Diabetes Education Program Tied to 57% Lower Early Death Risk in Major Slovak Cohort

People with type 2 diabetes who completed a structured education program run by a Slovak health insurer were dramatically less likely to die over the following years than comparable patients who never took the classes, according to a large retrospective cohort study published in Diabetes Therapy. The analysis, drawing on administrative healthcare records from 2013 to 2024, found that participants in the education arm of a disease management program operated by Dôvera Health Insurance Company had a 57 percent lower risk of death from any cause during the first two years after starting the program, and a 32 percent lower risk in the years that followed. The findings arrive at a moment when health systems worldwide are searching for scalable, low-cost interventions that can complement the wave of powerful new diabetes drugs, and they suggest that one of the oldest tools in diabetes care—teaching patients how to manage their own disease—may be among the most powerful.

The study was possible because Slovakia offers an unusual natural experiment. Since the end of 2015, Dôvera, one of the country’s three mandatory health insurance companies and the insurer of roughly 1.77 million people, has run a disease management program for adults with type 2 diabetes. Enrollment in the program was required to access its benefits, but participation in the educational activities themselves was voluntary. That distinction turned out to be scientifically crucial. The program’s education, delivered by trained nurses using the Diabetes Conversation Maps tool alongside self-prepared materials, originally consisted of six to eight structured sessions spaced two to three weeks apart, each lasting about two hours and covering everything from dietary behavior and physical activity to complication prevention, medication adherence, and goal setting. From 2018 onward, the model was condensed to three or four sessions, with refresher contacts at one and two years, making education a continuous, long-term process rather than a one-off lecture series.

To isolate the effect of education, the researchers constructed three comparison groups from the insurer’s administrative database: patients who underwent education within the program, patients treated by diabetologists who participated in the program but whose patients never enrolled in it, and patients whose doctors were entirely outside the program. Using propensity score matching—a statistical technique that pairs participants with non-participants who share similar baseline characteristics—the team formed 4,702 matched triplets comprising 14,106 patients. All patients in the education group were successfully matched to controls from both reference groups. After matching, standardized mean differences for all measured covariates fell below 0.1, indicating that the groups were well balanced on the characteristics the data could capture. The matched cohort was 54.6 percent female, with a mean age of just over 60 years, and patients were followed for a median of 5.04 years, during which 1,695 deaths occurred.

The statistical analysis had to confront a complication that often arises in long-term observational studies: the effect of the intervention was not constant over time. Standard Cox regression assumes that the ratio of hazards between groups remains proportional across the follow-up period, but Schoenfeld residual tests rejected that assumption here. The researchers therefore turned to a piecewise Cox model, splitting follow-up into two pre-specified intervals: the first two years and everything after. The two-year cut-point was not arbitrary. In routine program practice, educators typically contact patients up to about two years after the initial course to review their status and offer re-education, marking a natural point at which the effect of the initial training might begin to fade—or be reinforced. Splitting the timeline also allowed the team to estimate effects in a period less contaminated by short-term selection dynamics, since neither patients nor physicians can reliably predict health trajectories far into the future when deciding whether to enroll in education.

The results were striking. Compared with patients receiving usual care outside the program, those who completed education had a 57 percent lower mortality risk in the first two years of follow-up, a hazard ratio of 0.43 with a 95 percent confidence interval of 0.30 to 0.60. Beyond two years, the protective effect attenuated but remained substantial and statistically significant, at 32 percent lower risk. Crucially, the second comparison group told a very different story. Patients treated by program-participating diabetologists who never took the education showed no survival benefit at all relative to the general reference population; in the weighted analysis their mortality risk was actually slightly higher. This is perhaps the study’s most provocative finding: the system-level machinery of a disease management program—financial incentives for physicians, feedback reports, specialist booking systems, shared data platforms—apparently did nothing for survival on its own. What mattered was the patient-facing component.

To guard against the possibility that the result was an artifact of the matching strategy, the researchers ran a parallel analysis using inverse probability of treatment weighting, a technique that reweights the entire sample to estimate what would happen if the whole insured population were offered education. The pattern held. Under this population-wide scenario, education was associated with a 60 percent lower mortality risk in the first two years and a 31 percent lower risk thereafter, both highly significant. Subgroup analyses found no meaningful interactions with sex or age across four age bands, suggesting the benefit was broadly distributed rather than concentrated in any particular demographic. The concordance across two independent statistical approaches strengthens confidence that the association is not simply a product of model specification.

The Slovak results fit into a growing international evidence base. Austria’s Therapie Aktiv program reported a 43 percent reduction in four-year mortality and a sustained 30 percent reduction over eight years in its early cohorts. Hong Kong’s RAMP-DM program documented a 55 percent reduction in all-cause mortality over a median follow-up of 9.5 years, alongside reductions in both macrovascular and microvascular complications. Randomized evidence points in the same direction: a meta-analysis of 42 trials including 13,017 patients found a 26 percent mortality reduction from self-management education, and a 2026 scoping review reported a 45 percent reduction when restricted to studies in which mortality was the primary outcome. The new study extends this evidence to a Central European population, adding geographic and demographic diversity to a literature the review’s authors noted is dominated by high-income countries.

Why would classroom sessions with nurses translate into fewer deaths? The authors argue the mechanisms are likely multifactorial. Prior research consistently shows that disease management programs improve glycemic control, lipid profiles, and blood pressure, and increase adherence to guideline-recommended screening and drug therapy. A 2012 meta-analysis of more than 140 quality improvement studies documented more frequent foot examinations, retinopathy screening, and kidney function monitoring, along with lower HbA1c, LDL cholesterol, and blood pressure. The researchers’ own earlier work on the same Slovak program showed increased frequency of HbA1c monitoring, suggesting improved process quality of care. Education may also reduce diabetes-related distress, which the team’s previous studies observed, potentially improving adherence to and understanding of needed lifestyle changes. The study period also overlapped with the introduction of SGLT2 inhibitors and GLP-1 receptor agonists, drugs with proven cardiovascular and renal benefits, and the program’s incentives may have eased administrative barriers to prescribing them—though the authors note that causes of death were unavailable, so any effect on COVID-19 era mortality cannot be confirmed.

The time-dependent pattern of the benefit offers its own lessons. A recent dose-response meta-analysis found that HbA1c improves rapidly—by about 1.3 percent within the first 30 weeks of face-to-face education—before gradually declining, mirroring the mortality trajectory seen here. Motivation is highest immediately after the course; behavioral effects fade without reinforcement; and after two years the program relies on patients initiating contact themselves. The authors also candidly acknowledge the specter of residual confounding. Participation was voluntary, and healthier or more motivated patients may be more likely to attend. Frail patients or those with subtle health deterioration may be quietly steered away from education by physicians even when they do not meet formal exclusion criteria—a hidden selection effect the researchers argue is most potent in the first two years and likely negligible beyond them, which is partly why the persistence of a significant benefit after the two-year mark is so reassuring.

The study’s limitations are real but bounded. Laboratory values were absent from the insurance data, precluding direct measurement of metabolic change; unmeasured factors such as frailty, functional status, and health literacy could still shape who enrolls and who survives; and only a randomized trial could fully eliminate confounding. Yet the strengths are considerable: a dataset covering a third of Slovakia’s population, follow-up stretching to eight years, rigorous matching with a strict intention-to-treat approach, and consistent results across complementary analytical methods. The practical implication is hard to escape. As new glucose-lowering drugs command headlines and budgets, the cheapest intervention in the diabetes toolkit—structured, sustained, nurse-led patient education—may deliver survival benefits that no amount of physician-side incentives or system engineering can match. Health systems designing the next generation of chronic disease programs, the authors suggest, should place education at the center rather than the periphery.

Subject of Research: Association between structured diabetes self-management education in a disease management program and all-cause mortality in patients with type 2 diabetes

Article Title: Impact of Education in Disease Management Program for Patients with Diabetes Mellitus on All-Cause Mortality: A Propensity Score-Matched Cohort Study

Article References: Selvek, M., Saal, B., Mužik, R., & Tkáč, I. (2026). Impact of Education in Disease Management Program for Patients with Diabetes Mellitus on All-Cause Mortality: A Propensity Score-Matched Cohort Study. Diabetes Therapy. https://doi.org/10.1007/s13300-026-01922-z

Image Credits: AI Generated

DOI: 10.1007/s13300-026-01922-z

Keywords: type 2 diabetes, disease management program, diabetes self-management education, all-cause mortality, propensity score matching, inverse probability of treatment weighting, Cox regression, Slovakia, health insurance data, nurse-led education, observational cohort study, Diabetes Therapy