https://orcid.org/0000-0001-9109-0811
Figures
Abstract
Background
The Kidney Failure Risk Equation (KFRE) and Kaiser Permanente Northwest (KPNW) models have been proposed to predict progression to ESKD among adults with CKD within 2 and 5 years. We evaluated the utility of these equations to predict the 1-year risk of ESKD in a contemporary, ethnically diverse CKD population.
Methods
We conducted a retrospective cohort study of adult members of Kaiser Permanente Northern California (KPNC) with CKD Stages 3–5 from January 2008-September 2015. We ascertained the onset of ESKD through September 2016, and calculated stage-specific estimates of model discrimination and calibration for the KFRE and KPNW equations.
Results
We identified 108,091 eligible adults with CKD (98,757 CKD Stage 3; 8,384 CKD Stage 4; and 950 CKD Stage 5 not yet receiving kidney replacement therapy), with mean age of 75 years, 55% women, and 37% being non-white. The overall 1-year risk of ESKD was 0.8% (95%CI: 0.8–0.9%). The KFRE displayed only moderate discrimination for CKD 3 and 5 (c = 0.76) but excellent discrimination for CKD 4 (c = 0.86), with good calibration for CKD 3–4 patients but suboptimal calibration for CKD 5. Calibration by CKD stage was similar to KFRE for the KPNW equation but displayed worse calibration across CKD stages for 1-year ESKD prediction.
Conclusions
In a large, ethnically diverse, community-based CKD 3–5 population, both the KFRE and KPNW equation were suboptimal in accurately predicting the 1-year risk of ESKD within CKD stage 3 and 5, but more accurate for stage 4. Our findings suggest these equations can be used in1-year prediction for CKD 4 patients, but also highlight the need for more personalized, stage-specific equations that predicted various short- and long-term adverse outcomes to better inform overall decision-making.
Citation: Zheng S, Parikh RV, Tan TC, Pravoverov L, Patel JK, Horiuchi KM, et al. (2023) CKD stage-specific utility of two equations for predicting 1-year risk of ESKD. PLoS ONE 18(11): e0293293. https://doi.org/10.1371/journal.pone.0293293
Editor: Bo Hu, Cleveland Clinic, UNITED STATES
Received: February 25, 2023; Accepted: October 9, 2023; Published: November 1, 2023
Copyright: © 2023 Zheng et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting information files.
Funding: The study was supported by a research grant from the Kaiser Permanente Northern California Community Health fund (S.Z.). Dr. Go has received funding through his institution from the National Institute of Diabetes, Digestive and Kidney Diseases as well as the National Heart, Lung and Blood Institute. The other authors have no relevant financial disclosures. The funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study.
Competing interests: The study was supported by a research grant from the Kaiser Permanente Northern California Community Health fund (S.Z.). Dr. Go has received funding through his institution from the National Institute of Diabetes, Digestive and Kidney Diseases as well as the National Heart, Lung and Blood Institute. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Introduction
An estimated 15% of Americans are estimated to have CKD, in 2019, 134,608 individuals were newly diagnosed with end-stage kidney disease (ESKD) [1, 2]. For many patients initiating chronic dialysis, the process is not optimally coordinated, and a significant proportion of patients feel under-prepared for dialysis. Unfortunately, the transition period immediately after dialysis initiation has a very high risk of death and hospitalization [1, 3]. Improving the transition from CKD to ESKD represents an important opportunity to improve outcomes and potentially lower healthcare costs. There is also greater appreciation that decline in eGFR is often not linear, with many patients staying in an earlier CKD stage (i.e., stage 3) much longer than in later stages (i.e., stage 4–5) [4], and some progressing rapidly due to occurrence of an episode of acute kidney injury (AKI) [5, 6]. In CKD Stages 4–5, a linear decline in eGFR may occur in <40% of patients [5]. Therefore, accurate ESKD prediction is crucial to delineate patients with a high short-term risk of progressing to ESKD rapidly.
Two efforts to improve ESKD risk prediction include the Kidney Failure Risk Equation (KFRE) and the Kaiser Permanente Northwest (KPNW) equation, which focus only on 2- and 5-year estimates for combined CKD Stages 3–5 [7, 8]. However, both equations have limitations that affect their utility in contemporary practice. For example, the development and validation efforts have been performed in older, primarily white/European individuals, so it remains unclear how generalizable the results are to the increasingly diverse populations in the U.S. KFRE uses four variables (age, sex, eGFR and urine albumin-to-creatinine ratio) [7] and the KPNW equation uses eight variables (age, sex, eGFR, proteinuria/albuminuria, systolic blood pressure, use of antihypertensive medications, hemoglobin level and a diabetes severity index) [8], which may not include all relevant risk factors to accurately estimate the absolute risk of ESKD for individual patients [9]. Given the models’ 2- or 5-year timeframe, they may be mostly applicable in the primary care setting, but potentially less useful for nephrologists when personalizing planning for patients most likely to progress to ESKD within a shorter time frame. The Kidney Disease: Improving Global Outcomes (KDIGO) guidelines recommend timely nephrology referral for planning renal replacement therapy (RRT) in persons whose risk of kidney failure within 1 year is 10–20% or higher [10]. The 2019 update of KDOQI Clinical Practice Guideline for Vascular Access emphasizes a patient-centered care approach that promotes optimal dialysis access management based on the best available evidence and patient’s life expectance and preference. The "ESKD Life-Plan" is achieved by creating a "P-L-A-N" (Patient Life-Plan First then Access Needs) for each patient that considers the patient’s life plan and corresponding access needs [11]. This highlights the importance of having validated and well-calibrated 1-year ESKD risk prediction tools [12].
To address this key knowledge gap, we evaluated the performance of ESKD prediction models for predicting 1-year risk of ESKD by CKD stage within a large, ethnically diverse community-based population.
Methods
Source population
The source population was based within Kaiser Permanente Northern California (KPNC), a large, integrated health care delivery system currently caring for >4.6 million members in Northern and Central California. Its membership is highly diverse and representative of the local surrounding and statewide population across demographic characteristics [13]. Nearly all aspects of care are also captured through KPNC’s comprehensive electronic health records (EHR).
The study was approved by the Kaiser Permanente Northern California institutional review board. A waiver of informed consent was obtained due to the retrospective, data-only nature of the study.
Study sample
We identified all members who were ≥18 years old with at least 2 measurements of outpatient, non-emergency department estimated glomerular filtration rate (eGFR) <60 mL/min/1.73m2 calculated using the original serum creatinine-based CKD-EPI equation [14] between January 2008 and September 2015. We defined the index date as the earliest qualifying eGFR measurement during the study period after which all subsequent measurements were <60 mL/min/1.73m2 to ensure the patient had CKD. We also required patients to have a qualifying eGFR measurement within 90–365 days after their index date. We excluded patients who received chronic RRT before their index date using data from a comprehensive health system ESKD Registry (i.e., chronic hemodialysis, chronic peritoneal dialysis, or receipt of renal transplant) [15–17]. Lastly, we excluded members with <12 months of continuous health plan membership and pharmacy benefit before their index date to ensure more complete capture of relevant comorbidities, prescriptions, and baseline laboratory measurements.
To assess the validity of the KFRE and the KPNW equation in adult KPNC members with CKD, we applied additional exclusion criteria specific to each equation. For KFRE, we excluded patients with no baseline proteinuria measurements (urine albumin-to-creatinine ratio [uACR], urine protein-to-creatinine ratio [uPCR], or urine dipstick). For the KPNW equation, we additionally excluded patients without baseline outpatient hemoglobin or systolic blood pressure values.
Follow-up and outcomes
The primary outcome was occurrence of ESKD during up to 1-year follow-up, with censoring at the end of follow-up or earlier if they disenrolled from the health plan or died, based on comprehensive health plan administrative databases, proxy reporting, Social Security Administration vital status files and California state death certificate files [18, 19]. Secondary outcomes included all-cause death based on the data resources described, as well as receipt of hospice services ascertained from the EHR.
Covariates
Relevant covariates were obtained from EHR-based data that has been standardized and linked at the patient level into the KPNC Virtual Data Warehouse (VDW) [20–25]. Outpatient eGFR was obtained within one year before index date, and urine protein measures up to four years before index date. For patients with missing uACR results, we imputed uACR from available uPCR and urine dipstick values, if available [26]. Demographic characteristics and tobacco use were ascertained from health plan databases [27, 28]. Prior AKI was defined using serum creatinine-based KDIGO criteria [29] within hospitalizations in the 5 years before index date. Relevant comorbidities and cardiac procedures were defined based on inpatient and ambulatory diagnoses and procedures using International Classification of Disease, Ninth and Tenth Edition (ICD-9 and ICD-10) codes (codes available upon request), relevant laboratory test results, and pharmacy dispensings within 5 years before index date [20–25]. We calculated a Diabetes Complications Severity Index based on diagnostic ICD-9 codes as required for the KPNW equation [8, 30]. We ascertained diabetes mellitus based on diagnoses or procedures using ICD-9 and ICD-10 codes, laboratory results, and filled outpatient prescriptions from hospitalization, ambulatory visit, laboratory, and pharmacy databases. We obtained baseline vital signs from the most recent outpatient clinic visit within one year before index date. For patients with missing baseline hemoglobin and vital signs, we created separate categories for unknown values.
Statistical approach
All analyses were conducted using SAS, version 9.4 (Cary, N.C.). We compared baseline characteristics across outcome categories. Given the large sample size, we compared standardized mean and proportion differences between groups using Cohen’s D for continuous and binary variables and Cramer’s V for categorical variables. Differences were considered potentially clinically significant if the effect size estimate was ≥0.10. We calculated crude risks and associated 95% confidence intervals for ESKD, use of hospice services, and all-cause death, overall and stratified by CKD stage.
We used the four-variable KFRE (age, eGFR, sex, and uACR) to calculate the 1-year predicted risk of ESKD [7]. For the KPNW equation, we calculated a risk score using age, sex, eGFR, uACR, systolic blood pressure, anti-hypertensive medication usage, hemoglobin, and diabetes complications severity index as categorical variables. KPNW 2-year predicted risk of ESKD was generated from these risk scores using published thresholds [8] and a point estimate conversion table (E. Schroeder, personal communication). To generate a KPNW 1-year risk estimate, we assumed a constant hazard and divided the 2-year risk estimates in half.
Equation discrimination was evaluating using c statistics, and calibration assessed through plots of observed vs. predicted 1-year risk of ESKD within each decile of predicted risk as well as the Brier score [9].
Results
Study cohort
We identified 140,010 eligible CKD patients, with 108,091 (91.4% CKD 3, 7.8% CKD 4 and 0.9% CKD 5) eligible for KFRE and 84,449 (90.6% CKD 3, 8.4% CKD 4 and 1.0% CKD 5) eligible for the KPNW equation (Fig 1). Among the 108,091 patients eligible for the KFRE, mean age was 75 years, 54.9% were women, 36.8% were of non-white race, and 12.4% were of Hispanic ethnicity.
Patients who progressed to ESKD within 1 year were more likely to be younger; men; of a minority race; Hispanic; current smokers; have a prior history of AKI, acute myocardial infarction, heart failure, cardiac procedures, diabetes, and chronic liver disease; and have a higher baseline systolic blood pressure and lower baseline hemoglobin than patients who did not progress to ESKD (Table 1). In addition, patients who progressed to ESKD were less likely to have known atrial fibrillation or flutter, thyroid disease, chronic lung disease, and diagnosed dementia and depression (Table 1). Characteristics of the cohort eligible for the KPNW equation are shown in S1 Table.
One-year risks of ESKD, receipt of hospice services and death
During 1-year follow-up among 108,091 patients in the KFRE cohort, 879 (0.81%, 95% CI:0.76–0.87%) progressed to ESKD, 1,133 (1.05%, 95% CI:0.99–1.11%) received hospice services before ESKD or death, and 4,635 (4.29%, 95%CI:4.23–4.35%) died before ESKD (Table 2). As expected, the risk of each outcome was higher with more severe baseline CKD, particularly for the outcome of ESKD (Table 2).
The KPNW equation cohort displayed similar patterns across CKD stages in all outcomes, with slightly higher absolute risks than in the KFRE cohort (Table 2). During 1-year follow-up among 84,449 patients in the KPNW equation cohort, 757 (0.90%, 95% CI:0.83–0.96%) progressed to ESKD, 1,042 (1.23%, 95% CI:1.16–1.31%) received hospice services before ESKD or death, and 4,055 (4.80%, 95% CI:4.66–4.95%) died before ESKD (Table 2).
CKD stage-specific utility for predicting 1-year ESKD risk
KFRE displayed only moderate discrimination for CKD Stage 3 and Stage 5 patients (c = 0.76 for both groups), with much better discrimination for CKD Stage 4 patients (c = 0.86) (Fig 2). Overall equation calibration of the KFRE was relatively good for CKD Stages 3 and 4, with modest overestimation of risk, but KFRE consistently underestimated the actual 1-year risk of ESKD in CKD Stage 5 patients. Furthermore, in CKD Stage 3 and 4 patients, the overestimation of actual risk increased as the predicted risk increased, while in CKD Stage 5 patients, the highest underestimation was in patients with predicted risk between 20% and 32% (Fig 2).
The KPNW equation displayed very similar discrimination as the KFRE in CKD Stages 3 and 4 patients (c = 0.77 and 0.85, respectively), with poorer discrimination for CKD Stage 5 (c = 0.72). There was slight underestimation of actual 1-year ESKD risk in CKD Stage 3 patients but greater overestimation of ESKD risk in CKD Stage 4 patients. Calibration was notably worse in CKD Stage 5 patients, with significant underestimation of risk of up to 15.8% among patients with a predicted risk between 20% and 24% (Fig 3).
We also calculated Brier scores for each equation in each CKD stage to assess calibration (Table 3). Within each CKD stage, KFRE had better calibration than the KPNW equation.
Discussion
Within a large, contemporary, ethnically diverse population with CKD Stages 3–5, we found notably higher 1-year risks of ESKD with higher CKD stage, as expected, with a very low overall 1-year risk of ESKD in all CKD Stage 3 patients (0.25–0.27%) compared with more than one in four CKD Stage 5 patients within 1 year (27.7–28.2%). Across these ranges of absolute risk, it is important to identify more accurately the subset of patients who are at high risk for initiating RRT within the next year. Importantly, the KFRE and KPNW equation showed only moderate ability to discriminate between relatively higher vs. lower risk patients, particularly for CKD Stage 3 (c = 0.76–0.77) and Stage 5 (c = 0.72–0.76), and these estimates are notably lower than those previously reported among combined CKD 3–5 populations (c range: 0.84–0.91) [7, 8, 26, 31].
With regards to predicting absolute 1-year ESKD risk in CKD Stage 3 patients, equation calibration was of variable accuracy which reflects the challenge of predicting a rare event in this low-risk population, especially given the significantly higher crude risk of all-cause death and the higher rate of hospice care before any RRT. Importantly, the KFRE underestimated actual 1-year risk of ESKD among the lower deciles of predicted risk but overestimated ESKD risk among the higher deciles of predicted risk. In contrast, the KPNW equation underestimated the 1-year risk of ESKD across the entire spectrum of predicted risk.
Among CKD Stage 4 patients, the overall 1-year risk of ESKD was 4.3%, which was higher than the probability of entering hospice care but lower than the crude risk of dying before initiating RRT. Both KFRE and the KPNW equation had significantly better discrimination (c = 0.86 and 0.85, respectively), with good calibration across the spectrum of predicted risk and only modest overestimation of actual 1-year ESKD risk. Notably, the additional variables used in the KPNW equation (i.e., systolic blood pressure, anti-hypertensive medication usage, hemoglobin level, and presence of diabetes with complications) did not appear to provide incremental improvement in discrimination or calibration for 1-year ESKD above the demographic characteristics and kidney measures (eGFR, uACR) that were common between KPNW and KFRE equations.
Not surprisingly, CKD Stage 5 patients had the highest crude risk of initiating chronic RRT within 1 year (28.2%), which was much higher than the crude risks of either receiving hospice care or dying before initiating RRT. These data also highlight that a large fraction of CKD Stage 5 patients remain alive and not receiving RRT or hospice care within one year, supporting the need to better risk stratify these patients with eGFR <15 ml/min/1.73 m2 in informing decisions around issues like timing of vascular access or peritoneal dialysis catheter placement. Focusing resources on the subset of CKD Stage 5 patients most likely to need RRT in the short-term could also be a more cost-effective population strategy. Given that both the KFRE and KPNW equation had only modest discrimination (0.76 and 0.72, respectively) and significant underestimation of the actual risk of initiating RRT, even more accurate prediction equations are needed.
Current KDIGO guidelines [10] recommend timely referral for planning RRT in people with progressive CKD in whom the risk of kidney failure within 1 year is 10–20% or higher, as determined by validated risk prediction tools. Currently, there are no validated 1-year ESKD prediction models [32] to assist physicians and patients in facilitating timelier referral for transplant evaluation, earlier identification of potential living donors, promoting early education of home dialysis options, and more appropriate, timely referral for permanent vascular access or peritoneal catheter placement. Identifying the subset of patients in each CKD stage at highest short-term risk of developing ESKD would also help with targeting optimization of the use and dosing of angiotensin-converting enzyme inhibitors/angiotensin II receptor blocker therapy, and sodium-glucose cotransporter-2 (SGLT-2) Inhibitors and non-steroidal mineralocorticoid receptor antagonists, along with avoidance of nephrotoxin exposures as well as reducing placement of futile vascular access. Accurate short-term ESKD risk information would also promote timely, relevant discussions with high-risk patients about their preferences for different RRT options vs. conservative management approaches, including decisions about end-of-life care.
Given the observed limitations of KFRE and KPNW equations, we propose several approaches to improve the utility of risk prediction equations for current clinical practice. First, CKD stage-specific prediction models should be developed and validated given that there are substantial differences in the comorbidity burden and duration of time spent in different stages of CKD severity. This is highlighted by the variability in performance of the KFRE and KPNW equations across CKD stages, with their best performance only in CKD Stage 4 patients. Second, models should account for patients with missing values of covariates to facilitate clinical management in parallel with efforts to obtain the missing data. In our cohort, 31,919 patients (22.7%) were excluded from the KFRE analytic cohort because of missing proteinuria measurements, and an additional 25,338 (18.1%) were excluded from the KPNW equation analytic cohort because of missing hemoglobin or systolic blood pressure measurements (Fig 1). These represent a large proportion of all eligible CKD patients, some of whom may be at high risk for progression to ESKD and would otherwise not be identified even with partially available information on other variables. Lastly, prediction models should consider other potentially important risk factors for progression to ESKD beyond the limited candidate variables examined in the development of KFRE and KPNW equations. For example, in our cohort, differences were noted for characteristics such as prior AKI and chronic liver disease between patients who did or did not initiate RRT within one year (Table 1). However, the 8-variable KPNW equation did not have improved predictive ability compared to the 4-variable KFRE, suggesting that not all risk factors will result in meaningful performance improvements across diverse populations. With the increasingly widespread use of EHR systems nationally, ascertaining data on complex comorbidity and other patient characteristics is becoming more feasible on a large scale and in real time to facilitate more accurate, timely short-term ESKD risk prediction.
Finally, we observed that patients in CKD 3 and CKD 4 had a higher 1-year risk of death than ESKD, which is consistent with another study showing death is more common than reaching dialysis in CKD stages 2, 3, and 4 patients over a 5-year observation period [33]. In contrast, among CKD 5 patients, 1-year risk of ESKD was fourfold higher than death. This suggests that clinical management and patient education efforts should focus differently in CKD 3 and 4 vs. CKD 5. For example, the priority should likely be on reducing cardiovascular risk over ESKD risk for CKD stage 3 and 4 patients but more on preparing for KRT in CKD 5 patients in parallel with cardiovascular risk reduction.
Our study has several strengths. We studied a large, socio-demographically diverse population of eligible patients across the spectrum of CKD severity which allowed us to examine the utility (as reflected by estimates of discrimination and calibration) of both models within each CKD stage in the same analysis population. Unlike previous studies, we applied stricter eGFR criteria by requiring multiple outpatient, non-emergency department eGFR measurements <60 mL/min/1.73m2 to ensure patients had CKD. Given our comprehensive EHR system, we were also able to capture information on relevant comorbidities and laboratory measurements to characterize CKD patients who did or did not progress to ESKD within a short time frame.
Our study also had several limitations. Our method of calculating 1-year risks from the KPNW equation involved assuming a constant hazard for 2-year risks and dividing that estimate in half. Given that non-linear trajectories in kidney function can occur in certain CKD patients [34], our assumption may contribute to a less accurate estimate. However, we believe that this approach reflects how the KPNW equation would be used in clinical practice if a physician was trying to estimate a patient’s ESKD risk at 1 year. This also supports the need to develop and validate separate and more accurate equations across a range of follow-up periods to allow for more effective individual-level and population-level management strategies. This is particularly relevant given the expansion nationally in the use of CKD population management programs [35], along with the recent focus on more optimal KRT starts [36]. Imputation of uACR values may cause equations to be less accurate than originally intended, although the presence of missing data may reflect performance in real world EHR settings. Finally, we conducted our study among CKD patients receiving care within an integrated healthcare delivery system in Northern California, so our results may not be fully generalizable to uninsured patients or to other types of practice settings or geographic areas.
In conclusion, we found that existing ESKD risk prediction equations have variable utility by stage of CKD for predicting 1-year risk of initiating RRT within a diverse community-based population. Additional efforts are needed to provide even more accurate, personalized, stage-specific estimates of the risk of ESKD across short- and long-term time periods to facilitate more optimal CKD management and planning for individuals and populations.
Supporting information
S1 Table. Baseline characteristics of 84,449 adults with chronic kidney disease eligible for the KPNW equation, January 1, 2008-September 30, 2015, overall and stratified by initiation of renal replacement therapy (RRT) within 1 year of follow-up.
https://doi.org/10.1371/journal.pone.0293293.s001
(DOCX)
References
- 1. Saran R, Robinson B, Abbott KC, Agodoa LYC, Bhave N, Bragg-Gresham J, et al. US Renal Data System 2017 Annual Data Report: Epidemiology of Kidney Disease in the United States. Am J Kidney Dis. 2018;71(3 Suppl 1):A7. Epub 2018/02/27. pmid:29477157
- 2.
2021 USRDS Annual Data Report: Epidemiology of kidney disease in the United States. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD: United States Renal Data System, 2021. Report No.
- 3. Chan KE, Maddux FW, Tolkoff-Rubin N, Karumanchi SA, Thadhani R, Hakim RM. Early outcomes among those initiating chronic dialysis in the United States. Clin J Am Soc Nephrol. 2011;6(11):2642–9. Epub 2011/10/01. pmid:21959599
- 4. Ku E, Johansen KL, McCulloch CE. Time-Centered Approach to Understanding Risk Factors for the Progression of CKD. Clin J Am Soc Nephrol. 2018;13(5):693–701. Epub 2018/04/01. pmid:29602779
- 5. Caravaca-Fontan F, Azevedo L, Luna E, Caravaca F. Patterns of progression of chronic kidney disease at later stages. Clin Kidney J. 2018;11(2):246–53. Epub 2018/04/13. pmid:29644066
- 6. Hsu RK, Chai B, Roy JA, Anderson AH, Bansal N, Feldman HI, et al. Abrupt Decline in Kidney Function Before Initiating Hemodialysis and All-Cause Mortality: The Chronic Renal Insufficiency Cohort (CRIC) Study. Am J Kidney Dis. 2016;68(2):193–202. pmid:26830447
- 7. Tangri N, Stevens LA, Griffith J, Tighiouart H, Djurdjev O, Naimark D, et al. A predictive model for progression of chronic kidney disease to kidney failure. JAMA. 2011;305(15):1553–9. Epub 2011/04/13. pmid:21482743.
- 8. Schroeder EB, Yang X, Thorp ML, Arnold BM, Tabano DC, Petrik AF, et al. Predicting 5-Year Risk of RRT in Stage 3 or 4 CKD: Development and External Validation. Clin J Am Soc Nephrol. 2017;12(1):87–94. pmid:28028051
- 9. Alba AC, Agoritsas T, Walsh M, Hanna S, Iorio A, Devereaux PJ, et al. Discrimination and Calibration of Clinical Prediction Models: Users’ Guides to the Medical Literature. JAMA. 2017;318(14):1377–84. pmid:29049590.
- 10. Levin A, Stevens PE, Bilous RW, Coresh J, De Francisco ALM, De Jong PE, et al. Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group: KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney International. 2013;3(1):1–150. Epub January 2013.
- 11. Lok CE, Huber TS, Lee T, Shenoy S, Yevzlin AS, Abreo K, et al. KDOQI Clinical Practice Guideline for Vascular Access: 2019 Update. Am J Kidney Dis. 2020;75(4 Suppl 2):S1–S164. Epub 2020/08/12. pmid:32778223.
- 12. Summary of Recommendation Statements. Kidney Int Suppl (2011). 2013;3(1):5–14. Epub 2013/01/01. pmid:25598998
- 13.
Gordon NP. Characteristics of Adult Health Plan Members in the Northern California Region Membership, as Estimated from the 2011 Member Health Survey. Division of Research, Kaiser Permenante Medical Care Program, Oakland, CA. 2013.
- 14. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–12. pmid:19414839
- 15. Hsu CY, Chertow GM, McCulloch CE, Fan D, Ordonez JD, Go AS. Nonrecovery of kidney function and death after acute on chronic renal failure. Clin J Am Soc Nephrol. 2009;4(5):891–8. pmid:19406959
- 16. Lo LJ, Go AS, Chertow GM, McCulloch CE, Fan D, Ordonez JD, et al. Dialysis-requiring acute renal failure increases the risk of progressive chronic kidney disease. Kidney Int. 2009;76(8):893–9. pmid:19641480
- 17. Hsu CY, Ordonez JD, Chertow GM, Fan D, McCulloch CE, Go AS. The risk of acute renal failure in patients with chronic kidney disease. Kidney Int. 2008;74(1):101–7. pmid:18385668
- 18. Arellano MG, Petersen GR, Petitti DB, Smith RE. The California Automated Mortality Linkage System (CAMLIS). Am J Public Health. 1984;74(12):1324–30. pmid:6507683
- 19. Go AS, Chertow GM, Fan D, McCulloch CE, Hsu CY. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004;351(13):1296–305. Epub 2004/09/24. pmid:15385656.
- 20. Go AS, Fang MC, Udaltsova N, Chang Y, Pomernacki NK, Borowsky L, et al. Impact of proteinuria and glomerular filtration rate on risk of thromboembolism in atrial fibrillation: the anticoagulation and risk factors in atrial fibrillation (ATRIA) study. Circulation. 2009;119(10):1363–9. Epub 2009/03/04. pmid:19255343
- 21. Go AS, Iribarren C, Chandra M, Lathon PV, Fortmann SP, Quertermous T, et al. Statin and beta-blocker therapy and the initial presentation of coronary heart disease. Ann Intern Med. 2006;144(4):229–38. Epub 2006/02/24. pmid:16490908.
- 22. Go AS, Lee WY, Yang J, Lo JC, Gurwitz JH. Statin therapy and risks for death and hospitalization in chronic heart failure. JAMA. 2006;296(17):2105–11. Epub 2006/11/02. pmid:17077375.
- 23. Go AS, Magid DJ, Wells B, Sung SH, Cassidy-Bushrow AE, Greenlee RT, et al. The Cardiovascular Research Network: a new paradigm for cardiovascular quality and outcomes research. Circ Cardiovasc Qual Outcomes. 2008;1(2):138–47. Epub 2009/12/25. pmid:20031802.
- 24. Go AS, Yang J, Ackerson LM, Lepper K, Robbins S, Massie BM, et al. Hemoglobin level, chronic kidney disease, and the risks of death and hospitalization in adults with chronic heart failure: the Anemia in Chronic Heart Failure: Outcomes and Resource Utilization (ANCHOR) Study. Circulation. 2006;113(23):2713–23. Epub 2006/06/07. pmid:16754803.
- 25. Go AS, Yang J, Gurwitz JH, Hsu J, Lane K, Platt R. Comparative effectiveness of different beta-adrenergic antagonists on mortality among adults with heart failure in clinical practice. Arch Intern Med. 2008;168(22):2415–21. Epub 2008/12/10. pmid:19064823.
- 26. Tangri N, Grams ME, Levey AS, Coresh J, Appel LJ, Astor BC, et al. Multinational Assessment of Accuracy of Equations for Predicting Risk of Kidney Failure: A Meta-analysis. JAMA. 2016;315(2):164–74. pmid:26757465
- 27. Carretta HJ, Mick SS. Geocoding public health data. Am J Public Health. 2003;93(5):699; author reply -700. Epub 2003/05/02. pmid:12721119
- 28. Stewart SL, Swallen KC, Glaser SL, Horn-Ross PL, West DW. Comparison of methods for classifying Hispanic ethnicity in a population-based cancer registry. Am J Epidemiol. 1999;149(11):1063–71. Epub 1999/06/04. pmid:10355383.
- 29. Kellum JA, Lameire N, Aspelin P, Barsoum RS, Burdmann EA, Goldstein SL, et al. Kidney disease: Improving global outcomes (KDIGO) acute kidney injury work group. KDIGO clinical practice guideline for acute kidney injury. Kidney International. 2012;2(1):1–138.
- 30. Young BA, Lin E, Von Korff M, Simon G, Ciechanowski P, Ludman EJ, et al. Diabetes complications severity index and risk of mortality, hospitalization, and healthcare utilization. Am J Manag Care. 2008;14(1):15–23. Epub 2008/01/17. pmid:18197741
- 31. Grams ME, Li L, Greene TH, Tin A, Sang Y, Kao WH, et al. Estimating time to ESRD using kidney failure risk equations: results from the African American Study of Kidney Disease and Hypertension (AASK). Am J Kidney Dis. 2015;65(3):394–402. pmid:25441435
- 32. Eckardt KU, Bansal N, Coresh J, Evans M, Grams ME, Herzog CA, et al. Improving the prognosis of patients with severely decreased glomerular filtration rate (CKD G4+): conclusions from a Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference. Kidney Int. 2018;93(6):1281–92. pmid:29656903
- 33. Keith DS, Nichols GA, Gullion CM, Brown JB, Smith DH. Longitudinal follow-up and outcomes among a population with chronic kidney disease in a large managed care organization. Arch Intern Med. 2004;164(6):659–63. pmid:15037495.
- 34. Li L, Astor BC, Lewis J, Hu B, Appel LJ, Lipkowitz MS, et al. Longitudinal progression trajectory of GFR among patients with CKD. Am J Kidney Dis. 2012;59(4):504–12. Epub 2012/01/31. pmid:22284441
- 35. Mendu ML, Waikar SS, Rao SK. Kidney Disease Population Health Management in the Era of Accountable Care: A Conceptual Framework for Optimizing Care Across the CKD Spectrum. Am J Kidney Dis. 2017;70(1):122–31. pmid:28132720.
- 36. Crooks PW, Thomas CO, Compton-Phillips A, Leith W, Sundang A, Zhou YY, et al. Clinical outcomes and healthcare use associated with optimal ESRD starts. Am J Manag Care. 2018;24(10):e305–e11. pmid:30325191.