Defining measures of kidney function in observational studies using routine health care data: methodological and reporting considerations.

The availability of electronic health records and access to a large number of routine measurements of serum creatinine and urinary albumin enhance the possibilities for epidemiologic research in kidney disease. However, the frequency of health care use and laboratory testing is determined by health status and indication, imposing certain challenges when identifying patients with kidney injury or disease, when using markers of kidney function as covariates, or when evaluating kidney outcomes. Depending on the specific research question, this may influence the interpretation, generalizability, and/or validity of study results. This review illustrates the heterogeneity of working definitions of kidney disease in the scientific literature and discusses advantages and limitations of the most commonly used approaches using 3 examples. We summarize ways to identify and overcome possible biases and conclude by proposing a framework for reporting definitions of exposures and outcomes in studies of kidney disease using routinely collected health care data.

Association between practice coding of chronic kidney disease (CKD) in primary care and subsequent hospitalisations and death: a cohort analysis using national audit data.

OBJECTIVE: To examine the association between practice percentage coding of chronic kidney disease (CKD) in primary care with risk of subsequent hospitalisations and death. DESIGN: Retrospective cohort study using linked electronic healthcare records. SETTING: 637 general practitioner (GP) practices in England. PARTICIPANTS: 167 208 patients with CKD stages 3-5 identified by 2 measures of estimated glomerular filtration rate <60 mL/min/1.73 m2, separated by at least 90 days, excluding those with coded initiation of renal replacement therapy. MAIN OUTCOME MEASURES: Hospitalisations with cardiovascular (CV) events, heart failure (HF), acute kidney injury (AKI) and all-cause mortality RESULTS: Participants were followed for (median) 3.8 years for hospital outcomes and 4.3 years for deaths. Rates of hospitalisations with CV events and HF were lower in practices with higher percentage CKD coding. Trends of a small reduction in AKI but no substantial change in rate of deaths were also observed as CKD coding increased. Compared with patients in the median performing practice (74% coded), patients in practices coding 55% of CKD cases had a higher rate of CV hospitalisations (HR 1.061 (95% CI 1.015 to 1.109)) and HF hospitalisations (HR 1.097 (95% CI 1.013 to 1.187)) and patients in practices coding 88% of CKD cases had a reduced rate of CV hospitalisations (HR 0.957 (95% CI 0.920 to 0.996)) and HF hospitalisations (HR 0.918 (95% CI 0.855 to 0.985)). We estimate that 9.0% of CV hospitalisations and 16.0% of HF hospitalisations could be prevented by improving practice CKD coding from 55% to 88%. Prescription of antihypertensives was the most dominant predictor of a reduction in hospitalisation rates for patients with CKD, followed by albuminuria testing and use of statins. CONCLUSIONS: Higher levels of CKD coding by GP practices were associated with lower rates of CV and HF events, which may be driven by increased use of antihypertensives and regular albuminuria testing, although residual confounding cannot be ruled out.

A systematic review of statistical methodology used to evaluate progression of chronic kidney disease using electronic healthcare records.

BACKGROUND: Electronic healthcare records (EHRs) are a useful resource to study chronic kidney disease (CKD) progression prior to starting dialysis, but pose methodological challenges as kidney function tests are not done on everybody, nor are tests evenly spaced. We sought to review previous research of CKD progression using renal function tests in EHRs, investigating methodology used and investigators' recognition of data quality issues. METHODS AND FINDINGS: We searched for studies investigating CKD progression using EHRs in 4 databases (Medline, Embase, Global Health and Web of Science) available as of August 2021. Of 80 articles eligible for review, 59 (74%) were published in the last 5.5 years, mostly using EHRs from the UK, USA and East Asian countries. 33 articles (41%) studied rates of change in eGFR, 23 (29%) studied changes in eGFR from baseline and 15 (19%) studied progression to binary eGFR thresholds. Sample completeness data was available in 44 studies (55%) with analysis populations including less than 75% of the target population in 26 studies (33%). Losses to follow-up went unreported in 62 studies (78%) and 11 studies (14%) defined their cohort based on complete data during follow up. Methods capable of handling data quality issues and other methodological challenges were used in a minority of studies. CONCLUSIONS: Studies based on renal function tests in EHRs may have overstated reliability of findings in the presence of informative missingness. Future renal research requires more explicit statements of data completeness and consideration of i) selection bias and representativeness of sample to the intended target population, ii) ascertainment bias where follow-up depends on risk, and iii) the impact of competing mortality. We recommend that renal progression studies should use statistical methods that take into account variability in renal function, informative censoring and population heterogeneity as appropriate to the study question.

Feasibility of evaluation of the natural history of kidney disease in the general population using electronic healthcare records.

BACKGROUND: Knowledge about the nature of long-term changes in kidney function in the general population is sparse. We aim to identify whether primary care electronic healthcare records capture sufficient information to study the natural history of kidney disease. METHODS: The National Chronic Kidney Disease Audit database covers ∼14% of the population of England and Wales. Availability of repeat serum creatinine tests was evaluated by risk factors for chronic kidney disease (CKD) and individual changes over time in estimated glomerular filtration rate (eGFR) were estimated using linear regression. Sensitivity of estimation to method of evaluation of eGFR compared laboratory-reported eGFR and recalculated eGFR (using laboratory-reported creatinine), to uncover any impact of historical creatinine calibration issues on slope estimation. RESULTS: Twenty-five per cent of all adults, 92% of diabetics and 96% of those with confirmed CKD had at least three creatinine tests, spanning a median of 5.7 years, 6.2 years and 6.1 years, respectively. Median changes in laboratory-reported eGFR (mL/min/1.73 m2/year) were -1.32 (CKD) and -0.60 (diabetes). Median changes in recalculated eGFR were -0.98 (CKD) and -0.11 (diabetes), underestimating decline. Magnitude of underestimation (and between-patient variation in magnitude) decreased with deteriorating eGFR. For CKD Stages 3, 4 and 5 (at latest eGFR), median slopes were -1.27, -2.49 and -3.87 for laboratory-reported eGFR and -0.89, -2.26 and -3.75 for recalculated eGFR. CONCLUSIONS: Evaluation of long-term changes in renal function will be possible in those at greatest risk if methods are identified to overcome creatinine calibration problems. Bias will be reduced by focussing on patients with confirmed CKD.

How do primary care doctors in England and Wales code and manage people with chronic kidney disease? Results from the National Chronic Kidney Disease Audit.

Background: In the UK, primary care records are electronic and require doctors to ascribe disease codes to direct care plans and facilitate safe prescribing. We investigated factors associated with coding of chronic kidney disease (CKD) in patients with reduced kidney function and the impact this has on patient management. Methods: We identified patients meeting biochemical criteria for CKD (two estimated glomerular filtration rates <60 mL/min/1.73 m2 taken >90 days apart) from 1039 general practitioner (GP) practices in a UK audit. Clustered logistic regression was used to identify factors associated with coding for CKD and improvement in coding as a result of the audit process. We investigated the relationship between coding and five interventions recommended for CKD: achieving blood pressure targets, proteinuria testing, statin prescription and flu and pneumococcal vaccination. Results: Of 256 000 patients with biochemical CKD, 30% did not have a GP CKD code. Males, older patients, those with more severe CKD, diabetes or hypertension or those prescribed statins were more likely to have a CKD code. Among those with continued biochemical CKD following audit, these same characteristics increased the odds of improved coding. Patients without any kidney diagnosis were less likely to receive optimal care than those coded for CKD [e.g. odds ratio for meeting blood pressure target 0.78 (95% confidence interval 0.76-0.79)]. Conclusion: Older age, male sex, diabetes and hypertension are associated with coding for those with biochemical CKD. CKD coding is associated with receiving key primary care interventions recommended for CKD. Increased efforts to incentivize CKD coding may improve outcomes for CKD patients.

Accounting for overdispersion when determining primary care outliers for the identification of chronic kidney disease: learning from the National Chronic Kidney Disease Audit.

Background: Early diagnosis of chronic kidney disease (CKD) facilitates best management in primary care. Testing coverage of those at risk and translation into subsequent diagnostic coding will impact on observed CKD prevalence. Using initial data from 915 general practitioner (GP) practices taking part in a UK national audit, we seek to apply appropriate methods to identify outlying practices in terms of CKD stages 3-5 prevalence and diagnostic coding. Methods: We estimate expected numbers of CKD stages 3-5 cases in each practice, adjusted for key practice characteristics, and further inflate the control limits to account for overdispersion related to unobserved factors (including unobserved risk factors for CKD, and between-practice differences in coding and testing). Results: GP practice prevalence of coded CKD stages 3-5 ranges from 0.04 to 7.8%. Practices differ considerably in coding of CKD in individuals where CKD is indicated following testing (ranging from 0 to 97% of those with and glomerular filtration rate <60 mL/min/1.73 m 2 ). After adjusting for risk factors and overdispersion, the number of 'extreme' practices is reduced from 29 to 2.6% for the low-coded CKD prevalence outcome, from 21 to 1% for high-uncoded CKD stage and from 22 to 2.4% for low total (coded and uncoded) CKD prevalence. Thirty-one practices are identified as outliers for at least one of these outcomes. These can then be categorized into practices needing to address testing, coding or data storage/transfer issues. Conclusions: GP practice prevalence of coded CKD shows wide variation. Accounting for overdispersion is crucial in providing useful information about outlying practices for CKD prevalence.

Bayesian sample sizes for exploratory clinical trials comparing multiple experimental treatments with a control.

In this paper, a Bayesian approach is developed for simultaneously comparing multiple experimental treatments with a common control treatment in an exploratory clinical trial. The sample size is set to ensure that, at the end of the study, there will be at least one treatment for which the investigators have a strong belief that it is better than control, or else they have a strong belief that none of the experimental treatments are substantially better than control. This criterion bears a direct relationship with conventional frequentist power requirements, while allowing prior opinion to feature in the analysis with a consequent reduction in sample size. If it is concluded that at least one of the experimental treatments shows promise, then it is envisaged that one or more of these promising treatments will be developed further in a definitive phase III trial. The approach is developed in the context of normally distributed responses sharing a common standard deviation regardless of treatment. To begin with, the standard deviation will be assumed known when the sample size is calculated. The final analysis will not rely upon this assumption, although the intended properties of the design may not be achieved if the anticipated standard deviation turns out to be inappropriate. Methods that formally allow for uncertainty about the standard deviation, expressed in the form of a Bayesian prior, are then explored. Illustrations of the sample sizes computed from the new method are presented, and comparisons are made with frequentist methods devised for the same situation.