Clinical coding capacity and hospital-reported quality across hospital tiers during China's transition to DRG-based payment: a cross-sectional survey of 299 hospitals in Anhui Province

Abstract Background Diagnosis-related group (DRG) payment relies on coded medical records to assign cases and determine reimbursement. China has rapidly expanded case-based payment, but the coding workforce and infrastructure available across hospital tiers remain poorly described. Methods We conducted a province-wide cross-sectional survey of hospital medical record departments in Anhui Province, China, in June and July 2026; quantitative responses referred to the year ending 31 December 2025. Participation was required for secondary and higher-level hospitals and voluntary for primary and unclassified institutions. Of 311 returned questionnaires, 299 were retained. Measures covered coding workforce, certification and training, workload, coding software, coding dictionaries, physician participation, hospital-reported coding accuracy, and reported error types. Hospitals were grouped by official tier, and tier comparisons used chi-square tests. Results The 299 hospitals employed 1,470 coders. All measured capacity indicators declined across descending hospital tiers. National coder certification was present in 93.5% of Tier 3A hospitals and 21.6% of Tier 1 and unclassified institutions; a regularly updated coding dictionary in 74.2% and 21.6%, respectively; and coding assistance software rated as having good functionality in 58.1% and 5.4% (all P < 0.001). Overall, 54.8% of hospitals had no coding assistance software and 24.7% had no coding dictionary. Hospital-reported primary diagnosis accuracy of at least 95% was reported by 83.9% of Tier 3A hospitals and 48.6% of Tier 1 and unclassified institutions. The most frequently reported problems were principal diagnosis selection errors (72.2%), imprecise or non-specific coding (68.9%), and omission of complications, comorbidities, or other secondary diagnoses (60.2%). Conclusions Clinical coding capacity showed a marked tier gradient, whereas the gradient in hospital-reported performance was shallower. Because coding quality and reimbursement were not independently audited, the study cannot establish payment effects. Nevertheless, uneven coding capacity may disadvantage lower-tier institutions under uniform case-based payment rules. Coding readiness should be monitored during payment reform, with targeted training, shared technical infrastructure, and proportionate audit support.