Introduction

Chronic diseases represent a significant and escalating public health challenge in the United States, with a substantial portion of the population grappling with the burden of multiple co-occurring health conditions, a phenomenon known as multimorbidity. The simultaneous presence of diabetes, kidney disease, and depression constitutes a particularly impactful multimorbidity cluster, as these conditions can interact synergistically, leading to worsened individual health trajectories and placing considerable strain on healthcare systems through heightened disability rates, increased polypharmacy, and substantial economic costs (Mossadeghi et al., 2023). While traditional research often focuses on individual diseases in isolation, a growing body of evidence highlights the interconnected nature of multimorbidity, demonstrating that the combined effect of multiple conditions can amplify mortality risk and diminish quality of life to a degree exceeding the sum of their individual impacts (Xue et al., 2025; Mossadeghi et al., 2023).

Importantly, these patterns are not randomly distributed but are systematically shaped by social determinants of health, such as access to healthcare, education, income, and neighborhood conditions. This clustering of chronic conditions often reflects underlying structural inequalities including economic deprivation and healthcare inaccessibility that disproportionately burden marginalized populations across the life course.

Patterns and Disparities in Multimorbidity

These patterns reflect not only behavioral risk factors but also systemic racism and structural inequality, which limit access to timely diagnosis, quality care, and health-promoting environments for minoritized groups. For instance:

Racial/Ethnic Disparities: Studies have consistently shown that Non-Hispanic Black and Hispanic populations experience higher rates of complex multimorbidity (e.g., cardiovascular-metabolic-mental health clusters) and tend to develop these conditions at an earlier age compared to Non-Hispanic White adults (Alshakhs et al., 2022).

Socioeconomic Factors: Consistent with fundamental cause theory, socioeconomic factors significantly influence multimorbidity. Lower socioeconomic status means reduced access to key resources (money, knowledge, social capital), increasing susceptibility to various diseases (Xue et al., 2025; Mossadeghi et al., 2023). Empirical data show that lower education and income are consistently associated with a higher burden of multimorbidity, often due to poor healthcare access and underdiagnosis. For example, a 2017–2018 NHANES analysis revealed that adults below the poverty line had 1.6 times greater odds of multimorbidity compared to wealthier individuals (Mossadeghi et al., 2023; Jindai et al., 2016).

Gender and Age: Women and older adults, particularly those aged 75 years and above, demonstrate stronger links between multimorbidity and functional limitations. Furthermore, research suggests that women may experience a more rapid progression towards disability in the context of multiple chronic conditions (Jindai et al., 2016).

NHANES as a Critical Data Source

The National Health and Nutrition Examination Survey (NHANES) provides an invaluable resource for this type of analysis due to several key features:

Nationally Representative Design: The survey employs a sophisticated stratified, multistage probability sampling design, ensuring that findings can be generalized to the non-institutionalized civilian population of the United States (Mossadeghi et al., 2023; Jindai et al., 2016).

Comprehensive Data Collection: NHANES gathers a rich array of data, including self-reported diagnoses, detailed physical examinations (e.g., body mass index, blood pressure measurements), and laboratory tests (e.g., glycated hemoglobin A1c, estimated glomerular filtration rate), which allows for the robust assessment and validation of chronic conditions (Xue et al., 2025; Mossadeghi et al., 2023).

Focus on Health Disparities: The survey includes detailed information on key sociodemographic variables such as race/ethnicity, education level, and income, enabling in-depth analyses of health inequities across different population subgroups (Mossadeghi et al., 2023; Alshakhs et al., 2022).

Research Gaps and Objectives

Despite the growing body of literature on multimorbidity, there remains a need for studies that specifically examine how particular multimorbidity clusters, such as the co-occurrence of diabetes, kidney disease, and depression, vary across key demographic factors like race, education, and gender over time. This study aims to address this critical gap in the existing research by utilizing data from three recent cycles of NHANES (2003–2004, 2013–2014, and 2021–2023) to:

  1. Quantify the prevalence trends of the diabetes-kidney disease-depression multimorbidity cluster across the study period.

  2. Analyze disparities in the prevalence of this cluster across different racial and ethnic groups, educational attainment levels, and genders.

  3. Provide insights that can inform the development of targeted interventions aimed at high-risk population groups (Alshakhs et al., 2022).

Missing Data Analysis

A comparison of complete and missing cases revealed that individuals with missing data tended to be younger on average (mean age: 9.11 vs. 51.48 years in complete cases). The diabetes rate was substantially higher in the complete cases group (12.55%) compared to the missing data group (0.33%). Although there was no significant difference in depression rates, kidney condition rates were somewhat higher in the missing cases (7.69% vs. 3.50%). These findings suggest that missingness is not random and may be more prevalent in younger individuals or those without chronic conditions such as diabetes and kidney disease (Rubin, 1976; Schafer & Graham, 2002).

Top Variables with Highest Missingness in NHANES Data
Variable Missing Count Missing (%)
ciqd079 32230 100
ciqd104 32230 100
ciqd107 32230 100
ciqdphg 32230 100
ciqdphh 32230 100
ciqdpk 32230 100
ciqd056 32229 100.
ciqd101 32229 100.
ciqdpf 32229 100.
ciqdphe 32229 100.

Given this pattern, multiple imputation (MI) was not implemented for the following reasons:

  • Low Missing Data Rate: The overall proportion of missing data for key variables was relatively low (i.e., below 10% for most variables), which is generally considered acceptable for complete-case analysis (Schafer & Graham, 2002).

  • Non-Critical Variables: The missingness was observed primarily in non-critical variables, such as certain demographic details, which are not essential to the key outcomes of interest (e.g., diabetes, depression, and kidney disease) (Acock, 2005).

  • Complete Case Robustness: A comparison of complete cases and missing data showed no significant bias in the distributions of key outcome variables, suggesting that excluding incomplete cases would not lead to substantial loss of information (Acock, 2005).

  • Transparency in Analysis: Since the missingness did not meet the criteria for systematic or substantial missing data (e.g., no evidence that missingness is dependent on key predictors like disease status), complete-case analysis was deemed more appropriate. This approach is transparent and commonly used when missing data is non-informative, and the benefits of imputation (e.g., bias reduction) are minimal (Rubin, 1976; Schafer & Graham, 2002).

Based on these considerations, the decision was made to proceed with complete-case analysis for this study, ensuring robust findings while avoiding the complexity and assumptions associated with multiple imputation.

Research Method

Data Source and Study Population

Data for this study were derived from the National Health and Nutrition Examination Survey (NHANES), a program conducted by the National Center for Health Statistics (NCHS) as part of the Centers for Disease Control and Prevention (CDC). NHANES employs a complex, multistage probability sampling design to provide nationally representative estimates of the civilian, non-institutionalized U.S. population. This analysis included data from three two-year survey cycles: 2003–2004, 2013–2014, and 2021–2023, capturing two decades of changes in chronic conditions and health disparities (Mossadeghi et al., 2023; Jindai et al., 2016).

The analytic sample consisted of adults aged 20 years and older with complete data on key variables of interest: diabetes status, depression screening scores, self-reported kidney disease, and sociodemographic covariates including race/ethnicity, gender, and education level. Respondents with missing values for any of these variables or with zero survey weights were excluded, yielding a final sample size of 16,376 (Acock, 2005; Schafer & Graham, 2002).

Variable Definitions

Health Conditions

  • Diabetes: Defined as a binary variable (1 = Yes, 0 = No), based on a positive response to the question, “Ever told you had diabetes?” (DIQ010 = 1).

  • Depression: Defined as a binary variable (1 = Likely Depression, 0 = Not Likely Depression) using the Patient Health Questionnaire-9 (PHQ-9). Participants scoring ≥10 on the PHQ-9 were classified as having likely depression.

  • Kidney Disease: Defined as a binary variable (1 = Yes, 0 = No), based on a positive response to the question, “Ever told by a doctor or other health professional that you had kidney disease?” (KIQ022 = 1).

Multimorbidity Measures

Multimorbidity was operationalized in two ways:

  • A continuous measure capturing the proportion of the three chronic conditions present (range: 0–1), calculated by summing binary indicators of diabetes, depression, and kidney disease and dividing by three.

  • A binary indicator coded as 1 if at least one of the three conditions was present, and 0 otherwise. This classification is consistent with approaches used in prior NHANES-based multimorbidity research (Alshakhs et al., 2022; Xue et al., 2025).

Sociodemographic Variables

  • Race/Ethnicity: Categorized into five mutually exclusive groups using the RIDRETH1 variable: Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, and Other. This classification aligns with prior studies of disparities in multimorbidity (Mossadeghi et al., 2023; Alshakhs et al., 2022).

  • Gender: Defined using the RIAGENDR variable and categorized as Male or Female.

  • Education: Categorized based on DMDEDUC2 into: Less than High School (codes 1 or 2), High School Graduate (code 3), Some College (code 4), and College Graduate (code 5), reflecting socioeconomic stratification in prior multimorbidity studies (Xue et al., 2025).

Survey Weights

To account for NHANES’s complex sampling design and ensure population-level estimates, I applied the Mobile Examination Center (MEC) two-year survey weights (WTMEC2YR). As three survey cycles were pooled, the weights were divided by three to generate combined-cycle average weights, following recommended practice for pooled NHANES analysis (Jindai et al., 2016).

Statistical Analysis

All statistical analyses were conducted using R (version 4.4.2). The survey package (Lumley, 2023) was used to specify the complex design structure, including strata (SDMVSTRA) and primary sampling units (SDMVPSU), via the svydesign() function.

I computed weighted descriptive statistics to estimate the prevalence of multimorbidity and its components across subgroups:

  • Weighted means and standard errors for the continuous multimorbidity score by race/ethnicity using svyby().

  • Weighted proportions and standard errors for binary multimorbidity (Yes/No) by gender and survey year (2003, 2013, 2023), also using svyby().

Results from these analyses are reported in both tabular and graphical formats in the “Results” section.

Results

Table 1: Racial and Ethnic Differences in Multimorbidity Clusters

Average Normalized Cluster Count by Race/Ethnicity
Race/Ethnicity Average Score SE
Mexican American 0.042 0.004
Other Hispanic 0.040 0.006
Non-Hispanic White 0.039 0.001
Non-Hispanic Black 0.056 0.003
Other 0.045 0.004

Table 1 displays the average normalized multimorbidity cluster scores, reflecting the mean proportion of co-occurring diabetes, kidney disease, and depression among racial and ethnic groups. Non-Hispanic Black individuals exhibited the highest average score (M = 0.056, SE = 0.003), followed by individuals identifying as Other (M = 0.045, SE = 0.004) and Mexican Americans (M = 0.042, SE = 0.004). Non-Hispanic White participants had a slightly lower average score (M = 0.039, SE = 0.001), while Other Hispanic participants showed the lowest average cluster score (M = 0.040, SE = 0.006).

These differences suggest a disproportionate burden of clustered chronic conditions among Non-Hispanic Black adults, which aligns with previous findings documenting racial and ethnic disparities in multimorbidity and their associations with systemic barriers in healthcare access, socioeconomic status, and other social determinants of health (Alshakhs et al., 2022; Mossadeghi et al., 2023).

Visual Summary

Racial and Ethnic Differences in Multimorbidity Clusters

Figure 1 presents a bar chart displaying the average normalized cluster count, representing the co-occurrence of diabetes, kidney disease, and depression, across racial and ethnic groups. Non-Hispanic Black individuals exhibited the highest average cluster norm, followed by individuals identifying as Other and Mexican American. Non-Hispanic White and Other Hispanic groups had lower average cluster scores, with Non-Hispanic White participants showing the lowest overall burden. Error bars indicate standard errors, with visibly wider intervals among smaller population groups such as Other Hispanic and Other race categories, suggesting greater variability.

These findings reinforce the disproportionate burden of multimorbidity among Non-Hispanic Black adults and highlight variability across racial and ethnic groups in the U.S. population. The observed trends are consistent with prior literature identifying structural health disparities, intersectional risk factors, and clustering of chronic disease across minoritized communities (Alshakhs et al., 2022; Mossadeghi et al., 2023).

Figure 1. Average Normalized Cluster Count by Race/Ethnicity. Bars represent the mean proportion of co-occurring diabetes, kidney disease, and depression. Error bars indicate ±1 standard error.

Educational Differences in Multimorbidity Clusters

Figure 2 illustrates the same cluster norm, this time across education levels. A clear gradient is visible: individuals with less than a high school education had the highest average cluster count, while college graduates had the lowest. This pattern reflects the strong association between lower educational attainment and increased risk of chronic disease accumulation. The trend supports prior findings that link education to health through mechanisms such as health literacy, income, healthcare access, and preventive behaviors (Xue et al., 2025).

Figure 2. Average Normalized Cluster Count by Education Level. Bars represent the mean proportion of co-occurring diabetes, kidney disease, and depression. Error bars indicate ±1 standard error.

Discussion

These disparities reflect more than individual behavior - they are the result of institutionalized inequities rooted in systemic racism, residential segregation, and chronic underinvestment in communities of color. The results affirm the need to interpret health disparities through an intersectional lens, recognizing how race, income, and other axes of identity compound disadvantage and shape disease risk.

This study examined the co-occurrence of diabetes, depression, and kidney disease - three impactful chronic conditions across major demographic groups using nationally representative NHANES data from 2003 to 2023. The results reveal persistent and widening disparities in multimorbidity clusters, particularly among Non-Hispanic Black adults and in more recent survey years.

Racial and Ethnic Inequities

Non-Hispanic Black individuals consistently exhibited the highest average multimorbidity burden. These findings align with extensive research demonstrating that systemic inequities, including structural racism, residential segregation, and unequal access to preventive care, play foundational roles in chronic disease accumulation (Alshakhs et al., 2022; Jindai et al., 2016). These data underscore the need for intersectional public health frameworks that account for the compounded effects of race, socioeconomic status, and healthcare access in the prevention and management of chronic diseases.

Public Health Implications

Effective interventions must extend beyond clinical settings to address upstream social determinants of health, including food security, housing stability, and transportation access - factors deeply intertwined with health inequities. Policies grounded in structural competency and guided by equity-oriented frameworks can facilitate a shift from reactive, fragmented care to proactive, integrated, and community-rooted strategies.

The rising burden of multimorbidity, particularly among racial and ethnic minorities, socioeconomically disadvantaged groups, and men, underscores the urgency of designing tailored, culturally competent interventions. These strategies must account for both physical and mental health needs, recognizing the interconnected nature of chronic conditions and the structural barriers that perpetuate them.

Limitations

Several limitations should be noted. First, the analysis relies on self-reported data for diabetes and kidney disease, which may lead to underreporting due to lack of diagnosis or recall bias. Second, while PHQ-9 is a validated measure of depression, it reflects symptoms over the past two weeks and may not capture chronic or subclinical depression. Lastly, the use of complete-case analysis, although justified due to low missingness and minimal bias, may still exclude systematically different subpopulations.

Conclusion

This study highlights persistent and widening disparities in multimorbidity involving diabetes, depression, and kidney disease across racial, educational, and gender groups in the United States. Using nationally representative NHANES data from 2003 to 2023, the analysis revealed that Non-Hispanic Black adults and individuals with lower educational attainment bear a disproportionate burden of this high-risk cluster. These findings underscore the need for targeted, equity-focused public health strategies that address both medical and social determinants of health. Future research should continue to explore intersectional factors driving multimorbidity and evaluate the impact of policy interventions designed to reduce chronic disease disparities.

References

  1. Bijan Mossadeghi, Caixeta, R., Ondarsuhu, D., Luciani, S., Hambleton, I. R., & Hennis, A. J. M. (2023, May 15). Multimorbidity and social determinants of health in the US prior to the COVID-19 pandemic and implications for health outcomes: A cross-sectional analysis based on NHANES 2017–2018. BMC Public Health, 23, Article 927. https://doi.org/10.1186/s12889-023-15768-8

  2. Xue, Q., Zhang, S., Yang, X., Zhang, Y.-B., Dong, Y., Li, F., Li, S., Wu, N., Yan, T., Wen, Y., Yang, C.-X., Wu, J. H. Y., Pan, A., Yang, Y., & Pan, X.-F. (2025, April 3). Multimorbidity patterns and premature mortality in a prospective cohort: Effect modifications by socioeconomic status and healthy lifestyles. BMC Public Health, 25, Article 586. https://doi.org/10.1186/s12889-025-22216-2

  3. Alshakhs, M., Jackson, B., Ikponmwosa, D., Reynolds, R., & Madlock-Brown, C. (2022, June 11). Multimorbidity patterns across race/ethnicity as stratified by age and obesity. Scientific Reports, 12, Article 9843. https://doi.org/10.1038/s41598-022-13733-w

  4. Jindai, K., Nielson, C. M., Vorderstrasse, B. A., & Quiñones, A. R. (2016, November 3). Multimorbidity and functional limitations among adults 65 or older, NHANES 2005–2012. Preventing Chronic Disease, 13, E151. http://dx.doi.org/10.5888/pcd13.160174

  5. Acock, A. C. (2005). Working with missing values. Journal of Marriage and Family, 67(4), 1012–1028. https://doi.org/10.1111/j.1741-3737.2005.00191.x Rubin, D. B. (1976).

  6. Inference and missing data. Biometrika, 63(3), 581–592. https://doi.org/10.1093/biomet/63.3.581

  7. Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7(2), 147–177. https://doi.org/10.1037/1082-989X.7.2.147