In UK primary care, there are payment incentives to test for chronic kidney disease (CKD) and to add CKD-specific diagnosis codes to the electronic patient health record to establish a register of stage 3-5 CKD patients. The National Chronic Kidney Disease Audit (NCKDA) in England and Wales found that not all patients with biochemical evidence of CKD stages 3-5 were coded; those who were not coded had less cardiovascular preventative care than those who were coded. Practice-level coding is less likely to be influenced by individual patient characteristics and is therefore potentially less confounded than analyses using patient-specific coding decisions. We investigated the association between practice-level coding of CKD and individual adverse outcomes to quantify the effects of coding in patients with CKD. We conducted a retrospective cohort study linking NCKDA primary care records with Hospital Episode Statistics (HES) admissions and Office for National Statistics (ONS) mortality data in England. Patients with confirmed CKD (2 x estimated glomerular filtration rate < 60 ml/min/1.73m2, separated by ≥ 90 days, with the latest measure within 2 years prior to data extraction) were included, excluding those with coded initiation of renal replacement therapy. Cox proportional hazards regression models with splines, adjusted for practice characteristics, were used to investigate the association between practice percentage of confirmed CKD cases with a stage 3-5 CKD code and patient outcomes: hospitalisation for cardiovascular (CV) events, hospitalisation for heart failure (HF), hospitalisation with acute kidney injury (AKI) and all-cause mortality. 167,208 patients from 637 primary care practices in England were included in analysis, with a median follow up time of 3.8 years for hospitalisations and 4.3 years for deaths. Patient characteristics were well balanced in the two-thirds of most typically performing practices (55% to 88% coded CKD), but with slight differences at the extremes. Rates of CV and HF admissions were significantly reduced as practice CKD coding performance increased (Figure 1). In practices coding 60% of CKD cases, patients had a 4.8% (0.9%, 9.0%) higher hazard of CV hospitalisations and an 8.6% (1.5%, 16.1%) higher hazard of HF hospitalisations than the average practice (74% coded). Trends of a small reduction in AKI but no substantial change in risk of deaths were also observed as CKD coding increased. Figure 1. Percent difference in hazard rates of outcomes according to practice percent coded CKD, compared to median practice coding (74% coded) among patients with confirmed CKD in the two-thirds of most typically performing practices (55% to 88% coded). Adjusted for practice characteristics: mean age, percent male, median rank of index of multiple deprivation, diabetes prevalence, hypertension prevalence, CVD prevalence, percent of CKD cases at stages 3b-5, percent of CKD patients admitted for COPD in last 3 years, percent of CKD patients admitted for cancer in last 3 years, percent GFR test in last year in diabetes, percent GFR test in last year in CKD, percent of adult population with CKD Increased CKD coding in primary care and subsequent care efforts appear to play a role in improving patient outcomes, although residual confounding due to unobserved practice characteristics cannot be ruled out.
Read full abstract