Navigating Complexity in Hospital Construction: Diagnosing Performance Regimes With a Hierarchical Fuzzy Diagnostic Framework

Hospital construction involves tightly coupled clinical functions, specialist engineering systems, multi‐disciplinary interfaces, regulatory approvals and commissioning dependencies, yet repeated project‐level diagnosis of changing complexity–capacity conditions remains limited. This study longitudinally audits a prospectively specified hierarchical fuzzy diagnostic framework linking nine complexity dimensions and managerial decision‐making capacity (DMC) to time‐indexed performance regimes. Of 72 Iranian hospital projects screened, 30 met pre‐defined criteria; four waves at approximately 6‐month intervals from December 2023 to July 2025 yielded 120 project‐wave observations and 90 adjacent transitions. Before Wave 1 responses or outcomes were available, the hierarchical fuzzy inference system (HFIS), membership functions, 162‐rule base and 0.40/0.70 thresholds were fixed prospectively. The model‐implied project performance index (PPI; 0–1) classified 32 observations as Regime I, 61 as Regime II and 27 as Regime III. Separately scored respondent‐rated project performance (RPP; 1–5), excluded from HFIS inputs and PPI computation, averaged 3.605 in Regime I and 2.526 in Regime III. Within‐project fixed‐effects estimates associated higher complexity with poorer RPP ( b  = −0.521, 95% CI [−0.675, −0.367], p   < 0.001) and higher DMC with better RPP ( b  = 0.146, 95% CI [0.029, 0.264], p = 0.014). A 4999‐replicate project‐level wild‐cluster bootstrap supported both main associations but not the complexity × DMC interaction. DMC remained supported across several lifecycle adjustments but attenuated with contemporaneous progress gaps. PPI corresponded concurrently with RPP ( r  = 0.846), commissioning readiness ( r  = 0.816), schedule variance ( r  = −0.795) and cost variance ( r  = −0.762); 54 of 90 transitions changed regime. Post‐hoc monotonicity‐constrained recalibration changed no observed regime or transition assignment, whereas alternative hierarchies and wider membership overlap materially altered classifications. The framework is therefore an auditable, context‐bounded diagnostic prototype rather than a validated forecasting or intervention tool; external replication and lagged validation remain necessary.