The prevalence of chronic kidney disease (CKD) is increasing in the United States [1, 2] and is associated with an increased risk of cardiovascular events, end-stage renal disease (ESRD) and mortality [3–5]. Early identification and appropriate care of CKD patients may delays progression to ESRD and decreases mortality, morbidities, and cost [6, 7]. Yet, the level of CKD awareness is low in both the general U.S. population (24%) and among high risk patients (less than 10% among individuals with Stage 3 CKD) [8, 9].

The National Kidney Foundation's Kidney Disease Outcome Quality Initiative (KDOQI) evidence-based clinical practice guidelines for CKD recommend periodic screening of high risk individuals for CKD [10]. Specifically, this entails a clinical evaluation (eg. assessment of symptoms, risk factors, medical history, family history), measurement of blood pressure, and laboratory evaluation (eg. serum creatinine, protein-to-creatinine ratio, imaging of the kidneys).

Despite these recommendations, there is evidence from national population surveys that high risk patients are often not screened for CKD and that the presence of CKD may go unidentified [8, 11]. Moreover, CKD identification by health care providers within health maintenance organizations have not been fully examined, and predictors of provider awareness of CKD in a high risk patient population have not been reported.

While CKD documentation may not wholly reflect the provider's understanding of the individual patient's clinical status, incorporation of this information into clinical decision systems supported by claims data is important. This can be illustrated by the recent Centers for Medicare and Medicaid Services (CMS) contract awarded to Medicare Quality Improvement Organizations (QIO) in ten states in August of 2008 to reduce disparities in the early detection and care of diabetic CKD among Medicare beneficiaries [12]. A fundamental feature of the chronic disease model is using ongoing surveillance information, such as claims data, to develop CKD quality of care reports for providers and identify opportunities to improve care [13]. If these claims-based data are systematically biased or inaccurate due to incomplete documentation then one can anticipate marked inefficiencies in this and similar efforts to reduce the burden of CKD.

The aims of the present study were to explore these issues by examining the prevalence of a documented CKD diagnosis in high risk diabetic and hypertensive patients by primary and subspecialty providers in a health maintenance organization and to determine the nature and magnitude of predictors of CKD documentation.

Our primary findings are that, among patients with an eGFR 10-59 ml/min/1.73 m², documentation of CKD within a managed care organization is low (14.4%) and that CKD documentation is associated with male gender, increasing age, CKD severity, peripheral artery disease and congestive heart failure. In addition, CKD was more likely to be documented among patients with a combination of hypertension and diabetes.

The prevalence of CKD documented in our analysis is similar to the proportion of patients identified by their primary care physician as having CKD (13.9%) and reported in the pre-intervention period of an educational program performed at an outpatient family medicine practice [16].

Our findings suggest that CKD identification within a managed care organization is as low as in the non-institutionalized U.S. adult population. Indeed, studies assessing self-reported physician diagnosed kidney disease awareness among U.S. adults in the National Health and Nutrition Examination Survey (NHANES) found that CKD identification ranged from 2.9% (women with Stage 3 CKD) to 24.3% (individuals with a GFR of 15 to 59 ml/min per 1.73 m² and albuminuria) [8, 11]. Differences in the estimates reported from the NHANES study and our current study are notable since we assessed physician documentation of CKD using claims diagnoses while the NHANES study assessed self-reported CKD awareness by participants.

We found that CKD documentation by providers increased with the presence of hypertension or diabetes in both genders, but was consistently and considerably lower among women. While CKD was documented in one third of men with hypertension and diabetes, less than 20% of similarly affected women were documented to have CKD. The gender difference in CKD identification was even greater among participants without hypertension or diabetes, hypertension only, and diabetes only. Previous findings suggest that women are less likely to be aware of their CKD status, and our results extend these findings by reporting that providers are less likely to recognize CKD among women with a reduced GFR compared with men [8, 11, 17]. It has been suggested that health care providers are prone to overlook CKD in women who may have a serum creatinine value in the normal range, due to decreased muscle mass, but have a decline in GFR. Accordingly, the eGFR predictive equation is recommended to estimate the level of renal function from creatinine concentration [18].

We found that eGFR, hypertension, diabetes, congestive heart failure, peripheral artery disease, statin use, and older age were factors significantly associated with CKD diagnosis documentation. CKD patients with hypertension and/or diabetes were more likely to be identified as having CKD than were CKD participants without diabetes or hypertension.

While current guidelines recommend the screening of individuals with diabetes or hypertension for CKD,[6, 19] we found that patients with hypertension were more likely be documented as having CKD. Reasons for the greater impact of hypertension on CKD identification are unclear. It is possible that the lack or small number of endocrinologists within the HMO may delay CKD identification in diabetic enrollees, as other providers may not be as aware of the American Diabetes Association Guidelines recommending that serum creatinine is measured annually for eGFR in all adult diabetics [18]. The discrepancy between the effect of hypertension and diabetes has been reported in the general population, where a history of hypertension, but not diabetes, significantly increases the likelihood of CKD awareness by almost 3 times (aOR, 2.98; 95%CI, 1.39 to 6.39) [8].

Our study findings are consistent with previous observations that eGFR is a strong predictor of CKD identification [8, 11, 20]. We found a significant trend of CKD documentation with decreasing eGFR. Nevertheless, CKD documentation is only 6%, 14%, 31% and 50% among enrollees with eGFR 50-59, 40-49, 30-39 and 20-29 ml/min/1.73 m², respectively. One third of enrollees with an eGFR of 10-19 ml/min/1.73 m² were not documented as having CKD. This finding suggests CKD identification is poor at both early and advanced stages of CKD.

Although age is a key predictor of CKD, we found that CKD documentation decreased with age [21]. A major obstacle in the identification of CKD with increasing age may be the reliance on serum creatinine measurements as an estimation of eGFR,[22] resulting in an under-identification of CKD among the elderly. Our finding is supported by a recent study of Veterans Administration outpatients 70 years and older, revealing that among CKD Stage 2, 3, and 4 patients, the diagnosis of CKD was documented among only 1.2%, 20%, and 74.6% of patients, respectively, with nephrology consults requested in fewer than 5% of patients [23].

Finally, we found that statin medication use was predictive of CKD documentation, even after adjustment for diabetes, hypertension, and cardiovascular disease. While the predictive nature of statin use has not been previously reported in relationship to CKD care, statin use may be a surrogate marker of better quality of care, and has been proposed as a source of treatment bias in the evaluation of cardiovascular disase outcomes [24, 25]. Interestingly, ACEI use was not significantly associated with CKD identification after adjustment. Although ACEI has been shown to reduce the rate of progression of CKD in both diabetics and non-diabetics, [26–28] and although it is identified that ACEI is more effective when instituted early in the course of the CKD, ACEI treatment prevalence in our cohort of CKD HMO enrollees was poor, as described in other settings [29–31]. Furthermore, the prevalence of ACEI use was only slightly higher among enrollees with identified CKD than unidentified CKD. This suggests that factors other than CKD identification contribute to the low ACEI treatment prevalence observed among CKD patients. Contrary to a previous report which showed that CKD diabetic patients are not more likely to be prescribed ACEI than those without diabetes,[30] in our analysis, diabetics with CKD were nearly three-fold more likely to be prescribed ACEI than non-diabetic CKD members, confirming the finding of a previous HMO-base analysis [31].

This study has several limitations. CKD was determined on the basis of ICD 9 codes, leading to the possibility of misclassification. Although previous validation studies suggest that ICD-9 coding for disease identification has sensitivity and specificity greater than 80%,[32, 33] others suggest lower performance and methodological issues [34, 35]. However, our observations that CKD documentation increased with lower eGFRs, and NSAID prescription use was lower among enrollees with identified CKD, suggest that the use of ICD-CM codes is valid. Second we used ICD9-codes from event encounters which list only the presenting condition for that visit. Thus, in the absence of patient medical record abstraction, the estimate of CKD diagnosis documentation that we report may be an underestimate of the true CKD identification in this population. Third, we lacked data on patient race, and instead, used geocoding for estimating race, which then was used for the eGFR calculation. The validity of geocoding for estimating race has been recently explored. Fremont et al. reported that among 17,500 Medicare enrollees, 92% were successfully identified as either black or other using geocoding [36]. In another study, Glaeser et al also found that geocoding can produce accurate estimates of black race [37]. Experts in health services research suggest that in the absence of direct method, the use of block census geocoding is appropriate [38]. To calculate eGFR in the absence of race, others have assumed that no patients were black or alternatively, have used census data to impute an average constant proportion of black in their study population [9, 20]. To test the impact of estimating race using geocoding, we performed all our analysis using eGFR without race and it did not meaningfully change the nature and magnitude of the predictors. Since our cohort is limited to Kaiser Permanent Georgia, our results may not be generalizable to persons in other geographic settings. Although this study has limitations, it examines a large cohort of CKD patients. Finally, because our analyses applied to CKD members identified between January 1, 1999 through January 1, 2006, it is possible that current (2009) documentation of CKD is higher as awareness of CKD in the general population and/or among medical professionals may have increased thanks to advertising and published reports.

In conclusion, among a large cohort of HMO members, the prevalence of CKD diagnosis documentation is low, and similar to that of the general population. Several important predictors are associated with CKD diagnosis documentation among patients with an eGFR 10-59 ml/min/1.73 m², including age, male gender, severity of renal disease, hypertension, diabetes, congestive heart failure, peripheral artery disease, and statin use.

Emphasis on early detection of kidney disease among HMO members may slow the progression and complications of CKD, therefore strategies aimed to increase CKD early detection are warranted [19, 39, 40]. In addition, efforts that include focusing on patients in whom CKD identification is poor, such as women, and increasing the use of medications associated with a reduction in CKD progression may limit the number of patients who experience morbidity and mortality related to advanced chronic kidney disease.