Risk Governance of Payment Exceptions Under China’s DRG/DIP Reform: A Structured Policy-Text Analysis of Special Case Negotiation Rules Authors: Kedi Sun Published: 2026 ABSTRACT PurposeUnder China’s diagnosis-related group/diagnosis-intervention packet (DRG/DIP) payment reform, special case negotiation addresses clinically atypical cases that do not fit standard case-based payment rules but may also create payment misfit, fund leakage, discretionary review, delay, and accountability risks. This study examined whether policies translate these risks into enforceable governance rules.MethodsA structured policy-text analysis examined 20 documents within one provincial policy chain: one national document, one provincial document, and 18 municipal documents covering all prefecture-level jurisdictions in the province. The municipal documents constituted a complete census within the selected province rather than a nationwide sample of Chinese provinces. A 30-node, five-domain framework used a 0-1-2 scale. Two coders worked independently, and final analyses used adjudicated scores.ResultsInter-coder agreement was 95.0% (Cohen’s kappa, 0.899). Domain scores were highest for payment misfit risk control (0.797), discretion and capture risk control (0.786), and administrative burden and delay risk control (0.767), and lowest for accountability and equity risk control (0.512). Applicable scope, volume or proportion cap, and submission or review cycle were universally enforceable. The weakest or least consistently enforceable nodes concerned cross-area or fund responsibility, appeal or reconsideration, processing time limits, and special diseases, infectious diseases, or rare diseases. Vertical attenuation was highest for cross-area or fund responsibility (72.2%) and for processing time limits and special diseases, infectious diseases, or rare diseases (both 55.6%). Municipal documents formed balanced, boundary-heavy, procedure-heavy, and low-formalisation configurations.ConclusionSpecial case negotiation requires a minimum risk-governance package—a minimum set of enforceable procedural safeguards—combining eligibility boundaries, traceable review, appeal routes, and clear accountability. The analysis evaluates formal policy design rather than implementation outcomes. Functionally, the framework may inform risk audits of outlier, high-cost, new-technology, and other payment-exception mechanisms in DRG or other case-based payment systems internationally. Introduction The core risk in case-based provider payment is not only how to set prices, but how to govern the clinically legitimate cases that fall outside standard payment assumptions. International experience with DRG, DPC1,2 and value-based payment models3 shows that these payment systems can standardise reimbursement and improve efficiency, but their effects depend heavily on design and implementation context. China-based studies similarly show that patient-classification payment and regional global-budget arrangements can reduce costs4,5 and change hospital incentives.6 Evidence also shows heterogeneous effects on costs, length of stay, quality of care,7,8 and expenditure variation across hospitals and patient groups.9,10 Provider behaviour remains a central concern because payment rules can alter physicians’ service decisions,11–13 inpatient care patterns, and responses to case-based settlement rules.14 China has pursued a large-scale reform of provider payment through both diagnosis-related group (DRG) and diagnosis-intervention packet (DIP) systems. Recent studies have examined ordered delivery effects, single-disease payment, and DIP inpatient reform.15–17 Other analyses report disease-specific, regional, or hospital-level effects in Beijing, Sichuan, Guangzhou, and other settings.18–20 The same literature also identifies patient selection,21 unintended quality effects,22 cost-control mechanisms,23 and disease-specific cost variation as continuing risks under case-based payment.24–26 Additional evidence covers expense trends,27 negative impacts,28 pharmacy administration,29 Guangzhou DIP effects,30 and inpatient cost structure.31 These findings point to a shared design problem: regular payment groups cannot fully absorb every clinically legitimate variation in cost, complexity, and resource use. Special case negotiation is meant to address this payment misfit. It gives medical institutions an exception channel through which they may request case review, additional points, adjusted payment standards, project-based payment, or other settlement treatment for cases that depart substantially from standard DRG/DIP assumptions. Policy evaluation and implementation studies suggest that the quality of payment reform depends not only on the payment formula, but also on the clarity of local rules, implementation barriers, and organisational routines.32–34 At the policy level, special case negotiation protects hospitals from being penalised for treating complex or high-risk patients. In risk-management terms, though, the mechanism is not simply compensatory; it sits between cost control and clinical complexity. That dual role brings a further layer of risk. If the exception channel is too narrow, complex cases may remain under-compensated and hospitals may still have incentives to avoid them. If it is too broad or weakly controlled, it may encourage fund leakage, excessive applications, or strategic reclassification. Risk-adjustment research is relevant here because payment systems can under- or over-compensate providers when case complexity is not fully captured.35–37 Work on machine learning, regional funding allocation, and cost prediction also shows that technical correction mechanisms require transparent design and careful interpretation.38–40 Studies of risk-adjuster selection, coding incentives, and undercompensated groups further explain why exception mechanisms need safeguards against payment misfit and strategic gaming.41,42 Research on DRG/DIP reform has largely examined expenditure, length of stay, hospital behaviour, service volume, quality incentives, and provider responses under case-based payment. Far less attention has been paid to the formal design of exception mechanisms within DRG/DIP payment systems.17,32 This gap is important because procedural fairness in health financing depends on transparency, reason-giving, revisability, enforcement, and accountability, rather than on the mere existence of a formal decision channel.43–45 Recent qualitative research has begun to examine how Chinese physicians experience special case review under case-based payment.46 However, systematic evidence remains limited on how policy texts define eligibility, control application volume, organise expert review, specify administrative procedures, handle appeals, allocate cross-area responsibility, and convert broad governance principles into enforceable rules. In this study, risk governance refers to the formal rules, allocated responsibilities, decision procedures, controls, and corrective routes through which risks created by payment exceptions are managed. Policy-text analysis is useful for this problem because formal rules shape the institutional environment in which hospitals, health insurance administrative agencies, expert reviewers, and medical insurance administrations interact. Health-system governance research links quality, accountability, and institutional performance to the clarity of formal authority and decision structures.47–49 Collaborative, primary-care, local-health-system, and conflict-setting governance studies also show that coordination rules50 and responsibility allocation affect whether policy goals can be implemented across institutions and jurisdictions.51–53 Written provisions do not prove implementation effectiveness, but they define the authorised procedures, responsibilities, thresholds, and remedies available to actors. A policy that merely states that negotiation should be fair or standardised does not provide the same level of risk control as a policy that specifies eligible case categories, numerical thresholds, application deadlines, expert selection, recusal requirements, written review records, system-based submission, feedback channels, appeal periods, and fund responsibility.54 Broader work on healthcare accountability, safety, algorithmic governance, transparency, bias control, and adverse-event reporting reinforces the same governance principle: high-risk decision systems need traceable procedures, clear responsibility, and routes for review or correction. This study therefore reframes special case negotiation under China’s DRG/DIP reform as a risk-governance mechanism and examines its formal design across a linked set of national, provincial, and municipal policy documents. The analysis focuses on five domains of risk control: payment misfit risk control; fund leakage and strategic application risk control; discretion and capture risk control; administrative burden and delay risk control; and accountability and equity risk control.55 These domains reflect the main risks that arise when a fixed payment system creates an exception channel for clinically or financially atypical cases. The framework also distinguishes between simple mention of a risk-control issue and conversion of that issue into an enforceable requirement.54 This study therefore treats special case negotiation not merely as an exception channel, but as a risk-governance mechanism whose quality depends on whether payment misfit, fund leakage, discretion, administrative burden, and accountability risks are translated into enforceable procedural safeguards (Figure 1). In this study, a minimum governance package refers to the smallest set of formal safeguards needed for the mechanism to operate as governance rather than as a discretionary reimbursement supplement: clear eligibility boundaries, volume control, application and review procedures, traceable decision records,56–58 feedback or appeal routes,59,60 and accountability or responsibility allocation. Figure 1Conceptual framework of special case negotiation as a DRG/DIP risk-governance mechanism. Routine case-based payment may generate payment misfit for complex, high-cost, long-stay, ungroupable, or new-technology cases. Special case negotiation addresses this misfit through safeguards across five risk-control domains and a minimum governance package.The flowchart outlines DRG/DIP payment risk-control and governance. It starts with ′Routine DRG/DIP payment fixed groups, points and standards′ leading to ′Payment misfit risk′ for complex or new cases. This links to a ′Special case negotiation channel′ for review or adjustment. Key risk-control areas include ′Payment misfit′, ′Fund leakage′, ′Discretion and capture′, ′Administrative burden′ and ′Accountability and equity′. These inform a ′Minimum risk-governance package′ with ′Eligibility triggers′, ′Volume/CMI controls′, ′Expert pool′, ′Document checklist′, ′Online trace′, ′Processing time limit′, ′Outcome feedback′, ′Appeal′, ′Cross-area responsibility′ and ′Audit′. It concludes with a ′Governance quality criterion′ emphasizing clear, enforceable, traceable, appealable and accountable rules.A flowchart of DRG/DIP payment risk-control and governance mechanisms.Abbreviations: DRG, diagnosis-related group; DIP, diagnosis-intervention packet; RGS, Risk Governance Score; CMI, case mix index. Conceptual framework of special case negotiation as a DRG/DIP risk-governance mechanism. Routine case-based payment may generate payment misfit for complex, high-cost, long-stay, ungroupable, or new-technology cases. Special case negotiation addresses this misfit through safeguards across five risk-control domains and a minimum governance package. Internationally, payment systems use different mechanisms for cases or items that are poorly accommodated by standard case-based prices. In the United States, Medicare inpatient high-cost outlier payments supplement the basic prospective payment when case costs exceed a fixed-loss threshold, while the new technology add-on payment provides temporary additional payment for qualifying technologies that meet defined newness, cost, and clinical-improvement criteria.61,62 Germany’s NUB process allows local parties to negotiate time-limited, case-related payments or add-on payments for new diagnostic and treatment methods after an InEK information request.63 In England, the 2026/27 NHS Payment Scheme excludes listed high-cost drugs, devices, and procedures from prices and subjects them to specified payment arrangements.64 These arrangements are not institutionally equivalent to China’s special case negotiation. Their common functional problem is payment misfit: standard payment may not fully absorb extreme costs, complex cases, or new technology. Each therefore raises governance questions about eligibility, evidence, fiscal control, review transparency, decision timing, responsibility, and correction routes. The selected province offered a complete national–provincial–municipal policy chain, full coverage of all 18 prefecture-level jurisdictions, both DRG and DIP arrangements, and variation in document form and procedural formalisation. These characteristics made it suitable for examining vertical policy translation and within-province variation. This complete municipal census within one province should not be interpreted as a nationally representative sample of Chinese provincial systems. The study evaluates what policy texts formally authorise and require; it does not test implementation consistency, payment effects, provider behaviour, fund performance, or patient outcomes. The study has four objectives. First, it identifies which types of risk are addressed in special case negotiation policy texts. Second, it evaluates whether these risks are converted into enforceable procedural rules rather than remaining at the level of general principle. Third, it examines vertical attenuation between provincial guidance and municipal implementation rules, identifying where higher-level risk-control signals are weakened or omitted at the municipal level. Fourth, it classifies municipal policies by the balance between boundary control and procedural safeguards, thereby distinguishing local risk-governance models. To address these objectives, the study analyses 20 policy documents: one national document, one provincial document, and 18 municipal documents. A structured 30-node coding framework converts policy provisions into a transparent risk-governance matrix. The analysis covers inter-coder reliability, domain-level risk-control scores, node-level coverage and enforceability, document-level risk-governance scores, vertical attenuation rates, closed-loop completeness, and a boundary-procedure typology. This allows the study to identify where special case negotiation policies are relatively strong or weak as governance instruments. Methodologically, it provides a reproducible framework for evaluating exception mechanisms in health financing policy when implementation data are limited or unavailable. Methods Study Design This study used a structured comparative policy-text analysis to examine how special case negotiation mechanisms are formally designed as risk-governance instruments under China’s DRG/DIP payment reform. The analysis focused on the written rules contained in policy documents rather than on observed implementation outcomes. The unit of analysis was the policy document, and the coding unit was a policy provision or a group of related provisions that addressed a specific risk-control function. The methodological aim was to convert qualitative policy provisions into a transparent and reproducible coding matrix. This design was appropriate because the research question concerned the formalisation, enforceability, and procedural completeness of policy rules, not the statistical effect of DRG/DIP reform on expenditure, service use, or patient outcomes. No patient-level, hospital-level, or claims-level data were used. Data Sources and Document Selection The dataset contained 20 policy documents related to DRG/DIP payment reform and special case negotiation. These included one national document, one provincial document, and 18 municipal documents. The 18 municipal documents covered all prefecture-level municipal jurisdictions in the province, so the municipal-level analysis represents a complete within-province policy census rather than a sample of selected cities. The national document was a formal DRG/DIP 2.0 grouping implementation notice. The provincial document was a trial management measure for DRG/DIP payment. The municipal documents included implementation measures, operating procedures, management measures, review measures, workflows, and notices dealing with special case negotiation or special disease negotiation under DRG or DIP payment. Documents were included when they met three criteria: they addressed DRG or DIP payment reform; they contained provisions relevant to special case negotiation, special disease negotiation, special case review, or exception handling; and they belonged to the national, provincial, or municipal policy chain covered by the project dataset. General DRG/DIP payment documents were included only when they contained identifiable special case negotiation provisions. Documents that did not contain any special case negotiation or comparable exception mechanism were not included. The documents used several local labels for comparable exception-channel arrangements, including special disease negotiation and special case single-case review. For consistency, the manuscript uses special case negotiation as the umbrella term, except where original policy wording or document titles are described. To reduce identification risk and avoid presenting the analysis as a ranking of local performance, the 18 municipal documents were anonymised as M01–M18. The national and provincial documents were labelled N_National and N_Provincial. For each document, a profile table recorded policy level, year, payment type, document form, formal status, public availability, and the presence of application forms, expert review forms, or quota tables (Table 1).Table 1Profile of Included Policy DocumentsDocument IDLevelPayment TypeIssuing Body TypeDocument FormFormal StatusNotesN_NationalNationalDRG-DIPNational Healthcare Security Administration OfficeNoticeFormal noticeNational DRG/DIP 2.0 grouping implementation notice; includes general requirements for using the special-case negotiation mechanism, but no application or expert-review form.N_ProvincialProvincialDRG-DIPProvincial Healthcare Security AdministrationManagement measuresFormal; trialProvincial DRG/DIP management document; includes general special-case negotiation mechanism.M01MunicipalDRGSocial medical insurance centerBusiness handling procedureFormal; trialMunicipal DRG special-case negotiation procedure.M02MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes new technology filing form, application form, summary table, and review opinion form.M03MunicipalDRGSocial medical insurance centerProcedureFormal; revised; trialIncludes combined application and expert review form.M04MunicipalDIPHealthcare security administrationReview measuresFormal; trialMunicipal special-case review measure under multi-payment/DIP-related payment.M05MunicipalDRGSocial medical insurance centerWorkflowFormal; trialIncludes feedback opinion form and review/audit opinion form.M06MunicipalDIPHealthcare security administrationImplementation planFormalIncludes application form, summary table, review opinion form and process attachment.M07MunicipalDRGHealthcare security administrationManagement measuresFormal; trialMunicipal DRG special-case negotiation management measure.M08MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application table and application/expert review tableM09MunicipalDIPHealthcare security administrationInterim management measuresDraft for commentsDraft management document; special-case mechanism appears as part of broader DIP management.M10MunicipalDIPSocial medical insurance centerWork planFormal work planIncludes annual special-case application table and application/expert review tableM11MunicipalDIPHealthcare Security Administration and Health CommissionManagement measuresFormal; revisedJointly issued with health commission; includes application/expert review table and declaration list.M12MunicipalDIPHealthcare security administrationWork noticeFormal noticeIncludes summary table, application/expert review table, and application quantity tableM13MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application and expert review tableM14MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application and expert review tableM15MunicipalDIPMedical maternity insurance centerWork noticeFormal noticeShort notice; no attached form in available file.M16MunicipalDIPHealthcare security centerManagement measuresFormal; revised previous file abolishedSystem-based application and review; no attached form in available file.M17MunicipalDRGHealthcare security administrationProcedureFormal; trialIncludes application summary table and application/review opinion tableM18MunicipalDRGMedical insurance centerWork noticeFormal noticeIncludes application form and maximum application quantity table; expert review form not clearly attached.Note: Twenty policy documents are included: one national, one provincial, and eighteen municipal documents. Document IDs are anonymised; M01–M18 identify municipal documents. Table 1Profile of Included Policy DocumentsDocument IDLevelPayment TypeIssuing Body TypeDocument FormFormal StatusNotesN_NationalNationalDRG-DIPNational Healthcare Security Administration OfficeNoticeFormal noticeNational DRG/DIP 2.0 grouping implementation notice; includes general requirements for using the special-case negotiation mechanism, but no application or expert-review form.N_ProvincialProvincialDRG-DIPProvincial Healthcare Security AdministrationManagement measuresFormal; trialProvincial DRG/DIP management document; includes general special-case negotiation mechanism.M01MunicipalDRGSocial medical insurance centerBusiness handling procedureFormal; trialMunicipal DRG special-case negotiation procedure.M02MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes new technology filing form, application form, summary table, and review opinion form.M03MunicipalDRGSocial medical insurance centerProcedureFormal; revised; trialIncludes combined application and expert review form.M04MunicipalDIPHealthcare security administrationReview measuresFormal; trialMunicipal special-case review measure under multi-payment/DIP-related payment.M05MunicipalDRGSocial medical insurance centerWorkflowFormal; trialIncludes feedback opinion form and review/audit opinion form.M06MunicipalDIPHealthcare security administrationImplementation planFormalIncludes application form, summary table, review opinion form and process attachment.M07MunicipalDRGHealthcare security administrationManagement measuresFormal; trialMunicipal DRG special-case negotiation management measure.M08MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application table and application/expert review tableM09MunicipalDIPHealthcare security administrationInterim management measuresDraft for commentsDraft management document; special-case mechanism appears as part of broader DIP management.M10MunicipalDIPSocial medical insurance centerWork planFormal work planIncludes annual special-case application table and application/expert review tableM11MunicipalDIPHealthcare Security Administration and Health CommissionManagement measuresFormal; revisedJointly issued with health commission; includes application/expert review table and declaration list.M12MunicipalDIPHealthcare security administrationWork noticeFormal noticeIncludes summary table, application/expert review table, and application quantity tableM13MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application and expert review tableM14MunicipalDIPHealthcare security administrationImplementation measuresFormal; trialIncludes application and expert review tableM15MunicipalDIPMedical maternity insurance centerWork noticeFormal noticeShort notice; no attached form in available file.M16MunicipalDIPHealthcare security centerManagement measuresFormal; revised previous file abolishedSystem-based application and review; no attached form in available file.M17MunicipalDRGHealthcare security administrationProcedureFormal; trialIncludes application summary table and application/review opinion tableM18MunicipalDRGMedical insurance centerWork noticeFormal noticeIncludes application form and maximum application quantity table; expert review form not clearly attached.Note: Twenty policy documents are included: one national, one provincial, and eighteen municipal documents. Document IDs are anonymised; M01–M18 identify municipal documents. Profile of Included Policy Documents Note: Twenty policy documents are included: one national, one provincial, and eighteen municipal documents. Document IDs are anonymised; M01–M18 identify municipal documents. Development of the Risk-Governance Coding Framework The coding framework was developed deductively from the risk-management logic of DRG/DIP special case negotiation. Under case-based payment, fixed grouping and payment standards can create payment misfit for complex, severe, long-stay, high-cost, ungroupable, or new-technology cases. Special case negotiation can operate as a corrective mechanism, but it also introduces its own risks, including strategic application, fund leakage, expert discretion, administrative burden, delay, weak appeal mechanisms, and unclear responsibility allocation. The 30-node framework was developed deductively for this study as a theory-informed coding instrument and was not treated as a previously validated measurement scale. No separate formal pilot study or external validation exercise was conducted before full coding. On this basis, 30 scored risk-control nodes were organised into five domains. Payment misfit risk control contained eight nodes: applicable scope, high-cost trigger, length-of-stay trigger, critical or complex cases, ungroupable or unstable group cases, combined surgery or transfer cases, new technologies or new services, and special diseases, infectious diseases, or rare diseases. Fund leakage and strategic application risk control contained five nodes: volume or proportion cap, case mix index (CMI) or complexity adjustment, over-quota handling, deduction of unreasonable costs, and audit or penalty for problematic cases. Discretion and capture risk control contained seven nodes: review organisation or expert panel, expert source or selection mechanism, number or composition of experts, conflict-of-interest recusal, collective decision or voting, review standards or evidence base, and written confirmation and archiving. Administrative burden and delay risk control contained six nodes: submission or review cycle, deadline or overdue rule, document checklist, information system or format, multi-level review process, and processing time limit. Accountability and equity risk control contained four nodes: outcome feedback or disclosure, appeal or reconsideration, finality of decision, and cross-area or fund responsibility. The complete codebook is reported in Table 2.Table 2Risk-Governance Coding FrameworkNode CodeRisk-Control DomainRisk-Control NodeScore 2: Enforceable RequirementR1Payment misfit risk controlApplicable scope and eligible entitiesScore 2 when eligible institutions, cases, payment type, or settlement scenario are explicitly defined.R2Payment misfit risk controlHigh-cost trigger thresholdScore 2 when a cost multiple, fixed amount, payment-standard multiple, or comparable measurable threshold is specified.R3Payment misfit risk controlLength-of-stay trigger thresholdScore 2 when a specific number of days or a multiple of average length of stay is specified.R4Payment misfit risk controlCritical or complex casesScore 2 when critical, severe, complex, ICU, rescue, or comparable categories are clearly defined.R5Payment misfit risk controlUngroupable or unstable group casesScore 2 when ungroupable, unstable-group, high-outlier, or grouping-abnormal cases are operationally specified.R6Payment misfit risk controlCombined surgery or transfer casesScore 2 when combined surgery, multiple operations, multidisciplinary surgery, or transfer conditions are specified.R7Payment misfit risk controlNew technologies or new servicesScore 2 when filing, recognition body, separate accounting, payment period, or payment treatment is specified.R8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseasesScore 2 when disease list, rare-disease catalogue, new/outbreak infectious disease, or similar eligible category is specified.R9Fund leakage and strategic application risk controlVolume or proportion capScore 2 when a clear cap or formula is specified, such as 5‰, 3%, 2% × CMI, or another numeric rule.R10Fund leakage and strategic application risk controlCMI or complexity adjustmentScore 2 when CMI, grade coefficient, complexity formula, or comparable adjustment is specified.R11Fund leakage and strategic application risk controlOver-quota handlingScore 2 when over-quota cases are invalid, rejected, accepted, deferred, or otherwise handled by a defined rule.R12Fund leakage and strategic application risk controlDeduction of unreasonable costsScore 2 when unreasonable cost deduction, excluded cost, payment treatment, or deduction method is specified.R13Fund leakage and strategic application risk controlAudit or penalty for problematic casesScore 2 when audit, penalty, deduction, rejection, inspection, or supervision consequence is specified.R14Discretion and capture risk controlReview organisation or expert panelScore 2 when responsible body, expert panel, working group, or specific organisational duty is defined.R15Discretion and capture risk controlExpert source or selection mechanismScore 2 when expert pool, random selection, third-party expert, local/non-local expert, or comparable source rule is specified.R16Discretion and capture risk controlNumber or composition of expertsScore 2 when expert number, odd-number requirement, group leader, composition, or specialty mix is specified.R17Discretion and capture risk controlConflict-of-interest recusalScore 2 when recusal is specified, such as excluding experts from the applicant institution.R18Discretion and capture risk controlCollective decision or votingScore 2 when collective decision, majority rule, group leader responsibility, or final expert opinion rule is specified.R19Discretion and capture risk controlReview standards or evidence baseScore 2 when review criteria, clinical/economic evidence, reasonableness checks, or assessment dimensions are specified.R20Discretion and capture risk controlWritten confirmation and archivingScore 2 when written signature, confirmation, archiving, or record-retention requirement is specified.R21Administrative burden and delay risk controlSubmission or review cycleScore 2 when monthly, quarterly, semi-annual, annual, or other specific review cycle is defined.R22Administrative burden and delay risk controlDeadline or overdue ruleScore 2 when application deadline, overdue rejection, waiver, or non-acceptance rule is specified.R23Administrative burden and delay risk controlDocument checklistScore 2 when required materials, forms, records, settlement lists, cost details, or supporting evidence are specified.R24Administrative burden and delay risk controlInformation system or formatScore 2 when online system, file format, platform, PDF/Excel template, or comparable submission format is specified.R25Administrative burden and delay risk controlMulti-level review processScore 2 when institutional, county, municipal, preliminary review, re-review, or expert-review sequence is specified.R26Administrative burden and delay risk controlProcessing time limitScore 2 when an explicit number of days, working days, or fixed processing window is specified.R27Accountability and equity risk controlOutcome feedback or disclosureScore 2 when feedback, disclosure, notification, public reporting, or system query of results is specified.R28Accountability and equity risk controlAppeal or reconsiderationScore 2 when objection, appeal, reconsideration, grievance submission, deadline, authority, or review route is specified.R29Accountability and equity risk controlFinality of decisionScore 2 when finality, non-appealability after review, or binding decision status is specified.R30Accountability and equity risk controlCross-area or fund responsibilityScore 2 when settlement locality, fund-bearing responsibility, out-of-area care, or cross-area handling rule is specified.Note: Final scoring uses 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. The full scoring logic is defined in the Methods section. Table 2Risk-Governance Coding FrameworkNode CodeRisk-Control DomainRisk-Control NodeScore 2: Enforceable RequirementR1Payment misfit risk controlApplicable scope and eligible entitiesScore 2 when eligible institutions, cases, payment type, or settlement scenario are explicitly defined.R2Payment misfit risk controlHigh-cost trigger thresholdScore 2 when a cost multiple, fixed amount, payment-standard multiple, or comparable measurable threshold is specified.R3Payment misfit risk controlLength-of-stay trigger thresholdScore 2 when a specific number of days or a multiple of average length of stay is specified.R4Payment misfit risk controlCritical or complex casesScore 2 when critical, severe, complex, ICU, rescue, or comparable categories are clearly defined.R5Payment misfit risk controlUngroupable or unstable group casesScore 2 when ungroupable, unstable-group, high-outlier, or grouping-abnormal cases are operationally specified.R6Payment misfit risk controlCombined surgery or transfer casesScore 2 when combined surgery, multiple operations, multidisciplinary surgery, or transfer conditions are specified.R7Payment misfit risk controlNew technologies or new servicesScore 2 when filing, recognition body, separate accounting, payment period, or payment treatment is specified.R8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseasesScore 2 when disease list, rare-disease catalogue, new/outbreak infectious disease, or similar eligible category is specified.R9Fund leakage and strategic application risk controlVolume or proportion capScore 2 when a clear cap or formula is specified, such as 5‰, 3%, 2% × CMI, or another numeric rule.R10Fund leakage and strategic application risk controlCMI or complexity adjustmentScore 2 when CMI, grade coefficient, complexity formula, or comparable adjustment is specified.R11Fund leakage and strategic application risk controlOver-quota handlingScore 2 when over-quota cases are invalid, rejected, accepted, deferred, or otherwise handled by a defined rule.R12Fund leakage and strategic application risk controlDeduction of unreasonable costsScore 2 when unreasonable cost deduction, excluded cost, payment treatment, or deduction method is specified.R13Fund leakage and strategic application risk controlAudit or penalty for problematic casesScore 2 when audit, penalty, deduction, rejection, inspection, or supervision consequence is specified.R14Discretion and capture risk controlReview organisation or expert panelScore 2 when responsible body, expert panel, working group, or specific organisational duty is defined.R15Discretion and capture risk controlExpert source or selection mechanismScore 2 when expert pool, random selection, third-party expert, local/non-local expert, or comparable source rule is specified.R16Discretion and capture risk controlNumber or composition of expertsScore 2 when expert number, odd-number requirement, group leader, composition, or specialty mix is specified.R17Discretion and capture risk controlConflict-of-interest recusalScore 2 when recusal is specified, such as excluding experts from the applicant institution.R18Discretion and capture risk controlCollective decision or votingScore 2 when collective decision, majority rule, group leader responsibility, or final expert opinion rule is specified.R19Discretion and capture risk controlReview standards or evidence baseScore 2 when review criteria, clinical/economic evidence, reasonableness checks, or assessment dimensions are specified.R20Discretion and capture risk controlWritten confirmation and archivingScore 2 when written signature, confirmation, archiving, or record-retention requirement is specified.R21Administrative burden and delay risk controlSubmission or review cycleScore 2 when monthly, quarterly, semi-annual, annual, or other specific review cycle is defined.R22Administrative burden and delay risk controlDeadline or overdue ruleScore 2 when application deadline, overdue rejection, waiver, or non-acceptance rule is specified.R23Administrative burden and delay risk controlDocument checklistScore 2 when required materials, forms, records, settlement lists, cost details, or supporting evidence are specified.R24Administrative burden and delay risk controlInformation system or formatScore 2 when online system, file format, platform, PDF/Excel template, or comparable submission format is specified.R25Administrative burden and delay risk controlMulti-level review processScore 2 when institutional, county, municipal, preliminary review, re-review, or expert-review sequence is specified.R26Administrative burden and delay risk controlProcessing time limitScore 2 when an explicit number of days, working days, or fixed processing window is specified.R27Accountability and equity risk controlOutcome feedback or disclosureScore 2 when feedback, disclosure, notification, public reporting, or system query of results is specified.R28Accountability and equity risk controlAppeal or reconsiderationScore 2 when objection, appeal, reconsideration, grievance submission, deadline, authority, or review route is specified.R29Accountability and equity risk controlFinality of decisionScore 2 when finality, non-appealability after review, or binding decision status is specified.R30Accountability and equity risk controlCross-area or fund responsibilityScore 2 when settlement locality, fund-bearing responsibility, out-of-area care, or cross-area handling rule is specified.Note: Final scoring uses 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. The full scoring logic is defined in the Methods section. Risk-Governance Coding Framework Note: Final scoring uses 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. The full scoring logic is defined in the Methods section. Coding Rules Each document-node combination was coded on a three-point ordinal scale. A score of 0 indicated that the document did not contain a relevant provision. A score of 1 indicated that the document mentioned the issue only in general or principle-based terms. A score of 2 indicated that the document converted the issue into an enforceable requirement. A provision was treated as enforceable when it specified an operational element such as a responsible actor, eligible case category, numerical threshold, ratio, time limit, application material, review process, expert requirement, system requirement, appeal rule, payment handling rule, or accountability consequence. Coding used a conservative evidence rule. Broad wording such as “standardised management”, “fairness”, “timely handling”, or “reasonable control of costs” was not scored as 2 unless it was accompanied by an operational rule. For example, a statement that high-cost cases may be considered was coded as 1, whereas a rule specifying that cases exceeding a defined cost multiple or payment-standard multiple may apply was coded as 2. Similarly, a general reference to expert review was coded as 1 unless the document specified the expert source, expert group, number of experts, random selection, recusal, collective decision, written confirmation, or comparable procedural safeguards. When a single policy provision supported more than one node, it was coded for each relevant node. However, within each document-node combination, only one final score was assigned. If several provisions in the same document addressed the same node, the highest evidence-supported score was used. This avoided double counting while allowing the coding to capture dispersed provisions within a document. Distinguishing a principle-level statement from an enforceable requirement involved interpretive judgement. To limit this subjectivity, score 2 required an explicit operational element, both coders worked independently before reconciliation, reliability was calculated before adjudication, and disagreements were resolved against the original policy text. Inter-Coder Reliability and Adjudication Two coders independently coded all 20 documents against the 30-node codebook, producing 600 initial coding units. The coders used the same definitions and scoring rules but completed the coding independently. Inter-coder reliability was assessed before final adjudication using exact percent agreement, Cohen’s kappa, linear weighted kappa, and quadratic weighted kappa. Weighted kappa was included because the coding scale was ordinal, and disagreement between 0 and 2 is substantively larger than disagreement between adjacent categories. After reliability assessment, all disagreements were checked against the original policy text and the codebook definitions. Each disputed unit was then assigned a final adjudicated score. A higher score was retained only when the original text clearly met the relevant criterion; otherwise, the lower evidence-supported score was used. All descriptive statistics, domain scores, attenuation analysis, and typology construction were based on the adjudicated coding matrix. The reliability statistics and disagreement records are reported in Table 3 and the adjudication records.Table 3Inter-Coder Reliability and Adjudication SummaryPanel A. Reliability MetricsMetricValueInterpretationTotal coded units60020 documents × 30 risk-control nodesExact agreements570Same 0/1/2 score by both codersPercent agreement95.0%Exact agreements/total coded unitsCohen’s kappa0.899Nominal agreement adjusted for chanceLinear weighted kappa0.932Ordinal agreement; adjacent disagreements weighted less severelyQuadratic weighted kappa0.959Ordinal agreement; larger disagreements weighted more severelyDisagreements adjudicated30All disagreements were manually adjudicatedPanel B. Coder 1 by Coder 2 Confusion MatrixCoder 1 Score\Coder 2 Score012Row Total0953001251071071200404404Column total95101404600Note: Reliability was calculated from the two independent 0/1/2 codes before adjudication. Final analytical tables used adjudicated final scores. Cell values in Panel B are counts of coding units before adjudication. Table 3Inter-Coder Reliability and Adjudication SummaryPanel A. Reliability MetricsMetricValueInterpretationTotal coded units60020 documents × 30 risk-control nodesExact agreements570Same 0/1/2 score by both codersPercent agreement95.0%Exact agreements/total coded unitsCohen’s kappa0.899Nominal agreement adjusted for chanceLinear weighted kappa0.932Ordinal agreement; adjacent disagreements weighted less severelyQuadratic weighted kappa0.959Ordinal agreement; larger disagreements weighted more severelyDisagreements adjudicated30All disagreements were manually adjudicatedPanel B. Coder 1 by Coder 2 Confusion MatrixCoder 1 Score\Coder 2 Score012Row Total0953001251071071200404404Column total95101404600Note: Reliability was calculated from the two independent 0/1/2 codes before adjudication. Final analytical tables used adjudicated final scores. Cell values in Panel B are counts of coding units before adjudication. Inter-Coder Reliability and Adjudication Summary Note: Reliability was calculated from the two independent 0/1/2 codes before adjudication. Final analytical tables used adjudicated final scores. Cell values in Panel B are counts of coding units before adjudication. Construction of Risk-Governance Indicators Descriptive indicators were derived from the adjudicated coding matrix. The final adjudicated score for document d on risk-control node n was denoted as x_dn. Scores could take three values: 0 for absence, 1 for principle-level mention, and 2 for an enforceable requirement. In the formulas below, I(·) denotes an indicator function, equal to 1 when the condition in parentheses is satisfied and 0 otherwise; N_k denotes the set of nodes in risk-control domain k; D denotes the number of documents included in the relevant analysis; and |·| denotes the number of nodes in a set. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${{\mathrm{x}}_{{\mathrm{dn}}}} \in {\mathrm{\{ 0,1,2\} }}{\mathrm{.}}$$\end{document} For each document, the overall Risk Governance Score (RGS) was calculated as the sum of the 30 final node scores divided by the maximum possible score of 60. The resulting 0–1 index captured overall risk-governance formalisation. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{RG}}{{\mathrm{S}}_{\mathrm{d}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{n=1}}}^{{\mathrm{30}}} {{\mathrm{x}}_{{\mathrm{dn}}}}} \over {{\mathrm{60}}}}{\mathrm{.}}$$\end{document} Domain-specific scores divided the observed score within each domain by the maximum possible score for that domain, allowing comparison across domains with different numbers of nodes. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{DomainScor}}{{\mathrm{e}}_{{\mathrm{dk}}}}{\mathrm{ = }}{{\mathop \sum \nolimits_{{\mathrm{n}} \in {{\mathrm{N}}_{\mathrm{k}}}} {{\mathrm{x}}_{{\mathrm{dn}}}}} \over {{\mathrm{2}}\left| {{{\mathrm{N}}_{\mathrm{k}}}} \right|}}{\mathrm{.}}$$\end{document} Third, the coverage rate and enforceability conversion rate were calculated for each document. Coverage measured whether a document mentioned a broad range of risk-control issues. Enforceability conversion measured how often mentioned issues were converted into enforceable requirements. Conversion was calculated only when the denominator was greater than zero. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Coverag}}{{\mathrm{e}}_d} = {{\mathop \sum \nolimits_{n = 1}^{30} I\left({{x_{dn}} \ge 1} \right)} \over {30}}.$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Conversio}}{{\mathrm{n}}_{\mathrm{d}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{n=1}}}^{{\mathrm{30}}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}}{\mathrm{=2}}} \right)} \over {\mathop \sum \nolimits_{{\mathrm{n=1}}}^{{\mathrm{30}}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}} \ge {\mathrm{1}}} \right)}}{\mathrm{.}}$$\end{document} Fourth, node-level indicators were calculated to support the heat-map analysis of coverage, enforceability, mean normalised score, and enforceability conversion. These indicators used D = 20 for all-document analysis and D = 18 for municipal-only analysis. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Coverag}}{{\mathrm{e}}_{\mathrm{n}}}{\mathrm{=}}\frac{{{\mathop \sum \nolimits_{{\mathrm{d=1}}}^{\mathrm{D}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}} \ge {\mathrm{1}}} \right)}} \over {{\mathrm{D}}}}{\mathrm{.}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{{Enforceabilit}}{{\mathrm{y}}_{\mathrm{n}}}{\mathrm{=}}\frac{{\mathop \sum \nolimits_{{\mathrm{d=1}}}^{\mathrm{D}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}}{\mathrm{=2}}} \right)}}\over {{\mathrm{D}}}}{\mathrm{.}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{MeanScor}}{{\mathrm{e}}_{\mathrm{n}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{d=1}}}^{\mathrm{D}} {{\mathrm{x}}_{{\mathrm{dn}}}}} \over {{\mathrm{2D}}}}{\mathrm{.}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Conversio}}{{\mathrm{n}}_{\mathrm{n}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{d=1}}}^{\mathrm{D}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}}{\mathrm{=2}}} \right)} \over {\mathop \sum \nolimits_{{\mathrm{d=1}}}^{\mathrm{D}} {\mathrm{I}}\left({{{\mathrm{x}}_{{\mathrm{dn}}}} \ge {\mathrm{1}}} \right)}}{\mathrm{.}}$$\end{document} Fifth, closed-loop completeness was assessed across eight procedural stages: eligibility criteria, quota or volume control, provider application, preliminary or multi-level review, expert review, decision or payment handling, outcome feedback or disclosure, and appeal, audit, accountability, or responsibility allocation. A stage score c_dj was coded as 0, 1, or 2 for document d and closed-loop stage j, using the same absence, principle-level mention, and enforceable-requirement logic. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{CL}}{{\mathrm{P}}_{\mathrm{d}}}{\mathrm{ = }}{{\sum\nolimits_{{{j = 1}}}^{\mathrm{8}} {\mathrm{I}} \left({{{\mathrm{c}}_{{\mathrm{dj}}}} \ge {\mathrm{1}}} \right)} \over{\mathrm{8}}}.$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{CL}}{{\mathrm{E}}_{\mathrm{d}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{j=1}}}^{\mathrm{8}} {\mathrm{I}}\left({{{\mathrm{c}}_{{\mathrm{dj}}}}{\mathrm{=2}}} \right)}\over{\mathrm{8}}}{\mathrm{.}}$$\end{document} These indicators were selected because they correspond to the main points at which payment exceptions can either reduce payment misfit risk or create discretionary, delay, and accountability risks. Vertical Attenuation Analysis Vertical attenuation analysis assessed whether risk-control signals in the provincial document were retained, weakened, or omitted in municipal documents. The provincial document was used as the baseline because it represented the intermediate policy layer between national guidance and municipal implementation rules. Let p_n denote the provincial baseline score for node n. For each node n and municipal document m, an attenuation indicator a_mn was coded as 1 when the provincial baseline was stronger than the municipal rule. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${{\mathrm{a}}_{{\mathrm{mn}}}}{\mathrm{=I}}\left({\left({{{\mathrm{p}}_{\mathrm{n}}}{\mathrm{=2}} \wedge {{\mathrm{x}}_{{\mathrm{mn}}}}{\mathrm{\lt\,2}}} \right) \vee \left({{{\mathrm{p}}_{\mathrm{n}}}{\mathrm{=1}} \wedge {{\mathrm{x}}_{{\mathrm{mn}}}}{\mathrm{=0}}} \right)} \right){\mathrm{.}}$$\end{document} Here, ∧ denotes “and” and ∨ denotes “or”. \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Attenuatio}}{{\mathrm{n}}_{\mathrm{n}}}{\mathrm{=}}{{\mathop \sum \nolimits_{{\mathrm{m=1}}}^{{\mathrm{18}}} {{\mathrm{a}}_{{\mathrm{mn}}}}} \over {{\mathrm{18}}}}{\mathrm{.}}$$\end{document} For each node, the attenuation rate therefore represented the proportion of municipal documents that weakened a provincial baseline provision. Node-level attenuation rates were used to identify high-risk breakpoints where provincial policy signals were not consistently translated into municipal rules. The attenuation results are reported in the Results section. Boundary-Procedure Typology To identify municipal governance models, a boundary-procedure typology was constructed for the 18 municipal documents. The boundary score combined payment misfit risk control and fund leakage and strategic application risk control. This dimension captured how strongly a document defined entry conditions, exception boundaries, application caps, and financial control mechanisms. The procedure score combined discretion and capture risk control, administrative burden and delay risk control, and accountability and equity risk control. This dimension captured how strongly a document formalised expert review, application procedures, time rules, feedback, appeal, decision finality, and responsibility allocation. Here, B denotes the boundary-control node set, covering payment misfit risk control and fund leakage/strategic application risk control (R1–R13; |B| = 13). P denotes the procedure-safeguard node set, covering discretion/capture, administrative burden/delay, and accountability/equity risk control (R14–R30; |P| = 17). \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Boundar}}{{\mathrm{y}}_{\mathrm{m}}}{\mathrm{ = }}{{\mathop \sum \nolimits_{{\mathrm{n}} \in {\mathrm{B}}} {{\mathrm{x}}_{{\mathrm{mn}}}}} \over {{\mathrm{2}}\left| {\mathrm{B}} \right|}}{\mathrm{.}}$$\end{document} \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Procedur}}{{\mathrm{e}}_{\mathrm{m}}}{\mathrm{ = }}{{\mathop \sum \nolimits_{{\mathrm{n}} \in {\mathrm{P}}} {{\mathrm{x}}_{{\mathrm{mn}}}}} \over {{\mathrm{2}}\left| {\mathrm{P}} \right|}}{\mathrm{.}}$$\end{document} Municipal documents were classified using the median boundary score and median procedure score as empirical cut-off points. Let M_B = 0.788 denote the municipal median boundary score and M_P = 0.765 denote the municipal median procedure score. The classification was descriptive and was not intended as a causal ranking of municipal performance. Balanced: \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Boundar}}{{\mathrm{y}}_{\mathrm{m}}} \ge {{\mathrm{M}}_{\mathrm{B}}}{\mathrm{andProcedur}}{{\mathrm{e}}_{\mathrm{m}}} \ge {{\mathrm{M}}_{\mathrm{P}}}{\mathrm{.}}$$\end{document} Boundary-heavy: \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Boundar}}{{\mathrm{y}}_{\mathrm{m}}} \ge {{\mathrm{M}}_{\mathrm{B}}}{\mathrm{andProcedur}}{{\mathrm{e}}_{\mathrm{m}}}{\mathrm{ < }}{{\mathrm{M}}_{\mathrm{P}}}{\mathrm{.}}$$\end{document} Procedure-heavy: \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Boundar}}{{\mathrm{y}}_{\mathrm{m}}}{\mathrm{ < }}{{\mathrm{M}}_{\mathrm{B}}}{\mathrm{andProcedur}}{{\mathrm{e}}_{\mathrm{m}}} \ge {{\mathrm{M}}_{\mathrm{P}}}{\mathrm{.}}$$\end{document} Low-formalisation: \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$${\mathrm{Boundar}}{{\mathrm{y}}_{\mathrm{m}}}{\mathrm{ < }}{{\mathrm{M}}_{\mathrm{B}}}{\mathrm{andProcedur}}{{\mathrm{e}}_{\mathrm{m}}}{\mathrm{ < }}{{\mathrm{M}}_{\mathrm{P}}}{\mathrm{.}}$$\end{document} Analytical Scope and Reporting Conventions The analysis was descriptive and separated formal policy-design evidence from implementation-outcome evidence. Because the dataset consisted of 20 policy documents and did not include implementation outcomes, regression modelling or causal inference was not appropriate. The results therefore describe formal policy design, not actual implementation, compliance, payment adjustment, fund expenditure, hospital behaviour, or patient outcomes. All statistics were calculated from the adjudicated coding matrix. Percentages were calculated using the relevant denominator for each analysis: 20 documents for overall document-node frequencies, 18 documents for municipal-only analysis, and the number of present provisions for enforceability conversion rates. Scores were rounded to three decimal places in tables and to one decimal place when reported as percentages in the text. Municipal identifiers were kept anonymous throughout the analysis. The design supports analysis of vertical translation within one provincial policy system but does not establish generalisability across all Chinese provinces. Anonymised English translations of relevant policy provisions are provided in Supplementary Material (Supplementary Note S1; Supplementary Texts S1–S20). The complete original adjudicated coding matrix is provided in Supplementary Table S1, the full coding framework and scoring criteria in Supplementary Table S2, the original inter-coder reliability summary in Supplementary Table S3, and the original disagreement-resolution record in Supplementary Table S4. All coding decisions were based on the original Chinese policy texts rather than the English translations. Results Inter-Coder Reliability and Adjudicated Coding Matrix The final coding dataset contained 20 policy documents and 30 risk-governance nodes, producing 600 coding units. The two coders reached exact agreement on 570 units, giving an overall percent agreement of 95.0%. Overall Cohen’s kappa was 0.899, with a linear weighted kappa of 0.932 and a quadratic weighted kappa of 0.959 (Table 3). Reliability was also high within domains. Exact agreement was 100.0% for payment misfit risk control, 95.0% for fund leakage and strategic application risk control, 98.6% for discretion and capture risk control, 92.5% for administrative burden and delay risk control, and 82.5% for accountability and equity risk control. Accountability and equity risk control showed the lowest domain-level agreement, with Cohen’s kappa of 0.717, because feedback, appeal, decision finality, and cross-area responsibility provisions required more interpretive judgement than eligibility or quota rules. All 30 initial disagreements were adjudicated before the final risk-governance scores were calculated. Overall Distribution of Risk-Governance Provisions Across all 20 documents, the highest normalised mean score was for payment misfit risk control (0.797), followed by discretion and capture risk control (0.786), administrative burden and delay risk control (0.767), fund leakage and strategic application risk control (0.720), and accountability and equity risk control (0.512) (Figure 2). The municipal subset showed a similar imbalance. Discretion and capture risk control had the highest municipal mean score (0.825), followed by payment misfit risk control (0.802), administrative burden and delay risk control (0.792), fund leakage and strategic application risk control (0.744), and accountability and equity risk control (0.507). Figure 2Domain-level formalisation of risk-governance provisions. Dark bars represent all 20 documents; light bars represent the 18 municipal documents. The dashed vertical line marks a normalised score of 0.50. Scores range from 0 to 1, with higher values indicating stronger formalisation of coded risk-control provisions.A horizontal bar graph showing normalized risk governance scores by risk control domain. Domain-level formalisation of risk-governance provisions. Dark bars represent all 20 documents; light bars represent the 18 municipal documents. The dashed vertical line marks a normalised score of 0.50. Scores range from 0 to 1, with higher values indicating stronger formalisation of coded risk-control provisions. The distribution of provisions showed that the policy texts were more developed in defining eligibility boundaries, review arrangements, and application procedures than in specifying downstream remedies and responsibility allocation. Table 4 summarises node-level coverage and enforceability, while Table 5 summarises domain-level counts and scores. Payment misfit risk control had 136 present provisions out of 160 possible document-node combinations, of which 119 were enforceable. Accountability and equity risk control had only 43 present provisions out of 80 possible combinations, although 39 of these were enforceable. Therefore, the main weakness of the accountability and equity domain was not that provisions were usually vague once present, but that such provisions were absent from many documents.Table 4Frequency and Enforceability of Risk-Control NodesNode CodeRISK DOMAINRisk-Control NodeDocuments with Score ≥1% Documents with Score ≥1Documents with Score =2% Documents with Score =2Mean Final ScoreEnforceability ConversionMunicipal Documents with Score ≥1% Municipal Score ≥1Municipal Documents with score =2% Municipal Score =2Risk ImplicationR1Payment misfit risk controlApplicable scope and eligible entities20100.0%20100.0%2.000100.0%18100.0%18100.0%Identifies whether payment exceptions protect clinically or financially atypical cases.R2Payment misfit risk controlHigh-cost trigger threshold20100.0%1260.0%1.60060.0%18100.0%1161.1%Identifies whether payment exceptions protect clinically or financially atypical cases.R3Payment misfit risk controlLength-of-stay trigger threshold1365.0%945.0%1.10069.2%1161.1%950.0%Identifies whether payment exceptions protect clinically or financially atypical cases.R4Payment misfit risk controlCritical or complex cases1995.0%1995.0%1.900100.0%1794.4%1794.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R5Payment misfit risk controlUngroupable or unstable group cases1890.0%1890.0%1.800100.0%1794.4%1794.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R6Payment misfit risk controlCombined surgery or transfer cases1995.0%1785.0%1.80089.5%1794.4%1688.9%Identifies whether payment exceptions protect clinically or financially atypical cases.R7Payment misfit risk controlNew technologies or new services1890.0%1575.0%1.65083.3%1688.9%1372.2%Identifies whether payment exceptions protect clinically or financially atypical cases.R8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseases945.0%945.0%0.900100.0%844.4%844.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R9Fund leakage and strategic application risk controlVolume or proportion cap20100.0%20100.0%2.000100.0%18100.0%18100.0%Controls overuse of the exception channel and protects fund integrity.R10Fund leakage and strategic application risk controlCMI or complexity adjustment1470.0%1365.0%1.35092.9%1372.2%1266.7%Controls overuse of the exception channel and protects fund integrity.R11Fund leakage and strategic application risk controlOver-quota handling1995.0%210.0%1.05010.5%18100.0%211.1%Controls overuse of the exception channel and protects fund integrity.R12Fund leakage and strategic application risk controlDeduction of unreasonable costs1575.0%1365.0%1.40086.7%1477.8%1372.2%Controls overuse of the exception channel and protects fund integrity.R13Fund leakage and strategic application risk controlAudit or penalty for problematic cases1680.0%1260.0%1.40075.0%1583.3%1161.1%Controls overuse of the exception channel and protects fund integrity.R14Discretion and capture risk controlReview organisation or expert panel20100.0%1785.0%1.85085.0%18100.0%1583.3%Constrains expert judgement and reduces discretionary or capture risk.R15Discretion and capture risk controlExpert source or selection mechanism20100.0%1575.0%1.75075.0%18100.0%1583.3%Constrains expert judgement and reduces discretionary or capture risk.R16Discretion and capture risk controlNumber or composition of experts1575.0%1470.0%1.45093.3%1477.8%1372.2%Constrains expert judgement and reduces discretionary or capture risk.R17Discretion and capture risk controlConflict-of-interest recusal1470.0%1260.0%1.30085.7%1477.8%1266.7%Constrains expert judgement and reduces discretionary or capture risk.R18Discretion and capture risk controlCollective decision or voting1680.0%1365.0%1.45081.2%1477.8%1372.2%Constrains expert judgement and reduces discretionary or capture risk.R19Discretion and capture risk controlReview standards or evidence base1785.0%1470.0%1.55082.4%1583.3%1477.8%Constrains expert judgement and reduces discretionary or capture risk.R20Discretion and capture risk controlWritten confirmation and archiving1785.0%1680.0%1.65094.1%1794.4%1688.9%Constrains expert judgement and reduces discretionary or capture risk.R21Administrative burden and delay risk controlSubmission or review cycle20100.0%20100.0%2.000100.0%18100.0%18100.0%Makes application, review, and processing more predictableR22Administrative burden and delay risk controlDeadline or overdue rule1785.0%1260.0%1.45070.6%1794.4%1266.7%Makes application, review, and processing more predictableR23Administrative burden and delay risk controlDocument checklist1680.0%1575.0%1.55093.8%1688.9%1583.3%Makes application, review, and processing more predictableR24Administrative burden and delay risk controlInformation system or format1575.0%1575.0%1.500100.0%1477.8%1477.8%Makes application, review, and processing more predictableR25Administrative burden and delay risk controlMulti-level review process20100.0%1680.0%1.80080.0%18100.0%1477.8%Makes application, review, and processing more predictableR26Administrative burden and delay risk controlProcessing time limit1050.0%840.0%0.90080.0%844.4%738.9%Makes application, review, and processing more predictableR27Accountability and equity risk controlOutcome feedback or disclosure1575.0%1575.0%1.500100.0%1372.2%1372.2%Supports feedback, challenge, responsibility allocation, and equitable handling.R28Accountability and equity risk controlAppeal or reconsideration840.0%525.0%0.65062.5%738.9%527.8%Supports feedback, challenge, responsibility allocation, and equitable handling.R29Accountability and equity risk controlFinality of decision1365.0%1260.0%1.25092.3%1372.2%1266.7%Supports feedback, challenge, responsibility allocation, and equitable handling.R30Accountability and equity risk controlCross-area or fund responsibility735.0%735.0%0.700100.0%527.8%527.8%Supports feedback, challenge, responsibility allocation, and equitable handling.Notes: Counts and rates are based on adjudicated final scores across all 20 policy documents. Scores use 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. R1–R30 = coded risk-control nodes. Table 5Domain-Level Risk-Governance SummaryRisk DomainNumber of NodesObserved Score (all docs)Maximum Possible Score (All Docs)Normalised Domain ScoreCoverage CountCoverage RateEnforceable CountEnforceable RateEnforceability ConversionMunicipal Observed ScoreMunicipal Maximum scoreMunicipal Normalised ScorePayment misfit risk control82553200.79713685.0%11974.4%87.5%231.000288.0000.802Fund leakage and strategic application risk control51442000.7208484.0%6060.0%71.4%134.000180.0000.744Discretion and capture risk control72202800.78611985.0%10172.1%84.9%208.000252.0000.825Administrative burden and delay risk control61842400.7679881.7%8671.7%87.8%171.000216.0000.792Accountability and equity risk control4821600.5124353.8%3948.8%90.7%73.000144.0000.507Notes: Normalized scores are observed score divided by maximum possible score in the same domain. Table 4Frequency and Enforceability of Risk-Control NodesNode CodeRISK DOMAINRisk-Control NodeDocuments with Score ≥1% Documents with Score ≥1Documents with Score =2% Documents with Score =2Mean Final ScoreEnforceability ConversionMunicipal Documents with Score ≥1% Municipal Score ≥1Municipal Documents with score =2% Municipal Score =2Risk ImplicationR1Payment misfit risk controlApplicable scope and eligible entities20100.0%20100.0%2.000100.0%18100.0%18100.0%Identifies whether payment exceptions protect clinically or financially atypical cases.R2Payment misfit risk controlHigh-cost trigger threshold20100.0%1260.0%1.60060.0%18100.0%1161.1%Identifies whether payment exceptions protect clinically or financially atypical cases.R3Payment misfit risk controlLength-of-stay trigger threshold1365.0%945.0%1.10069.2%1161.1%950.0%Identifies whether payment exceptions protect clinically or financially atypical cases.R4Payment misfit risk controlCritical or complex cases1995.0%1995.0%1.900100.0%1794.4%1794.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R5Payment misfit risk controlUngroupable or unstable group cases1890.0%1890.0%1.800100.0%1794.4%1794.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R6Payment misfit risk controlCombined surgery or transfer cases1995.0%1785.0%1.80089.5%1794.4%1688.9%Identifies whether payment exceptions protect clinically or financially atypical cases.R7Payment misfit risk controlNew technologies or new services1890.0%1575.0%1.65083.3%1688.9%1372.2%Identifies whether payment exceptions protect clinically or financially atypical cases.R8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseases945.0%945.0%0.900100.0%844.4%844.4%Identifies whether payment exceptions protect clinically or financially atypical cases.R9Fund leakage and strategic application risk controlVolume or proportion cap20100.0%20100.0%2.000100.0%18100.0%18100.0%Controls overuse of the exception channel and protects fund integrity.R10Fund leakage and strategic application risk controlCMI or complexity adjustment1470.0%1365.0%1.35092.9%1372.2%1266.7%Controls overuse of the exception channel and protects fund integrity.R11Fund leakage and strategic application risk controlOver-quota handling1995.0%210.0%1.05010.5%18100.0%211.1%Controls overuse of the exception channel and protects fund integrity.R12Fund leakage and strategic application risk controlDeduction of unreasonable costs1575.0%1365.0%1.40086.7%1477.8%1372.2%Controls overuse of the exception channel and protects fund integrity.R13Fund leakage and strategic application risk controlAudit or penalty for problematic cases1680.0%1260.0%1.40075.0%1583.3%1161.1%Controls overuse of the exception channel and protects fund integrity.R14Discretion and capture risk controlReview organisation or expert panel20100.0%1785.0%1.85085.0%18100.0%1583.3%Constrains expert judgement and reduces discretionary or capture risk.R15Discretion and capture risk controlExpert source or selection mechanism20100.0%1575.0%1.75075.0%18100.0%1583.3%Constrains expert judgement and reduces discretionary or capture risk.R16Discretion and capture risk controlNumber or composition of experts1575.0%1470.0%1.45093.3%1477.8%1372.2%Constrains expert judgement and reduces discretionary or capture risk.R17Discretion and capture risk controlConflict-of-interest recusal1470.0%1260.0%1.30085.7%1477.8%1266.7%Constrains expert judgement and reduces discretionary or capture risk.R18Discretion and capture risk controlCollective decision or voting1680.0%1365.0%1.45081.2%1477.8%1372.2%Constrains expert judgement and reduces discretionary or capture risk.R19Discretion and capture risk controlReview standards or evidence base1785.0%1470.0%1.55082.4%1583.3%1477.8%Constrains expert judgement and reduces discretionary or capture risk.R20Discretion and capture risk controlWritten confirmation and archiving1785.0%1680.0%1.65094.1%1794.4%1688.9%Constrains expert judgement and reduces discretionary or capture risk.R21Administrative burden and delay risk controlSubmission or review cycle20100.0%20100.0%2.000100.0%18100.0%18100.0%Makes application, review, and processing more predictableR22Administrative burden and delay risk controlDeadline or overdue rule1785.0%1260.0%1.45070.6%1794.4%1266.7%Makes application, review, and processing more predictableR23Administrative burden and delay risk controlDocument checklist1680.0%1575.0%1.55093.8%1688.9%1583.3%Makes application, review, and processing more predictableR24Administrative burden and delay risk controlInformation system or format1575.0%1575.0%1.500100.0%1477.8%1477.8%Makes application, review, and processing more predictableR25Administrative burden and delay risk controlMulti-level review process20100.0%1680.0%1.80080.0%18100.0%1477.8%Makes application, review, and processing more predictableR26Administrative burden and delay risk controlProcessing time limit1050.0%840.0%0.90080.0%844.4%738.9%Makes application, review, and processing more predictableR27Accountability and equity risk controlOutcome feedback or disclosure1575.0%1575.0%1.500100.0%1372.2%1372.2%Supports feedback, challenge, responsibility allocation, and equitable handling.R28Accountability and equity risk controlAppeal or reconsideration840.0%525.0%0.65062.5%738.9%527.8%Supports feedback, challenge, responsibility allocation, and equitable handling.R29Accountability and equity risk controlFinality of decision1365.0%1260.0%1.25092.3%1372.2%1266.7%Supports feedback, challenge, responsibility allocation, and equitable handling.R30Accountability and equity risk controlCross-area or fund responsibility735.0%735.0%0.700100.0%527.8%527.8%Supports feedback, challenge, responsibility allocation, and equitable handling.Notes: Counts and rates are based on adjudicated final scores across all 20 policy documents. Scores use 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. R1–R30 = coded risk-control nodes. Frequency and Enforceability of Risk-Control Nodes Notes: Counts and rates are based on adjudicated final scores across all 20 policy documents. Scores use 0 = absent, 1 = principle-level/general statement, and 2 = enforceable requirement. R1–R30 = coded risk-control nodes. Table 5Domain-Level Risk-Governance SummaryRisk DomainNumber of NodesObserved Score (all docs)Maximum Possible Score (All Docs)Normalised Domain ScoreCoverage CountCoverage RateEnforceable CountEnforceable RateEnforceability ConversionMunicipal Observed ScoreMunicipal Maximum scoreMunicipal Normalised ScorePayment misfit risk control82553200.79713685.0%11974.4%87.5%231.000288.0000.802Fund leakage and strategic application risk control51442000.7208484.0%6060.0%71.4%134.000180.0000.744Discretion and capture risk control72202800.78611985.0%10172.1%84.9%208.000252.0000.825Administrative burden and delay risk control61842400.7679881.7%8671.7%87.8%171.000216.0000.792Accountability and equity risk control4821600.5124353.8%3948.8%90.7%73.000144.0000.507Notes: Normalized scores are observed score divided by maximum possible score in the same domain. Domain-Level Risk-Governance Summary Notes: Normalized scores are observed score divided by maximum possible score in the same domain. Strong and Weak Risk-Control Nodes Three nodes were universally present and enforceable across all 20 documents: applicable scope and eligible entities (R1), volume or proportion cap (R9), and submission or review cycle (R21). In the municipal subset, all 18 documents also scored 2 on these three nodes. This indicates that the basic entry point, application volume control, and cyclical organisation of special case negotiation had become the most consistently formalised components of the policy design (Table 4; Figure 3). Figure 3Node-level coverage, enforceability, and conversion. Rows represent R1–R30. Columns show the proportion of documents with score ≥1 (coverage), the proportion with score =2 (enforceable), the mean final score divided by 2 (mean score), and the proportion of present provisions that were enforceable (conversion). The color scale ranges from 0 to 1; darker blue indicates a higher value. Colored strips identify the five risk-control domains.The heatmap displays data for rows R1 to R30, with columns Coverage, Enforceable, Mean score and Conversion. Each cell value ranges from 0.0 to 1.0. Notable patterns include rows R1, R9 and R21 having all values at 1.00, indicating maximum scores across all metrics. R11 has the lowest Enforceable score at 0.10. Conversion scores of 1.00 are seen in rows R4, R5, R8, R24, R27 and R30. The data highlights variations in performance across different nodes, with some achieving high scores in all categories, while others show mixed results.A heatmap of node-level coverage, enforceability, mean score and conversion for R1 to R30 with cell values.Abbreviations: CMI, case mix index. Node-level coverage, enforceability, and conversion. Rows represent R1–R30. Columns show the proportion of documents with score ≥1 (coverage), the proportion with score =2 (enforceable), the mean final score divided by 2 (mean score), and the proportion of present provisions that were enforceable (conversion). The color scale ranges from 0 to 1; darker blue indicates a higher value. Colored strips identify the five risk-control domains. Fund leakage and strategic application risk control showed a mixed pattern. The volume or proportion cap (R9) was enforceable in all 20 documents. CMI or complexity adjustment (R10) appeared in 14 documents and was enforceable in 13. Audit or penalty for problematic cases (R13) appeared in 16 documents and was enforceable in 12. Deduction of unreasonable costs (R12) appeared in 15 documents and was enforceable in 13. Over-quota handling (R11), by contrast, showed a clear enforceability gap: it appeared in 19 documents, but only 2 documents converted it into an enforceable requirement. Many documents recognised over-quota handling without attaching a strong operational consequence. Expert-review procedural safeguards were relatively well developed. Review organisation or expert panel (R14) appeared in all 20 documents and was enforceable in 17. Expert source or selection mechanism (R15) also appeared in all 20 documents and was enforceable in 15. Number or composition of experts (R16), conflict-of-interest recusal (R17), collective decision or voting (R18), review standards or evidence base (R19), and written confirmation and archiving (R20) were not universal, but where these provisions appeared they were usually enforceable. The weakest or least consistently enforceable nodes were concentrated in accountability, remedies, and time-bound procedural safeguards. Cross-area or fund responsibility (R30) appeared in only 7 of 20 documents and in only 5 of 18 municipal documents. Appeal or reconsideration (R28) appeared in 8 documents and in 7 municipal documents. Processing time limit (R26) appeared in 10 documents and in 8 municipal documents. These nodes mark the main residual gaps in the formal risk-governance design. Document-Level Variation in Risk-Governance Scores Document-level risk-governance scores varied substantially. The national document had an overall RGS of 0.533, whereas the provincial document had an overall RGS of 0.600. The moderate national score reflected the nature of the national text as a general DRG/DIP 2.0 implementation notice rather than a detailed operating procedure. The provincial document scored especially high in payment misfit risk control (0.938) and fund leakage and strategic application risk control (0.700), but lower in discretion and capture risk control (0.357) and administrative burden and delay risk control (0.417) (Table 6).Table 6Document-Level Risk-Governance ScoresDocument IDLevelYearPayment TypeTotal ScoreOverall RGSCoverage RateEnforceability ConversionPayment Misfit ScoreFund leakage scoreDiscretion ScoreAdministrative Burden ScoreAccountability/equity ScoreBoundary scoreProcedure ScoreBoundary–Procedure MismatchClosed-Loop PresenceClosed-Loop EnforceabilityN_NationalNational2024DRG-DIP320.53366.7%60.0%0.5620.3000.5000.6670.6250.4620.588−0.127100.0%87.5%N_ProvincialProvincial2025DRG-DIP360.60070.0%71.4%0.9380.7000.3570.4170.5000.8460.4120.434100.0%100.0%M01Municipal2024DRG520.86790.0%92.6%0.8751.0000.7860.7501.0000.9230.8240.100100.0%87.5%M02Municipal2024DIP440.73376.7%91.3%0.8750.9001.0000.5830.0000.8850.6180.26787.5%87.5%M03Municipal2023DRG510.85093.3%82.1%0.7500.8001.0001.0000.6250.7690.912−0.143100.0%87.5%M04Municipal2023DIP460.76780.0%91.7%0.6880.9000.8571.0000.2500.7690.7650.00587.5%87.5%M05Municipal2021DRG440.73386.7%69.2%0.8120.5000.9290.7500.5000.6920.765−0.072100.0%75.0%M06Municipal2023DIP480.80086.7%84.6%0.7500.7001.0000.9170.5000.7310.853−0.122100.0%87.5%M07Municipal2022DRG410.68376.7%78.3%0.8120.8000.4290.6670.7500.8080.5880.219100.0%87.5%M08Municipal2025DIP510.85090.0%88.9%1.0000.9000.9290.7500.5000.9620.7650.19787.5%87.5%M09Municipal2025DIP400.66773.3%81.8%1.0000.9000.5000.3330.5000.9620.4410.520100.0%100.0%M10Municipal2025DIP500.83393.3%78.6%0.8750.8000.8570.7500.8750.8460.8240.023100.0%100.0%M11Municipal2024DIP510.85086.7%96.2%0.7500.7001.0001.0000.7500.7310.941−0.210100.0%100.0%M12Municipal2025DIP390.65070.0%85.7%0.6880.6000.8570.8330.0000.6540.6470.00787.5%87.5%M13Municipal2025DIP510.85090.0%88.9%0.6881.0001.0000.8330.7500.8080.882−0.075100.0%100.0%M14Municipal2022DIP450.75080.0%87.5%0.7500.9001.0000.5830.3750.8080.7060.102100.0%100.0%M15Municipal2025DIP370.61766.7%85.0%0.8750.3000.4290.8330.5000.6540.5880.06687.5%87.5%M16Municipal2024DIP530.88390.0%96.3%1.0000.7001.0001.0000.5000.8850.8820.002100.0%100.0%M17Municipal2022DRG430.71773.3%95.5%0.6250.5001.0000.8330.5000.5770.824−0.24787.5%87.5%M18Municipal2025DRG310.51760.0%72.2%0.6250.5000.2860.8330.2500.5770.4710.10675.0%75.0%Notes: RGS = total final score/60. Boundary score and procedure score are normalized composites. M01–M18 identify anonymised municipal documents. Table 6Document-Level Risk-Governance ScoresDocument IDLevelYearPayment TypeTotal ScoreOverall RGSCoverage RateEnforceability ConversionPayment Misfit ScoreFund leakage scoreDiscretion ScoreAdministrative Burden ScoreAccountability/equity ScoreBoundary scoreProcedure ScoreBoundary–Procedure MismatchClosed-Loop PresenceClosed-Loop EnforceabilityN_NationalNational2024DRG-DIP320.53366.7%60.0%0.5620.3000.5000.6670.6250.4620.588−0.127100.0%87.5%N_ProvincialProvincial2025DRG-DIP360.60070.0%71.4%0.9380.7000.3570.4170.5000.8460.4120.434100.0%100.0%M01Municipal2024DRG520.86790.0%92.6%0.8751.0000.7860.7501.0000.9230.8240.100100.0%87.5%M02Municipal2024DIP440.73376.7%91.3%0.8750.9001.0000.5830.0000.8850.6180.26787.5%87.5%M03Municipal2023DRG510.85093.3%82.1%0.7500.8001.0001.0000.6250.7690.912−0.143100.0%87.5%M04Municipal2023DIP460.76780.0%91.7%0.6880.9000.8571.0000.2500.7690.7650.00587.5%87.5%M05Municipal2021DRG440.73386.7%69.2%0.8120.5000.9290.7500.5000.6920.765−0.072100.0%75.0%M06Municipal2023DIP480.80086.7%84.6%0.7500.7001.0000.9170.5000.7310.853−0.122100.0%87.5%M07Municipal2022DRG410.68376.7%78.3%0.8120.8000.4290.6670.7500.8080.5880.219100.0%87.5%M08Municipal2025DIP510.85090.0%88.9%1.0000.9000.9290.7500.5000.9620.7650.19787.5%87.5%M09Municipal2025DIP400.66773.3%81.8%1.0000.9000.5000.3330.5000.9620.4410.520100.0%100.0%M10Municipal2025DIP500.83393.3%78.6%0.8750.8000.8570.7500.8750.8460.8240.023100.0%100.0%M11Municipal2024DIP510.85086.7%96.2%0.7500.7001.0001.0000.7500.7310.941−0.210100.0%100.0%M12Municipal2025DIP390.65070.0%85.7%0.6880.6000.8570.8330.0000.6540.6470.00787.5%87.5%M13Municipal2025DIP510.85090.0%88.9%0.6881.0001.0000.8330.7500.8080.882−0.075100.0%100.0%M14Municipal2022DIP450.75080.0%87.5%0.7500.9001.0000.5830.3750.8080.7060.102100.0%100.0%M15Municipal2025DIP370.61766.7%85.0%0.8750.3000.4290.8330.5000.6540.5880.06687.5%87.5%M16Municipal2024DIP530.88390.0%96.3%1.0000.7001.0001.0000.5000.8850.8820.002100.0%100.0%M17Municipal2022DRG430.71773.3%95.5%0.6250.5001.0000.8330.5000.5770.824−0.24787.5%87.5%M18Municipal2025DRG310.51760.0%72.2%0.6250.5000.2860.8330.2500.5770.4710.10675.0%75.0%Notes: RGS = total final score/60. Boundary score and procedure score are normalized composites. M01–M18 identify anonymised municipal documents. Document-Level Risk-Governance Scores Notes: RGS = total final score/60. Boundary score and procedure score are normalized composites. M01–M18 identify anonymised municipal documents. Among the 18 municipal documents, the mean overall RGS was 0.756, with a median of 0.758 and a range from 0.517 to 0.883. The mean coverage rate was 81.3%, and the mean enforceability conversion rate was 85.9%. Closed-loop presence ranged from 75.0% to 100.0%, with a municipal mean of 94.4%, while closed-loop enforceability ranged from 75.0% to 100.0%, with a municipal mean of 90.3%. In practical terms, most municipal policies included the main procedural stages of special case negotiation, but the level of enforceable detail still differed across jurisdictions. Because the maximum raw score was 60, a normalised RGS difference of 0.10 corresponds to six raw coding points. This conversion is provided only to aid interpretation of textual formalisation and does not represent a clinical, fiscal, statistical, or policy-significance threshold. The highest municipal overall RGS was observed in M16 (0.883), followed by M01 (0.867), and M03, M08, and M11 (all 0.850). The lowest municipal scores were observed in M18 (0.517), M15 (0.617), M12 (0.650), M09 (0.667), and M07 (0.683). Variation was especially large in accountability and equity risk control, where municipal scores ranged from 0 to 1. Discretion and capture risk control also ranged widely, from 0.286 to 1.000. The main difference between municipal documents was therefore not simply whether they referred to special case negotiation, but the extent to which they formalised later-stage procedural and accountability safeguards. Some municipal documents combined relatively high overall scores with weak accountability and equity provisions. For example, M02 had an overall RGS of 0.733 but scored 0 on accountability and equity risk control. M12 also scored 0 on accountability and equity risk control. These cases show that a high or moderate overall score may still coexist with no provisions on appeal, feedback, decision finality, or cross-area responsibility. Vertical Attenuation from Provincial Baseline to Municipal Rules The strongest attenuation was found in cross-area or fund responsibility (R30). The provincial document scored 2 on this node, but 13 of the 18 municipal documents were weaker than the provincial baseline, giving an attenuation rate of 72.2%. Thirteen municipal documents had no provision on this node, and only five municipal documents contained an enforceable provision. Figure 4 visualises the highest-attenuation nodes and shows that attenuation was concentrated in back-end responsibility, time-bound processing, special disease coverage, high-cost and length-of-stay triggers, audit or penalty, and CMI or complexity adjustment. The full node-level attenuation results are reported in Table 7.Table 7Vertical Attenuation from Provincial Baseline to Municipal RulesNode CodeRisk DomainRisk-Control NodeNational ScoreProvincial Baseline ScoreMunicipal Mean ScoreMunicipal Documents Weaker Than Provincial BaselineAttenuation RateMunicipal IDs with AttenuationRisk InterpretationR1Payment misfit risk controlApplicable scope and eligible entities222.00000.0%No municipal attenuation against provincial baselineR2Payment misfit risk controlHigh-cost trigger threshold121.611738.9%M03, M05, M06, M07, M12, M13, M18Moderate vertical attenuationR3Payment misfit risk controlLength-of-stay trigger threshold111.111738.9%M02, M03, M05, M11, M12, M14, M17Moderate vertical attenuationR4Payment misfit risk controlCritical or complex cases221.88915.6%M17Moderate vertical attenuationR5Payment misfit risk controlUngroupable or unstable group cases021.88915.6%M10Moderate vertical attenuationR6Payment misfit risk controlCombined surgery or transfer cases121.833211.1%M04, M06Moderate vertical attenuationR7Payment misfit risk controlNew technologies or new services221.611527.8%M03, M04, M12, M13, M18Moderate vertical attenuationR8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseases020.8891055.6%M01, M04, M06, M07, M11, M13, M14, M15, M17, M18High vertical attenuationR9Fund leakage and strategic application risk controlVolume or proportion cap222.00000.0%No municipal attenuation against provincial baselineR10Fund leakage and strategic application risk controlCMI or complexity adjustment021.389633.3%M05, M06, M11, M12, M15, M16Moderate vertical attenuationR11Fund leakage and strategic application risk controlOver-quota handling011.11100.0%No municipal attenuation against provincial baselineR12Fund leakage and strategic application risk controlDeduction of unreasonable costs101.5000Not applicable: provincial baseline absent.R13Fund leakage and strategic application risk controlAudit or penalty for problematic cases021.444738.9%M03, M05, M07, M10, M15, M17, M18Moderate vertical attenuationR14Discretion and capture risk controlReview organisation or expert panel221.833316.7%M10, M15, M18Moderate vertical attenuationR15Discretion and capture risk controlExpert source or selection mechanism111.83300.0%No municipal attenuation against provincial baselineR16Discretion and capture risk controlNumber or composition of experts201.5000Not applicable: provincial baseline absent.R17Discretion and capture risk controlConflict-of-interest recusal001.4440Not applicable: provincial baseline absent.R18Discretion and capture risk controlCollective decision or voting111.500422.2%M07, M09, M15, M18Moderate vertical attenuationR19Discretion and capture risk controlReview standards or evidence base111.611316.7%M04, M07, M18Moderate vertical attenuationR20Discretion and capture risk controlWritten confirmation and archiving001.8330Not applicable: provincial baseline absent.R21Administrative burden and delay risk controlSubmission or review cycle222.00000.0%No municipal attenuation against provincial baselineR22Administrative burden and delay risk controlDeadline or overdue rule001.6110Not applicable: provincial baseline absent.R23Administrative burden and delay risk controlDocument checklist001.7220Not applicable: provincial baseline absent.R24Administrative burden and delay risk controlInformation system or format201.5560Not applicable: provincial baseline absent.R25Administrative burden and delay risk controlMulti-level review process221.778422.2%M01, M05, M06, M07Moderate vertical attenuationR26Administrative burden and delay risk controlProcessing time limit210.8331055.6%M01, M02, M08, M09, M10, M12, M13, M14, M15, M18High vertical attenuationR27Accountability and equity risk controlOutcome feedback or disclosure221.444527.8%M02, M04, M08, M12, M18Moderate vertical attenuationR28Accountability and equity risk controlAppeal or reconsideration100.6670Not applicable: provincial baseline absent.R29Accountability and equity risk controlFinality of decision001.3890Not applicable: provincial baseline absent.R30Accountability and equity risk controlCross-area or fund responsibility220.5561372.2%M02, M03, M04, M05, M06, M07, M11, M12, M13, M14, M15, M16, M17High vertical attenuationNotes: A municipal document is weaker when the provincial baseline is 2 and municipal score is 0/1, or baseline is 1 and municipal score is 0. R1–R30 = coded risk-control nodes; M01–M18 identify anonymised municipal documents. Figure 4High-attenuation risk-control nodes at municipal level. Attenuation means that a municipal score was weaker than the provincial baseline. Bar labels show the attenuation rate and count of weaker municipal documents out of 18. Colors identify risk-control domains; higher rates indicate greater loss of formalisation during municipal translation.A horizontal bar graph displays categories on the y-axis and Municipal attenuation rate (%) on the x-axis, ranging from 0 to 80. The y-axis lists eight nodes: R30 Cross area/fund responsibility, R26 Processing time limit, R8 Special/infectious/rare diseases, R13 Audit or penalty, R2 High cost trigger, R3 Length of stay trigger, R10 CMI/complexity adjustment, R27 Outcome feedback/disclosure. Bar end labels show percentages and counts out of 18: R30 72.2% (13/18), R26 55.6% (10/18), R8 55.6% (10/18), R13 38.9% (7/18), R2 38.9% (7/18), R3 38.9% (7/18), R10 33.3% (6/18), R27 27.8% (5/18). Legend entries include: Payment misfit, Fund leakage/strategic use, Administrative burden/delay, Accountability/equity.A horizontal bar graph showing municipal attenuation rates across risk control nodes.Abbreviations: CMI, case mix index. Table 7Vertical Attenuation from Provincial Baseline to Municipal RulesNode CodeRisk DomainRisk-Control NodeNational ScoreProvincial Baseline ScoreMunicipal Mean ScoreMunicipal Documents Weaker Than Provincial BaselineAttenuation RateMunicipal IDs with AttenuationRisk InterpretationR1Payment misfit risk controlApplicable scope and eligible entities222.00000.0%No municipal attenuation against provincial baselineR2Payment misfit risk controlHigh-cost trigger threshold121.611738.9%M03, M05, M06, M07, M12, M13, M18Moderate vertical attenuationR3Payment misfit risk controlLength-of-stay trigger threshold111.111738.9%M02, M03, M05, M11, M12, M14, M17Moderate vertical attenuationR4Payment misfit risk controlCritical or complex cases221.88915.6%M17Moderate vertical attenuationR5Payment misfit risk controlUngroupable or unstable group cases021.88915.6%M10Moderate vertical attenuationR6Payment misfit risk controlCombined surgery or transfer cases121.833211.1%M04, M06Moderate vertical attenuationR7Payment misfit risk controlNew technologies or new services221.611527.8%M03, M04, M12, M13, M18Moderate vertical attenuationR8Payment misfit risk controlSpecial diseases, infectious diseases, or rare diseases020.8891055.6%M01, M04, M06, M07, M11, M13, M14, M15, M17, M18High vertical attenuationR9Fund leakage and strategic application risk controlVolume or proportion cap222.00000.0%No municipal attenuation against provincial baselineR10Fund leakage and strategic application risk controlCMI or complexity adjustment021.389633.3%M05, M06, M11, M12, M15, M16Moderate vertical attenuationR11Fund leakage and strategic application risk controlOver-quota handling011.11100.0%No municipal attenuation against provincial baselineR12Fund leakage and strategic application risk controlDeduction of unreasonable costs101.5000Not applicable: provincial baseline absent.R13Fund leakage and strategic application risk controlAudit or penalty for problematic cases021.444738.9%M03, M05, M07, M10, M15, M17, M18Moderate vertical attenuationR14Discretion and capture risk controlReview organisation or expert panel221.833316.7%M10, M15, M18Moderate vertical attenuationR15Discretion and capture risk controlExpert source or selection mechanism111.83300.0%No municipal attenuation against provincial baselineR16Discretion and capture risk controlNumber or composition of experts201.5000Not applicable: provincial baseline absent.R17Discretion and capture risk controlConflict-of-interest recusal001.4440Not applicable: provincial baseline absent.R18Discretion and capture risk controlCollective decision or voting111.500422.2%M07, M09, M15, M18Moderate vertical attenuationR19Discretion and capture risk controlReview standards or evidence base111.611316.7%M04, M07, M18Moderate vertical attenuationR20Discretion and capture risk controlWritten confirmation and archiving001.8330Not applicable: provincial baseline absent.R21Administrative burden and delay risk controlSubmission or review cycle222.00000.0%No municipal attenuation against provincial baselineR22Administrative burden and delay risk controlDeadline or overdue rule001.6110Not applicable: provincial baseline absent.R23Administrative burden and delay risk controlDocument checklist001.7220Not applicable: provincial baseline absent.R24Administrative burden and delay risk controlInformation system or format201.5560Not applicable: provincial baseline absent.R25Administrative burden and delay risk controlMulti-level review process221.778422.2%M01, M05, M06, M07Moderate vertical attenuationR26Administrative burden and delay risk controlProcessing time limit210.8331055.6%M01, M02, M08, M09, M10, M12, M13, M14, M15, M18High vertical attenuationR27Accountability and equity risk controlOutcome feedback or disclosure221.444527.8%M02, M04, M08, M12, M18Moderate vertical attenuationR28Accountability and equity risk controlAppeal or reconsideration100.6670Not applicable: provincial baseline absent.R29Accountability and equity risk controlFinality of decision001.3890Not applicable: provincial baseline absent.R30Accountability and equity risk controlCross-area or fund responsibility220.5561372.2%M02, M03, M04, M05, M06, M07, M11, M12, M13, M14, M15, M16, M17High vertical attenuationNotes: A municipal document is weaker when the provincial baseline is 2 and municipal score is 0/1, or baseline is 1 and municipal score is 0. R1–R30 = coded risk-control nodes; M01–M18 identify anonymised municipal documents. Vertical Attenuation from Provincial Baseline to Municipal Rules Notes: A municipal document is weaker when the provincial baseline is 2 and municipal score is 0/1, or baseline is 1 and municipal score is 0. R1–R30 = coded risk-control nodes; M01–M18 identify anonymised municipal documents. High-attenuation risk-control nodes at municipal level. Attenuation means that a municipal score was weaker than the provincial baseline. Bar labels show the attenuation rate and count of weaker municipal documents out of 18. Colors identify risk-control domains; higher rates indicate greater loss of formalisation during municipal translation. Two nodes had attenuation rates of 55.6%: special diseases, infectious diseases, or rare diseases (R8), and processing time limit (R26). For R8, 10 municipal documents were weaker than the provincial baseline and 10 had no provision. For R26, 10 municipal documents were also weaker than the provincial baseline and 10 had no provision, while only 7 municipal documents contained an enforceable processing time limit. Municipal rules often translated the general idea of special case negotiation into eligibility and review provisions, but they did not consistently retain the provincial-level signals on special disease coverage and time-bound processing. Moderate attenuation was also observed in several other nodes. High-cost trigger threshold (R2), length-of-stay trigger threshold (R3), and audit or penalty for problematic cases (R13) were each weaker than the provincial baseline in 7 of 18 municipal documents, corresponding to an attenuation rate of 38.9%. CMI or complexity adjustment (R10) was weaker in 6 municipal documents (33.3%). New technologies or new services (R7) and outcome feedback or disclosure (R27) were each weaker in 5 municipal documents (27.8%). Multi-level review process (R25) and collective decision or voting (R18) were each weaker in 4 municipal documents (22.2%). Not every node showed attenuation. Applicable scope and eligible entities (R1), volume or proportion cap (R9), and submission or review cycle (R21) were fully retained at the municipal level, with all 18 municipal documents scoring 2. Several procedural nodes had low or no attenuation because the provincial document itself did not define a strong operational baseline. In these areas, some municipal documents supplied more detailed rules than the provincial text, particularly for expert composition, recusal, written confirmation, archiving, deadlines, and document checklists. Boundary-Procedure Mismatch and Municipal Typology The municipal typology was constructed by comparing two composite scores: the boundary score, which combined payment misfit risk control and fund leakage and strategic application risk control, and the procedure score, which combined discretion and capture risk control, administrative burden and delay risk control, and accountability and equity risk control. The median thresholds were 0.788 for boundary score and 0.765 for procedure score. This classification produced four municipal risk-governance models (Table 8; Figure 5).Table 8Municipal Boundary-Procedure TypologyDocument IDMunicipal Anonymous IDYearPayment TypeBoundary ScoreProcedure ScoreBoundary–Procedure MismatchTypologyInterpretationM01M012024DRG0.9230.8240.100Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M02M022024DIP0.8850.6180.267Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M03M032023DRG0.7690.912−0.143Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M04M042023DIP0.7690.7650.005Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M05M052021DRG0.6920.765−0.072Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M06M062023DIP0.7310.853−0.122Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M07M072022DRG0.8080.5880.219Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M08M082025DIP0.9620.7650.197Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M09M092025DIP0.9620.4410.520Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M10M102025DIP0.8460.8240.023Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M11M112024DIP0.7310.941−0.210Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M12M122025DIP0.6540.6470.007Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.M13M132025DIP0.8080.882−0.075Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M14M142022DIP0.8080.7060.102Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M15M152025DIP0.6540.5880.066Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.M16M162024DIP0.8850.8820.002Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M17M172022DRG0.5770.824−0.247Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M18M182025DRG0.5770.4710.106Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.Notes: Typology uses municipal medians as cutoffs: boundary median = 0.788; procedure median = 0.765. M01–M18 identify anonymised municipal documents. Figure 5Municipal boundary–procedure configurations. Each point represents one anonymised municipal document (M01–M18). The horizontal axis is the boundary score for R1–R13; the vertical axis is the procedure score for R14–R30. Dashed lines mark the municipal medians (boundary, 0.788; procedure, 0.765). Point colors denote balanced, boundary-heavy, procedure-heavy, and low-formalisation models. Higher scores indicate stronger formalisation; the quadrants describe policy-design configurations rather than implementation performance.A scatter plot features points M01 to M18 with a legend titled ′Municipal typology.′ The x-axis, ′Boundary score,′ ranges from 0.6 to 1.0 and the y-axis, ′Procedure score,′ ranges from 0.4 to 0.9. Dashed lines indicate medians: horizontal at ′Procedure median = 0.765′ and vertical at ′Boundary median = 0.788.′ Quadrants are labeled: top left ′Procedure heavy,′ top right ′Balanced,′ bottom left ′Low formalisation,′ bottom right ′Boundary heavy.′ Legend categories: Balanced risk governance, Boundary heavy, Low formalisation, Procedure heavy. Points: Balanced model includes M13 (0.80, 0.88), M16 (0.88, 0.88), M10 (0.85, 0.82), M01 (0.92, 0.82), M08 (0.96, 0.76). Boundary heavy model includes M14 (0.80, 0.70), M07 (0.80, 0.59), M02 (0.89, 0.62), M09 (0.96, 0.44). Low formalisation model includes M12 (0.66, 0.64), M15 (0.66, 0.59), M18 (0.58, 0.47). Procedure heavy model includes M17 (0.59, 0.82), M11 (0.74, 0.94), M03 (0.77, 0.91), M06 (0.74, 0.85), M05 (0.69, 0.76), M04 (0.77, 0.76).A scatter plot showing municipal typology by boundary score and procedure score. Table 8Municipal Boundary-Procedure TypologyDocument IDMunicipal Anonymous IDYearPayment TypeBoundary ScoreProcedure ScoreBoundary–Procedure MismatchTypologyInterpretationM01M012024DRG0.9230.8240.100Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M02M022024DIP0.8850.6180.267Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M03M032023DRG0.7690.912−0.143Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M04M042023DIP0.7690.7650.005Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M05M052021DRG0.6920.765−0.072Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M06M062023DIP0.7310.853−0.122Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M07M072022DRG0.8080.5880.219Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M08M082025DIP0.9620.7650.197Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M09M092025DIP0.9620.4410.520Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M10M102025DIP0.8460.8240.023Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M11M112024DIP0.7310.941−0.210Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M12M122025DIP0.6540.6470.007Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.M13M132025DIP0.8080.882−0.075Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M14M142022DIP0.8080.7060.102Boundary-heavy modelEligibility and quota controls are comparatively stronger than procedural safeguards.M15M152025DIP0.6540.5880.066Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.M16M162024DIP0.8850.8820.002Balanced risk-governance modelBoundary controls and procedural safeguards are both comparatively strong.M17M172022DRG0.5770.824−0.247Procedure-heavy modelProcedural safeguards are comparatively stronger than eligibility/quota controls.M18M182025DRG0.5770.4710.106Low-formalisation modelBoth boundary controls and procedural safeguards are comparatively weak.Notes: Typology uses municipal medians as cutoffs: boundary median = 0.788; procedure median = 0.765. M01–M18 identify anonymised municipal documents. Municipal Boundary-Procedure Typology Notes: Typology uses municipal medians as cutoffs: boundary median = 0.788; procedure median = 0.765. M01–M18 identify anonymised municipal documents. Municipal boundary–procedure configurations. Each point represents one anonymised municipal document (M01–M18). The horizontal axis is the boundary score for R1–R13; the vertical axis is the procedure score for R14–R30. Dashed lines mark the municipal medians (boundary, 0.788; procedure, 0.765). Point colors denote balanced, boundary-heavy, procedure-heavy, and low-formalisation models. Higher scores indicate stronger formalisation; the quadrants describe policy-design configurations rather than implementation performance. Five municipal documents were classified as balanced risk-governance models: M01, M08, M10, M13, and M16. These documents scored at or above the median on both boundary and procedure dimensions. Their boundary scores ranged from 0.808 to 0.962, and their procedure scores ranged from 0.765 to 0.882. M16 was the most balanced case, with a boundary score of 0.885, a procedure score of 0.882, and a boundary-procedure mismatch of only 0.002. Substantively, the balanced model indicates that eligibility and fund-control boundaries were paired with relatively strong procedural safeguards. Four municipal documents were classified as boundary-heavy models: M02, M07, M09, and M14. These documents had relatively stronger eligibility and fund-control provisions than procedural safeguards. M09 had the largest positive mismatch, with a boundary score of 0.962 and a procedure score of 0.441, producing a mismatch of 0.520. M02 also showed a clear boundary-heavy pattern, with a boundary score of 0.885 and a procedure score of 0.618. This model suggests that admission conditions, quotas, or payment boundaries were more formalised than feedback, appeal, processing time, and accountability safeguards. Six municipal documents were classified as procedure-heavy models: M03, M04, M05, M06, M11, and M17. These documents had stronger procedural provisions than boundary-control provisions. The largest negative mismatch occurred in M17, where the boundary score was 0.577 and the procedure score was 0.824, producing a mismatch of −0.247. M11 also showed a procedure-heavy pattern, with a boundary score of 0.731 and a procedure score of 0.941. This model indicates that review organisation, documentation, or procedural safeguards were more complete than the front-end definition of exception boundaries. Three municipal documents were classified as low-formalisation models: M12, M15, and M18. These documents scored below the median on both boundary and procedure dimensions. M18 had the lowest overall formalisation, with a boundary score of 0.577 and a procedure score of 0.471. This group shows that some municipal policies still relied on a relatively limited formal rule structure for governing special case negotiation. Overall, the results show a clear pattern: special case negotiation policies were strongest in defining who may apply, how many cases may be submitted, and how review cycles are organised. They were less consistent in defining appeal, processing time limits, cross-area responsibility, and other accountability-oriented safeguards. The main risk-governance gap therefore lies not in the absence of the special case negotiation mechanism itself, but in the uneven conversion of that mechanism into enforceable, traceable, and remediable procedural rules. The typology identifies different areas for policy improvement rather than ranking implementation performance. Discussion Main Findings This study examined special case negotiation under China’s DRG/DIP payment reform as a risk-governance mechanism rather than as a simple reimbursement supplement. The results show three central patterns. First, pre-adjudication inter-coder agreement was high, supporting the consistency of the coding process. Second, the formal policy design was uneven across risk domains: payment misfit, discretion, and administrative procedure were more formalised than accountability and equity. Third, municipal documents varied substantially in both overall Risk Governance Score (RGS) and boundary-procedure configuration. Figure 1 provides the mechanism logic; Figures 2 and 3 show that risk-control provisions were not converted into enforceable safeguards evenly across domains and nodes; Figure 4 shows that attenuation was concentrated in high-risk procedural and accountability nodes; and Figure 5 shows that municipalities adopted different boundary-procedure models. This interpretation is consistent with payment-reform studies showing that case-based payment is simultaneously a cost-control instrument,4,5 a behavioural incentive,11,14 and an implementation challenge.32,34 The strongest provisions were concentrated in front-end access control and review organisation. All documents specified applicable scope and eligible entities, volume or proportion caps, and submission or review cycles. Most municipal documents also specified critical or complex cases, ungroupable or unstable groups, combined surgery or transfer cases, new technologies, review organisation, expert source, and written confirmation or archiving. By contrast, the weakest provisions were located at the back end of the governance chain. Cross-area or fund responsibility, appeal or reconsideration, and processing time limits were much less consistently specified. This imbalance mirrors the procedural fairness distinction between defining eligibility and providing reasons, reviewability, remedy, and enforcement.43,44 RGS reflects the density and enforceability of written rules; it is neither a clinical outcome measure nor a ranking of policy or implementation performance. The practical meaning of an RGS difference depends on the composition of the missing safeguards: the same numerical gap may reflect absent eligibility thresholds, processing deadlines, written reasons, or appeal routes, each of which creates a different governance vulnerability. Accordingly, two documents with the same numerical difference may require different policy responses. Special Case Negotiation as a Risk-Governance Mechanism These results support a reframing of special case negotiation. Under DRG/DIP payment, routine reimbursement is organised around case groups, disease-value points, budgets, and standardised payment units. Such payment rules can improve predictability and cost control, but they also generate payment misfit risk for cases that fall outside ordinary cost or clinical patterns. Complex and critical illness, long hospital stays, high-cost cases, ungroupable cases, multidisciplinary or combined surgical cases, and new technologies may not be adequately accommodated by a fixed payment standard. In this sense, special case negotiation acts as a safety valve within case-based payment reform. Empirical studies of DRG/DIP and related payment reforms show that these systems can reduce average costs4,5 but produce heterogeneous effects across diseases,16,18 hospitals, and patient groups.7,9,21 An exception channel, however, also creates risks of its own. Loose eligibility boundaries may turn the mechanism into a strategic route for additional compensation and weaken fund control. Boundaries that are too strict may leave complex cases under-compensated, creating incentives for avoidance, delayed admission, undertreatment, or risk selection. Poorly constrained expert review may make decisions depend on local discretion, professional networks, or inconsistent judgement. A burdensome or slow procedure may exist on paper but fail to provide timely relief. Unclear rules on results, appeals, and cross-area responsibilities may cause similar cases to be handled differently across institutions or jurisdictions. These risks are closely related to the risk-adjustment problem: a payment system must recognise legitimate complexity35,36 without creating coding or selection incentives.41,42 The mechanism therefore sits at the intersection of healthcare risk management, provider payment, and policy implementation. Local policies have already converted some risks into enforceable rules: volume caps and submission cycles were universal, and expert-review arrangements were common. At the same time, the low accountability47,48 and equity score points to an incomplete governance architecture. A policy may be detailed in application scope and review organisation while leaving remedy, responsibility, transparency, and fairness under-specified. This pattern is consistent with healthcare governance research that links institutional quality to accountability, information governance,49 collaborative governance, and decision traceability.50 Eligibility Control Was Stronger Than Procedural Accountability A key finding is the uneven conversion of risk-control signals into enforceable safeguards. Eligibility-related provisions were generally more developed than accountability-related provisions. This pattern appears in the high scores for payment misfit risk control and in the universal presence of applicable scope, volume caps, and review cycles. Many municipal documents specified high-cost thresholds, length-of-stay triggers, critical or complex conditions, ungroupable cases, combined surgery or transfer cases, and new technologies. These provisions make the exception channel’s entry boundary more predictable and can reduce arbitrary or excessive applications. The distinction matters because formal mention of a governance issue is weaker than a rule that specifies who acts, when they act, what evidence is required, and how the decision can be reviewed.43,44 This front-end emphasis is understandable. Local payers need to protect the payment budget and prevent the exception mechanism from becoming a routine supplement to DRG/DIP payment. A clearly defined application scope, a maximum proportion or case quota, and CMI- or complexity-related adjustment rules35,36 can reduce strategic applications and preserve the integrity of the payment system. These rules also give medical institutions a clearer sense of which cases may be considered exceptional. From a provider-payment perspective, entry and quota rules preserve budget discipline while limiting predictable opportunities for strategic use of the exception pathway.32,34 Eligibility control alone, however, does not produce a complete risk-governance loop. The weakest areas were appeal or reconsideration, processing time limits, and cross-area or fund responsibility. These items are not peripheral procedural details. They determine whether a hospital can contest an unfavourable review outcome, whether a decision is delivered within a predictable time window, and whether responsibility is clear when cases involve cross-area settlement or fund allocation. Without such rules, the formal exception channel may remain exposed to uncertainty, delay, and perceived unfairness. Procedural fairness research treats appeal, revision, reason-giving, and enforcement as central to the legitimacy and usability of health financing decisions.43,44 The low accountability and equity score therefore has substantive meaning. It suggests that the most underdeveloped part of local policy design is not the recognition of exceptional cases, but the institutionalisation of remedy and responsibility. This finding should not be read as a simple criticism of local policy quality. Rather, it identifies the next stage of policy maturation. Early-stage special case negotiation policies may naturally focus on scope, quota, and review organisation. A more mature governance model needs to move beyond entry control and build a closed loop from eligibility to application, review, decision, feedback, appeal, audit, and responsibility allocation.47,48 Accountability and equity therefore mark the transition from a compensatory payment rule to a mature governance arrangement.55 Vertical Attenuation and the Loss of Enforceability at Municipal Level The vertical attenuation results show that provincial policy signals were not uniformly converted into municipal enforceable rules. The highest attenuation appeared in cross-area or fund responsibility, followed by special diseases, infectious diseases or rare diseases, processing time limits, high-cost trigger thresholds, length-of-stay trigger thresholds, audit or penalty provisions, and CMI or complexity adjustment. These nodes affect whether the special case negotiation mechanism can operate consistently across local jurisdictions.50,52 Similar implementation gaps have been reported in health-system governance and provider-payment reform research, where broad policy goals may lose operational specificity when translated into local rules.32,34 The clearest attenuation concerned cross-area or fund responsibility. This gap is important because DRG/DIP payment reform operates through local pooling arrangements, while patients, hospitals, and settlement responsibilities may cross administrative boundaries. Without clear rules on which locality accepts the application, which institution organises review, and which fund bears responsibility, special case negotiation may become uncertain at the point where coordination is most needed. This is a typical governance risk in decentralised implementation: higher-level policy may recognise the exception channel, while local rules may leave the administrative interface for cross-area cases insufficiently specified.50,52,53 Processing time limits were another major attenuation point. Although often treated as administrative details, they are central to delay risk in risk-governance terms. When a special case negotiation mechanism lacks predictable processing deadlines, hospitals face greater uncertainty and the policy’s incentive effect may weaken. If review outcomes arrive too late, the mechanism may become retrospective adjustment rather than timely governance. Procedurally, time limits operate as an enforcement mechanism because they reduce uncertainty and prevent review procedures from remaining open-ended.43,44 The attenuation of special disease and infectious disease provisions also warrants attention. Such cases often involve atypical clinical pathways, low frequency, high uncertainty, or public health relevance. If municipal policies do not consistently recognise them as exceptional categories, local handling may depend on discretionary interpretation. In a similar way, inconsistent use of CMI or complexity adjustment can make volume caps appear formally uniform while functioning differently across institutions with different case-mix structures. This attenuation pattern matters for risk adjustment because rare, severe, infectious, or otherwise atypical cases are precisely the types of cases that standard grouping systems may handle less well.35,36,42 Municipal Boundary-Procedure Configurations and Governance Trade-Offs The municipal typology provides a more interpretable view of local variation. Rather than ranking cities by a single total score, the boundary-procedure typology distinguishes different governance configurations. Five municipal documents were classified as balanced risk-governance models, four as boundary-heavy models, six as procedure-heavy models, and three as low-formalisation models. This classification is useful because the same overall level of formalisation can hide different risk profiles. The approach reflects a governance perspective in which policy design is evaluated not only by the presence of rules, but also by the configuration of safeguards across the decision chain.47–49 Balanced models had relatively strong eligibility and fund-control rules as well as relatively strong procedural safeguards. These documents are closer to a complete risk-governance architecture because they combine entry boundaries with review, documentation, feedback, and accountability-related provisions. Their strength is not simply that they contain more rules, but that the rules cover both sides of the exception channel: they protect the fund from uncontrolled applications while also making the procedure more predictable for medical institutions. Balanced models are closest to the procedural fairness logic of combining eligibility, evidence, reason-giving, reviewability, and enforceability in a single decision process.43,44 For balanced models, the priority is to monitor consistency of application, processing times, result feedback, and periodic rule updates rather than simply add more clauses. Boundary-heavy models emphasised application scope, high-cost or length-of-stay triggers, quotas, CMI-related adjustment, or other front-end controls, while their procedural safeguards were comparatively weaker. This configuration is administratively attractive because volume and eligibility are clearly controlled. It can also make the exception channel operate more as a gatekeeping tool than as a complete risk-governance mechanism. Where the front-end boundary is strong but appeal, disclosure, processing time, and responsibility rules are weak, hospitals may know when to apply while remaining less certain about how decisions will be made or challenged. Payment-reform evidence shows that strong financial controls20,23 can alter provider behaviour11,14 and may create unintended incentives if not matched by quality and accountability safeguards.22,28 Their main policy priority is to add written reasons, result feedback, reconsideration or appeal routes, processing deadlines, and explicit responsibility allocation. Procedure-heavy models showed the opposite pattern. These documents had relatively developed review and procedural arrangements but weaker boundary or fund-control scores. This configuration may improve procedural predictability, but it can also leave more room for variation in which cases enter the channel. If review structures, archiving, expert arrangements, and feedback are relatively strong while eligibility and fund-control provisions are less precise, the mechanism may rely heavily on downstream judgement. Risk-adjustment research similarly suggests that downstream correction can reduce mismatch35,37 only when the correction process is transparent, consistent, and insulated from discretionary distortion.40–42 Their main policy priority is to clarify eligibility conditions, measurable thresholds, volume controls, CMI or complexity adjustment, and the treatment of over-quota applications. Low-formalisation models had weaker boundary and procedural scores. These documents are not necessarily ineffective in practice, because informal routines, internal administrative guidance, or case-by-case coordination may exist outside the written text. However, from a formal risk-governance perspective, low formalisation increases dependence on local discretion and reduces external predictability. For a payment exception mechanism, this is a vulnerability. If the text does not clearly define both entry conditions and procedural consequences, implementation may vary with local capacity, administrative habits, and institutional negotiation.47,50,52 Low formalisation can therefore weaken auditability, comparability, appealability, and reproducibility. Their main policy priority is a minimum complete template that establishes both entry boundaries and procedural safeguards. International Implications for Risk Governance The cross-country relevance of these findings is functional rather than institutional. Medicare outlier payments and NTAP, Germany’s NUB process, and England’s excluded-item arrangements differ from China’s special case negotiation in legal authority, payment unit, review process, and responsible organisations.61–64 They nevertheless address a similar payment-misfit problem: a standard case-based price may inadequately cover unusually costly or complex cases or newly introduced technologies. An exception mechanism that is too narrow may heighten provider concern about accepting complex patients; one that is too permissive may expose the payer to uncontrolled spending or strategic applications. Case-specific review can also generate discretion, delay, and inconsistent decisions. The international value of the 30-node framework is therefore a set of functional risk-audit questions, not a model for institutional transplantation. A payer can ask whether eligibility is explicit, thresholds measurable, volumes controlled, evidence requirements specified, reviewers sufficiently independent, conflicts subject to recusal, decisions recorded in writing, processing time-bound, reconsideration available, finality defined, payer or jurisdictional responsibility allocated, and review experience fed back into payment recalibration. Other countries need not reproduce China’s organisational arrangements to use these questions when auditing outlier, high-cost, new-technology, or other payment-exception mechanisms. Policy Implications: A Minimum Risk-Governance Package The results point to a practical policy implication: special case negotiation should be designed around a minimum risk-governance package rather than around isolated eligibility lists. The package should include at least eight components. First, policies should specify core eligibility triggers, including high-cost cases, unusually long stays, complex or critical illness, ungroupable cases, combined surgery or transfer cases, new technologies, and special disease categories when relevant. Second, they should define volume-control rules, including proportion caps, CMI or case-mix adjustment, and the treatment of over-quota applications. Third, they should specify material requirements, including medical records, settlement lists, cost details, evidence for new technologies, and application forms. The package links special case negotiation to risk-adjustment and procedural fairness principles by recognising legitimate complexity, controlling strategic use,35,42 clarifying evidence, and preserving an enforceable remedy.43,44 Fourth, expert-review rules should be explicit, including expert source, random selection where feasible, expert number and composition, conflict-of-interest recusal, collective decision-making, written confirmation, and archiving. Fifth, information-system and traceability requirements should be strengthened. Online submission, system-based review, standardised file formats, and traceable decision records can reduce administrative ambiguity and improve auditability. Sixth, time limits should cover application, initial review, expert review, feedback, and appeal. These requirements align with information governance,49 accountability, and safety literature,56,59 which emphasises documentation, transparency, bias control, and clear allocation of authority in high-risk decision systems.57,58 Seventh, policy texts should specify how results are communicated and how remedies can be pursued.43,44 Hospitals should know whether reasons are provided, how long they have to raise objections, which authority handles appeals, and whether the appeal decision is final. Eighth, cross-area responsibility and fund responsibility should be clarified, especially where patients receive care outside their insured locality or where payment responsibility depends on multiple administrative actors.50,52 Appeal and cross-area responsibility are not secondary details; they are parts of the governance loop that make the mechanism revisable, equitable, and enforceable.55 These implications do not require all municipalities to adopt identical operational details. Local adaptation is necessary because case mix, hospital capacity, information systems, and fund conditions vary. The point is that local flexibility should operate within a minimum set of risk-governance safeguards.43,44 Policy standardisation should focus less on making all local texts identical and more on ensuring that every local policy contains the essential elements of a complete governance loop. This approach preserves local administrative discretion while reducing avoidable variation in procedural fairness, transparency, and accountability.47,49 Practical Challenges of Accountability and Appeals Adding accountability and appeal safeguards carries costs. Complex cases may require supplementary clinical and cost evidence, and an independent second review requires experts who were not involved in the first decision. Repeated objections may be used strategically, while an open-ended appeal can delay settlement and increase administrative workload. The opposite design—prohibiting any reconsideration—may reduce workload but weaken procedural fairness, especially when a denial turns on disputed coding, clinical complexity, or incomplete evidence. Cross-area cases create an additional coordination risk because hospitals, treatment-place authorities, insured-place authorities, and different funds may seek to shift responsibility. An operational compromise is a bounded review route. Policy texts can identify who may appeal and within what period, require written reasons for the initial decision, allow one independent reconsideration, and make the reconsidered decision final. Separate deadlines can govern material screening, expert review, decision, and feedback. A lead department should be named for cross-area cases, together with an explicit rule allocating fund liability. These limits protect a meaningful correction route without leaving payment settlement indefinitely open. Methodological Implications This study also makes a methodological contribution by adapting policy-text analysis to healthcare risk management. The 0-1-2 framework does more than record whether a policy clause is present: it distinguishes absence, principle-level mention, and enforceable requirement. That distinction matters because healthcare governance depends heavily on enforceability. A policy that states a principle such as fairness, standardisation, or timely handling does not necessarily create a usable rule. By contrast, a clause that names the responsible actor, the application condition, the material requirement, the time limit, the expert rule, or the consequence of non-compliance can be treated as a stronger risk-control provision. The same distinction is consistent with broader policy and complex-intervention evaluation guidance, where a formal statement of intent is not equivalent to an operationalised mechanism.54 The use of domain-specific scores, enforceability conversion, vertical attenuation, closed-loop completeness, and boundary-procedure typology allows policy texts to be analysed as governance instruments rather than as narrative documents. This approach can be transferred to other healthcare payment settings where exception mechanisms, discretionary review, or special reimbursement channels are used. It is especially relevant when administrative data on implementation outcomes are unavailable but policy design itself is an important object of risk assessment. The high inter-coder reliability suggests that the framework can be applied with acceptable consistency when coding rules are explicit and evidence-based adjudication is used. Limitations Several limitations should be acknowledged. First, because this study analyses formal policy texts, it cannot determine how the rules were implemented in practice. A policy may contain detailed provisions that are weakly implemented, and a policy with sparse written rules may be supported by strong informal routines or internal administrative guidance. The findings therefore describe formal policy design rather than actual implementation performance. Second, the analysis draws on one national document, one provincial document, and eighteen municipal documents from a single provincial context. This design provides complete coverage of municipal-level policy documents within one provincial policy environment, making it suitable for examining vertical conversion and municipal variation inside that province; however, it does not allow direct generalisation to all Chinese provinces. Other provinces may differ in payment histories, fund pressures, information systems, and administrative capacities. Future research should apply the framework across multiple provinces to test whether the same risk-control gaps recur. Third, some policy documents were partly non-public or obtained through project sources, which limits full external reproducibility. To mitigate this limitation, the study used document metadata, a structured codebook, double coding, inter-coder reliability statistics, adjudication records, translated excerpts, and the complete original adjudicated coding matrix. These steps improve transparency, although they cannot fully replace public access to all original documents. Fourth, the coding framework necessarily involves judgement, especially when distinguishing principle-level statements from enforceable requirements. Although reliability was high, future studies could further validate the framework through expert review or by linking coded policy provisions to implementation data. Finally, this study does not assess patient outcomes, provider behaviour, medical expenditure, or fund performance after special case negotiation. Its purpose was to evaluate risk-governance design at the policy-text level. A useful next step would combine policy coding with claims data, hospital-level application records, review outcomes, interviews with health insurance administrative agencies and hospital administrators, and case-level examples. Such mixed evidence would allow researchers to examine whether stronger formal risk-governance provisions are associated with more predictable review outcomes, lower administrative burden, better accommodation of complex cases, or fewer equity disputes. Future research should first replicate the framework across multiple provinces and link policy-text scores to application, approval, and reconsideration records, processing times, claims data, payment adjustments, and fund expenditure. Such work could test whether formal safeguards are associated with provider behaviour, selection of complex patients, readmission, emergency revisits, mortality, or patient cost burden. Interviews with payer administrators, hospital medical-insurance offices, and review experts could also identify how written rules interact with workload, evidence standards, discretion, coordination, and informal practice. Conclusion Special case negotiation under DRG/DIP payment reform should be understood as a risk-governance mechanism. It compensates exceptional cases while also balancing cost control, clinical complexity, administrative feasibility, discretion control, and procedural fairness. The analysed policy texts show substantial progress in defining eligibility boundaries, application cycles, quotas, and expert-review arrangements. Their main weakness lies at the back end of the governance chain, especially appeal, processing time, cross-area responsibility, and accountability-related provisions. Future policy refinement should therefore move beyond listing eligible exceptional cases and build a minimum governance loop that is enforceable, traceable, appealable, and accountable. Special case negotiation can function as a minimum risk-governance package only when eligibility boundaries are paired with enforceable procedural safeguards, traceable review, appeal routes, and clear accountability rules. In this study, a “minimum risk-governance package” means a minimum set of enforceable procedural safeguards. The findings evaluate formal policy design, not implementation effectiveness. Used as a functional audit framework, the 30-node approach may help other DRG or case-based payment systems examine outlier, high-cost, new-technology, and other payment exceptions without reproducing China’s institutional arrangements. Future empirical studies should link policy design to claims data, application and approval records, processing times, payment adjustments, provider behaviour, and patient outcomes.