Assessment of Year-to-Year Patient-Specific Comorbid Conditions Reported in the Medicare Chronic Conditions Data Warehouse | Geriatrics | JAMA Network Open | JAMA Network
[Skip to Navigation]
Sign In
Table 1.  Single-Year Persistence for 51 Condition Flags
Single-Year Persistence for 51 Condition Flags
Table 2.  Multiyear Persistence for 51 Condition Flags
Multiyear Persistence for 51 Condition Flags
1.
 Chronic Conditions Data Warehouse. Centers for Medicare & Medicaid Services. Accessed April 30, 2020. https://www2.ccwdata.org/web/guest
2.
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.2020.2280PubMedGoogle Scholar
3.
Janssen  F, Kunst  AE.  ICD coding changes and discontinuities in trends in cause-specific mortality in six European countries, 1950-99.   Bull World Health Organ. 2004;82(12):904-913.PubMedGoogle Scholar
4.
Mainor  AJ, Morden  NE, Smith  J, Tomlin  S, Skinner  J.  ICD-10 coding will challenge researchers: caution and collaboration may reduce measurement error and improve comparability over time.   Med Care. 2019;57(7):e42-e46. doi:10.1097/MLR.0000000000001010PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Views 1,062
    Citations 0
    Research Letter
    Health Informatics
    October 1, 2020

    Assessment of Year-to-Year Patient-Specific Comorbid Conditions Reported in the Medicare Chronic Conditions Data Warehouse

    Author Affiliations
    • 1Microsoft Research, Microsoft, Redmond, Washington
    • 2CareJourney, Arlington, Virginia
    • 3White River Junction VA Outcomes Group, White River Junction, Vermont
    • 4The Dartmouth Institute for Health Policy and Clinical Practice, Dartmouth College, Lebanon, New Hampshire
    • 5Kellogg School of Management, Northwestern University, Evanston, Illinois
    • 6Tuck School of Business, Dartmouth College, Hanover, New Hampshire
    JAMA Netw Open. 2020;3(10):e2018176. doi:10.1001/jamanetworkopen.2020.18176
    Introduction

    The Centers for Medicare & Medicaid Services created the Chronic Conditions Data Warehouse (CCDW) for researchers and health policy analysts seeking to improve health care quality and reduce health care costs and use.1 Refreshed annually, the CCDW uses International Classification of Diseases (ICD) codes to identify people enrolled in fee-for-service Medicare who have specific chronic conditions, adding flags to denote these individuals.

    We sought to determine whether CCDW condition flags identified using codes from the International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) persist in a given patient across time. The CCDW algorithms use different numbers of sequential years, or lookback periods, to flag different conditions. We hypothesized that flags based on longer lookback periods (eg, 3 years for Alzheimer disease) and used for conditions unlikely to be cured (eg, schizophrenia) would be more likely to persist in the same patient than flags based on shorter lookback periods (eg, 1 year for anemia) and conditions with high cure rates (eg, hip and pelvic fracture).

    Methods

    The Solutions IRB institutional review board and the Centers for Medicare & Medicaid Services approved this cross-sectional study. They determined that informed consent was not required because Medicare enrollees consent to research analysis of their claims data on enrollment and because retroactive collection of informed consent from millions of patients would be impractical. Our work followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Using CCDW data from 2010 to 2017, we identified people enrolled in fee-for-service Medicare who had a chronic condition flag that did not exist in the prior year. We limited our analysis to 51 conditions relevant to adults, and we limited the analysis to individuals who survived and remained in fee-for-service Medicare the entire year after they received a chronic condition flag. Using a serial cross-sectional design, we calculated condition persistence, defined as the proportion of enrollees who retained the chronic condition flag the year after they received the flag. We calculated year-to-year and sequential multiyear (up to 5 years) condition persistence. Analyses were conducted using SAS statistical software version 9.4 (SAS Institute) between March 3 and April 27, 2020.

    Results

    This study included 70 871 475 Medicare fee-for-service beneficiaries (38 285 908 [54.0%] women); the mean (SD) age was 70.9 (12.2) years. Year-to-year condition persistence varied considerably across conditions. Among 188 252 beneficiaries with an acute myocardial infarction flag in 2011, 12 773 beneficiaries retained the flag in 2012 (6.8%). Of 187 664 beneficiaries with hip/pelvic fracture in 2011, 19 110 beneficiaries retained the flag in 2012 (10.2%). Persistence was higher for those with chronic kidney disease, with 1 098 973 of 1 183 808 beneficiaries who had the flag in 2011 retaining it in 2012 (92.8%), and for those with congestive heart failure, with 1 053 944 of 1 096 806 who had the flag in 2011 retaining it in 2012 (96.1%). Among 50 315 beneficiaries with personality disorder who had the flag in 2011, 46 955 beneficiaries (93.3%) retained it in 2012, with similar numbers for numerous other neuropsychiatric disorders (Table 1). Condition flags with shorter lookback periods demonstrated lower year-to-year persistence compared with conditions with longer lookback periods. Flags for conditions amenable to cure (eg, hip or pelvic fracture) were less persistent than flags for conditions not amenable to cure (eg, chronic kidney disease, schizophrenia, and Alzheimer disease). Condition persistence did not differ in 2013 to 2014 (using ICD-9) vs 2016 to 2017 (using ICD-10), with 5 exceptions. Hip/pelvic fracture persistence decreased by 39.4%, from 9.7% (17 764 of 183 690 individuals) to 5.9% (9168 of 156 446 individuals). Alcohol use disorders persistence decreased by 12.3%, from 90.2% (136 830 of 151 778 individuals) to 79.0% (157 388 of 199 151 individuals). Acute myocardial infarction persistence increased by 11.0%, from 5.9% (10 544 of 179 033 individuals) to 6.5% (11 783 of 180 182 individuals). Endometrial cancer persistence increased by 13.0%, from 43.3% (13 097 of 30 251 individuals) to 48.9% (13 522 of 27 649 individuals). Osteoporosis persistence increased by 10.6%, from 34.2% (195 886 of 572 097 individuals) to 37.9% (164 950 of 435 424 individuals). Prostate cancer persistence increased by 10.3%, from 61.9% (78 037 of 126 140 individuals) to 68.2% (70 947 of 104 012 individuals).

    Multiyear analysis demonstrated a decrease in condition persistence after the lookback period, frequently followed by increasing persistence rates in later years (Table 2). Among 107 individuals who had a hip/pelvic fracture flag that was new in year 0 and who continued to meet inclusion and exclusion criteria for 4 consecutive years, 13 individuals retained that flag in the fifth year (12.1%). Of 1 314 240 individuals who had a hypertension flag that was new in year 0 and who continued to meet inclusion and exclusion criteria for 4 consecutive years, 926 539 individuals retained that flag in the fifth year (70.5%).

    Discussion

    Using CCDW data from 2010 to 2017, this cross-sectional study found that condition persistence varied considerably across conditions, was not associated with the transition to ICD-10, decreased after the condition-specific lookback period, and then increased over time. The lookback period–associated decrease in condition persistence suggests that a condition was likely to be flagged again if multiple years were used to define the flag.

    Our findings did not reveal changes in year-to-year diagnostic categorization when ICD versions changed, which other studies have found.2-4 However, our findings have broader implications for research and policies that relied on CCDW flags, even when ICD versions did not change. Specifically, we found that many individuals diagnosed with a chronic condition were in a CCDW flag–defined cohort 1 year but not in the next, raising concerns that the conclusions of CCDW-based research and policies regarding prevalence, reimbursement, risk adjustment, and quality are inaccurate.

    Our study was limited because it analyzed only people enrolled in fee-for-service Medicare and because it used condition flag definitions. Nonetheless, we recommend that researchers seeking to inform reimbursement, risk adjustment, and quality policy aim to conduct sensitivity analyses using multiple years of sequential, condition-specific flags to more confidently identify disease incidence and prevalence and to define cohorts of patients.

    Back to top
    Article Information

    Accepted for Publication: July 7, 2020.

    Published: October 1, 2020. doi:10.1001/jamanetworkopen.2020.18176

    Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License. © 2020 Weeks WB et al. JAMA Network Open.

    Corresponding Author: William B. Weeks, MD, PhD, MBA, Microsoft Research, Microsoft, 14820 NE 36th St, Redmond, WA 98052 (william.weeks@microsoft.com).

    Author Contributions: Ms Cao had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Weeks, Huynh, Weinstein.

    Acquisition, analysis, or interpretation of data: Weeks, Huynh, Cao, Smith.

    Drafting of the manuscript: Weeks, Weinstein.

    Critical revision of the manuscript for important intellectual content: All authors.

    Statistical analysis: Cao, Smith.

    Obtained funding: Weinstein.

    Administrative, technical, or material support: Huynh, Weinstein.

    Supervision: Weinstein.

    Conflict of Interest Disclosures: None reported.

    References
    1.
     Chronic Conditions Data Warehouse. Centers for Medicare & Medicaid Services. Accessed April 30, 2020. https://www2.ccwdata.org/web/guest
    2.
    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.2020.2280PubMedGoogle Scholar
    3.
    Janssen  F, Kunst  AE.  ICD coding changes and discontinuities in trends in cause-specific mortality in six European countries, 1950-99.   Bull World Health Organ. 2004;82(12):904-913.PubMedGoogle Scholar
    4.
    Mainor  AJ, Morden  NE, Smith  J, Tomlin  S, Skinner  J.  ICD-10 coding will challenge researchers: caution and collaboration may reduce measurement error and improve comparability over time.   Med Care. 2019;57(7):e42-e46. doi:10.1097/MLR.0000000000001010PubMedGoogle ScholarCrossref
    ×