The Role Of Social, Cognitive, And Functional Risk Factors In Medicare Spending For Dual And Nondual Enrollees

In an effort to improve quality and reduce costs in health care, the Centers for Medicare and Medicaid Services (CMS) has increasingly focused on value-based payment programs, which tie payment to performance on quality and cost measures.^1–4 These programs, which are currently in place for hospitals, dialysis facilities, skilled nursing facilities, and physician practices, are predicated on having accurate and fair measurement of provider performance. If quality and cost measures do not accurately reflect this performance, value-based payment programs will at best be ineffective and at worst reward and penalize the wrong providers.

It is well established that social risk factors such as poverty, social isolation, and low educational attainment are associated with worse outcomes and higher health care costs.^5–9 One group with particularly high levels of social risk is the population of Medicare beneficiaries who also qualify for Medicaid, which serves as an indicator of poverty. These dually enrolled beneficiaries tend to have lower incomes,¹⁰ less education,^11,12 and lower social support^10–12 and have worse outcomes on many value-based payment measures.^13–15 Indeed, prior research has shown that existing value-based payment programs inflict higher penalties on hospitals, physicians, and facilities that serve these groups.^{3,4,13,16–22} As a result, many high-profile health policy studies^13,23–26 have argued that dual enrollment or other social risk factors should be included in risk adjustment of value-based payment outcomes. However, this has been controversial, and opponents of social-risk-factor adjustment argue that it amounts to accepting lower quality for these groups.²⁷

An alternative solution, which may be more politically palatable, is to adjust quality and cost measures not for the social risk factors themselves, but for medical factors that mediate the link between social risk and poor outcomes. For example, compared to people not dually enrolled in Medicaid, dual enrollees have a greater prevalence of dementia,^11,28 mental health conditions,^28,29 and other chronic conditions;²⁸ worse functional status;^10–12 and greater use of health services.^28,30,31 It is therefore possible that worse clinical and cost outcomes among socially at-risk people are due to higher levels of medical risk—for example, a higher prevalence of diabetes, heart disease, or lung disease. It is also possible that the worse outcomes are related to less easily measured markers of medical risk such as functional and cognitive status.²² Finally, it is possible that the real drivers of worse outcomes are the social risk factors themselves and that even at the same level of medical comorbidity and functional status, social risk is predictive.

Understanding the factors that underlie the relationship between social risk and outcomes is important because it would facilitate appropriate risk adjustment of quality and cost measures and ensure that such measures reflect the quality of care delivered rather than underlying differences outside the control of individual providers. Therefore, in this study we set out to comprehensively assess differences in medical comorbidities, social risk, and functional/cognitive status between dually enrolled and non–dually enrolled Medicare beneficiaries; quantify differences in annual costs of care between these two groups; and determine both the extent to which medical comorbidities, social risk, and functional/cognitive status mediate the relationship between dual enrollment status and costs of care and whether adjusting for these factors attenuates observed differences in spending.

Study Data And Methods

Data Sources And Study Population

We conducted a retrospective observational study using cohort data from the Medicare Current Beneficiary Survey (MCBS) linked to respondents’ fee-for-service Medicare claims and administrative data for the years 2006–13. The MCBS is an annual, nationally representative survey of the Medicare population with a rotating four-year longitudinal cohort design.³² It includes information on patient-reported social, cognitive, and functional risk factors that are not available in Medicare claims data.

We limited the study population to beneficiaries who had at least two years of continuous enrollment in fee-for-service Medicare Parts A and B that overlapped with two years of enrollment in the survey, and at least one physician or clinic visit in the baseline year—identified by evaluation and management visits (Berenson-Eggers Type of Service codes M1A, M1B, M4A, M4B, M5C, M5D, and M6 and revenue center codes 0521, 0522, and 0525) and positive covered dollar amounts. This ensured that all beneficiaries in our study were treated by outpatient providers before the occurrence of Medicare spending outcomes for which providers are held accountable under value-based payment. It also provided us with twelve months of baseline data from which to predict following-year cost outcomes. We excluded beneficiaries without a US ZIP code because they are excluded from Medicare value-based payment measures. We used MCBS longitudinal survey weights to compute nationally representative estimates of the fee-for-service Medicare population. These weights account for the overall selection probability of each person sampled and include adjustments for stratified sampling design, survey nonresponse, and coverage error.

We identified dually enrolled beneficiaries as fee-for-service beneficiaries with at least one month of Medicaid coverage during the baseline year, defined as any state Part A or Part B buy-in.

Study Variables

Our dependent variable was following-year Medicare spending. We measured total costs per beneficiary as the annual Medicare reimbursed amount for Parts A and B services, an approach similar to the total per capita costs measure used by Medicare in the Merit-based Incentive Payment System. We decomposed these costs into five mutually exclusive spending categories: inpatient, skilled nursing facility, home health, outpatient or other (including hospice and durable medical equipment), and physician. Following the system’s methodology, we bottom-coded all costs at the first-percentile value and top-coded at the ninety-ninth-percentile value.³³ All costs were converted to 2013 US dollars using the gross domestic product deflator.³⁴

Our independent variables were measured in the baseline year. Medical claims risk factors included age, sex, original reason for Medicare eligibility (disabled, end-stage renal disease, or age sixty-five or older), institutionalization in long-term care, Hierarchical Condition Categories (HCC) risk score, and Charlson Comorbidity Index. We also included seventy binary indicators for HCC disease categories.

We measured social, cognitive, and functional risk factors using MCBS data. We assessed social risk factors using self-reported measures of material capital (annual income), human capital (highest level of education attained), and social support (marital status). For cognitive risk factors, we used patient self-report (or patient’s proxy respondent’s report) of having been diagnosed either with depression or with Alzheimer disease or dementia, or a confirmatory diagnosis in their medical claims using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), codes defined by CMS.³⁵ For functional status, we grouped counts of self-reported difficulty with activities of daily living (ADLs) and instrumental ADLs (IADLs) into no limitations (0), mild or moderate limitations (1–2), and severe limitations (3–6).^17,36

Statistical Analysis

First, we computed weighted descriptive statistics from the MCBS and linked claims data on our baseline independent variables and following-year cost variables. We used the Wald test to compare differences in proportions (or means) for dually enrolled versus non–dually enrolled beneficiaries on these risk factors in the baseline year as well as on following-year costs.

Second, we estimated five patient-level multivariable regression models that predicted following-year Medicare costs under different baseline risk-adjustment scenarios: using age, age squared, and sex; adding to the first scenario the seventy HCC disease indicators as well as indicators for disabled status, end-stage renal disease, and institutionalization in long-term care; adding to the second scenario the cognitive and functional risk factor variables; adding to the third scenario the social risk factor variables; and using an indicator for dual enrollment. For each of these modeling scenarios, we estimated two-part regression models. The first part predicted any health care use via logistic regression, and the second part predicted spending conditional on use via ordinary least squares regression. Each model also included year fixed effects to control for secular trends and adjusted for the complex survey design of the MCBS and intraperson correlation over time. We estimated each of these five regression models for total costs as well as separately for each of the five cost categories (inpatient, skilled nursing facility, home health, outpatient or other, and physician).

We then used these regression models to predict expected costs under the five risk-adjustment modeling scenarios. This enabled us to compare mean observed costs to expected costs under each scenario separately for dual and nondual enrollees. We computed observed-to-expected cost ratios under these scenarios to determine which most accurately predicted the observed differences in following-year Medicare spending between dual and nondual enrollees.

Limitations

Our study had several limitations. First, we focused only on Medicare spending for dually enrolled beneficiaries and did not measure their spending in Medicaid. Medicaid spending for dual enrollees is concentrated in home, community-based, and institutional care services,²⁸ which are not included in Medicare value-based payment measures.

Second, we did not differentiate between beneficiaries who were fully and partially dually enrolled. Although the latter are demographically similar to the former, they tend to be healthier and use less medical care. Therefore, this could have biased our study to the null.³¹

Third, our study population was limited to fee-for-service Medicare beneficiaries for whom both medical claims and MCBS data were available. This excluded the 25 percent of the Medicare population enrolled in managed care during this period.³⁷

Fourth, and relatedly, we were able to identify only Medicare beneficiaries who participated in the survey. While survey and nonresponse weighting can significantly reduce bias related to this issue, our results might not be generalizable more broadly.

Fifth, Medicaid enrollment thresholds vary by state. Although this is less of an issue with Medicare beneficiaries because most states use the federal income standard for Supplemental Security Income to define eligibility within this population, minor differences between states in how enrollment and benefits are administered could have biased our results to the null via misclassification.

Study Results

Study Population And Baseline-Year Risk Factors

Of the 169,429,155 (weighted) patient-years eligible for inclusion in the study, we excluded 568,644 (0.3 percent) without a valid ZIP code (online appendix exhibit 1).³⁸ Our final study population consisted of 168,860,511 patient-years (42,557 unweighted patient-years). Of these, 31,095,750 (18.4 percent) were for dually enrolled beneficiaries and 137,764,761 (81.6 percent) were for non–dually enrolled beneficiaries.

Dually enrolled beneficiaries were significantly more likely to be younger than age sixty-five and to receive Medicare entitlement due to disability or end-stage renal disease, compared to non–dually enrolled beneficiaries (exhibit 1). Dual enrollees were also significantly more likely to be female and to be institutionalized in long-term care. In addition, dual enrollees exhibited greater medical complexity, as evidenced in their Charlson Comorbidity Index and HCC risk scores. Dual enrollees had double or nearly double the prevalence for the HCC categories of HIV/AIDS; septicemia or shock; diabetes with specific complications; chronic obstructive pulmonary disease; dialysis status; liver disease; hepatitis; seizure disorders; substance abuse disorders; schizophrenia; and major depressive, bipolar, and paranoid disorders (appendix exhibit 2).³⁸

The greatest difference between dually enrolled and non–dually enrolled beneficiaries was in social risk factors. Ninety-three percent of dual enrollees had annual incomes of less than $25,000, compared to 33 percent of nondual enrollees (exhibit 1). Similarly, over half of dual enrollees lacked a high school education, compared to 17 percent of nondual enrollees. Lastly, the proportion of dual enrollees who were unmarried was nearly double that of nondual enrollees (82 percent versus 42 percent).

Dually enrolled beneficiaries also exhibited a higher prevalence of cognitive and functional risk factors than nondual enrollees did. Forty-one percent of dual enrollees had depression, and 15 percent had Alzheimer disease or dementia, compared to 18 percent and 7 percent, respectively, for nondual enrollees. Twenty-seven percent of dual enrollees reported difficulty with 3–6 ADLs, and 40 percent reported difficulty with 3–6 IADLs, compared to 10 percent and 13 percent, respectively, for nondual enrollees.

Following-Year Cost Outcomes

The mean total Parts A and B costs for dually enrolled and non–dually enrolled beneficiaries were $11,928 and $8,310, respectively (exhibit 2). Total cost and costs for each of the five subcategories were significantly higher for dually enrolled beneficiaries. Inpatient, skilled nursing facility, home health, and outpatient or other costs were more than 50 percent greater for dual enrollees than for nondual enrollees. Physician cost was the only cost area that showed a relatively small difference (9 percent) between the two groups.

Impact Of Adjusting For Medical, Cognitive Or Functional, And Social Risk Factors

The observed-to-expected cost ratios for each of the five risk-adjustment scenarios for total costs and for the five cost subcategories are shown for dually enrolled beneficiaries in exhibit 3 and appendix exhibit 3, respectively, and full results with 95% confidence intervals are shown in appendix exhibit 4.³⁸ The same results are shown for non–dually enrolled beneficiaries in exhibit 4 and appendix exhibits 5 and 6.³⁸ An observed-to-expected ratio of 1 implies a perfect match between the prediction of following-year costs and observed following-year costs. From a value-based payment perspective, an observed-to-expected ratio greater than 1 would imply poor performance on the cost measures.

SOURCE Authors’ analysis of data for 2006–13 from the Medicare Current Beneficiary Survey. NOTES An observed-to-expected cost ratio of 1 indicates a match between predicted and observed costs in the following year. A ratio of less than 1 indicates that the model overpredicted costs, and a ratio of greater than 1 indicates that the model underpredicted costs—implying poor performance on cost measures. The risk-adjustment scenarios are explained in the text. HCC is Hierarchical Condition Categories.

SOURCE Authors’ analysis of data for 2006–13 from the Medicare Current Beneficiary Survey. NOTES The cost ratios are explained in the notes to exhibit 3. The risk-adjustment scenarios are explained in the text. HCC is Hierarchical Condition Categories.

Not surprisingly, the risk-adjusted predictions that included dual enrollment as an indicator consistently resulted in observed-to-expected ratios equal to 1 for both dual enrollees and nondual enrollees (exhibits 3 and 4). That result was expected because these models explicitly used the presence of baseline dual enrollment to predict following-year costs based on dual enrollment. Thus, these models are the gold standard against which we compared the observed-to-expected ratios from the other four risk-adjustment modeling scenarios to determine how accurately they predicted cost differences between dual and nondual enrollees.

For total Parts A and B costs for dual enrollees, the observed-to-expected ratio was 1.35 for the age-sex model, 1.12 for the HCC model, 1.07 for the cognitive and functional factors plus HCC model, 1.06 for the social, cognitive, and functional factors plus HCC model, and 1.00 for the dual eligibility model. For total Parts A and B costs for nondual enrollees, the observed-to-expected ratios were 0.92, 0.97, 0.98, 0.98, and 1.00, respectively. This implies that the HCC model systematically underpredicted costs for dual enrollees and overpredicted costs for non–dual enrollees. Furthermore, the risk-adjusted predictions that included social, cognitive, and functional risk factors were best able to discriminate between the differing risk profiles of dual enrollees and nondual enrollees in predicting their actual observed costs. The model for total costs that included social risk factors was only marginally better than the model that included cognitive and functional but not social risk factors. For the cost subcategories of inpatient, skilled nursing facility, and home health, however, adding social factors did meaningfully improve the predictive models.

Discussion

In this nationally representative study of Medicare beneficiaries, we found that dually enrolled Medicare beneficiaries had strikingly higher levels of medical, functional, and cognitive comorbidities, as well as social needs. Dual enrollees had significantly higher annual costs of care than their nondual enrollee counterparts had. Including functional, cognitive, and social factors in cost prediction, in addition to risk factors derived from medical claims, best predicted observed differences between the two groups of beneficiaries.

Our findings confirmed prior work showing that dually enrolled beneficiaries have higher levels of risk^{10–12,28,29} and higher spending^28,30,31 than nondual enrollees do, and we extended the prior work to a national sample with rich patient-reported data. Furthermore, our findings suggest that cognitive and functional factors, which differed markedly between the two groups, may be important areas for further research into targeted interventions to improve health outcomes and reduce disparities. For example, home-based primary care has shown significant promise for reducing costs and adverse health events in frail elders.^39–41 Such programs might also benefit dual enrollees with poor functional status, many of whom may also be at risk for frailty.

We also found that adjusting cost measures for factors that may mediate the relationship between dual enrollment status and utilization—including cognitive, functional, and social factors—notably improved the match between expected and observed costs for both dually enrolled and non–dually enrolled beneficiaries. Within the dual enrollee group, expected costs were higher after adding these factors, bringing the observed-to-expected ratio lower and closer to 1. Within the nondual enrollee group, expected costs were lower after adding these factors, bringing the observed-to-expected ratio higher and closer to 1. This suggests that such adjustment would not only reduce potentially inappropriate penalties among providers that disproportionately care for vulnerable populations but would also reduce inappropriate bonuses for providers that care for less complex populations.

Adding functional, cognitive, and social factors to risk models is often not done because these data elements are not routinely collected on all Medicare beneficiaries, and the effort required to do so would be prohibitive logistically and financially. However, there is increasing interest in using claims data to detect impaired functional status and frailty, and a number of recent publications suggest that validated claims-based indices perform reasonably well.^42–44 If similar efforts were undertaken for cognitive status, such adjustment might be broadly feasible without additional resources. Social factors, such as marital status and educational attainment, are harder to determine from claims. However, the move from ICD-9 to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), introduced a number of diagnostic codes for social risk factors⁴⁵ such as homelessness, unemployment, income, education, and literacy. If risk-adjustment and payment schemes included these factors, it is probable that clinicians would increase their coding of them commensurately.⁴⁶ Such coding could allow for broader adjustment of outcomes for social factors where appropriate.

Finally, dual enrollment in Medicare and Medicaid appears to be a very good marker for social, cognitive, and functional risk factors that are generally unmeasured in Medicare beneficiaries. Beginning in 2019 CMS began to adjust hospital readmissions in the Hospital Readmissions Reduction Program for the dual enrollment status of patients by stratifying hospitals by the proportion of Medicare patients who are dual enrollees.⁴⁷ Given the current lack of availability of these social, cognitive, and functional risk factors in Medicare claims data, our findings suggest that the use of an indicator of dual enrollment for risk adjustment of value-based payment cost outcomes by Medicare is a good first step toward adjusting for beneficiaries’ risk profiles. However, Medicare should exercise caution in implementing adjustment for dual status because Medicaid enrollment thresholds vary by state, and such adjustment could systematically penalize providers in states with less generous thresholds.

Conclusion

We found that dually enrolled Medicare beneficiaries had much higher levels of medical, functional, cognitive, and social risk factors and had higher spending, compared to beneficiaries not dually enrolled. Adding these factors to cost measures improved prediction and decreased differences between the two groups of beneficiaries. Medicare could consider using such adjustment to improve accuracy and fairness in value-based payment programs in the future.

ACKNOWLEDGMENTS

Kenton Johnston received support from Saint Louis University, which purchased and provided access to the data used in this study. Karen Joynt Maddox is supported by the National Heart, Lung, and Blood Institute (Grant No. K23-HL109177-03). Joynt Maddox does work under contract with the Office of the Assistant Secretary for Planning and Evaluation in the Department of Health and Human Services. The authors gratefully acknowledge Leslie Hinyard for her thoughtful review of the manuscript.

NOTES

• 1 Fontanarosa PB, McNutt RA. Revisiting hospital readmissions. JAMA. 2013;309(4):398–400. Crossref, Medline, Google Scholar
• 2 Joynt KE, Jha AK. Characteristics of hospitals receiving penalties under the Hospital Readmissions Reduction Program. JAMA. 2013;309(4):342–3 Crossref, Medline, Google Scholar
• 3 Buntin MB, Ayanian JZ. Social risk factors and equity in Medicare payment. N Engl J Med. 2017;376(6):507–10. Crossref, Medline, Google Scholar
• 4 Joynt KE, De Lew N, Sheingold SH, Conway PH, Goodrich K, Epstein AM. Should Medicare value-based purchasing take social risk into account? N Engl J Med. 2017;376(6):510–3. Crossref, Medline, Google Scholar
• 5 Franks P, Tancredi DJ, Winters P, Fiscella K. Including socioeconomic status in coronary heart disease risk estimation. Ann Fam Med. 2010;8(5):447–53. Crossref, Medline, Google Scholar
• 6 Braveman PA, Cubbin C, Egerter S, Williams DR, Pamuk E. Socioeconomic disparities in health in the United States: what the patterns tell us. Am J Public Health. 2010;100(Suppl 1):S186–96. Crossref, Medline, Google Scholar
• 7 Galea S, Tracy M, Hoggatt KJ, Dimaggio C, Karpati A. Estimated deaths attributable to social factors in the United States. Am J Public Health. 2011;101(8):1456–65. Crossref, Medline, Google Scholar
• 8 Muennig P, Fiscella K, Tancredi D, Franks P. The relative health burden of selected social and behavioral risk factors in the United States: implications for policy. Am J Public Health. 2010;100(9):1758–64. Crossref, Medline, Google Scholar
• 9 Bernheim SM, Spertus JA, Reid KJ, Bradley EH, Desai RA, Peterson EDet al. Socioeconomic disparities in outcomes after acute myocardial infarction. Am Heart J. 2007;153(2):313–9. Crossref, Medline, Google Scholar
• 10 Komisar HL, Feder J, Kasper JD. Unmet long-term care needs: an analysis of Medicare-Medicaid dual eligibles. Inquiry. 2005;42(2):171–82. Crossref, Medline, Google Scholar
• 11 Allen SM, Piette ER, Mor V. The adverse consequences of unmet need among older persons living in the community: dual-eligible versus Medicare-only beneficiaries. J Gerontol B Psychol Sci Soc Sci. 2014;69(Suppl 1):S51–8. Crossref, Medline, Google Scholar
• 12 Medicare Payment Advisory Commission, Medicaid and CHIP Payment and Access Commission. Data book: beneficiaries dually eligible for Medicare and Medicaid [Internet]. Washington (DC): MACPAC; 2018 Jan [cited 2019 Feb 6]. Available from: https://www.macpac.gov/wp-content/uploads/2017/01/Jan18_MedPAC_MACPAC_DualsDataBook.pdf Google Scholar
• 13 Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation. Report to Congress: social risk factors and performance under Medicare’s Value-Based Purchasing Programs [Internet]. Washington (DC): ASPE; 2016 Dec [cited 2019 Feb 6]. Available from: https://aspe.hhs.gov/system/files/pdf/253971/ASPESESRTCfull.pdf Google Scholar
• 14 Samson LW, Chen LM, Epstein AM, Maddox KEJ. Dually enrolled beneficiaries have higher episode costs on the Medicare spending per beneficiary measure. Health Aff (Millwood). 2018;37(1):86–94. Go to the article, Google Scholar
• 15 Walsh EG, Wiener JM, Haber S, Bragg A, Freiman M, Ouslander JG. Potentially avoidable hospitalizations of dually eligible Medicare and Medicaid beneficiaries from nursing facility and Home- and Community-Based Services waiver programs. J Am Geriatr Soc. 2012;60(5):821–9. Crossref, Medline, Google Scholar
• 16 Joynt Maddox KE. Social and behavioral determinants of spending. JAMA Intern Med. 2017;177(10):1431–2. Crossref, Medline, Google Scholar
• 17 Meddings J, Reichert H, Smith SN, Iwashyna TJ, Langa KM, Hofer TPet al. The impact of disability and social determinants of health on condition-specific readmissions beyond Medicare risk adjustments: a cohort study. J Gen Intern Med. 2017;32(1):71–80. Crossref, Medline, Google Scholar
• 18 Barnett ML, Hsu J, McWilliams JM. Patient characteristics and differences in hospital readmission rates. JAMA Intern Med. 2015;175(11):1803–12. Crossref, Medline, Google Scholar
• 19 Calvillo-King L, Arnold D, Eubank KJ, Lo M, Yunyongying P, Stieglitz Het al. Impact of social factors on risk of readmission or mortality in pneumonia and heart failure: systematic review. J Gen Intern Med. 2013;28(2):269–82. Crossref, Medline, Google Scholar
• 20 Kind AJ, Jencks S, Brock J, Yu M, Bartels C, Ehlenbach Wet al. Neighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study. Ann Intern Med. 2014;161(11):765–74. Crossref, Medline, Google Scholar
• 21 Oronce CIA, Shao H, Shi L. Disparities in 30-day readmissions after total hip arthroplasty. Med Care. 2015;53(11):924–30. Crossref, Medline, Google Scholar
• 22 Johnston KJ, Wen H, Hockenberry JM, Joynt Maddox KE. Association between patient cognitive and functional status and Medicare total annual cost of care: implications for value-based payment. JAMA Intern Med. 2018;178(11):1489–97. Crossref, Medline, Google Scholar
• 23 Roberts ET, Zaslavsky AM, McWilliams JM. The Value-Based Payment Modifier: program outcomes and implications for disparities. Ann Intern Med. 2018;168(4):255–65. Crossref, Medline, Google Scholar
• 24 Chen LM, Epstein AM, Orav EJ, Filice CE, Samson LW, Joynt Maddox KE. Association of practice-level social and medical risk with performance in the Medicare Physician Value-Based Payment Modifier Program. JAMA. 2017;318(5):453–61. Crossref, Medline, Google Scholar
• 25 Markovitz AA, Ellimoottil C, Sukul D, Mullangi S, Chen LM, Nallamothu BKet al. Risk adjustment may lessen penalties on hospitals treating complex cardiac patients under Medicare’s bundled payments. Health Aff (Millwood). 2017;36(12):2165–74. Go to the article, Google Scholar
• 26 Durfey SNM, Kind AJH, Gutman R, Monteiro K, Buckingham WR, DuGoff EHet al. Impact of risk adjustment for socioeconomic status on Medicare Advantage plan quality rankings. Health Aff (Millwood). 2018;37(7):1065–72. Go to the article, Google Scholar
• 27 Krumholz HM, Bernheim SM. Considering the role of socioeconomic status in hospital outcomes measures. Ann Intern Med. 2014;161(11):833–4. Crossref, Medline, Google Scholar
• 28 Kasper J, Watts MO, Lyons B. Chronic disease and co-morbidity among dual eligibles: implications for patterns of Medicaid and Medicare service use and spending [Internet]. Washington (DC): Kaiser Commission on Medicaid and the Uninsured; 2010 Jul [cited 2019 Feb 7]. (Issue Paper). Available from: https://kaiserfamilyfoundation.files.wordpress.com/2013/01/8081.pdf Google Scholar
• 29 Kuramoto-Crawford J, Zodet M. Behavioral health conditions and health care expenditures of adults aged 18 to 64 dually eligible for Medicaid and Medicare [Internet]. Rockville (MD): Substance Abuse and Mental Health Services Administration; 2014 Jul 15 [cited 2019 Feb 7]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK384683/pdf/Bookshelf_NBK384683.pdf Google Scholar
• 30 Moon S, Shin J. Health care utilization among Medicare-Medicaid dual eligibles: a count data analysis. BMC Public Health. 2006;6(1):88. Crossref, Medline, Google Scholar
• 31 Congressional Budget Office. Dual-eligible beneficiaries of Medicare and Medicaid: characteristics, health care spending, and evolving policies [Internet]. Washington (DC): CBO; 2013 Jun [cited 2019 Feb 7]. Available from: https://www.cbo.gov/sites/default/files/113th-congress-2013-2014/reports/44308dualeligibles2.pdf Google Scholar
• 32 Centers for Medicare and Medicaid Services. Technical documentation for the Medicare Current Beneficiary Survey. In: CMS. Medicare Current Beneficiary Survey data tables [Internet]. Baltimore (MD): CMS; 2003 [cited 2019 Feb 8]. p. 227–38. Available from: https://www.cms.gov/Research-Statistics-Data-and-Systems/Research/MCBS/Downloads/OLD/HHC2003appendixA.pdf Google Scholar
• 33 Centers for Medicare and Medicaid Services. 2016 measure information about the Per Capita Costs for All Attributed Beneficiaries measure, calculated for the 2018 Value-Based Payment Modifier program [Internet]. Baltimore (MD): CMS; [last updated 2017 Aug; cited 2019 Feb 7]. Available from: https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/PhysicianFeedbackProgram/Downloads/2016-TPCC-MIF.pdf Google Scholar
• 34 Johnston L, Williamson SH. What was the U.S. GDP then? [Internet]. Chicago (IL): MeasuringWorth; 2019 [cited 2019 Feb 8]. Available from: https://www.measuringworth.com/datasets/usgdp/ Google Scholar
• 35 Chronic Conditions Data Warehouse. CCW chronic condition algorithms [Internet]. Baltimore (MD): Centers for Medicare and Medicaid Services; [last revised 2018 Jun; cited 2019 Feb 7]. Available for download from: https://www.ccwdata.org/web/guest/condition-categories Google Scholar
• 36 Ivanova JI, Mytelka DS, Duh MS, Birnbaum HG, Cummings AK, San Roman AMet al. Evaluating the utility of existing patient-reported outcome scales in novel patient populations with pancreatic cancer, lung cancer, and myeloproliferative neoplasms using Medicare Current Beneficiary Survey data. Patient. 2013;6(3):189–200. Crossref, Medline, Google Scholar
• 37 Medicare Payment Advisory Commission. A data book: health care spending and the Medicare program. Washington (DC): MedPAC; 2013 Jun [cited 2019 Feb 7]. Available from: https://c.ymcdn.com/sites/www.michbio.org/resource/resmgr/Policy_-_MedicareB/MedPac_A_Data_Book_Health_Ca.pdf Google Scholar
• 38 To access the appendix, click on the Details tab of the article online.
• 39 De Jonge KE, Jamshed N, Gilden D, Kubisiak J, Bruce SR, Taler G. Effects of home-based primary care on Medicare costs in high-risk elders. J Am Geriatr Soc. 2014;62(10):1825–31. Crossref, Medline, Google Scholar
• 40 Edes T, Kinosian B, Vuckovic NH, Nichols LO, Becker MM, Hossain M. Better access, quality, and cost for clinically complex veterans with home-based primary care. J Am Geriatr Soc. 2014;62(10):1954–61. Crossref, Medline, Google Scholar
• 41 Stanhope SA, Cooley MC, Ellington LF, Gadbois GP, Richardson AL, Zeddes TCet al. The effects of home-based primary care on Medicare costs at Spectrum Health/Priority Health (Grand Rapids, MI, USA) from 2012–present: a matched cohort study. BMC Health Serv Res. 2018;18(1):161. Crossref, Medline, Google Scholar
• 42 Kim DH, Schneeweiss S, Glynn RJ, Lipsitz LA, Rockwood K, Avorn J. Measuring frailty in Medicare data: development and validation of a claims-based frailty index. J Gerontol A Biol Sci Med Sci. 2018;73(7):980–7. Crossref, Medline, Google Scholar
• 43 Segal JB, Chang H-Y, Du Y, Walston JD, Carlson MC, Varadhan R. Development of a claims-based frailty indicator anchored to a well-established frailty phenotype. Med Care. 2017;55(7):716–22. Crossref, Medline, Google Scholar
• 44 Joynt KE, Figueroa JF, Beaulieu N, Wild RC, Orav EJ, Jha AK. Segmenting high-cost Medicare patients into potentially actionable cohorts. Healthc (Amst). 2017;5(1-2):62–7. Crossref, Medline, Google Scholar
• 45 American Hospital Association. ICD-10-CM coding for social determinants of health [Internet]. Chicago (IL): AHA; 2018 Apr [cited 2019 Feb 7]. Available from: http://www.ahacentraloffice.org/PDFS/2018PDFS/value-initiative-icd-10-code-sdoh-0418.pdf Google Scholar
• 46 Reidhead M. Decoding social determinants of health [Internet]. Jefferson City (MO): Missouri Hospital Association; 2018 Sep [cited 2019 Feb 7]. (Policy Brief). Available from: https://www.mhanet.com/mhaimages/Policy_Briefs/PolicyBrief_SDOH.pdf Google Scholar
• 47 Centers for Medicare and Medicaid Services. New Stratified Methodology Hospital-Level Impact File user guide: Hospital Readmissions Reduction Program [Internet]. Baltimore (MD): CMS; 2017 Nov [cited 2019 Feb 7]. Available from: https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Downloads/HRRP_StratMethod_ImpctFile_UG.PDF Google Scholar