Chronic Kidney Disease and Major Risk Factors in Patients from Two Large Healthcare Systems: The CURE-CKD Registry

IMPORTANCE Chronic kidney disease (CKD) is serious and common, yet recognition and public health responses are limited. OBJECTIVE To describe clinical features of, prevalence of, major risk factors for, and care for CKD among patients treated in 2 large US health care systems. DESIGN, SETTING, AND PARTICIPANTS This cohort study collected data from the Center for Kidney Disease Research, Education, and Hope (CURE-CKD) registry, an electronic health record– based registry jointly curated and sponsored by Providence St Joseph Health and the University of California, Los Angeles. Patients were adults and children with CKD (excluding end-stage kidney disease) and adults at risk of CKD (ie, with diabetes, hypertension, or prediabetes) identified by laboratory values, vital signs, prescriptions, and administrative codes. Data were collected from January 2006 through December 2017, with analyses performed from March 2019 through November 2019. a (interquartile 70 (59-81) 338785 (55.9%) and 434474 non-Latino white individuals (71.7%). A total of 12591 children (0.4%) with CKD a median (IQR) age of 6 (1-13) years, with 7079 girls (56.2%) and 6653 non-Latino white children (52.8%). Median (IQR) estimated glomerular CONCLUSIONS AND RELEVANCE This registry-based cohort study revealed a burgeoning number of patients with CKD and its major risk factors. Rates of identification and use of kidney protective agents were low, while potential nephrotoxin use was widespread, underscoring the pressing need for practice-based improvements in CKD prevention, recognition, and treatment.


Introduction
Chronic kidney disease (CKD) is a serious and common disease, and it eventuates in multiple complications, including premature mortality and end-stage kidney disease (ESKD). [1][2][3] An estimated 1 in 7 to 10 adults worldwide have CKD, with only approximately 10% surviving to ESKD and only half of survivors receiving dialysis or a kidney transplant because of lack of access or high costs. 3 From 1990 to 2016, the prevalence of CKD increased by 90%, and related deaths, mainly due to cardiovascular diseases and infections, nearly doubled in the United States and globally. [4][5][6] In highincome countries, 2% to 3% of annual health care costs are devoted to the 0.03% of the population with ESKD. 7 The increasing prevalence of CKD is closely tied to the increase of at-risk populations with diabetes, hypertension, and prediabetes. Indeed, diabetes is the leading cause of CKD and a global health emergency, with 425 million individuals affected worldwide in 2017 and a projected 629 million individuals affected by 2045. [8][9][10] Hypertension is the second most frequent cause of CKD, affecting nearly one-third of US adults and 1.13 billion people globally in 2015. 11,12 The estimated population size for prediabetes was 78.5 million among adults in the United States between 2011 and 2014, and nearly one-tenth have been reported with CKD. 13 Even so, awareness of CKD and its major risk factors remains strikingly low among health care professionals and patients alike. [14][15][16] The Advancing American Kidney Health initiative was recently launched by a US executive order calling for new approaches to prevent and treat CKD, with a goal of reducing ESKD incidence 25% by 2030. 17 Real-world data from electronic health records (EHRs) provide an ideal platform to answer this call by improving CKD detection, tracking, and public health responses. The Center for Kidney Disease Research, Education, and Hope (CURE-CKD) registry contains detailed patient-level EHR data from more than 2.6 million adults and children with CKD and at risk of CKD during 12 inclusive years. 18 The objective of this study was to describe baseline clinical features of, prevalence of, major risk factors for, and care for CKD based on data from the CURE-CKD registry.

Methods
The study was approved by Providence St Joseph Health (PSJH; Washington, Montana, Oregon, Alaska, and California) and the University of California, Los Angeles (UCLA; California) Health institutional review boards, which determined that written informed consent was not required for this limited data set. Data use agreements between PSJH and UCLA Health formed the framework for data sharing, stewardship, and security. This study was conducted according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. 19

Study Design
Detailed methodology for CURE-CKD has been previously published. 18 The formation of CURE-CKD was supported by institutional funding from PSJH and UCLA Health. Both health care systems use Epic EHRs (Epic Systems). The first phase of CURE-CKD created a data repository with patient information from EHRs with at least 1 measure indicating CKD, diabetes, prediabetes, or The second phase crafted an EHR-based registry of participants with CKD and at risk for CKD derived from the repository. The first 90 days a patient was included in the registry were considered the baseline period. Registry criteria were based on established clinical practice guidelines for CKD (eTable 1 in the Supplement). Adults (ie, aged Ն18 years) were included with 2 or more of the following laboratory measurements recorded at least 90 days apart: estimated glomerular filtration rate (eGFR) less than 60 mL/min/1.73 m 2 , calculated from serum creatinine levels using the Chronic Kidney Disease Epidemiology equation; urine albumin to creatinine ratio greater than 30 mg/g; and urine protein to creatinine ratio greater than 150 mg/g. 20,21 Children (ie, aged <18 years) with CKD were identified using the same criteria, except the bedside Schwartz equation was used to calculate eGFR from serum creatinine levels. 22 We identified CKD categories 1 and 2 by an administrative code, urine albumin to creatinine ratio greater than 30 mg/g, and/or urine protein to creatinine ratio greater than 150 mg/g. We identified CKD categories 3 to 5 based on eGFR and/or administrative          (Figure 1).

Prescription Medication Use and Temporal Trends in Adults With CKD
Angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs) were prescribed to 127 574 adults (20.5%) with CKD, with slightly higher use of these agents among those Temporal trends in prescription medications were determined for participants with CKD determined by 2 eGFR measurements of less than 60 mL/min/1.73 m 2 at least 90 days apart for the 3 periods. Use rates of ACE inhibitors, ARBs, NSAIDs, and PPIs across CKD categories 3a to 5 all

Discussion
More than 2.6 million adults and children who received care at PSJH and UCLA Health from 2006 to 2017 had CKD or were at risk of CKD. Overall, CKD prevalence among adults in the health care systems was 4.8%, as determined by a combination of eGFR, albuminuria and proteinuria measures, and administrative code criteria. However, adult CKD prevalence adjusted for age, sex, and race/ ethnicity was 22.6% based on persistently low eGFR alone. Adults with CKD were more likely to be older, women, and non-Latino white individuals. In this study, CKD category 3 was most frequent, with a clear drop-off in prevalence at more advanced categories. Kidney protective agents (ie, reninangiotensin system inhibitors) were prescribed to approximately one-fifth of adults with CKD, while potential nephrotoxins (ie, NSAIDs and PPIs) were prescribed to more than one-third of adults with CKD. Albuminuria and proteinuria testing for CKD assessment was rarely reported.
The CURE-CKD registry is among the most comprehensive CKD registries worldwide. A unique feature is the extensive amount of patient-level data on laboratory measures, prescriptions, and vital signs, combined with administrative codes, to identify CKD and major risk factors according to guideline-based criteria. 21,24,25 Previous registries were restricted by containing primarily administrative data, ESKD, primary care practices, single health care systems, older adults, or men. 4,[26][27][28][29][30][31][32] In contrast, CURE-CKD participants represent the life span, from children to adults, and include women and men and a wide spectrum of races and ethnicities across an expansive region of the western United States that has not been previously involved in large-scale epidemiologic studies of CKD. Moreover, PSJH and UCLA Health care for patients in a variety of settings that include academic, primary care, and specialty practices as well as community health and safety-net systems.
Rural patients were well represented in the geography covered by PSJH. Thus, CURE-CKD provides in-depth identification of patients with and at risk for CKD in contemporary US health care systems.
In CURE-CKD, the progressive increase in adult CKD prevalence was largely driven by diagnostic coding. Among adults with persistently low eGFR, use of CKD administrative codes increased from 3.2% to 52.3% between the periods of 2006 to 2009 and 2014 to 2017, while overall CKD prevalence estimates, adjusted for age, sex, and race/ethnicity, were essentially stable between Although nearly two-thirds of the adults with CKD had diabetes, hypertension, or prediabetes, rates of laboratory testing for albuminuria or proteinuria and of prescribing ACE inhibitors or ARBs were low. Potentially nephrotoxic agents (ie, NSAIDs and PPIs) were used more commonly than renin-angiotensin system inhibitors. Given the most common cause of death in CKD is cardiovascular disease, the low use of preventive agents, such as statins and aspirin, is also concerning. 33,34 Compared with participants in the National Health and Nutrition Examination Survey, patients with CKD in CURE-CKD received ACE inhibitors or ARBs much less often during approximately the same period. 35 Although CURE-CKD found an increase in uptake of renin-angiotensin system inhibitors in adults with CKD categories 3a to 5, NSAID and PPI use also increased over time. However, these prescription rates were lower than in the overall CKD cohort, perhaps because of concerns about adverse effects with more advanced CKD. While this may seem counterintuitive for renin-angiotensin system inhibitors, these agents may be avoided because of fear of complications such as hyperkalemia or acute kidney injury, especially in acute care settings. In Ontario, Canada, primary care practices reported ACE inhibitor or ARB use in three-fourths of patients with CKD, but the metric was confined to those with diabetes and albuminuria or adults older than 66 years. 30,31 Nevertheless, rates of albuminuria testing in the overall CKD population were comparably low, although avoidance of NSAIDs was better among patients in Canada than in CURE-CKD. 30 A recent Canada-wide study 32 from an EHR-based surveillance system in primary care found that only 4 of 12 quality indicators for CKD care were met, with ACE inhibitor or ARB use among approximately one-third of patients with diabetes or proteinuria. Contrasts exist between reports from health care systems, community screenings, primary care practices, and countries, but they consistently illuminate major gaps in CKD care and the need for more comprehensive surveillance to uncover actionable trends.
In comparison with patients treated at PSJH and UCLA Health, the Kidney Early Evaluation Program and the National Health and Nutrition Examination Survey have reported lower frequencies of individuals at risk for CKD in community screenings. 36 Moreover, associations between risk factors and CKD are remarkably complex. For example, although a primary contributor to CKD is diabetes, CKD in patients with diabetes greatly amplifies cardiovascular risks. 37 Additionally, nearly one-fifth of patients with CKD in the CURE-CKD registry had prediabetes. The prediabetes phenotype of CKD appears less severe than the diabetes phenotype of CKD, as reflected by fewer patients with advanced CKD categories, albuminuria, or proteinuria and prediabetes. Nevertheless, consistent with findings from the National Health and Nutrition Examination Survey, findings from CURE-CKD support the observation that subdiabetic hyperglycemia may contribute to kidney damage before overt diabetes ensues. 13 The CURE-CKD registry contains abundant longitudinal data that will be invaluable for elucidating CKD incidence among individuals at risk as well as progression and complications in those with CKD. Given its vast scope, CURE-CKD is also ideally suited to generate and validate CKD risk prediction models. 38

Strengths and Limitations
Strengths of the CURE-CKD registry include the large sample size, long observation duration, and curated patient-level data from 2 US health care systems. However, this study has limitations. First, CURE-CKD is limited by differences in documentation methods across and between health care systems and varying attrition rates based on insurance, socioeconomic factors, and geolocation.
Variation in platforms even within a common EHR system also presents a limitation to the creation of interinstitutional registries, highlighting the importance of collaboration in identifying data elements, structures, and synchronization. Lack of information on over-the-counter medications underestimates the usage rates for NSAIDs, PPIs, and other potential nephrotoxins. Data in CURE-CKD on sodium-glucose cotransporter 2 inhibitor use, recently recommended for diabetes and CKD, came from an era before this new indication. It will be important to follow this trend to ensure sodium-glucose cotransporter 2 inhibitors are delivered to patients who may benefit. Another limitation of EHR-based registries is undercoding and miscoding. To mitigate this limitation, patientlevel data for laboratory values, vital signs, and prescriptions were used to classify CURE-CKD registry participants and their care, which allowed for the use of guideline-based criteria for persistence of low eGFR or elevated albuminuria or proteinuria levels. Although CURE-CKD produced a lower range estimate of overall CKD prevalence compared with other US reports, this prevalence rate is similar to that found in Canadian primary care. 4,32 A higher range estimate for CKD based solely on eGFR could be because of more frequent testing in patients with higher risk who were treated by both specialty and primary care practices at PSJH and UCLA Health. Ascertainment bias is an inherent limitation of EHR-based registries, and information about CKD will also be missed from patients receiving care elsewhere or not receiving testing. The actual prevalence of overall CKD likely lies between the low (4.8%) and high (22.6%) range estimations from CURE-CKD. Nevertheless, these detailed prevalence estimates are strengths that represent the complexity and composition of patients treated in typical US health care systems.

Conclusions
In conclusion, the CURE-CKD registry reveals a burgeoning number of patients with CKD and major risk factors of diabetes, hypertension, and prediabetes. Rates of identification and use of kidney protective agents were low, while nephrotoxin use was widespread, underscoring the pressing need for practice-based improvement in CKD prevention, recognition, and treatment. These real-world data lay the groundwork for the development of more effective strategies to deliver care that enhances wellness and survival for patients with and at risk for CKD.