China’s fragmented primary health care system has struggled to meet the growing burden of noncommunicable diseases. The authors note that, despite substantial investment, fewer than half of patients with diabetes in China achieve adequate glycaemic control. The Tianjin reform aimed to address fragmentation by aligning financing, governance and digital clinical support to strengthen community‑level care for people with diabetes.
Tianjin is an urban health jurisdiction serving about 15 million residents through 177 hospitals and 266 community health centres. Chronic disease management in the city was characterised as fragmented: the Health Commission set care standards while the Insurance Bureau controlled funding, creating separation between regulation and financing that limited integrated primary care delivery.
Between 2020 and 2023 Tianjin implemented a public–private partnership model in which the private digital health company WeDoctor was contracted to manage community health centres under a monthly prepaid capitation arrangement. The capitation covered all outpatient services related to diabetes, creating a single per‑patient payment intended to encourage primary care provision and proactive chronic disease management.
A core component of the model was third‑party governance combining claims auditing with clinical decision support driven by artificial intelligence (AI). The approach used AI to assist in auditing claims and to provide clinical decision support to clinicians at the community health centres, intended to improve adherence to care standards and to detect inappropriate claims or gaps in care. The abstract reports the presence of these AI and auditing elements but does not detail specific algorithms, performance metrics, or clinical decision pathways.
The reform included dedicated health managers responsible for coordinating care for patients with diabetes. Services at community health centres were redesigned to incorporate complication screening and digital medication management. These changes were intended to improve continuity of care, early detection of complications and medication adherence through digital tools integrated into the capitation arrangement.
A pilot study evaluated the model in a cohort of 494,945 patients with diabetes. The pilot compared three care models over 2022–2023: (1) WeDoctor–community health centre care under the capitation and governance model; (2) hospital‑based care; and (3) usual care. The abstract reports comparisons in service use, expenditure, financial outcomes and patient satisfaction across these groups. Specifics of patient selection, risk adjustment, baseline characteristics and statistical methods were not described in the abstract.
The pilot found differing trends in diabetes visits across models during 2022–2023. Visits for diabetes at WeDoctor community health centres increased by 2.6% (0.7/26.6). In contrast, diabetes visits decreased by 10.6% (−3.8/35.7) for hospital‑based care and by 2.3% (−0.8/34.1) for usual care. All three groups reduced outpatient expenditure during the study period. The abstract does not report clinical endpoint changes such as glycaemic control or complication rates.
Financially, the WeDoctor capitation model generated a surplus of US$ 37.62 million. The model increased diabetes‑related revenue for community health centres by 65% (US$ 154,577/237,805) and was associated with a 30% increase in physicians’ annual salaries (US$ 5,172/17,241). These figures indicate that the capitation model shifted resources toward primary care providers and improved financial incentives for clinicians working in community settings.
Patient experience measures reported in the abstract indicate that more than three quarters of patients expressed satisfaction with and trust in the WeDoctor model. The abstract does not provide the exact survey instrument, response rate, or subgroup analyses.
The authors conclude that integrating capitation financing with third‑party governance and AI clinical support can strengthen primary health care, contain costs and enhance patient‑centred care. Key demonstrated effects in the pilot included increased use of community health centre services, reduced outpatient expenditure across groups, financial surplus generation and increased provider income and patient satisfaction. The abstract does not report detailed clinical outcome measures (for example, changes in glycaemic control), nor does it present granular methodological details such as matching, confounding control or the internal functioning and validation of the AI tools. These gaps limit assessment of clinical effectiveness beyond utilisation, financial and reported satisfaction outcomes.
Clinicians and policymakers considering similar reforms should note the multi‑component nature of the intervention—capitated payments, third‑party management, AI‑enabled auditing and decision support, dedicated care managers and redesigned services—and that the pilot evidence presented focuses on utilisation, fiscal flows and patient satisfaction rather than reported clinical biomarkers in the abstract. Further reporting of detailed methods and clinical endpoints would be necessary to evaluate health outcomes and replicability in other settings.