We conducted an exploratory cross-sectional analysis using the AoU Controlled Tier version 7 data release in the secure, cloud-based AoU Researcher Workbench platform (date of first access for research purposes: September 1, 2023; date of last access: June 16, 2024). Registered researchers can access the data through the AoU Research Hub [7]. This study was exempt from ethical approval for human subjects research as only deidentified data were analyzed.

Using the Cohort Builder tool, we identified all adult participants 18 years of age or older who consented to share electronic health record (EHR) data and had at least one ICD-10 (International Statistical Classification of Diseases, Tenth Revision) code documented for any healthcare purpose (ie, not limited to a chronic condition-related code) to ensure the availability of EHR data (n=242,828). For all participants meeting these eligibility criteria, we extracted relevant ICD-10 codes and the corresponding date of documentation for 58 chronic conditions (Multimedia Appendix 1) cataloged in the Centers for Medicare & Medicaid Services (CMS) Chronic Conditions Data Warehouse [8]. CMS supports a Chronic Conditions Data Warehouse that contains computable phenotypes for “common chronic conditions” plus “other chronic or potentially disabling conditions,” which include additional chronic health, mental health, and substance abuse conditions [9]. We operationalized the presence of a chronic condition as documentation of one or more ICD-10 codes for the respective condition (Multimedia Appendix 1). Due to reliance on ICD-10 codes for both cohort identification and as the source of condition information, we limited participant visits to on or after October 1, 2015, that is, initiation of ICD-10 implementation. Thus, data included in this analysis are from October 1, 2015, to July 1, 2022, that is, latest date available in the version 7 release.

We used the Python programming language in Jupyter Notebook to chronologically order the chronic condition ICD-10 codes by date of documentation for each participant. By distinguishing the onset date of the second unique chronic condition, we were able to identify participants with MCC. We calculated the number of participants with 0, 1, 2, 3, 4, 5, and 6+ chronic conditions in the overall cohort and stratified by age category, that is, early adulthood (18-39 years), middle adulthood (40-49 years), late middle adulthood (50-64 years), late adulthood (65-74 years), and advanced old age (75-89 years) [10,11]. Age categorization supports methodological requirements of analyses, promotes interpretability of results for clinicians and policy makers, and helps capture changes in physical health and social roles throughout adulthood. We used the age at first documentation of the second chronic condition for categorization. We used chi-square tests to assess differences in demographic variables between participants with 0 or 1 chronic condition and participants with MCC. This approach tested for differences in the frequencies of race, ethnicity, gender, marital status, education, health insurance status, and employment status. We used P<.05 to indicate statistical significance. Finally, we calculated the frequency of unique MCC combinations by the number of chronic conditions (ie, 2, 3, 4, 5, and 6+) and age category.

This study analyzed deidentified data from participants enrolled in the AoU Research Program. Participants in the AoU Research Program were recruited either in-person at AoU enrollment sites or on the AoU web-based portal [12]. Participants gave informed consent when they joined the program, agreeing that their anonymized data could be used broadly for biomedical research [12]. Enrollment and data collection of participants in the AoU program is approved by the AoU institutional review board (IRB) [13].

We analyzed the EHR data in the secure, cloud-based AoU Researcher Workbench. All data used in this analysis were deidentified without personal identification including shifted dates for categorization of age. Researchers included in this analysis were registered within the AoU Research Program and the Researcher Workbench, completed the required training for data access, and signed the Data User Code of Conduct. All analyses were performed following AoU policies and guidelines to ensure confidentiality and responsible use of data [14]. Analyses for this study did not require subsequent IRB approval because the study does not constitute research involving human subjects according to the AoU Research Program IRB [12]. The data used in this analysis do not include the elements necessary to identify individuals so the study is not subject to the Common Rule protections 45 CFR part 46 [15].

Approximately 76% (n=183,753) of participants were diagnosed with MCC (Figure 1), with over 40% of participants (n=98,885) diagnosed with 6 or more conditions (Table 1). Of the remaining participants, 11% (n=27,122) were diagnosed with a single chronic condition, and 13% (n=31,953) did not have a chronic condition documented. Overall, participants (Table 2) were on average 49.8 (SD 16.7) years of age and primarily women (n=147,991, 60.9%), White (n=137,039, 54.6%), not Hispanic or Latino (n=187,586, 77.3%), married or partnered (n=121,282, 49.9%), highly educated (n=166,815, 68.7% with some college or higher), insured (n=223,842, 92.2%), and not currently employed (n=111,229, 47.7%). Compared to participants diagnosed with 0 or 1 conditions, those diagnosed with MCC were older (age: mean 52.8 years vs 40.4 years). In addition, a higher percentage of participants with MCC were men (n=68,776, 37.4% vs n=19,956, 33.8%), White (n=107,725, 58.6% vs n=29,314, 49.6%), not Hispanic or Latino (n=145,393, 79.1% vs n=42,193, 71.4%), insured (n=170,985, 93.1% vs n=52,857, 89.5%), ever married (n=139,118, 75.7% vs n=39,264, 66.5%), and not currently employed (n=96,339, 52.4% vs n=19,484, 33.0%), with a much greater frequency of retired participants (n=48,565, 26.4% vs n=6025, 10.2%), compared to participants diagnosed with 0 or 1 conditions. The percentage of participants within an age category diagnosed with 6 or more chronic conditions increased with age (Table 1), from 33.78% in early adulthood to 68.04% in advanced old age.

The top 5 most frequently diagnosed MCC combinations in each age category are detailed in Multimedia Appendix 1. The most frequently occurring combination in each age categorization was as follows: early adulthood: anxiety disorders and depression, bipolar disorder, or other depressive mood disorders (n=845); middle and late middle adulthood: fibromyalgia, chronic pain and fatigue, and rheumatoid arthritis or osteoarthritis (n=204 and n=457, respectively); and late adulthood and advanced old age: hyperlipidemia and hypertension (n=200 and n=59, respectively).

In this study, we estimated MCC prevalence in adults living in the United States from early adulthood to advanced old age using ICD-10 codes from participants enrolled in the AoU program. We found that approximately 76% of participants have MCC, with higher prevalence as age increases. The most frequent MCC combinations varied notably by age category. Mental health conditions (ie, anxiety disorders and depression, bipolar disorder, or other depressive mood disorders) were most common in early adulthood, whereas combinations of physical health conditions (eg, fibromyalgia, chronic pain, fatigue, arthritis, hyperlipidemia, and hypertension) predominated in middle and older adulthood. These findings emphasize the value of the AoU program in identifying the MCC burden throughout adulthood. We believe the percentage of adults with MCC in this study is higher than that reported previously due to our robust inclusion of 58 chronic conditions, selected based on the CMS Chronic Conditions Data Warehouse.

Our results highlight distinct patterns in MCC combinations that vary by age category. Given the notable prevalence of MCC across adulthood, there is a continued need for innovative care modalities and public health initiatives to address the multifaceted nature of MCC [16]. MCC are associated with clinical complexity characterized by increased symptom burden, high health care usage and cost, and extensive caregiver strain [17]. The Agency for Healthcare Research and Quality states that care for people with MCC requires alignment among available resources, policies, and the workforce. For every dollar of health care spending in the United States, US $0.71 goes to treating individuals with MCC [3]. Subgroups at greater risk for MCC can benefit from early detection and prevention strategies, which may reduce hospital admissions and associated health care costs [3]. Ongoing research efforts that seek to both reduce the burden of MCC on the health care system and the development of person- and family-centered care strategies across the health care continuum are needed [17]. Public health leaders should coordinate efforts with clinicians, administrators, and communities to support changes in care management and outcomes across the lifespan.

We found that the most frequent MCC combinations vary by age category. Notably, a combination of mental, rather than physical, health conditions (ie, anxiety disorders and depression, bipolar disorder, or other depressive mood disorders) was the most common MCC combination in early adulthood. Following early adulthood, combinations of physical health conditions were the most common (eg, hyperlipidemia and hypertension in late adulthood and advanced old age). For precision health, characterizing MCC profiles and the most frequently co-occurring conditions in different age categories can inform targeted prevention and treatment approaches and allow clinicians to better anticipate and manage comorbidities in their patients [18].

We acknowledge, however, that this initial analysis is exploratory and descriptive in nature. The characteristics of adult participants who are enrolled in the AoU program and consented to share EHR data—ie, women, White individuals, non-Hispanic/Latino individuals, married/partnered individuals, highly educated individuals, insured individuals, and those not currently employed—must also be considered when interpreting the reported estimates and generalizing results to the broader US, and worldwide, population. Another limitation of our analysis is the reliance on ICD-10 codes for identifying chronic conditions, which could introduce inaccuracies. Computable phenotyping challenges, such as accurate identification and measurement of chronic conditions, are not unique to this analysis but rather an inherent aspect of analysis using EHR data. Computable phenotypes for chronic conditions could be enhanced to reduce the risk of misdiagnosis and refine prevalence estimates. One approach to support the validity of chronic condition phenotypes would be comparison of CMS-derived diagnoses to EHR computable phenotypes from other repositories, such as the Observational Health Data Sciences and Informatics (OHDSI [19]) Phenotype Library and Phenotype KnowledgeBase (PheKB [20]). OHDSI and PheKB phenotypes use value sets including and beyond codes (eg, laboratory values and medications). Another approach to enhance the validity of chronic condition diagnoses would be to compare the computable phenotype–derived diagnoses to self-reported diagnoses on AoU surveys. Without a true gold standard, triangulation of EHR and survey sources may be an ideal strategy to achieve diagnosis accuracy in future work.

This brief report fills an important knowledge gap and provides a foundation for future studies on MCC correlates, predictors, and outcomes by communicating current MCC prevalence estimates, which capture a wider range of chronic conditions, across the adult lifespan. The high prevalence of MCC in adults from early adulthood to advanced old age underscores the need for resource prioritization, targeted prevention, and improved treatment and care plans in individuals with co-occurring conditions.