Predicting Chronic Kidney Disease Outcomes: Are Two Estimated Glomerular Filtration Rates Better Than One?

Abstract

The National Kidney Foundation's (NKF) Kidney Disease Outcomes Quality Initiative (KDOQI) definition of chronic kidney disease (CKD), stages 3–5, requires 2 estimated glomerular filtration rates (eGFRs) <60 ml/min/1.73 m² more than 3 months apart. By requiring 2 eGFRs, the NKF definition reduced identification of people without chronic disease, which may have decreased identification of individuals with early CKD, but increased identification of those who ultimately have progression of CKD or require renal replacement therapy (RRT). Our objective was to determine whether 2 eGFR tests were better than 1 eGFR as a predictor of RRT, CKD progression, or death. This retrospective incident cohort study evaluates outcomes in adults with an initial eGFR < 60 ml/min/1.73 m² and a second eGFR after 90 days by examining a third follow-up eGFR. For the 2086 patients in this study, the mean initial eGFR was 50.7 ml/min/1.73 m² and the mean second eGFR was 59.3 ml/min/1.73 m². More than 40% of the population (925) did not have CKD based upon their second eGFR. The initial eGFR was the best predictor of the third eGFR. There was no material difference in the ability to predict outcome measures between 1 versus 2 eGFR tests, regardless of eGFR value or associated comorbidities. Identifying patients with CKD is a critical step when beginning to implement population management strategies for those patients. Our findings illustrate some of the trade-offs in strategies inherent in methods that might be used to identify patients with CKD; 1 eGFR will identify patients about 5 months sooner, allowing additional time for nephrologist and other therapeutic intervention, but approximately doubles the population to be managed. (Population Health Management 2012;15:113–118)

Introduction

C hronic kidney disease (CKD) is a major public health problem that is estimated to affect up to 20 million people nationwide.¹ In an effort to reduce the number of patients progressing to renal replacement therapy (RRT) the National Kidney Foundation (NKF) developed the Kidney Disease Outcomes Quality Initiative (KDOQI) definition of CKD.²

KDOQI recommends an estimated glomerular filtration rate (eGFR) to assess kidney function using the Modification of Diet and Renal Disease Study equation.² Estimated GFR varies depending on age, sex, serum creatinine, and other variables. The KDOQI workgroup based the definition of CKD on eGFRs <60 ml/min/1.73 m².

Serum creatinine levels vary among individuals, based upon diet, muscle mass, and physical activity, and among laboratories. For these reasons, the KDOQI workgroup chose a cutoff of more than 3 months (90 days) between 2 eGFR assessments (each less than 60 ml/min/1.73²) to define CKD.² By requiring 2 eGFRs (rather than 1), the NKF definition may have decreased identification of individuals with early CKD or other physiologic changes like fluid imbalances, but increased identification of a larger proportion of those who ultimately have progression of CKD.

A previous cohort analysis examined the outcomes of using 1 versus 2 eGFR values for CKD definition and recognition. The study found a single point-in-time eGFR identified twice as many CKD cases as 2 eGFRs but, among those identified, only 10%–20% ultimately progressed to end-stage renal disease (ESRD) over 1 year. Both the number of patients requiring RRT and the rate of RRT were smaller among patients with 1 eGFR (n=6 vs. 39; 1.3 cases per 1000 CKD patients vs. 14.4 cases per 1000 CKD patients).³ These results raise a question: Does an increase in the early identification of future ESRD patients diagnosed with single eGFR testing warrant management of approximately twice as many patients? If the use of 1 eGFR improves early identification of patients with CKD and the number likely to require RRT, using 1 eGFR rather than 2 may be preferable in the implementation of early intervention strategies.

The development of a clinical definition should aim to minimize errors (ie, maximize the sum of sensitivity and specificity). To this end, we utilized sensitivity and specificity values to measure the area under the receiver operating characteristic curve (AUROC), which provides a global assessment of the diagnostic accuracy⁴ of 1 and 2 eGFRs.

Methods

Setting and study design

We conducted a retrospective study, designed to simulate an incidence cohort. We evaluated the incidence of RRT and CKD progression at Kaiser Permanente Northwest (KPNW) from the year 2000 to 2006. KPNW is a group-model health maintenance organization (HMO), which serves the Portland, Oregon and the Vancouver, Washington metropolitan area. KPNW has an annual membership of approximately 450,000 people with demographics similar to the community it serves.⁵ KPNW's electronic medical record (EMR), EpicCare/HealthConnect, has served as the sole medical record at all clinics since January 1997. KPNW as a research setting has been described in detail elsewhere.⁶ The study was reviewed and approved by KPNW's human subjects committee.

Using medical encounter data to compare patients with 1 vs. 2 eGFRs presents inherent biases: (1) patients with a single eGFR may be less likely to have a second eGFR checked because they never received follow-up care, or (2) patients with lower eGFRs may be more likely to have a second one checked, biasing the sample toward sicker individuals. In order to circumvent these potential biases, we utilized the same patients to create both the first eGFR cohort and the second eGFR cohort. We compared testing strategies, namely whether the rate at which patients who had an adverse outcome that would have been identified with a single eGFR vs. 2 eGFRs differed, and how quickly that identification might occur with the 2 approaches.

Participants and eligibility criteria

We identified all KPNW members 18 years of age or older who were enrolled in the HMO for 1 year prior to cohort entry (to ensure no prior diagnosis of CKD or preexisting abnormal eGFR values). Those included in the study population had an initial eGFR (first eGFR, which defined cohort entry) value < 60 ml/min/1.73 m² completed during the year 2000 and a subsequent eGFR (second eGFR) value after 90 days (consistent with the current KDOQI definition) and within 270 days. The 270-day time limitation was chosen to minimize the occurrence of random measurements assessed for reasons other than diagnosing CKD. Each participant had to be a member of the health plan for at least 5 years following the second eGFR, unless they died or underwent RRT. A third eGFR assessment was required during the 5-year follow-up period. All serum creatinine levels were collected in the outpatient setting because of the potential effect of hospitalization or acute illness on serum creatinine. Patient characteristics such as age, sex, smoking status, blood pressure, proteinuria, and body mass index were obtained from the EMR. Congestive heart failure and coronary artery disease were identified via International Classification of Diseases, Ninth Revision (ICD-9) coding (410–414, 440, 441–445, 424–425.9, 428–429.9). Angiotensin receptor blocker and angiotensin-converting enzyme inhibitor use was obtained from pharmacy records. Patients with diabetes were identified via the HMO's diabetes registry. Hypertension was defined as systolic blood pressure > 140 mmHg or diastolic blood pressure > 90 mmHg. Exclusion criteria included prior renal transplantation or a CKD defining diagnosis such as polycystic kidney disease, glomerulonephritis, or congenital renal defect identified via ICD-9 code (223.0, 236.91, 250.4, 271.4, 274.1, 283.11, 403.x1, 404.x2, 404.x3, 440.1, 442.1, 447.3, 572.4, 580–588, 591, 642.1, 646.2, 753.12–753.17, 753.19, 753.2, 794.4).

Outcomes

Each patient existed in both the first eGFR and second eGFR cohorts. The first eGFR cohort patients were followed from the date of their first eGFR; likewise they were followed from the date of their second eGFR in the second eGFR cohort. In order to determine the most accurate method of CKD identification utilizing first eGFR vs. second eGFR measurements with specific outcomes, we used 3 separate outcomes for the CKD population: (1) death, (2) RRT, and (3) CKD progression (defined as a > 50% decline in eGFR). Assessment of CKD progression was analyzed by taking the lowest recorded eGFR during the 5-year follow-up period and comparing it with the first eGFR (in the first eGFR cohort) and, separately, the second eGFR assessment (in the second eGFR cohort). RRT was identified through the medical record and verified by chart review: Death was identified through the HMO's membership database. Those who progressed to RRT were censored from the mortality analysis.

Statistical analysis

The 2 cohorts were followed for 5 years or until death or RRT after their second eGFR measurement. We calculated the sensitivity, specificity, and positive predictive value of the 2 cohorts stratified by several characteristics. We compared the 2 testing strategies for characteristics between those patients younger and older than 60 years of age. Sixty years was used as a cutoff to remain consistent with KDOQI guidelines for high-risk individuals (Part 8, Guideline 3).² AUROC curves were used to compare the 2 testing strategies and their accuracy. We performed the statistical analyses using Stata, version 9 (StataCorp LP, College Station, TX) and SAS, version 8.2 (SAS Institute Inc., Cary, NC).

Results

Table 1 displays the baseline characteristics of the 2086 subjects who were included in the study population. By study design, all 2086 subjects had an initial eGFR < 60 ml/min/1.73 m²; however, only 1161 (56%) subjects had a second eGFR < 60 ml/min/1.73 m².

Table 1.

Baseline Characteristics

Number of patients	2086
Age
mean (sd)	70 (12.2)
Range	19–98
<60 yrs (%)	396 (19.0)
>60 yrs (%)	1690.0 (81.0)
Sex
n (%) Male	832 (39.9)
Race
N (%) White	1822 (87.3)
Missing N (%)	125 (6.0)
eGFR (ml/min/1.73 m²)
Mean (sd)
First eGFR	50.7 (8.9)
Second eGFR	59.3 (17.7)
First eGFR (ml/min/1.73 m²)
N (%)
45—60	1685 (80.8)
30—45	314 (15.1)
15—30	79 (3.8)
<15	8 (0.4)
Second eGFR (ml/min/1.73 m²)
N (%)
>60	925 (44.3)
45—60	786 (37.7)
30—45	297 (14.2)
15—30	69 (3.3)
<15	9 (0.4)
Diabetes
N (%)	570 (27.3)
Hypertension
N (%)	889 (43.1)
Proteinuria (+1 or greater) of 1121 with U/A available
N (%)	176(15.7)
Missing N (%)	965(46.3)
Smoker
N (%)	296 (14.7)
Coronary artery disease
N (%)	598 (28.7)
Congestive heart failure
N (%)	403 (19.3)
Angiotensin-converting enzyme/angiotensin receptor blocker
N (%)	987 (47.3)
Body mass index
Mean (sd)	30.3 (7.4)
Missing N (%)	215 (10.3)
Obese (body mass index >30)
N (%)	819 (43.8)
Missing N (%)	215 (10.3)

eGFR, estimated glomerular filtration rate; U/A, urinalysis.

Of the initial 2086 subjects, 1752 (84%) had a third eGFR < 60 ml/min/1.73 m² (Table 2).

Table 2.

Cohort Results: Crude Occurrences

Number of patients	2086
Third eGFR
Mean, ml/min/1.72 m² (SD)	44.3 (17.8)
Third eGFR (ml/min/1.72 m²)
N (%)
>60	334 (16.0)
45–60	683 (32.7)
30–45	613 (29.4)
15–30	366 (17.6)
<15	90 (4.3)
Mean time intervals between eGFR evaluations,
Days (SD)
First–Second eGFR	157.1 (59.1)
Second–Third eGFR	774.7 (523.3)
First–Third eGFR	931.8 (527.4)
Range	96–1826
Absolute difference in eGFR,
ml/min/1.72 m² (SD)
First vs. Third eGFR	−6.4 (17.3)
Second vs. Third eGFR	−15.0 (16.6)
RRT
N (%)	41 (2.0)
CKD progression
N (%)
eGFR 1 to 3	235 (11.3)
eGFR 2 to 3	304 (14.5)
Death
N (%)	621 (29.8)

CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; RRT, renal replacement therapy.

Table 3 summarizes sensitivity and specificity of 1 eGFR vs. 2 eGFRs by using the AUROC curves subdivided into patients older than or younger than 60 years of age with a cutoff eGFR of < 45 ml/min/1.73 m². Among the population older than 60 years of age, the sensitivity of 1 eGFR for CKD progression and RRT was 17.7% and 61.5% and specificity was 81.1% and 81.9% (data incorporated into AUROC). Too few events were observed to determine significance among those younger than 60 years of age. When we applied the second eGFR assessment (ie, both first and second eGFR had to be < 60 ml/min/1.73 m²) to those older than 60 years of age, the sensitivity of CKD progression and RRT was 15.4% and 69.2% with specificity of 80.4% and 81.7%. The AUROC curves for all 3 outcomes were not significantly different when comparing 1 eGFR to 2 eGFRs regardless of diabetes, hypertension, or age (Table 3).

Table 3.

Area Under the Receiver Operating Characteristic Curve with 95% Confidence Intervals as a Test for CKD Progression, Renal Replacement Therapy, and Death for eGFR < 45ml/min/1.73 m ²

	Area under curve (95% CI)
Population	1 eGFR > 60 years	2 eGFR	1 eGFR < 60 years	2 eGFR
CKD progression
Overall	0.49(0.46, 0.52)	0.48(0.45, 0.50)	0.51(0.45, 0.57)	0.54(0.49, 0.59)
Diabetes	0.49(0.44, 0.54)	0.49(0.44, 0.53)	0.51(0.42, 0.60)	0.49(0.42, 0.56)
No diabetes	0.50(0.46, 0.53)	0.48(0.45, 0.51)	0.50(0.41, 0.59)	0.57(0.50, 0.65)
Hypertension	0.50(0.46, 0.55)	0.47(0.43, 0.52)	0.53(0.42, 0.64)	0.54(0.45, 0.63)
No hypertension	0.48(0.44, 0.52)	0.48(0.45, 0.51)	0.50(0.42, 0.58)	0.54(0.48, 0.60)
RRT
Overall	0.72(0.62, 0.81)	0.76(0.66, 0.85)	0.70(0.57, 0.83)	0.64(0.51, 0.77)
Diabetes	0.74(0.61, 0.87)	0.81(0.69, 0.92)	0.68(0.52, 0.83)	0.61(0.45, 0.76)
No diabetes	0.70(0.55, 0.84)	0.70(0.55, 0.85)	0.78(0.53, 1.00)	0.69(0.41, 0.98)
Hypertension	0.76(0.61, 0.91)	0.74(0.59, 0.89)	0.70(0.50, 0.90)	0.65(0.45, 0.85)
No hypertension	0.68(0.55, 0.81)	0.76(0.63, 0.88)	0.71(0.53, 0.89)	0.63(0.44, 0.81)
Death
Overall	0.54(0.52, 0.56)	0.54(0.52, 0.56)	0.55(0.49, 0.61)	0.54(0.49, 0.60)
Diabetes	0.52(0.48, 0.56)	0.50(0.47, 0.54)	0.53(0.44, 0.62)	0.55(0.49, 0.63)
No diabetes	0.55(0.52, 0.57)	0.56(0.53, 0.58)	0.56(0.47, 0.65)	0.53(0.46, 0.60)
Hypertension	0.56(0.52, 0.59)	0.58(0.54, 0.61)	0.50(0.40, 0.61)	0.53(0.42, 0.63)
No hypertension	0.53(0.50, 0.56)	0.53(0.50, 0.55)	0.57(0.49, 0.65)	0.55(0.48, 0.62)

CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; RRT, renal replacement therapy.

We repeated the analysis using a lower test cutoff: eGFR < 30 ml/min/1.73 m². Table 4 summarizes sensitivity and specificity using AUROC curves among patients with an eGFR < 30ml/min/1.73 m². As in the preceding analysis, when using a cutoff for the first eGFR < 45 ml/min/1.73 m², there were no consistent changes in AUROC curves with 1 eGFR vs. 2 eGFRs regardless of age or other comorbidities. When comparing the eGFR cutoffs (<45 ml/min/1.73 m² and < 30 ml/min/1.73 m²) between Tables 3 and 4 there is no consistent difference despite the initial eGFR value.

Table 4.

Area Under the Receiver Operating Characteristic Curve with 95% Confidence Intervals as a Test for CKD Progression, Renal Replacement Therapy, and Death for eGFR < 30 ml/min/1.73 m ²

	Area under curve (95% CI)
Population	1 eGFR > 60 years	2 eGFR	1 eGFR < 60years	2 eGFR
CKD progression
Overall	0.50(0.48, 0.51	0.50(0.49, 0.52)	0.52(0.47, 0.56)	0.53(0.50, 0.57)
Diabetes	0.49(0.47, 0.51)	0.50(0.48, 0.52)	0.50(0.43,0.56)	0.50(0.46, 0.55)
No diabetes	0.51(0.49, 0.53)	0.51(0.49, 0.53)	0.51(0.46, 0.56)	0.55(0.50, 0.61)
Hypertension	0.51(0.49, 0.54)	0.49(0.47, 0.51)	0.52(0.46, 0.58)	0.55(0.48, 0.62)
No hypertension	0.48(0.47,0.49)	0.51(0.49, 0.53)	0.52(0.46, 0.57)	0.52(0.48, 0.56)
RRT
Overall	0.62(0.53, 0.70)	0.68(0.58, 0.77)	0.64(0.52, 0.77)	0.69(0.56, 0.82)
Diabetes	0.63(0.50, 0.75)	0.74(0.60, 0.88)	0.64(0.49, 0.79)	0.67(0.52, 0.82)
No diabetes	0.61(0.48, 0.74)	0.61(0.48, 0.73)	0.61(0.37, 0.86)	0.74(0.46, 1.00)
Hypertension	0.64(0.49, 0.79)	0.73(0.57, 0.90)	0.64(0.46, 0.82)	0.70(0.51, 0.90)
No hypertension	0.58(0.48, 0.69)	0.62(0.50, 0.73)	0.66(0.48, 0.84)	0.68(0.50, 0.86)
Death
Overall	0.51(0.50, 0.53)	0.51(0.50, 0.52)	0.54(0.49, 0.58)	0.50(0.47, 0.53)
Diabetes	0.50(0.48, 0.52)	0.50(0.48, 0.51)	0.54((0.47, 0.62)	0.49(0.45, 0.53)
No diabetes	0.52(0.51, 0.53)	0.51(0.50, 0.52)	0.52(0.47, 0.57)	0.50(0.47, 0.54)
Hypertension	0.52(0.50, 0.54)	0.52(0.50, 0.54)	0.49(0.47, 0.50)	0.51(0.45, 0.58)
No hypertension	0.51(0.50, 0.53)	0.50(0.49, 0.51)	0.56(0.50, 0.62)	0.50(0.47, 0.52)

CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; RRT, renal replacement therapy.

We compared the prevalence of eGFRs < 45 ml/min/1.73 m² following the first or second eGFR < 60 ml/min/1.73 m² to assess whether 1 or 2 eGFRs were more likely to predict significant progression. We found little difference. Among patients with an initial eGFR between 45–59 ml/min/1.73 m², 49.9% had a second eGFR >60 ml/min/1.73 m². Among those with a first eGFR between 30–44 ml/min/1.73 m², 21.3% had a second eGFR >60ml/min/1.73 m². Even those with a first eGFR between 15–29 ml/min/1.73 m² had a false positive rate (ie, second eGFR >60ml/min/1.73 m²).

Discussion

Our study suggests that eGFR (1 or 2) is generally a poor test for predicting death and CKD progression but may be better at predicting future RRT. There was little difference in the ability to predict outcome measures between strategies relying upon 1 eGFR and 2 eGFR regardless of the eGFR cutoff and associated comorbidities. Although there may be a lack of important differences between 1 and 2 eGFRs in predicting outcomes, strategies utilizing 2 eGFRs may result in delayed diagnosis and treatment. We found that requiring the second eGFR would delay identification of patients at risk for adverse outcomes by 157 days; whether the increased time to identification warrants doubling the number of potential diagnoses (including a presumably large number of false positive diagnoses) depends on the interventions taken to mitigate progression. The trade-offs in these 2 strategies to identify patients with CKD should be balanced against each other; 1 eGFR will identify patients about 5 months sooner, allowing additional time for nephrologist and other therapeutic intervention, but approximately doubles the population to be managed. To this end, future work might usefully develop a decision analytic model that could enumerate and value the costs and benefits of each strategy.

We assessed the sensitivity and specificity of 1 and 2 eGFRs utilizing AUROC curves to measure the 2 testing strategies. We utilized 2 cutoff eGFR values (<45 ml/min/1.73 m² and < 30 ml/min/1.73 m²; Tables 3 and 4). Several consistent trends were noted when evaluating the sensitivity and specificity of these testing strategies. The ability of both testing strategies to identify true positives (sensitivity) and discern true negatives (specificity) was poor. The AUROC did not improve in clinically important ways using 2 eGFRs to predict outcome measures (when compared with 1 eGFR). Estimated GFR, whether 1 or 2, was a poor test to predict 5-year death and CKD progression. Estimated GFR (1 or 2) does a better job of predicting RRT. No difference in AUROC was seen between 1 eGFR vs. 2 eGFR assessments in predicting any of the outcomes. None of the comorbidities assessed (diabetes, hypertension, age) had any impact on AUROC curves nor did our predetermined eGFR cutoffs (<45 and < 30 ml/min/1.73 m²).

We evaluated changes in eGFR (first, second, and third). Among patients with an initial eGFR of 45–59 ml/min/1.73 m², 191 (11%) had progressive CKD (defined as a 50% decrease in eGFR). Among those with both a first and second eGFR ≤ 60 ml/min/1.73 m², 162 (18%) had progressive CKD. By adhering to the KDOQI guidelines of 2 eGFR evaluations, these 162 patients would have had a delayed diagnosis of CKD relying upon their second eGFR > 60 ml/min/1.73 m². Although many of these patients likely would have obtained further eGFR assessments, one must consider the potential time lost for CKD management and nephrology referral. Curtis et al described the impact of late nephrology referral and lack of complete CKD investigation by primary care providers prior to nephrology referral.⁷ Given the lack of improved prediction of outcomes using 2 eGFRs vs. 1 eGFR, the potential consequences of late referral and lost to follow-up occurrences in these patients should be considered.

Our study may provide direction for population managers seeking the most effective means of identifying patients at risk of developing ESRD or progressive CKD. Importantly, neither approach (1 or 2 eGFRs) was effective enough to warrant a recommendation that might be applied to individual patients. Further studies are needed to understand the utility of using 1 or 2 eGFRs (or other strategies) and how to better identify patients with CKD early in the course of their disease with greater sensitivity and specificity. Rather than simply using 1 characteristic, namely eGFR, as done in this analysis, others have suggested using several characteristics simultaneously to predict patient risk of poor outcomes.⁸ One such “risk score” (that used the characteristics of age, sex, eGFR, and whether the patient had diabetes, anemia, or hypertension) showed appreciably better performance characteristics than the single characteristic approach we report here.⁹ Although requiring more input from clinicians to calculate patient risk, the multiple factor risk score approach may be implemented in an automated fashion using electronic health record data to minimize clinician burden.

These data have implications for population research. Large studies analyzing the prevalence and incidence of CKD generally have relied on a single eGFR measurement. Although the KDOQI guidelines recommend 2 eGFR assessments 90 days apart prior to diagnosis of CKD,² most prevalence studies using only 1 eGFR may overestimate the prevalence of CKD. A recent study of CKD prevalence using National Health and Nutrition Surveys data collected during 1999–2004, and 1988–1994 found the prevalence of CKD stages 1–4 increased from 10.1% to 13.1%. Among the subgroup of CKD patients for whom diagnosis was based upon eGFR alone (stages 3 and 4), the prevalence increased from 5.6% to 8.0%.¹⁰ Our results suggest that using a single eGFR measure (rather than requiring 2) may be a reasonable approach in research studies that seek to classify patients into categories of renal risk.

Limitations

There are a number of limitations to our study. It is unclear whether our 5-year follow-up period is long enough to have captured all potential adverse outcomes. Additionally, by conditioning a 5-year KPNW membership requirement, we may have produced a biased absolute outcome rate. Another limitation may be use of the lowest eGFR measurement recorded for outcome identification. Use of the lowest measurement may be more clinically relevant because clinicians may respond to a lower, rather than higher, eGFR; however, the lowest measurement may be less reflective of true kidney disease. Finally, the inability of eGFR to predict outcome measures may be related to the age of our patient population. Elderly populations with CKD defined by the Modification of Diet and Renal Disease Study eGFR do not require RRT or suffer CKD progression as frequently as younger populations. By using an elderly population we may have minimized our outcome occurrences because older patients are less likely to undergo dialysis.⁹ Finally, by requiring that all patients have 3 eGFRs, these outcomes may be relevant only to those patients who have been tested more than once. Many patients with CKD may be unrecognized and many more may have had only a single eGFR measured, leading to selection bias in our study.

Conclusions

Estimated GFR is generally a poor predictor of all-cause mortality, RRT, or CKD progression. Use of a single eGFR to identify CKD may increase the number of patients identified with the condition, but utilizing a second eGFR does little to improve identification of patients likely to suffer adverse CKD-related outcomes. Furthermore, our data suggest that the major epidemiological studies on CKD using only 1 eGFR to determine prevalence are generally accurate in assessing the burden CKD imposes on our population. Given the poor performance of eGFR to predict renal outcomes, other variables should be included to identify patients with CKD who are likely to suffer from adverse outcomes.

Footnotes

Author Disclosure Statement

Drs. DeVille, Smith, Johnson, Weiss, and Thorp, Ms. Yang, and Ms. Petrik disclosed no conflicts of interest.

References

Schoolwerth

, Engelgau

, Hostetter

et al. Chronic kidney disease: A public health problem that needs a public health action plan. Prev Chronic Dis, 2006; 3:A57.

National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification, and stratification. Am J Kidney Dis, 2002; 39:S1–S266.

Thorp

, Weinstein

, Deville

et al. Comparison of renal replacement therapy and mortality using 1 versus 2 estimated glomerular filtration rates. Popul Health Manag, 2009; 12:149–155.

Altman

, Bland

. Diagnostic tests 3: Receiver operating characteristic plots. BMJ, 1994; 309:188.

Selby

, Smith

, Johnson

, Rabel

, Friedman

, McFarland

. Kaiser Permanente medical care program. Strom

. Pharmacoepidemiology, 4th. New York: John Wiley & Sons, 2005; 241–260.

Greenlick

, Freeborn

, Pope

. Health Care Research in an HMO: Two Decades of Discovery. Baltimore: Johns Hopkins University Press, 1998.

Curtis

, Barrett

, Djurdjev

et al. Evaluation and treatment of CKD patients before and at their first nephrologist encounter in Canada. Am J Kidney Dis, 2007; 50:733–742.

Taal

, Brenner

. Predicting initiation and progression of chronic kidney disease: Developing renal risk scores. Kidney Int, 2006; 70:1694–1705.

Johnson

, Thorp

, Platt

, Smith

. Predicting the risk of dialysis and transplant among patients with CKD: A retrospective cohort study. Am J Kidney Dis, 2008; 52:653–660.

10.

Coresh

, Selvin

, Stevens

et al. Prevalence of chronic kidney disease in the United States. JAMA, 2007; 298:2038–2047.