National Trends in the Prevalence of Chronic Kidney Disease Among Racial/Ethnic and Socioeconomic Status Groups, 1988-2016

Key Points Question What were the trends in prevalence of chronic kidney disease among all major sociodemographic groups in recent years? Findings In this repeated cross-sectional study, gaps in chronic kidney disease prevalence across levels of socioeconomic status persisted over 28 years. Chronic kidney disease prevalence almost doubled among Mexican American persons between 2003-2004 and 2015-2016; other racial/ethnic and socioeconomic groups did not experience appreciable increases in the prevalence of chronic kidney disease. Meaning There is a need to renew efforts to effectively mitigate persistent disparities in the prevalence of chronic kidney disease.


Introduction
Chronic kidney disease (CKD) is associated with a disproportionately high risk of cardiovascular disease, progression to kidney failure requiring dialysis or transplantation, acute kidney injury, and death. [1][2][3][4][5][6] Chronic kidney disease affects about 15% of the US adult population. 7 Previous research has examined overall trends in CKD prevalence in the United States and has found no appreciable increase in CKD prevalence in recent years. 8 Although this overall trend is encouraging, it is unclear whether CKD rates have stabilized across major sociodemographic groups for whom kidney failure rates have been high, such as those defined by race/ethnicity and socioeconomic status.
Moreover, Healthy People 2010 and 2020, the US national blueprint for public health goals, aimed to eliminate health disparities, including those associated with kidney disease, and achieve health equity. To examine progress toward this aim, the assessment of disparities in CKD prevalence is pivotal. Healthy People 2020 explicitly highlighted the importance of assessing health disparities by major sociodemographic factors in the US population in association with chronic conditions. 9 In this study using a large, nationally representative sample of the US population, we aimed to investigate trends in CKD prevalence by race/ethnicity and socioeconomic status from 1988 to 2016 and investigate whether disparities persist.

Study Population
The National Health and Nutrition Examination Surveys (NHANES) 10 are nationally representative surveys of the noninstitutionalized US population. In brief, NHANES III was conducted from 1988 to 1994; beginning in 1999, NHANES collected data continuously and released it in 2-year data cycles. In this study, we examined data from NHANES III (1988)(1989)(1990)(1991)(1992)(1993)(1994) through NHANES 2015-2016.
Participants 20 years or older with available information on their serum creatinine measurements, race/ethnicity, and socioeconomic status were included for analysis. Participants with CKD stage 5 (estimated glomerular filtration rate [eGFR] <15 mL/min/1.73 m 2 ) were excluded owing to the small sample size and lack of information on participants' dialysis history (eTable 1 in the Supplement). The research ethics boards of the National Center for Health Statistics approved NHANES. All participants provided written informed consent. Our report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. 11

Sociodemographic Factors
Race/ethnicity and socioeconomic status (commonly assessed using educational level and poverty income ratio 12 ) were considered as major sociodemographic variables of interest. Information on these 3 factors was self-reported. Race/ethnicity was categorized into 4 groups: non-Hispanic white, non-Hispanic black, Mexican American, and other. Educational level was categorized into 3 categories: less than high school (low), high school or equivalent (medium), and more than high school (high). 13 Poverty income ratio (ratio of total household income divided by federal poverty income threshold corrected for household size and inflation) was categorized into less than 1.0 (low income), 1.0-3.9 (middle income), and 4.0 or more (high income). 14

Chronic Kidney Disease
The prevalence of stage 3 and 4 CKD (ie, eGFR 15-59 mL/min/1.73 m 2 ) 15 was the main outcome of interest. Glomerular filtration rate was estimated from serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration equation. 16 The calibrations recommended by the NHANES for serum creatinine measurements across time periods was used. 17

Statistical Analysis
Statistical analysis was conducted from May 1, 2017, to April 6, 2020. Following the NHANES' data analysis protocol, our analyses included the recommended sample weights. Prevalence estimates were adjusted for age, sex, and race/ethnicity (except when analyzing race/ethnicity). In adjusted analyses, restricted cubic splines were used to flexibly model trends over time. To allow changes during the study period  and give the needed flexibility in shape, we used 3 predetermined knots at years 2000, 2004, and 2007. 8 Temporal trends in CKD prevalence, with stratification by race/ethnicity, educational levels, and poverty income ratio, were examined with marginal-effect estimation. 18 In this method, for example, the estimate of CKD prevalence across racial/ethnic groups over time was obtained by holding age and sex fixed at their observed values and varying race/ethnicity and survey year. Predicted probabilities (of having CKD for each year and for each person) obtained from fitted models were averaged to obtain the mean CKD risk for a given racial/ethnic group and year. The standard Wald tests from the sample survey logistic regression analysis in Stata, release 14 (StataCorp LLC) were used to calculate P values. To test for interactions of temporal trends across sociodemographic factors, we tested the combined interaction of the spline terms with race/ethnicity, educational level, and poverty income ratio. Values for graphing the spline fits and their SEs were derived using the "margins" command in Stata, release 14. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05.
To assess the robustness of our results, a number of sensitivity analyses were performed. First, we examined trends in the prevalence of CKD redefined to include eGFR of 15 to 59 mL/min/1.73 m 2 or a one-time urinary albumin to creatinine ratio of 30 mg/g or more. Second, we also assessed trends when using a more stringent cutoff to define CKD (ie, eGFR, 15-44 mL/min/1.73 m 2 ). Third, trends in CKD prevalence were examined when additionally adjusting for diabetes and systolic blood pressure. Fourth, in cases in which a sociodemographic characteristic showed an interaction with time, we excluded the subgroup(s) that showed different trends and repeated the test for interaction. Characteristics of the study population are presented by survey period ( Table 1). 16 In later survey years, the mean (SE) age was higher (1988)(1989)(1990)(1991)(1992)(1993)(1994)44 The proportion of the population with less than a high school education was also lower in later years (1988-1994, 23.7%; 2015-2016, 13.5%). Demographic characteristics of the study population by racial/ethnic groups are presented by survey period in eTable 2 in the Supplement. In later survey years, the mean age was higher, and there was a reduction in the percentage of those with less than a high school education across all racial/ethnic groups. Over the years, the trend in the proportion of those living below the poverty line was similar across racial/ethnic groups.
Crude CKD prevalence differed significantly by race/ethnicity (P = .04 for interaction). Among non-Hispanic white and non-Hispanic black persons, CKD prevalence increased between 1988-1994     Overall, % (95% CI) 4.6 (4.0-5. In 1988-1994, individuals with less than high school education (9.1%) and those with low incomes (5.5%) had the highest CKD prevalence, while those with more than high school education (2.6%) and high income (3.0%) had the lowest CKD prevalence (P < .001 for difference in CKD prevalence between low and high educational levels and between low and high income). This trend was similar in 2015-2016.

Adjusted CKD Prevalence
The Trends in adjusted CKD prevalence among adults with less than high school education, high school education, and more than high school education followed the trend in CKD prevalence for the overall population (P = .38 for interaction), with similar results for low-income, middle-income, and high-income adults (P = .38 for interaction). The differences in CKD prevalence between those with less than high school education and those with more than high school education (Figure,

Discussion
In this national study, we examined temporal trends in CKD prevalence for major US  19 Observed results for Mexican American persons are unlikely to be due to racial/ethnic differences in the burden of diabetes and hypertension, as there were statistically significant differences in CKD prevalence even in analyses adjusting for these conditions. Migration patterns (immigration and emigration) of Mexican American persons may be a factor associated with the results. With the increasing racial/ ethnic diversity of the US population, there is a need to identify and address the underlying causes of this trend.
Our observation of higher CKD prevalence at any time point among persons with low educational or low income levels updates the findings of prior work. 20,21 However, the observation demonstrates consistent gaps in CKD prevalence across these groups over nearly 3 decades.
Although the aim of this study was not to elucidate causes, it is conceivable that consistent inequality in CKD prevalence could be in part owing to health-promoting interventions not targeting specific at-risk groups or not being specifically designed to be capable of eliminating or reducing disparities. [22][23][24][25] For instance, clustering of interventions (eg, bicycle-supporting infrastructure promoting physical activity) in neighborhoods whose residents have high median incomes and/or high employment rates may worsen health disparities experienced by residents of less advantaged communities. 26 Our study has several research and public health implications. Our findings highlight the need to focus on the increasing CKD prevalence in Mexican American individuals. There may also be a need for enhanced efforts aimed at narrowing racial/ethnic and socioeconomic disparities in CKD prevalence. Some strategies have proven successful in decreasing rates of CKD among racial/ ethnic minority populations (eg, the Indian Health Service program for lowering rates of end-stage kidney disease among persons with diabetes) 27,28 and individuals with low socioeconomic status (eg, interventions targeting education and early childhood) 29 and should be explored for their effectiveness and broader implementation. Although the overall CKD prevalence has stabilized in recent years, CKD awareness continues to remain very low and is even lower in some age and racial/ ethnic groups than others. 30

Strengths and Limitations
Our study had several strengths. The data are nationally representative, which allows them to be generalizable to the entire US noninstitutionalized adult population. The data were collected using standardized procedures on a wide range of factors, including demographic characteristics and comorbid conditions, and in a uniform manner throughout the study years. In addition, we performed NHANES-recommended calibration of serum creatinine measurements for a comparable assessment of kidney function across survey years. 16 Our study also has some limitations, including a lack of sufficient data on other racial/ethnic groups in the United States, including American Indian, Asian, and Pacific Islander individuals, as well as other subpopulations of Hispanic persons. Thus, we were not able to examine these subgroups.
In addition, because there was only 1 measurement of serum creatinine or albuminuria in NHANES, as well as the use of estimated rather than measured glomerular filtration rate, there was a possibility of misclassification of CKD status. However, misclassification is likely to be nondifferential across survey years, making it unlikely to have resulted in a biased assessment of temporal trends. Also, evidence suggests that there is no appreciable bias in CKD staging based on only the first measurement compared with more than 1 measurement. 31 Our primary conclusion is based on CKD prevalence defined by eGFR alone, which also meant that CKD prevalence was lower than 10%, which is the prevalence when CKD is defined using both eGFR and albuminuria. This finding was owing to the lack of a standardized assay and inaccuracy of assays for albuminuria, which raises the potential for bias, especially when assays are performed years apart and the aim is to assess temporal trends. Because NHANES does not include data on duration and history of CKD, it was not possible to investigate any difference in sociodemographic factors between subgroups of patients with CKD of various duration and history.

Conclusions
The prevalence of CKD in the United States has stabilized in recent years among most racial/ethnic groups but not among Mexican American persons. More important, gaps in CKD prevalence by racial/ ethnic groups and socioeconomic status were largely persistent. To achieve health equity, renewed efforts are needed to effectively mitigate persistent disparities in CKD prevalence.