Lab-based and diagnosis-based chronic kidney disease recognition and staging concordance

Chronic kidney disease (CKD) is a costly and increasingly prevalent condition that results in poor health outcomes for many individuals [1]. Documented CKD in Medicare beneficiaries has steadily increased from 2.7% in 2000 to 10.0% in 2011 and 11.0% in 2014, with stage 3 CKD being the fastest growing subgroup [2]. Prior work has shown disparities in CKD prevalence, such that Black or elderly individuals have increased risk for CKD development and progression [2‐6].

However, CKD prevalence is likely underestimated [7, 8], especially by stage, with over one-third of beneficiaries with documented CKD lacking specific staging [2]. Moreover, estimates of CKD prevalence are typically based on International Classification of Diseases (ICD)-9 codes, which have low sensitivity and do not identify a majority of individuals with CKD [9‐13] compared to laboratory-based measurement of kidney function.

A limited number of studies have assessed patterns of both CKD-related labs and diagnoses. Winkelmayer and colleagues examined lab values and diagnoses of 1852 Medicare beneficiaries hospitalized for myocardial infarction in 1999–2000 in Pennsylvania and found the sensitivity of a range of diagnosis-based identification strategies peaked at 20.7%, while specificity and positive predictive values were consistently above 96% [9]. In a small cohort of patients receiving laboratory testing in the Midwest in 2002–2003, Stevens and colleagues found 11% sensitivity and 96% specificity of diagnostic codes compared to lab-indicated CKD [7]. Ferris and colleagues compared medical chart, laboratory reported, and diagnostic codes of hospitalized patients in a single center and found diagnostic code sensitivity of 50 and 88% specificity [8]. In the cohort of Medicare beneficiaries in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study, only 15.5% of patients with lab-identified CKD had Medicare claims with a CKD diagnosis [10]. These studies demonstrate poor identification of CKD from diagnostic coding under specific clinical contexts, but laboratory values have not been comprehensively linked to Medicare claims. Further, limited research has focused on identification of factors independently associated with recognition of CKD or concordance of lab and diagnosis-based CKD identification.

With the availability of 2011 laboratory results linked to Medicare fee-for-service (FFS) claims [11], we can examine the potential value of improving upon population surveillance of CKD in two ways. First, we can identify the extent of clinically unrecognized CKD in a large cohort of Medicare beneficiaries from 10 states in 2011 and patient factors associated with clinical recognition. Given the prior literature noted above, we expected to find significant under-recognition of CKD from claims-based diagnosis codes. Second, we can examine the concordance in CKD staging as reported from lab-based estimated glomerular filtration rates (eGFR) sources and ICD-9 diagnosis codes among those with CKD identified from both sources. We expected to find similar CKD staging from both data sources (i.e. concordance) or a later CKD stage in lab data than diagnosis data. We also examined factors associated with concordance in CKD staging in the two sources.

Results from this analysis provide insights about approaches to CKD recognition, which have implications for the value of comprehensive laboratory data for population surveillance and management. Prior work has demonstrated that CKD is under-recognized in non-Medicare populations, but the extent of under-recognition in Medicare beneficiaries is important to evaluate because 18.5% of Medicare beneficiaries were diagnosed with CKD in 2015 [14]. CKD is costly and puts older adults at risk for ESRD, and Medicare does not obtain laboratory results despite being the largest payer of laboratory testing in the United States. Given significant disparities in the prevalence and management of CKD [4, 15‐17], lab-based identification of CKD can improve recognition and may help mitigate known disparities in CKD care. Further, as Centers for Medicare and Medicaid Services (CMS) integrates social determinants of health into value-based purchasing models (including Black race, Hispanic ethnicity, dual eligibility, disability and rural residence) [18], identifying and addressing disparities in these factors is increasingly important.

This is the first study to examine the extent to which CKD is clinically unrecognized via diagnoses in a cohort of Medicare FFS beneficiaries with lab-identified CKD by examining concordance between CKD stages reported from lab values and ICD-9 diagnoses in Medicare FFS claims. Such a comparison has not been possible heretofore because lab results had never been linked to FFS claims. This analysis found that only 11.8% of beneficiaries with lab-identified CKD had a corresponding diagnosis despite having lab values indicating CKD, which is consistent with prior work in non-Medicare populations that CKD is under-diagnosed [22, 23]. While a majority of beneficiaries with clinically unrecognized CKD had earlier stages of disease, there was a large percentage (37.8%) with unrecognized moderate to severe CKD (stages 3–5) - levels of kidney function for which intervention could slow progression or enable timely preparation for renal replacement therapy.

Despite epidemiologic evidence of worsened outcomes among racial/ethnic minorities with CKD compared to white counterparts [5, 24], our study found CKD recognition via diagnostic codes to be higher among Hispanic and non-Hispanic black beneficiaries than non-Hispanic whites. Others have similarly noted recognition of CKD to be higher among minorities, possibly due to providers’ heightened recognition of risk among individuals with higher prevalence of CKD risk factors (e.g., diabetes and hypertension) [25]. However, recognition alone, without concomitant adherence to CKD guideline recommendations, has not improved clinical outcomes among high risk groups [25, 26]. Further, our study found CKD recognition to be lower among elderly beneficiaries compared to non-elderly beneficiaries. While distinguishing between true kidney dysfunction and age-related renal decline remains challenging, overall CKD prevalence rises in parallel with age [27]. Therefore, as life expectancies continues to lengthen, efforts must be directed toward attenuating the sustained increased risk of adverse clinical outcomes, such as cardiovascular disease, progression to end-stage renal disease (ESRD), and death in the growing elderly population with CKD compared to those without CKD [2, 28].

Several factors may partly explain suboptimal provider recognition of CKD, including lack of provider understanding of CKD and CKD risk categories, and poor knowledge of appropriate CKD management [29, 30]. Despite availability of practice guidelines, providers are unfamiliar with them or prioritize management of co-morbid conditions over that of CKD [30‐32]. Although the automated reporting of eGFR modestly increased overall CKD recognition, recognition among high risk Black or older patients remained unaffected [33]. Our study found that being seen by a nephrologist in the prior year (2010) was strongly associated with having CKD clinically recognized in 2011, which is consistent with prior studies that found higher rates of appropriate testing, medication management, and preparation for renal replacement therapy among individuals engaged in nephrology care (versus not-engaged) [34‐36]. It is also possible that patients had competing clinical demands, limiting time and sufficient attention with providers to enable CKD diagnosis. Early identification and management of CKD is paramount to improving outcomes, particularly among those at highest risk of progression, and provides an opportunity for earlier patient education, avoidance of nephrotoxins and acute kidney injury, implementation of behavioral and medication management strategies to slow progression, and referral to specialist care [37‐39].

In the analysis of concordance between lab-based and diagnosis-based indicators of CKD, we found that most beneficiaries had similar CKD staging. We also found that older or male beneficiaries were significantly more likely to have indicator concordance than those < 65 or females, respectively. This latter finding is of particular interest because nationally, females constitute the majority of Medicare FFS beneficiaries [40] and have higher national prevalence estimates of pre-dialysis CKD compared to males. However, recognition and appropriate documentation of CKD in females is worse than in males [41, 42], suggesting a gender disparity in CKD care. Interestingly, there was no relation between race/ethnicity and concordance after adjustment, highlighting the potential for lab-based identification of disease to attenuate, to some extent, social influences historically associated with provider-driven disparities in diagnosis-based recognition of disease [43]. Nephrologists were also a significant predictor of concordance, such that beneficiaries seeing a nephrologist in 2010 had 25% higher odds of concordance in 2011 than those without a nephrologist visit.

This work demonstrates that laboratory data offer three distinct opportunities for identifying CKD on a population basis. First, serum creatinine values converted to eGFR values can identify CKD that is clinically unrecognized because it is not diagnosed. In population surveillance efforts, the addition of laboratory data will enable appropriate recognition of CKD. Further, lab-based identification can avoid under-treatment of beneficiaries with advanced CKD who are at risk for adverse health events and significant health care utilization. Second, lab-based identification of CKD can more accurately indicate CKD stage when providers use the diagnosis codes for CKD of unknown stage. This was less common than clinically unrecognized CKD, but is still important for identifying patients in need of closer management. Third, lab-based CKD stage was concordant with diagnosis-based CKD in the vast majority of patients identified in both data sources. When there was discordance, lab data often indicated a higher stage of CKD than diagnoses from Medicare claims. This may reflect providers not updating billing documentation, despite actually observing disease progression and managing risk, or initial CKD identification by providers without subsequent recognition of CKD progression.

There are several limitations that must be acknowledged. We cannot say with absolute certainty that lack of a diagnosis code indicates unrecognized CKD; it is feasible that there is clinical recognition that does not rise to the level of a formal diagnosis code but is notated elsewhere on a patient’s medical record problem list. In fact, receipt of a nephrology consultation is likely evidence of CKD recognition, although this cannot be truly confirmed without knowledge of the indication for referral. These results strongly suggest that use of ICD-9 based diagnosis for identifying patients with CKD is questionable to the point of being invalid. Further, our results may have limited generalizability to Medicare FFS beneficiaries outside of these 10 states examined or to years subsequent to 2011. The ability to examine disparities in CKD staging was limited because the lab results data was available from only one national laboratory vendor, albeit the vendor with the largest market share in these 10 states [11]. However, it should be noted that the sample size in the concordance analysis exceeds the prior two studies [9, 10] by a factor of 100. These results do not reflect all laboratory data however this single laboratory vendor’s market share for Medicare laboratory tests in these 10 states in 2011 ranged from 9 to 51%. However, data capture would not converge to 100% because of in-house laboratory and point-of-care laboratory testing reimbursed by Medicare. Serum creatinine values are only available in this sample if a physician ordered the test, a Medicare FFS beneficiary followed through and had the test done, and the test was processed by a clinical facility affiliated with this single lab vendor [44]. Similarly, the use of just one qualifying UACR measurement may misclassify individuals as having CKD Stage 1 instead of a transient urinary alteration, as can be seen with a urinary infection or acute kidney injury, rather than persistent albuminuria (> 3 months) consistent with CKD. Further, our list of included Medicare claims covariates likely do not capture all factors that may explain variation in CKD diagnosis, such as medication use patterns or iatrogenic factors. Last, staging was based on most recent creatinine values which may represent transient fluctuations in kidney function from superimposed acute kidney injury on CKD rather than stable CKD.

These results have research and policy implications. Possible linkage of longitudinal lab data to Medicare claims would allow exploration of the continuum of CKD progression and identify targets for early interventions for CKD disease management. As the single largest payer of laboratory tests in the United States [45], linkage of Medicare with lab data facilitates population surveillance. Further, investment in preventive strategies earlier in CKD care may offset the substantial downstream costs by slowing or preventing progression to ESRD, or by prioritization of management strategies and resources for individuals at greatest risk of adverse outcomes based on more than eGFR alone [46]. These results indicate that the availability of lab data has tremendous value for population surveillance of CKD by significantly reducing the under-recognition of CKD from claims-based diagnoses that are readily available but insufficient to the task.