efectiveness and cost efciency of this type of disease management program is critically important for diabetes control because of the long-term impact on patient health and healthcare cost. Te present study aimed to examine the long-term efects of a local disease management program focused on diabetic nephropathy prevention using administrative data under the universal public health insurance scheme in Japan to extend our previous fndings. Methods Study setting and disease management program Tis study took advantage of a public health insurance scheme, the National Community-based Health Insurance System in Japan. Te National Community-based Health Insurance System is a mandatory public health insurance scheme for local self-employed residents, retired citizens, and their dependents. Te insurance provides universal coverage of outpatient, inpatient, dental, and prescription services with 10%–30% copayment under a monthly upper limit charge [16]. More specifcally, the present study used a local disease management program setting driven by the local public health insurance authority for Kure City, a large city in West Japan with a population of about 240,000 people. Te insurance authority obtains administrative claim data linked by unique encrypted IDs containing detailed information on comorbidity diagnoses, prescription contents and interventional treatments provided, physical and cognitive functional statuses, and prognosis in electronic standardized record form [17, 18]. Using this information, the authority relied on a private information company to automatically screen candidate outpatients for the program with specifc targeting of those with stage 3 or 4 diabetic kidney disease [19]. Te company basically screened for outpatients who were diagnosed with type 2 diabetes and aged between 20 and 75 years. Because laboratory data were lacking, the screening relied on patterns of medication and treatment to assess clinical severity, the details of which were not made public. Patients with the following conditions were excluded: type 1 diabetes, renal replacement therapy, dementia, mental disease, Keywords: Diabetic nephropathy, Health outcomes, Self-management, Disease management, Electronic claims database Watanabe et al. BMC Endocrine Disorders (2022) 22:135 Page 3 of 9 terminal condition, chemotherapy, radiotherapy, severe hearing loss, intractable diseases, and certifcated for long-term social care. Te list of candidate outpatients was shared with local medical care providers. Te fnal decision on whether to invite patients to join the program was made by the attending physicians, based on their clinical evaluation of the disease stage and program feasibility. Once patients were invited and agreed to join the program, specially trained nurses provided a self-management educational program via telephone and in person to support the patients in acquiring self-management skills and making behavioral changes based on self-efcacy and an existing theoretical frame, [20–22] and conducted data monitoring in collaboration with the physicians for 6 months. Patient education was conducted by face-to-face interviews every 2 weeks for the frst 2 months and by telephone every month from the third to sixth months. Te nurses provided knowledge about self-management including diet, exercise, medication, stress management, and self-monitoring to the patients, and encouraged their practice. Tey also consulted with the patients to assess their health condition and risk factors and to make shared decisions for tailored action plans, in accordance with clinical practice guidelines and advice from endocrinologists and nephrologists as needed. During the period of 2011 to 2013, nearly 2,700 benefciary outpatients were selected as program candidates, of whom 159 patients joined and completed the program. Among them, 5-year follow-up data were available for 153 patients. Study design Given the non-random selection process described above, we chose to use propensity score matching to treat the selection bias when evaluating the program efectiveness. We prepared two subpopulations for the selection of control groups for this purpose. Te subpopulation for control group 1 (screened candidates for program) was selected by reference to a list of candidates screened by a preset algorithm (N=2,635). Te subpopulation for control group 2 (benefciaries meeting inclusion criteria) was selected from the whole benefciary pool in Kure City who had a diagnosis of type 2 diabetes, regardless of the severity of nephropathy, were aged between 20 and 75 years, and did not have the exclusion criteria described above (N=11,806) (Fig. 1). Although we intended to equalize the clinical and other background characteristics related to the selection decision between the patients undergoing the program and the control patients, we had limited access to the full information for the automatic screening