factors and cardiometabolic risk markers, followed by individualized feedback and motivational counseling aimed at promoting lifestyle changes relevant to cardiovascular and metabolic disease prevention. The program did not include structured long‑term follow‑up; participants returned to ordinary care and were advised to consult their general practitioner if undiagnosed conditions were suspected.
A population‑based matched cohort study design was used. The intervention cohort included 6,018 HAI participants enrolled between 2012 and 2022. Controls were drawn from the general Swedish population and matched 1:10 to participants on birth year, sex, and educational level, yielding 57,543 matched controls. Individuals with prevalent diabetes at or before the index date were excluded. The index date corresponded to the second HAI visit for participants and the same assigned date for matched controls. Mean follow‑up was 4.9 years (SD 2.9) in the HAI cohort and 4.8 years (SD 2.9) in controls.
The primary outcome was incident diabetes, defined as the first recorded diagnosis of diabetes in specialist inpatient or outpatient care (ICD‑10 code E11) or the first dispensed glucose‑lowering medication (ATC code A10), whichever occurred first. Outcomes were ascertained through nationwide linkages to the National Patient Register and the Prescribed Drug Register. Deaths were identified via the Cause of Death Register and used for censoring in standard survival analyses and as a competing event in sensitivity analyses.
Matching accounted for age, sex, and educational level. Additional baseline covariates ascertained from registers included country of birth, indicators of functional dependency (home care services), preexisting diagnoses reflecting cardiometabolic and other comorbidities, and selected medication use. The unadjusted association was estimated using Cox models stratified by matched set. Adjusted hazard ratios and cumulative incidence functions were derived using Royston–Parmar flexible parametric survival models with robust standard errors clustered by matched set; these models produced adjusted absolute risk estimates and allowed assessment of time‑varying effects. Bootstrap resampling was used to derive confidence intervals for absolute risk differences. Competing‑risk analyses used Fine–Gray subdistribution hazard models. Effect modification by predefined baseline characteristics was explored using interaction terms.
After exclusions for prevalent diabetes, the analytic sample comprised 63,561 individuals (6,018 HAI participants; 57,543 controls). During follow‑up, incident diabetes occurred in 335 (5.6%) HAI participants and 3,919 (6.8%) controls. Participation in the HAI was associated with a lower hazard of incident diabetes: adjusted hazard ratio 0.76 (95% confidence interval 0.68–0.85; p < 0.001). The association was reported as stable over time.
Using flexible parametric models, the estimated absolute risk reduction associated with participation was 1.46 percentage points (95% CI 0.91–1.98; p < 0.001) at 5 years and 3.42 percentage points (95% CI 2.13–4.75; p < 0.001) at 10 years.
The authors performed competing‑risk sensitivity analyses treating death as a competing event using Fine–Gray models; these analyses supported the robustness of the primary association. Subgroup analyses examining predefined baseline characteristics showed broadly consistent associations across most groups. The report indicates no data‑driven variable selection and that analyses were planned prior to data access.
Key limitations stem from the observational design. Voluntary participation in HAI may have introduced selection bias and a healthy‑volunteer effect, as participants tended to be healthier than nonparticipants in prior analyses. Residual confounding is possible despite adjustment for an extensive set of register‑based covariates. The lack of structured long‑term follow‑up within the program limits insight into the duration and content of any sustained behavior change attributable to the intervention. The authors explicitly note that causality cannot be established from these findings.
In this real‑world, routine‑care setting, participation in a low‑intensity, population‑based health assessment and counseling program was associated with lower incidence of diabetes among older adults. The magnitude of absolute risk reduction is modest at the individual level but could be meaningful at population scale. These results suggest that scalable prevention strategies embedded within routine care may complement intensive, high‑risk interventions for diabetes prevention in aging populations. However, confirmation from more robust designs would be needed before policy or clinical practice changes are recommended based solely on these observational data.
Outcome and covariate ascertainment relied on nationwide Swedish registries subject to data‑use agreements and ethical approval. Individual‑level data cannot be publicly shared due to data protection regulations; eligible researchers may request access via the relevant Swedish authorities. The study was approved by the Swedish Ethical Review Authority and participants in the HAI provided written informed consent for program participation and research use of collected outcome data. The authors report no competing interests and acknowledge the use of AI assistance for drafting statistical scripts and language, with full responsibility retained by the authors.