Insurance-Related Differences in Chronic Conditions Data Warehouse Comorbidities of Medicare Beneficiaries

ABSTRACT

Objectives: To demonstrate the prevalence of comorbidities documented by Chronic Conditions Data Warehouse (CCW) data for Medicare beneficiaries and to illustrate how failing to account for differences in reported comorbidities can result in information bias.

Study Design: Retrospective cohort study of Medicare beneficiaries who underwent coronary artery bypass grafting (CABG) between 2008 and 2019.

Methods: A total of 1,158,701 Medicare beneficiaries underwent CABG. The prevalence of CCW-reported comorbidities was compared between beneficiaries enrolled in Medicare Advantage (MA) or traditional Medicare (TM) plans. Median survival differences (with 95% CIs) were compared in unadjusted and risk-adjusted analyses using overlap propensity score weighting, with and without inclusion of CCW-reported comorbidities during risk adjustment.

Results: The proportion of MA-enrolled CABG recipients increased annually from 17.5% in 2008 to 38.3% in 2019. MA-enrolled beneficiaries had fewer CCW-reported comorbidities than TM-enrolled beneficiaries (average standardized mean difference across 27 CCW comorbidities, 0.431). After risk adjustment for demographics, median survival differed minimally between MA- and TM-enrolled beneficiaries (10.00 vs 10.05 years; difference, –15 [95% CI, –41 to 13] days). However, when CCW-reported comorbidity data were included in risk adjustment, MA-enrolled beneficiaries demonstrated substantially lower median survival (9.52 vs 10.91 years; difference, –507 [95% CI, –538 to –466] days).

Conclusions: The prevalence of CCW-reported comorbidities differs significantly between TM-enrolled and MA-enrolled beneficiaries who underwent CABG. These differences can introduce substantial bias in risk-adjusted analyses that erroneously assume equivalent CCW-reported comorbidity documentation across insurance types. Medicare outcomes research that relies on CCW-reported comorbidity data without accounting for insurance-related differences may produce biased treatment-effect estimates, potentially misinforming clinical or policy decisions.

Am J Manag Care. 2025;31(11):In Press

Takeaway Points

• Multiple methodologies exist for determining patient comorbidities using Medicare data.
• Comorbidity data derived from the Chronic Conditions segment of the Master Beneficiary Summary File have advantages over those derived from Hierarchical Condition Categories or Clinical Classifications Software algorithms, including enhanced temporal stability and stability across the transition from International Classification of Diseases, Ninth Revision to International Statistical Classification of Diseases, Tenth Revision.
• The prevalence of comorbidities reported by the Chronic Conditions Data Warehouse is substantially higher for beneficiaries with a traditional Medicare plan (for which fee-for-service data are available) than a Medicare Advantage plan (for which fee-for-service data are missing).
• Analyses using Chronic Conditions Data Warehouse comorbidity data that fail to account for insurance status are subject to information bias that may compromise findings.

CMS maintains a database of administrative claims that is increasingly used in clinical research.¹ Hospital claims submitted to CMS using International Classification of Diseases (ICD) codes are available in the Medicare Provider Analysis and Review (MedPAR) file for analysis by research groups.² For studies analyzing Medicare claims, correctly interpreting coding data to characterize beneficiary demographics, comorbidities, procedural characteristics, and outcomes is essential.³

Three common methodologies are used for deriving comorbidity data from Medicare administrative claims. First, a subset of the Master Beneficiary Summary File (MBSF) known as the Chronic Conditions segment identifies chronic conditions using a combination of ICD codes and fee-for-service claims and maintains this information in its Chronic Conditions Data Warehouse (CCW).⁴ Second, Medicare maintains a mapping of ICD codes to Hierarchical Condition Categories (HCCs), a system primarily intended to adjust federal payments to insurers and health systems based on expected spending.⁵ Third, the Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project has developed a tool, known as Clinical Classifications Software (CCS), that uses algorithms to translate ICD codes into comorbidity categories.^6,7

Given that CCW-reported comorbidities are stable across the transition from ICD, Ninth Revision (ICD-9) to ICD, Tenth Revision (ICD-10) and demonstrate high sensitivity rates in identifying chronic conditions,^8-11 we have preferred using this construct in our analyses of administrative claims. However, one of the strengths of CCW-reported data (its use/reliance on fee-for-service claims) can also represent a significant limitation. Because fee-for-service claims inform the determination of CCW-reported comorbidities, the prevalence of these comorbidities may be systematically underestimated for Medicare Advantage (MA)–enrolled beneficiaries whose health care provider claims are submitted to MA organizations rather than directly to CMS. As a result, MA-enrolled beneficiaries may have lower rates of recorded comorbidities in their CCW file vs beneficiaries with a traditional Medicare (TM) plan, not due to actual differences in health status but due to differential data availability.

Recent reports have demonstrated that MA-enrolled and TM-enrolled beneficiaries have similar (or minimally different) rates of comorbidities such as obesity, hypertension, chronic kidney disease, and diabetes when assessed using objective physical examination and laboratory measurements.¹² Therefore, the use of CCW-derived comorbidity data in analyses that include both MA- and TM-enrolled beneficiaries may introduce systematic bias due to differences in data completeness rather than actual clinical differences.¹³ As our research team focused on longitudinal outcomes following cardiac surgery, we initially observed differences in CCW-reported comorbidities between TM- and MA-enrolled beneficiaries while studying a cohort of Medicare beneficiaries undergoing coronary artery bypass grafting (CABG). This observation prompted us to systematically evaluate whether CCW-reported comorbidity documentation differed by insurance type in this population. Our objectives were 2-fold: first, to quantify differences in CCW-reported comorbidity burden between MA- and TM-enrolled CABG recipients, and second, to demonstrate the extent to which this information bias could influence risk-adjusted survival analyses in this cohort.

METHODS

Study Design

We retrospectively reviewed CMS administrative claims data from 2008 to 2019. The institutional review board (IRB) of Baylor Scott & White Research Institute approved this study (IRB approval 019-270 dated August 23, 2019).

Study Population

A 100% sample of Medicare beneficiaries with both Part A and Part B insurance who had been hospitalized for CABG was identified from the MedPAR file using ICD procedural codes (eAppendix Table 1 [eAppendix available at ajmc.com]). Beneficiaries who underwent concomitant cardiac surgery were excluded to ensure our analysis was limited to isolated CABG recipients (eAppendix Table 2). Beneficiaries missing adequate insurance documentation and those missing 9-digit zip code data or with a 9-digit zip code lacking a corresponding Area Deprivation Index (ADI) measure were also excluded from the analysis to allow for enhanced risk adjustment.¹⁴

Beneficiary demographics on the date of surgery were ascertained from the MBSF. Preoperative comorbidity data were obtained from the CCW; diagnoses were confirmed as preexisting if the earliest documented date of diagnosis occurred prior to the date of surgery. CCW data are recorded annually; diagnoses depend on that particular year’s coding data (or set of years depending on the look-back period) and thus are not necessarily stable over time.^9,10 If a comorbidity diagnosis was documented for a beneficiary in any of the annual CCW records dating back to 1999 (but before the date of surgery), the beneficiary was identified as having this condition.

Social Determinants of Health

Beneficiary race was ascertained from the MBSF using the CMS imputation algorithm.¹⁵ Imputed race codes have been shown to have improved sensitivity in identifying minority race based on third-party survey data.¹⁶

Neighborhood deprivation using the ADI metric has been commonly implemented as a composite measure of key social determinants of health impacting patient outcomes.¹⁴ National 2015 and 2019 ADI values were used to assess deprivation; beneficiaries who underwent surgery between 2008 and 2015 were assessed using 2015 values, and those with surgery between 2016 and 2019 were assessed using 2019 values. Critiques of ADI have emerged regarding its lack of standardization, which may result in it being reduced to a weighted average of income and home values.¹⁷ These findings have implications regarding whether ADI underestimates deprivation associated with patients living in urban (as opposed to rural) regions given the differences in income and home values.¹⁸ Our risk-adjustment models include ADI along with sex, race, insurance status, admission urgency, and state of residence; we have found that after adjusting for these other social determinants of health, ADI remains a strong predictor of both early and late outcomes following cardiac surgery.¹⁹ Because our models include ADI with a range of additional social determinants of health, we believe the aforementioned limitations of ADI are mitigated, especially because we explicitly account for state-level effects.

Hospital Medicare CABG Volumes

Annual hospital CABG volumes among Medicare beneficiaries were determined using the hospital CMS certification number associated with ICD procedural codes for CABG. Hospital volume was included as a variable for risk adjustment because it has been previously associated with survival after CABG.²⁰

Primary End Point

The primary end point was all-cause mortality. Vital status and date of death were determined from the MBSF.

Statistical Analysis

The prevalence rates of CCW-reported preexisting comorbidities among MA-enrolled and TM-enrolled beneficiaries were measured. Three survival analyses were undertaken to evaluate whether the prevalence of CCW-reported comorbidities could result in information bias affecting risk-adjusted survival estimates among MA-enrolled and TM-enrolled beneficiaries: (1) unadjusted; (2) risk-adjusted for demographics, social determinants of health, admission urgency, and hospital volume but not CCW-reported comorbidities; and (3) risk-adjusted for demographics, social determinants of health, admission urgency, hospital volume, and CCW-reported comorbidities.

Year-to-year trends in the distribution of beneficiary insurance status were evaluated using the Cochran-Armitage test. Overlap propensity score weighting was performed to risk adjust for imbalances between MA-enrolled and TM-enrolled beneficiary demographics, social determinants of health, dialysis dependence, admission urgency, year of surgery, hospital CABG volume, and, in the second weighted analysis only, CCW-reported comorbidities (eAppendix Methods).^21,22 Between-group differences were evaluated using standardized mean differences (SMDs) before and after risk adjustment.²³ Survival probabilities were estimated using the Kaplan-Meier method before and after risk adjustment, and 95% CIs for risk-adjusted Kaplan-Meier analyses and risk-adjusted median survival estimates were obtained using a nonparametric bootstrap procedure; all survival estimates are reported together with 95% CIs in parentheses.²⁴ Data management was conducted using SAS 9.4 (SAS Institute Inc), and statistical analyses were conducted with Stata/MP 18.0 (StataCorp LLC).

RESULTS

A total of 1,520,533 Medicare beneficiaries with both Part A and Part B plans were identified by ICD procedural codes as having undergone CABG between 2008 and 2019, of whom 1,158,701 underwent isolated CABG and had suitable data for analysis (Figure 1). Of these, 335,080 (28.9%) beneficiaries were MA enrolled and 823,621 (71.1%) were TM enrolled at the time of CABG. The percentage of MA-enrolled beneficiaries increased annually from 17.5% in 2008 to 38.3% in 2019 (Cochran-Armitage test P < .001) (Figure 2). Median (IQR) follow-up was 6.12 (2.97-9.27) years.

Compared with TM-enrolled beneficiaries, MA-enrolled beneficiaries were less commonly White, more commonly Hispanic, more likely to reside in the Western US, less likely to be dialysis dependent, and more likely to have undergone CABG more recently in the study period (Table 1). Neighborhood disadvantage (assessed by ADI) was similar between MA-enrolled and TM-enrolled beneficiaries.

The prevalence of CCW-reported preoperative comorbidities was substantially higher in TM-enrolled (vs MA-enrolled) beneficiaries, with an average SMD of 0.431 among the 27 documented comorbidities (Table 1). For reference, the prevalence of CCW-reported chronic kidney disease (CKD) was more than triple in TM-enrolled compared with MA-enrolled beneficiaries (32.3% vs 10.5%; SMD, 0.554). However, the prevalence of dialysis dependence (a variable maintained separately by CMS in the MedPAR file, independent of comorbidity data) in TM-enrolled beneficiaries was only modestly different compared with that in MA-enrolled beneficiaries (4.6% vs 2.6%; SMD, 0.106).

The prevalence of dialysis dependence and most CCW-reported comorbidities changed modestly over time for both our TM-enrolled and MA-enrolled beneficiary cohorts, without a substantial change across the ICD-9 to ICD-10 transition (Table 2). Most CCW-reported comorbidities had a low-magnitude trend of increasing prevalence, including hypertension, hyperlipidemia, diabetes, anemia, history of myocardial infarction, atrial fibrillation, asthma, hypothyroidism, benign prostatic hypertrophy, rheumatoid arthritis/osteoarthritis, depression, and dementia/Alzheimer disease. Conversely, we noted low-magnitude trends of decreasing prevalence in the comorbidities ischemic heart disease, chronic obstructive pulmonary disease, osteoporosis, stroke/transient ischemic attack, cataract, breast cancer, colorectal cancer, and lung cancer. The only large-magnitude trend noted was an increase in the prevalence of CKD over the study duration, from 22.3% to 45.9% and 8.1% to 14.3% in TM-enrolled and MA-enrolled beneficiaries, respectively.

Unadjusted survival analysis demonstrated a clinically insignificant survival advantage for MA-enrolled vs TM-enrolled beneficiaries undergoing CABG: median survival difference, 32 (95% CI, 4-56) days (10.04 vs 9.95 years, respectively)(Figure 3 [A]). Risk adjustment was used to account for potential confounding variables (demographics, neighborhood deprivation, admission urgency, year of surgery, hospital volume, and dialysis dependence) but not CCW-reported comorbidities (Table 1); no difference in risk-adjusted survival was noted between MA-enrolled and TM-enrolled beneficiaries (weighted median survival difference, –15 [95% CI, –41 to 13] days; 10.00 vs 10.05 years, respectively) (Figure 3 [B]). Risk adjustment was then repeated to also include CCW-reported comorbidities; this weighted analysis exhibited substantially reduced survival in MA-enrolled compared with TM-enrolled beneficiaries (median survival difference, –507 [95% CI, –538 to –466] days; 9.52 vs 10.91 years) (Figure 3 [C]). As a sensitivity analysis, separate univariate and multivariable Cox regression analyses of predictors of postoperative survival are provided for beneficiaries with a TM plan, those with an MA plan, and the entire cohort (eAppendix Table 3).

DISCUSSION

This analysis evaluated the prevalence of preexisting CCW-reported comorbidities and postoperative survival among a 100% sample of Medicare beneficiaries undergoing isolated CABG between 2008 and 2019. We report several key findings. First, the prevalence of CCW-reported comorbidities was substantially lower for MA-enrolled beneficiaries compared with TM-enrolled beneficiaries. Second, despite differences in CCW-reported comorbidities, both unadjusted and risk-adjusted survival (when adjusted for demographics but not comorbidities) were similar between MA-enrolled and TM-enrolled beneficiaries. Third, when CCW-reported comorbidities were included in risk adjustment, MA-enrolled beneficiaries exhibited significantly lower survival compared with their TM-enrolled counterparts.

Although our analysis focused on evaluating CCW-derived comorbidity data, it is important to recognize that researchers using CMS claims can choose between several methodologies to identify preexisting comorbidities and/or risk adjust their analyses, including the HCC and CCS constructs described previously. CCW-reported comorbidities, although available only through purchase, offer several features that enhance their utility in research applications. These include standardized formatting that eliminates the need for researchers to independently implement coding algorithms, documentation of the initial date of diagnosis, and the incorporation of fee-for-service claims, all of which improve both the sensitivity and specificity of reported diagnoses. Perhaps most notably, owing to their use of fee-for-service data, CCW-reported comorbidities remained stable across the ICD-9 to ICD-10 transition, a key advantage in longitudinal analyses.^9-11 For these reasons, CCW-reported comorbidities are often considered the gold standard for comorbidity identification in studies using administrative claims.¹¹

As described in a CCW white paper, our findings are an expected consequence of the methodology used to derive CCW-reported comorbidities, which rely in part on fee-for-service claims (data that are not available for MA-enrolled beneficiaries).²⁵ Our findings of improved risk-adjusted post-CABG survival in TM-enrolled beneficiaries (Figure 3 [C]) may appear counterintuitive, particularly in light of prior analyses demonstrating a higher prevalence of documented comorbidities among MA-enrolled beneficiaries when using alternative methodologies.²⁶ Additionally, the well-documented practice of MA organizations upcoding diagnoses to enhance reimbursement may lead to an expectation that documented comorbidity burdens should appear higher among MA-enrolled beneficiaries.²⁷ However, these expectations overlook the critical influence of data source and methodology on the apparent prevalence of comorbidities in administrative claims.

For example, data from the MedPAR file are derived from inpatient claims submitted directly by hospitals to Medicare Administrative Contractors, regardless of beneficiary insurance status. As such, researchers should expect minimal differences in MedPAR data between MA- and TM-enrolled beneficiaries. In contrast, data submitted by MA organizations for reimbursement purposes (through the CMS Risk Adjustment Processing System) as well as MA encounter data (which include separate files for inpatient and health care provider claims submitted by MA organizations as required by CMS since 2012) may be vulnerable to diagnosis upcoding and other documentation discrepancies.²⁸ Accordingly, data submitted to CMS by MA organizations require careful validation prior to integration with fee-for-service claims or other data submitted directly to CMS by hospitals or health care providers.²⁹

Interestingly, despite the substantial differences in CCW-reported comorbidity burdens among the MA- and TM-enrolled beneficiaries, we noted slow increases in the prevalence of most reported comorbidities over the decade-long study period, regardless of insurance status. The only substantial change in CCW-reported comorbidity prevalence over time was the increase in CKD, which we hypothesize is related to the improved ability of electronic health records to capture early-stage CKD. This is consistent with national data showing a stable age-adjusted prevalence of CKD, despite rising rates of end-stage kidney disease requiring dialysis or transplantation.^30,31

Despite our described differences in CCW-reported comorbidities at the time of CABG, we suspect that MA- and TM-enrolled beneficiaries undergoing CABG have a similar (or at least more closely balanced) burden of chronic disease. The results of our survival analyses support this hypothesis. Our unweighted survival analysis identified a statistically significant but clinically minimal post-CABG survival advantage for MA-enrolled compared with TM-enrolled beneficiaries (Figure 3 [A]). We then adjusted for beneficiary demographics, social determinants of health, dialysis dependence, admission urgency, year of surgery, and hospital CABG volume due to previous reports suggesting important demographic and geographic differences among beneficiaries electing to enroll in MA vs TM.³² After adjusting for these confounding variables, post-CABG survival was equivalent between MA- and TM-enrolled beneficiaries (Figure 3 [B]). Only with the additional adjustment of CCW-reported comorbidities did we identify a difference in median survival between the 2 groups.

Our hypothesis is supported by several recent reports. Oseran et al compared MA- and TM-enrolled beneficiaries who participated in the National Health and Nutrition Examination Survey and found that when using objective physical exam and laboratory data to ascertain diagnoses, the prevalence rates of comorbidities were largely similar between groups.¹² Second, prospectively collected registry data from the Society of Thoracic Surgeons Adult Cardiac Surgery Database (STS-ACSD), cross-linked to CMS over a similar time frame as our study, demonstrate a prevalence of comorbidities similar to CCW-derived comorbidities in our TM-enrolled cohort, but in sharp contrast to that of our MA-enrolled cohort.³³ For example, the STS-ACSD reported incidence rates of diabetes (45%), hypertension (90%), chronic lung disease (27%), and cerebrovascular disease (22%) are more similar to CCW-derived comorbidities in the TM-enrolled vs MA-enrolled beneficiaries included in our study (Table 1).

Limitations

Our study is subject to several limitations inherent to retrospective analyses of administrative claims, including susceptibility to unmeasured confounding, selection bias related to our cohort of only Medicare beneficiaries undergoing CABG, and our use of administrative claims to account for beneficiary comorbidities and procedural characteristics. Given our focus on CABG recipients, our findings may not be generalizable to other cohorts of Medicare beneficiaries; however, we can think of no apparent reason our findings would be limited only to the cohort under study. We lack a true gold standard for adjudicating the actual prevalence of comorbidities in our study population. However, CCW-reported comorbidities have been extensively validated over time and across the ICD-9 to ICD-10 transition, are regarded as having the highest sensitivity and specificity among claims-based comorbidity methodologies, and closely resemble clinical data of Medicare beneficiaries documented in the STS database.^9,12,34,35

CONCLUSIONS

Large-scale retrospective analyses of Medicare claims require meticulous study design, including thoughtful cohort selection and appropriate risk adjustment. Our findings demonstrate that CCW-reported comorbidities can systematically underestimate disease burden among MA-enrolled beneficiaries, leading to distorted risk-adjusted survival estimates when both MA- and TM-enrolled populations are analyzed concurrently. We suggest that the observed 507-day median survival advantage for TM-enrolled beneficiaries undergoing CABG reflects the magnitude of bias that may be introduced by differential comorbidity capture between insurance types. We recommend restricting analyses using CCW-reported comorbidity data to TM-enrolled beneficiaries. Although CMS has begun releasing MA encounter data for health care provider–submitted claims (akin to fee-for-service claims for TM-enrolled beneficiaries), it remains to be seen whether these data will be suitable for comorbidity determination in analyses that include both MA- and TM-enrolled beneficiaries.³⁶

Acknowledgments

The authors are especially grateful to Alessandro Gasparini, PhD, of Red Door Analytics, for his helpful feedback regarding the statistical methodologies.

Author Affiliations: Department of Cardiac Surgery, Baylor Scott & White, The Heart Hospital—Plano (JMS, JJS, AK, MJM, JMD), Plano, TX; Baylor Scott & White Research Institute, The Heart Hospital—Plano (JJS, AK, JKB, MJM), Plano, TX; Department of Cardiothoracic Surgery, Weill Cornell Medicine (MFLG), New York, NY.

Source of Funding:Data acquisition and effort of Drs Kluis and Banwait were supported by a philanthropic gift from Satish and Yasmin Gupta to Baylor Scott & White, The Heart Hospital, in Plano, TX.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (JMS, AK, JKB, MJM, JMD); acquisition of data (JMS, JJS, AK, JKB); analysis and interpretation of data (JMS, JJS, AK, JKB, MFLG, JMD); drafting of the manuscript (JMS, JJS, AK, JKB, MFLG, JMD); critical revision of the manuscript for important intellectual content (JMS, JJS, AK, MFLG, MJM, JMD); statistical analysis (JMS, JKB); provision of patients or study materials (JKB); obtaining funding (JMS); administrative, technical, or logistic support (JMS, JJS); and supervision (JMS, MJM, JMD).

Address Correspondence to: Justin M. Schaffer, MD, Baylor Scott & White, The Heart Hospital—Plano, 4708 Alliance Blvd, Plano, TX 75093. Email:
justin.schaffer@bswhealth.org.

REFERENCES

1. Mues KE, Liede A, Liu J, et al. Use of the Medicare database in epidemiologic and health services research: a valuable source of real-world evidence on the older and disabled populations in the US. Clin Epidemiol. 2017;9(9):276-277. doi:10.2147/CLEP.S105613

2. Cartwright DJ. ICD-9-CM to ICD-10-CM codes: what? why? how? Adv Wound Care (New Rochelle). 2013;2(10):588-592. doi:10.1089/wound.2013.0478

3. Hall M, Attard TM, Berry JG. Improving cohort definitions in research using hospital administrative databases—do we need guidelines? JAMA Pediatr. 2022;176(6):539-540. doi:10.1001/jamapediatrics.2022.0091

4. Goodman AR, Posner SF, Huang ES, Parekh AK, Koh HK. Defining and measuring chronic conditions: imperatives for research, policy, program, and practice. Prev Chronic Dis. 2013;10:E66. doi:10.5888/pcd10.120239

5. Weissman GE. Hierarchical Condition Categories for pulmonary diseases: population health management and policy opportunities. Chest. 2019;155(4):868-873. doi:10.1016/j.chest.2018.12.013

6. Clinical Classifications Software Refined (CCSR). Agency for Healthcare Research and Quality. Updated November 13, 2024. Accessed January 30, 2025. https://hcup-us.ahrq.gov/toolssoftware/ccsr/ccs_refined.jsp

7. Clinical Classifications Software (CCS) for ICD-9-CM. Agency for Healthcare Research and Quality. Updated March 6, 2017. Accessed January 30, 2025. https://hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp

8. Khera R, Dorsey KB, Krumholz HM. Transition to the ICD-10 in the United States: an emerging data chasm. JAMA. 2018;320(2):133-134. doi:10.1001/jama.2018.6823

9. Weeks WB, Huynh G, Cao ST, Smith J, Weinstein JN. Assessment of year-to-year patient-specific comorbid conditions reported in the Medicare Chronic Conditions Data Warehouse. JAMA Netw Open. 2020;3(10):e2018176. doi:10.1001/jamanetworkopen.2020.18176

10. Ellis RP, Hsu HE, Song C, et al. Diagnostic category prevalence in 3 classification systems across the transition to the International Classification of Diseases, Tenth Revision, Clinical Modification. JAMA Netw Open. 2020;3(4):e202280. doi:10.1001/jamanetworkopen.2020.2280

11. Tu N, Henderson M, Sundararajan M, Salas M. Discrepancies in ICD-9/ICD-10-based codes used to identify three common diseases in cancer patients in real-world settings and their implications for disease classification in breast cancer patients and patients without cancer: a literature review and descriptive study. Front Oncol. 2023;13:1016389. doi:10.3389/fonc.2023.1016389

12. Oseran AS, Aggarwal R, Figueroa J, Joynt Maddox KE, Landon BE, Wadhera RK. Prevalence of chronic medical conditions among Medicare Advantage and Traditional Medicare beneficiaries. Ann Intern Med. 2025;178(3):327-335. doi:10.7326/ANNALS-24-01531

13. Kesmodel US. Information bias in epidemiological studies with a special focus on obstetrics and gynecology. Acta Obstet Gynecol Scand. 2018;97(4):417-423. doi:10.1111/aogs.13330

14. Kind AJH, Buckingham WR. Making neighborhood-disadvantage metrics accessible—the Neighborhood Atlas. N Engl J Med. 2018;378(26):2456-2458. doi:10.1056/NEJMp1802313

15. Filice CE, Joynt KE. Examining race and ethnicity information in Medicare administrative data. Med Care. 2017;55(12):e170-e176. doi:10.1097/MLR.0000000000000608

16. Huang AW, Meyers DJ. Assessing the validity of race and ethnicity coding in administrative Medicare data for reporting outcomes among Medicare Advantage beneficiaries from 2015 to 2017. Health Serv Res. 2023;58(5):1045-1055. doi:10.1111/1475-6773.14197

17. Petterson S. Deciphering the Neighborhood Atlas Area Deprivation Index: the consequences of not standardizing. Health Aff Sch. 2023;1(5):qxad063. doi:10.1093/haschl/qxad063

18. Bertin M, Chevrier C, Pelé F, Serrano-Chavez T, Cordier S, Viel JF. Can a deprivation index be used legitimately over both urban and rural areas? Int J Health Geogr. 2014;13:22. doi:10.1186/1476-072X-13-22

19. Shih E, Squiers JJ, Banwait JK, et al. Race, neighborhood disadvantage, and survival of Medicare beneficiaries after aortic valve replacement and concomitant coronary artery bypass grafting. J Thorac Cardiovasc Surg. 2024;167(6):2076-2090.e19. doi:10.1016/j.jtcvs.2023.02.005

20. Peterson ED, Coombs LP, DeLong ER, Haan CK, Ferguson TB. Procedural volume as a maker for quality for CABG surgery. JAMA. 2004;291(2):195-201. doi:10.1001/jama.291.2.195

21. Cheng C, Li F, Thomas LE, Li FF. Addressing extreme propensity scores in estimating counterfactual survival functions via the overlap weights. Am J Epidemiol. 2022;191(6):1140-1151. doi:10.1093/aje/kwac043

22. Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA. 2020;323(23):2417-2418. doi:10.1001/jama.2020.7819

23. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083-3107. doi:10.1002/sim.3697

24. Rogawski ET, Westreich DJ, Kang G, Ward HD, Cole SR. Brief report: estimating differences and ratios in median times to event. Epidemiology. 2016;27(6):848-851. doi:10.1097/EDE.0000000000000539

25. Impact of Transition From 27 to 30 CCW Chronic Conditions. Chronic Conditions Warehouse; March 2022. Accessed April 1, 2025. https://www2.ccwdata.org/documents/10280/19002256/ccw-condition-categories-impact-of-transition-from-27-to-30.pdf

26. Cohen K, Ameli O, Chaisson CE, et al. Comparison of care quality metrics in 2-sided risk Medicare Advantage vs fee-for-service Medicare programs. JAMA Netw Open. 2022;5(12):e2246064. doi:10.1001/jamanetworkopen.2022.46064

27. Gilfillan R, Berwick DM. Medicare Advantage, direct contracting, and the Medicare ‘money machine,’ part 1: the risk-score game. Health Affairs Forefront. September 29, 2021. Accessed April 1, 2025. https://www.healthaffairs.org/do/10.1377/forefront.20210927.6239/

28. Chapter 7: ensuring the accuracy and completeness of Medicare Advantage encounter data. In: Report to the Congress: Medicare and the Health Care Delivery System. Medicare Payment Advisory Commission; June 2019:205-239. Accessed April 1, 2025. https://www.medpac.gov/wp-content/uploads/import_data/scrape_files/docs/default-source/reports/jun19_ch7_medpac_reporttocongress_sec.pdf

29. Agarwal R, Connolly J, Gupta S, Navathe AS. Comparing Medicare Advantage and traditional Medicare: a systematic review. Health Aff (Millwood). 2021;40(6):937-944. doi:10.1377/hlthaff.2020.02149

30. Weeks WB, Huynh G, Cao SY, Smith J, Weinstein JN. Assessment of year-to-year patient-specific comorbid conditions reported in the Medicare Chronic Conditions Data Warehouse. JAMA Netw Open. 2020;3(10):e2018176. doi:10.1001/jamanetworkopen.2020.18176

31. Burrows NR, Koyama A, Pavkov ME. Reported cases of end-stage kidney disease – United States, 2000-2019. MMWR Morb Mortal Wkly Rep. 2022;71(11):412-415. doi:10.15585/mmwr.mm7111a3

32. Neuman P, Jacobson GA. Medicare Advantage checkup. N Engl J Med. 2018;379(22):2163-2172. doi:10.1056/NEJMhpr1804089

33. Kelly JJ, Han JJ, Desai ND, et al. Coronary endarterectomy: analysis of the Society of Thoracic Surgeons Adult Cardiac Surgery Database. Ann Thorac Surg. 2022;114(3):667-674. doi:10.1016/j.athoracsur.2022.01.067

34. Schneider KM, O’Donnell BE, Dean D. Prevalence of multiple chronic conditions in the United States’ Medicare population. Health Qual Life Outcomes. 2009;7:82. doi:10.1186/1477-7525-7-82

35. DuGoff EH, Canudas-Romo V, Buttorff C, Leff B, Anderson GF. Multiple chronic conditions and life expectancy: a life table analysis. Med Care. 2014;52(8):688-694. doi:10.1097/MLR.0000000000000166

36. Brennan N, Ornstein C, Frakt AB. Time to release Medicare Advantage claims data. JAMA. 2018;319(10):975-976. doi:10.1001/jama.2017.21519