Multi-level clinical inertia in chronic kidney disease management in primary care: A retrospective cross-sectional analysis of coding, reclassification and risk stratification.

Abstract

Aims/hypothesis: Accurate coding, longitudinal stage review and complete risk stratification underpin chronic kidney disease (CKD) care in primary care, especially in people with diabetes, where CKD modifies cardiovascular risk, prescribing safety and referral thresholds. We hypothesised that, beyond well-described inertia in pharmacological treatment intensification, primary care records would also exhibit inertia in diagnostic coding, longitudinal reclassification and albuminuria-based risk stratification, clustering with higher kidney and cardiovascular risk.

Methods: We performed a retrospective cross-sectional analysis of adult primary care records with coded CKD. The latest estimated glomerular filtration rate (eGFR) was compared with the coded stage. Where eGFR suggested reclassification but coding was unchanged, manual case review determined whether changes were sustained for ≥ 3 months. Risk was derived from the latest urine albumin-to-creatinine ratio (uACR) using KDIGO categories. Modifiable risk factors were predefined as blood pressure above CKD-specific thresholds and/or HbA1c ≥ 58 mmol/mol. A separate electronic search identified adults with uncoded CKD.

Results: Of 583 individuals with coded CKD, 581 were analysed. Among those with unchanged coding despite eGFR suggesting de-escalation, 47/87 (54.0%) had sustained improvement; of 47 with unchanged coding despite a latest eGFR suggesting escalation, 26 (55.3%) had sustained decline. uACR was missing in 175 cases (30.1%), including 12 with stage G4/G5 disease. Women were more likely than men to lack uACR data (37.7% vs 22.1%; OR 2.13, 95% CI 1.48, 3.08; χ² = 16.78, p < 0.001). KDIGO high or very high risk was present in 179 (30.8%) and a combined high-risk profile in 118 (20.3%). Combined high-risk status clustered in the missed-escalation group (16/47, 34.0%) compared with the rest of the cohort (83/447, 18.6%; RR 1.83, 95% CI 1.18, 2.85; OR 2.26, 95% CI 1.18, 4.33; p = 0.020). Diabetes (n = 198, 34.1%) was not associated with coding accuracy, but was associated with higher risk CKD (30.3% vs 15.1%; χ² = 18.53, p < 0.001; RR 2.00; Cramér's V = 0.18) and more complete uACR testing (82.8% vs 63.2%; χ² = 23.93, p < 0.001). Separate case finding identified 38 additional adults with biochemical evidence of uncoded CKD.

Conclusions/interpretation: Primary care CKD records exhibited multi-level clinical inertia across coding, stage review and albuminuria-based risk assessment. Missed reclassification, absent uACR data and uncoded disease clustered with higher-risk profiles and disparities disadvantaging women. Structured EHR-led CKD review processes that integrate longitudinal kidney function, mandatory uACR ordering, active code review and risk-based action may improve timeliness, equity of CKD care.