Association of Medicaid Eligibility With Surgical Readmission Among Medicare Beneficiaries | Health Disparities | JAMA Network Open | JAMA Network
[Skip to Content]
[Skip to Content Landing]
Figure 1.  Flow Diagram Showing Sample Exclusions
Flow Diagram Showing Sample Exclusions

aTo build the Medicare claims cohort, we extracted data for patients undergoing the set of procedures included in the Michigan Surgical Quality Collaborative (MSQC) clinical registry, including: appendectomy, cholecystectomy, colectomy, proctectomy, esophagectomy, gastrectomy, hepatectomy, hernia repair, pancreatectomy, antireflux surgery, hysterectomy, amputations, abdominal aortic aneurysm repair, aortoiliac, carotid endarterectomy, lower extremity bypass, splenectomy, thyroidectomy, adrenalectomy, and mastectomy.

Figure 2.  Association of Dual Eligibility With Hospital Profiling Using Risk-Adjusted Readmission Rates After Surgery at Michigan Hospitals
Association of Dual Eligibility With Hospital Profiling Using Risk-Adjusted Readmission Rates After Surgery at Michigan Hospitals

aC statistic = 6.6421.

bC statistic = 0.6436.

cC statistic = 0.6619.

dC statistic = 0.6620.

Table 1.  Characteristics of the Study Population
Characteristics of the Study Population
Table 2.  Association Between Dual Eligibility and Risk-Adjusted Readmission After Surgery at Michigan Hospitals in the Medicare Claims Cohort and MSQC Clinical Registry Cohort
Association Between Dual Eligibility and Risk-Adjusted Readmission After Surgery at Michigan Hospitals in the Medicare Claims Cohort and MSQC Clinical Registry Cohort
1.
Joynt Maddox  KE, Reidhead  M, Qi  AC, Nerenz  DR.  Association of stratification by dual enrollment status with financial penalties in the Hospital Readmissions Reduction Program.   JAMA Intern Med. 2019;179(6):769-776. doi:10.1001/jamainternmed.2019.0117 PubMedGoogle ScholarCrossref
2.
Barnett  ML, Hsu  J, McWilliams  JM.  Patient characteristics and differences in hospital readmission rates.   JAMA Intern Med. 2015;175(11):1803-1812. doi:10.1001/jamainternmed.2015.4660 PubMedGoogle ScholarCrossref
3.
Gaskin  DJ, Zare  H, Vazin  R, Love  D, Steinwachs  D.  Racial and ethnic composition of hospitals’ service areas and the likelihood of being penalized for excess readmissions by the Medicare program.   Med Care. 2018;56(11):934-943. doi:10.1097/MLR.0000000000000988 PubMedGoogle ScholarCrossref
4.
Sills  MR, Hall  M, Cutler  GJ,  et al.  Adding social determinant data changes children’s hospitals’ readmissions performance.   J Pediatr. 2017;186:150-157. doi:10.1016/j.jpeds.2017.03.056 PubMedGoogle ScholarCrossref
5.
Sills  MR, Hall  M, Colvin  JD,  et al.  Association of social determinants with children’s hospitals’ preventable readmissions performance.   JAMA Pediatr. 2016;170(4):350-358. doi:10.1001/jamapediatrics.2015.4440 PubMedGoogle ScholarCrossref
6.
Roberts  ET, Zaslavsky  AM, Barnett  ML, Landon  BE, Ding  L, McWilliams  JM.  Assessment of the effect of adjustment for patient characteristics on hospital readmission rates: implications for pay for performance.   JAMA Intern Med. 2018;178(11):1498-1507. doi:10.1001/jamainternmed.2018.4481 PubMedGoogle ScholarCrossref
7.
Meddings  J, Reichert  H, Smith  SN,  et 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. doi:10.1007/s11606-016-3869-x PubMedGoogle ScholarCrossref
8.
Hoffman  GJ, Hsuan  C, Braun  T, Ponce  N.  Health equity and hospital readmissions: does inclusion of patient functional and social complexity improve predictiveness?   J Gen Intern Med. 2019;34(1):26-28. doi:10.1007/s11606-018-4635-z PubMedGoogle ScholarCrossref
9.
Blum  AB, Egorova  NN, Sosunov  EA,  et al.  Impact of socioeconomic status measures on hospital profiling in New York City.   Circ Cardiovasc Qual Outcomes. 2014;7(3):391-397. doi:10.1161/CIRCOUTCOMES.113.000520 PubMedGoogle ScholarCrossref
10.
McCarthy  CP, Vaduganathan  M, Patel  KV,  et al.  Association of the new peer group–stratified method with the reclassification of penalty status in the Hospital Readmission Reduction Program.   JAMA Netw Open. 2019;2(4):e192987. doi:10.1001/jamanetworkopen.2019.2987 PubMedGoogle Scholar
11.
Joynt Maddox  KE, Reidhead  M, Hu  J,  et al.  Adjusting for social risk factors impacts performance and penalties in the Hospital Readmissions Reduction Program.   Health Serv Res. 2019;54(2):327-336. doi:10.1111/1475-6773.13133 PubMedGoogle ScholarCrossref
12.
Centers for Medicare & Medicaid Services. Hospital Readmissions Reduction Program (HRRP). Updated February 11, 2020. Accessed July 14, 2019. https://www.cms.gov/medicare/medicare-fee-for-service-payment/acuteinpatientpps/readmissions-reduction-program.html
13.
Ibrahim  AM, Nathan  H, Thumma  JR, Dimick  JB.  Impact of the Hospital Readmission Reduction Program on surgical readmissions among Medicare beneficiaries.   Ann Surg. 2017;266(4):617-624. doi:10.1097/SLA.0000000000002368 PubMedGoogle ScholarCrossref
14.
Glance  LG, Kellermann  AL, Osler  TM, Li  Y, Li  W, Dick  AW.  Impact of risk adjustment for socioeconomic status on risk-adjusted surgical readmission rates.   Ann Surg. 2016;263(4):698-704. doi:10.1097/SLA.0000000000001363 PubMedGoogle ScholarCrossref
15.
Anderson  JE, Li  Z, Romano  PS, Parker  J, Chang  DC.  Should risk adjustment for surgical outcomes reporting include sociodemographic status? a study of coronary artery bypass grafting in California.   J Am Coll Surg. 2016;223(2):221-230. doi:10.1016/j.jamcollsurg.2016.05.008 PubMedGoogle ScholarCrossref
16.
Martin  JR, Wang  TY, Loriaux  D,  et al.  Race as a predictor of postoperative hospital readmission after spine surgery.   J Clin Neurosci. 2017;46:21-25. doi:10.1016/j.jocn.2017.08.015PubMedGoogle ScholarCrossref
17.
Strombom  P, Widmar  M, Keskin  M,  et al.  Assessment of the value of comorbidity indices for risk adjustment in colorectal surgery patients.   Ann Surg Oncol. 2019;26(9):2797-2804. doi:10.1245/s10434-019-07502-9 PubMedGoogle ScholarCrossref
18.
Campbell  DA  Jr, Englesbe  MJ, Kubus  JJ,  et al.  Accelerating the pace of surgical quality improvement: the power of hospital collaboration.   Arch Surg. 2010;145(10):985-991. doi:10.1001/archsurg.2010.220PubMedGoogle ScholarCrossref
19.
Englesbe  MJ, Brooks  L, Kubus  J,  et al.  A statewide assessment of surgical site infection following colectomy: the role of oral antibiotics.   Ann Surg. 2010;252(3):514-519. doi:10.1097/SLA.0b013e3181f244f8 PubMedGoogle Scholar
20.
Portier  K, Ida  KK.  The ASA Physical Status Classification: what is the evidence for recommending its use in veterinary anesthesia?—a systematic review.   Front Vet Sci. 2018;5:204. doi:10.3389/fvets.2018.00204 PubMedGoogle ScholarCrossref
21.
Hu  J, Gonsahn  MD, Nerenz  DR.  Socioeconomic status and readmissions: evidence from an urban teaching hospital.   Health Aff (Millwood). 2014;33(5):778-785. doi:10.1377/hlthaff.2013.0816 PubMedGoogle ScholarCrossref
22.
Talutis  SD, Chen  Q, Wang  N, Rosen  AK.  Comparison of risk-standardized readmission rates of surgical patients at safety-net and non–safety-net hospitals using Agency for Healthcare Research and Quality and American Hospital Association data.   JAMA Surg. 2019;154(5):391-400. doi:10.1001/jamasurg.2018.5242 PubMedGoogle ScholarCrossref
23.
Gu  Q, Koenig  L, Faerberg  J, Steinberg  CR, Vaz  C, Wheatley  MP.  The Medicare Hospital Readmissions Reduction Program: potential unintended consequences for hospitals serving vulnerable populations.   Health Serv Res. 2014;49(3):818-837. doi:10.1111/1475-6773.12150 PubMedGoogle ScholarCrossref
24.
Bynum  J, Lewis  V.  Value-based payments and inaccurate risk adjustment—who is harmed?   JAMA Intern Med. 2018;178(11):1507-1508. doi:10.1001/jamainternmed.2018.4142 PubMedGoogle ScholarCrossref
25.
National Academies of Sciences, Engineering, and Medicine 2017. Kwan LY, Stratton K, Steinwach DM, eds.  Accounting for Social Risk Factors in Medicare Payment: Identifying Social Risk Factors. National Academies Press; 2016.
26.
Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation. 2013 Measures updates and specifications: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized mortality measure (version 7.0). Published March 2013. Accessed March 13, 2020. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Mortality_AMI-HF-PN_Measures_Updates_Report_FINAL_06-13-2013.pdf
27.
Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation. Hospital-level 30-day all-cause unplanned readmission following coronary artery bypass graft surgery. Published February 1, 2014. Accessed March 13, 2020. https://www.qualitynet.org/inpatient/measures/readmission
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
    Original Investigation
    Health Policy
    June 10, 2020

    在医疗保险受益人中,医疗补助资格与手术后再入院率的关联性

    Author Affiliations
    • 1University of Michigan Medical School, Ann Arbor
    • 2Center for Healthcare Outcomes and Policy, Ann Arbor, Michigan
    • 3Department of Health Management and Policy, University of Michigan School of Public Health, Ann Arbor
    • 4University of Michigan Institute for Healthcare Policy and Innovation, Ann Arbor
    • 5Department of Systems, Population, and Leadership, University of Michigan School of Nursing, Ann Arbor
    • 6Department of Surgery, University of Michigan Medical School, Ann Arbor
    JAMA Netw Open. 2020;3(6):e207426. doi:10.1001/jamanetworkopen.2020.7426
    关键点 español English

    问题  在手术后再入院率的风险调整模型中,医疗保险和医疗补助的双重资格是否与手术后 30 天全因再入院率相关?

    结果  这项队列研究涉及 55,651 名医疗保险受益人。研究发现,使用标准的、基于索赔的风险调整后,双重资格与再入院率相关,但在对整个州临床注册机构测量的其他临床风险因素进行调整后,这种关联性有所减弱。调整医疗保险和医疗补助的双重资格与医院基于风险调整后的再入院率的排名几乎没有关联。

    意义  双重资格状态可能反映了未衡量的临床风险,而不是与手术后再入院率相关的社会风险因素。

    Abstract

    Importance  The Centers for Medicare & Medicaid Services is beginning to consider adjusting for social risk factors, such as dual eligibility for Medicare and Medicaid, when evaluating hospital performance under value-based purchasing programs. It is unknown whether dual eligibility represents a unique domain of social risk or instead represents clinical risk unmeasured by variables available in traditional Medicare claims.

    Objective  To assess how dual eligibility for Medicare and Medicaid is associated with risk-adjusted readmission rates after surgery.

    Design, Setting, and Participants  A retrospective cohort study was conducted of 55 651 Medicare beneficiaries undergoing general, vascular, and gynecologic surgery at 62 hospitals in Michigan between January 1, 2014, and December 1, 2016. Representative cohorts were derived from traditional Medicare claims (n = 29 710) and the Michigan Surgical Quality Collaborative (MSQC) clinical registry (n = 25 941), which includes additional measures of clinical risk. Statistical analysis was conducted between April 10 and July 15, 2019. The association between dual eligibility and risk-adjusted 30-day readmission rates after surgery was compared between models inclusive and exclusive of additional measurements of clinical risk. The study also examined how dual eligibility is associated with hospital profiling using risk-adjusted readmission rates.

    Exposures  Dual eligibility for Medicare and Medicaid.

    Main Outcomes and Measures  Risk-adjusted all-cause 30-day readmission after surgery.

    Results  There were a total of 3986 dual-eligible beneficiaries in the Medicare claims cohort (2554 women; mean [SD] age, 72.9 [6.9] years) and 1608 dual-eligible beneficiaries in the MSQC cohort (990 women; mean [SD] age, 72.9 [6.8] years). In both data sets, higher proportions of dual-eligible beneficiaries were younger, female, and nonwhite than Medicare-only beneficiaries (Medicare claims cohort: female, 2554 of 3986 [64.1%] vs 12 879 of 25 724 [50.1%]; nonwhite, 1225 of 3986 [30.7%] vs 2783 of 25 724 [10.8%]; MSQC cohort: female, 990 of 1608 [61.6%] vs 12 578 of 24 333 [51.7%]; nonwhite, 416 of 1608 [25.9%] vs 2176 of 24 333 [8.9%]). In the Medicare claims cohort, dual-eligible beneficiaries were more likely to be readmitted (15.5% [95% CI, 13.7%-17.3%]) than Medicare-only beneficiaries (13.3% [95% CI, 12.7%-13.9%]; difference, 2.2 percentage points [95% CI, 0.4-3.9 percentage points]). In the MSQC cohort, after adjustment for more granular measures of clinical risk, dual eligibility was not significantly associated with readmission (difference, 0.6 percentage points [95% CI, –1.0 to 2.2 percentage points]). In both the Medicare claims and MSQC cohorts, adding dual eligibility to risk-adjustment models had little association with hospital ranking using risk-adjusted readmission rates.

    Conclusions and Relevance  This study suggests that dual eligibility for Medicare and Medicaid may reflect unmeasured clinical risk of readmission in claims data. Policy makers should consider incorporating more robust measures of social risk into risk-adjustment models used by value-based purchasing programs.

    Introduction

    The Centers for Medicare & Medicaid Services is beginning to consider adjusting for social risk factors when evaluating hospital performance under value-based purchasing (VBP) programs.1 Prior work suggests that accounting for social risk factors may help to address disproportionate penalties received by hospitals caring for a large proportion of patients with high social risk.2-9 For example, after the Hospital Readmissions Reduction Program (HRRP) began to adjust for social risk by stratifying hospitals by their proportion of patients with dual eligibility for Medicare and Medicaid, hospitals with a greater share of dual-eligible patients were less likely to be penalized.1,10,11 Annual per-hospital penalties decreased by as much as $436 000,1 and 14% of hospitals in the highest dual-eligibility quintile no longer received penalties.10

    However, research on the association between performance under VBP programs and social risk has focused primarily on patients with medical conditions.1-11 The effect of accounting for social risk factors in readmissions after surgical care is less clear. Surgical care is an important area of study because surgical populations are being incorporated into VBP programs with greater regularity12 and are already a common focus for quality improvement within hospitals.13 Social risk factors are established factors associated with surgical readmissions,14-16 but this work is limited to younger patient populations and may not be generalizable to the Medicare population in which alternative payment demonstrations often occur. There are also concerns about whether measures of social risk, such as dual eligibility, which is currently used by some VBP programs,1,10,11 may represent unmeasured clinical risk severity because variables traditionally available in Medicare claims (eg, comorbidity indices) are imperfectly associated with surgical outcomes.17 Instead of measuring true social risk, dual eligibility may be a proxy for clinical illness severity unmeasured by the variables available in traditional Medicare claims. Whether dual eligibility represents a unique domain of social risk, and thus represents the true consequences of adjusting for dual eligibility in risk-adjustment models of surgical readmission rates, is challenging to explore using claims data alone.

    In this context, we investigated how dual eligibility for Medicare and Medicaid is associated with hospital profiling using risk-adjusted surgical readmission rates at Michigan hospitals. We examined the association between dual eligibility and risk-adjusted readmission in complementary data sets: (1) Medicare claims and (2) a statewide clinical registry inclusive of additional measures to more fully account for clinical illness severity. We then compared between the data sets how the inclusion of dual eligibility into risk-adjustment models is associated with hospital profiling using risk-adjusted readmission rates. This assessed the extent to which dual eligibility, a commonly used proxy for social risk, is independent from clinical risk and therefore a unique domain of risk for readmission after surgery.

    Methods
    Study Population

    Our study included Medicare patients undergoing surgical procedures from 20 common surgical procedure categories at 62 Michigan hospitals between January 1, 2014, and December 1, 2016. Patients at hospitals with fewer than 10 readmissions during the study period were excluded from the analysis. We included the following procedure groups: appendectomy, cholecystectomy, colectomy, proctectomy, esophagectomy, gastrectomy, hepatectomy, hernia repair, pancreatectomy, antireflux surgery, hysterectomy, amputations, abdominal aortic aneurysm repair, aortoiliac, carotid endarterectomy, lower extremity bypass, splenectomy, thyroidectomy, adrenalectomy, and mastectomy. We verified each case by ensuring that Current Procedural Terminology codes corresponded to International Classification of Diseases, Ninth Revision or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes as appropriate. A flow diagram of sample exclusions is presented in Figure 1. Our study was deemed exempt by the University of Michigan Institutional Review Board, which waived the requirement for patient consent because the data were deidentified. This study was designed and reported in adherence to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

    Data Source–Medicare Claims

    We extracted data from the 100% Medicare Provider Analysis and Review files for all surgical procedures of interest during the study period. We excluded patients younger than 65 years at the time of surgery. Data collected for this analysis included patient demographic characteristics, baseline comorbidities, dual eligibility, and 30-day readmission rates.

    Data Source–Michigan Surgical Quality Collaborative

    We generated a similar surgical cohort using data from the Michigan Surgical Quality Collaborative (MSQC). The MSQC is a prospective clinical registry formed by a partnership between Blue Cross Blue Shield of Michigan and 73 Michigan hospitals. Data definitions and collection protocols have been described previously.18,19 Data were manually extracted from the electronic medical record by nurses at each hospital who were independent of financial incentives, and accuracy was ensured by rigorous training and data audits. Data collected for this analysis included patient demographic characteristics, baseline comorbidities, dual eligibility, 30-day readmission rates, and results of a preoperative clinical examination performed by the surgical team. This assessment included 2 clinical measures unavailable in traditional Medicare fee-for-service claims: functional status and American Society of Anesthesiologists (ASA) classification of physical status. Functional status measures patient independence on a 3-point ordinal scale: independent, partially dependent (requiring assistance with the instrumental activities of daily living), or totally dependent (requiring assistance with all activities of daily living). American Society of Anesthesiologists classification is widely used to describe perioperative risk and is associated with postsurgical morbidity and mortality.20 A higher ASA classification is associated with worse postsurgical outcomes.

    Exposure and Outcome

    The study outcome was all-cause 30-day readmission after surgery. Readmission was defined as any hospitalization within 30 days of the index surgery. Exposures were dual eligibility for Medicare and Medicaid and clinical risk. Dual eligibility is commonly used as a marker of social risk.10 In the Medicare data set, dual eligibility was defined as eligibility for Medicaid at any point during the year in which the surgery was performed. In the MSQC data set, dual eligibility was defined as eligibility for Medicaid at the time of surgery. Clinical risk was measured using functional status and ASA classification of physical status.

    Statistical Analysis

    Statistical analysis was conducted between April 10 and July 15, 2019. We determined the association between dual eligibility and risk-adjusted readmission by estimating the marginal effect of dual eligibility in the most comprehensive risk-adjustment model possible in each data set. In the Medicare data set, we modeled 30-day readmission using age, sex, case mix, dual eligibility, and Elixhauser comorbidities. We assessed the association between dual eligibility and risk-adjusted readmission by comparing the patient-level risk-adjusted readmission rate between dual-eligible beneficiaries and Medicare-only beneficiaries. We adjusted for case mix using indicator variables for each procedure group.

    To address concerns about unmeasured clinical risk severity, we conducted a similar analysis in the MSQC data set, which includes a more robust set of clinical variables. We developed a model as above, with additional inclusion of functional status and ASA classification of physical status. We determined the association between dual eligibility and risk-adjusted readmission as described above.

    Then, for each data set, we calculated hospital-level risk-adjusted readmission rates using models both inclusive and exclusive of dual eligibility. In each model, patients’ estimated probability of readmission was used to calculate each hospital’s expected readmission rate. For each hospital, we divided the observed number of readmissions by the expected number of readmissions. We then multiplied by the population readmission rate, inclusive of patients at all hospitals, to generate a risk-adjusted readmission rate for each hospital. We ranked hospitals by their risk-adjusted readmission rate and presented the data in a caterpillar plot with 95% CIs. Each caterpillar plot compared models inclusive and exclusive of dual eligibility, allowing us to examine how the inclusion of dual eligibility changed the distribution of the risk-adjusted readmission rate across hospitals.

    We conducted sensitivity analyses to confirm the robustness of our findings. We replicated our analyses while restricting the study population to patients undergoing 2 common procedures: cholecystectomy and colectomy (to represent surgical procedures with high and low readmission risk). Then, to ensure the generalizability of our results to the entire United States, we replicated our analysis of Michigan Medicare data using a national data set. Results of sensitivity analyses are presented in eTable 1, eTable 2, and the eFigure in the Supplement.

    All models used multiple logistic regression to estimate 30-day readmission rates. We used the Huber/White sandwich estimator to obtain robust variance estimates that adjust for within-hospital correlation of risk-adjusted readmission rates. We evaluated model performance using the C statistic (area under the receiver operating characteristic curve).

    Statistical analyses were performed using Stata, version 15.1 (StataCorp LLC). All P values were from 2-sided tests, and results were deemed statistically significant at P < .05.

    Results

    The Medicare claims cohort included 29 710 Medicare beneficiaries at 61 hospitals in Michigan, of whom 3986 were dual eligible (Table 1). Compared with Medicare-only beneficiaries, dual-eligible beneficiaries were younger (mean [SD], 72.9 [6.9] vs 75.7 [7.5] years; P < .001), and higher proportions of dual-eligible beneficiaries were female (2554 of 3986 [64.1%] vs 12 879 of 25 724 [50.1%]; P < .001), were nonwhite (1225 of 3986 [30.7%] vs 2783 of 25 724 [10.8%]; P < .001), and had more than 1 Elixhauser comorbidity (3671 of 3986 [92.1%] vs 21 760 of 25 724 [84.6%]; P < .001).

    The MSQC cohort included 25 941 Medicare beneficiaries at 62 hospitals in Michigan, of whom 1608 were dual eligible (Table 1). As in the Medicare claims cohort, dual-eligible beneficiaries were younger (mean [SD] age, 72.9 [6.8] vs 74.6 [7.1] years), and higher proportions were female (990 of 1608 [61.6%] vs 12 578 of 24 333 [51.7%]), were nonwhite (416 of 1608 [25.9%] vs 2176 of 24 333 [8.9%]), and had more than 1 Elixhauser comorbidity (1157 of 1608 [72.0%] vs 14 181 of 24 333 [58.3%]). Dual-eligible beneficiaries were more likely than Medicare-only beneficiaries to be ASA classification 4 (394 of 1608 [24.5%] vs 3580 of 24 333 [14.7%]; P < .001) and less likely to be ASA classification 2 (218 of 1608 [13.6%] vs 6070 of 24 333 [24.9%]; P < .001). Dual-eligible beneficiaries were less likely to have independent functional status than Medicare-only beneficiaries (1232 of 1608 [76.6%] vs 22 227 of 24 333 [91.3%]; P < .001).

    The association between dual eligibility and 30-day hospital readmission is presented in Table 2. In the Medicare claims cohort, in models adjusting for age, sex, and case mix, dual eligibility was associated with a 3.2-percentage point (pp) increase (95% CI, 1.4-5.0 pp) in readmission rate. With added adjustment for clinical comorbidities, dual eligibility was associated with a 2.2-pp increase (95% CI, 0.4-3.9 pp). In the MSQC cohort in models adjusting for age, sex, and case mix, dual eligibility was associated with a 2.0-pp increase (95% CI, 0.2-3.7 pp) in the risk-adjusted readmission rate. With added adjustment only for clinical comorbidities, the association between dual eligibility and risk-adjusted readmission was 1.2 pp (95% CI, –0.5 to 3.0 pp). With added adjustment only for clinical risk assessment, the association between dual eligibility and risk-adjusted readmission was attenuated further to 0.8 pp (95% CI, –0.9 to 2.4 pp). With added adjustment for both clinical comorbidities and clinical risk assessment, the association between dual eligibility and risk-adjusted readmission was 0.6 pp (95% CI, –1.0 to 2.2 pp).

    Adding dual eligibility to our risk-adjustment models had little association with hospital rankings in either data set (Figure 2). In Medicare claims, 27 of 61 hospitals (44.3%) did not experience a change in ranking by risk-adjusted readmission rate. Only 6 hospitals (9.8%) experienced a change greater than 1 in their ranking, with the greatest change being a decrease of 8. The average absolute change in risk-adjusted readmission rate across the 61 hospitals in our sample was 0.1 pp. The maximum decrease experienced by a single hospital was −0.8 pp, and the maximum increase experienced by a single hospital was 0.3 pp.

    In the MSQC registry, 45 of 62 hospitals (72.6%) did not experience a change in rank after dual eligibility was added to the risk adjustment model. Only 1 hospital experienced a change in ranking greater than 1. The average absolute change in risk-adjusted readmission rate across the 62 hospitals in our sample was 0.03 pp. The maximum decrease experienced by a single hospital was –0.2 pp, and the maximum increase experienced by a single hospital was 0.05 pp.

    Results were similar in sensitivity analyses for patients undergoing cholecystectomy or colectomy (eTable 1 in the Supplement). For patients undergoing cholecystectomy, dual eligibility was not significantly associated with 30-day hospital readmission for both cohorts. For patients undergoing colectomy, dual eligibility was associated with readmission only in the Medicare claims cohort. In both data sets, for all beneficiaries, the risk-adjusted readmission rate was higher for colectomy than for cholecystectomy.

    In the national Medicare data set, we identified 836 663 surgical patients at 1940 hospitals (eTable 2 in the Supplement). In the full risk-adjustment model, dual-eligible beneficiaries were 1.8 pp (95% CI, 1.6-2.0 pp) more likely to be readmitted than Medicare-only beneficiaries. The distribution of risk-adjusted readmission rates was similar to the Michigan Medicare data set (eFigure in the Supplement). Again, there was little change in the distribution when dual eligibility was added to the risk-adjustment models. The median absolute change in ranking was 23, equivalent to a 1.2 percentile change. The largest change in ranking experienced by a single hospital was 361 (18.6 percentile difference). The mean absolute change in risk-adjusted readmission rate across the hospitals in our sample was 0.2 pp. The maximum decrease experienced by a single hospital was –1.1 pp, and the maximum increase experienced by a single hospital was 0.4 pp.

    Discussion

    In this study of the association between social risk and 30-day readmission among surgical patients, we report 3 main findings. First, dual eligibility for Medicare and Medicaid was associated with a substantial increase in 30-day readmission after surgery. Second, this association was attenuated and nonsignificant when evaluated within a clinical registry with more detailed information on patient risk, suggesting that dual eligibility may be a proxy for clinical risk unmeasured by Medicare claims instead of capturing social determinants of surgical outcomes. Third, adjustment for social risk made little difference in hospitals’ relative performance on 30-day readmissions in either the Medicare claims or registry data.

    Our results align with those of prior work demonstrating that social risk factors may be associated with readmission rates after surgery. For example, Joynt Maddox et al1 found that penalties for excess readmission after coronary artery bypass grafting for hospitals in the highest quintile of proportion of dual-eligible patients changed when the HRRP began to stratify hospitals by their proportion of dual-eligible patients. Glance et al14 found that the proportion of dual-eligible patients accounted for the difference in risk-adjusted surgical readmission rates between safety-net hospitals and non–safety-net hospitals. Anderson et al15 found that adjusting for factors such as insurance status and median income based on zip code improved the discrimination of risk-adjustment models used to estimate surgical outcomes. However, we also suggest that unmeasured clinical risk severity could explain these findings. We question whether these adjustments were successfully capturing social risk factors that varied across hospital populations or were instead adjusting for clinical risk severity that was previously unmeasured. Our results are also consistent with prior findings14,15 that adjustment for dual eligibility has little association with hospital ranking. We demonstrate that the prior work is generalizable to the Medicare population and a surgical cohort inclusive of several general and vascular surgery procedures.

    Despite our application of robust risk-adjustment models inclusive of clinical risk, substantial variation in risk-adjusted surgical readmission rates persists across hospitals. With value-based or bundled payment models in mind, the Centers for Medicare & Medicaid Services and hospitals continue to seek explanations for this variation. For hospitals, social risk factors continue to be an important consideration when evaluating the quality of hospital care over time. The argument that social risk should be included in risk-adjustment models is valid for several reasons: patients with low socioeconomic status are more likely to be readmitted,14,15,21 surgical patients treated at safety-net hospitals experience higher readmission rates than patients treated at non–safety-net hospitals,22 and safety-net hospitals are more likely to be penalized under HRRP.22,23 There are also concerns about hospitals “cherry-picking” high–socioeconomic status patients to improve measured outcomes.24 Our findings suggest that variables available in claims data and currently available to programs such as HRRP (ie, dual eligibility) are insufficient in capturing these phenomena. Further research can explore measures of social risk11,25 beyond the scope of this study as possible explanations for variations in risk-adjusted surgical readmission rates. Besides dual eligibility, other measures of social risk may have more substantial associations with hospital profiling using risk-adjusted readmission rates. This work is warranted because programs such as HRRP are strongly associated with surgical readmissions—for example, HRRP began to have spillover effects on surgical conditions even before the program measured any surgical outcomes.13

    Social risk may play a lesser, and possibly different, role in modeling surgical readmissions vs medical readmissions. Several studies of medical conditions found that risk adjustment for social variables at the patient level had substantial association with risk-adjusted readmission rates.2,4-7,9 In addition, one study of the HRRP policy change mentioned above that examined only medical readmissions10 found substantial changes in penalties associated with stratification. Further work is necessary, however, to elucidate how dual eligibility is associated with medical and surgical readmission rates differently and to understand how programs such as HRRP may be affected by inclusion of more surgical conditions.

    Limitations

    Our study had some limitations. First, we did not link patient records between the Medicare and MSQC cohorts. However, these cohorts are quite similar because both data sets included the largest Michigan hospitals, and demographic and clinical variables were similar between cohorts. Second, we identified a smaller proportion of dual-eligible beneficiaries in the MSQC cohort than in the Medicare claims cohort. However, this was likely owing to the MSQC data set’s stricter definition of dual eligibility, and we are confident that the exclusion of patients was not associated with social risk. Also, demographic and clinical characteristics were similar between the dual-eligible patients identified in each data set. Third, our analysis accounts for dual eligibility at the individual level, while programs such as HRRP account for dual eligibility at the hospital level by stratifying based on hospitals’ proportions of dual-eligible patients. However, our goal was to generate results applicable to any VBP program by determining the extent to which a patient’s dual-eligibility status reflects their unmeasured clinical risk. Fourth, our analysis did not differentiate between planned and unplanned readmissions. However, planned readmissions are rare for most hospitalizations, particularly surgical procedures.26,27 Fifth, we identified lower readmission rates in the MSQC data set than in Medicare claims, perhaps because the MSQC data set included readmissions only to the same hospital. However, we identified similar patterns of associations between the 2 data sets, and our goal was not to directly compare readmission rates between the 2 data sets.

    Conclusions

    Our results demonstrated that adjusting for dual eligibility for Medicare and Medicaid may not account for social risk factors associated with readmission after surgery. Instead of representing social risk factors associated with surgical outcomes, dual eligibility was a proxy for clinical risk severity unmeasured by variables traditionally available in Medicare claims. In addition, dual eligibility had little association with hospital ranking using risk-adjusted readmission rates. Policy makers interested in accounting for the social determinants of health in risk-adjustment models used by VBP programs should consider incorporating additional measures of social risk.

    Back to top
    Article Information

    Accepted for Publication: April 3, 2020.

    Published: June 10, 2020. doi:10.1001/jamanetworkopen.2020.7426

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

    Corresponding Author: Benjamin A. Y. Cher, MS, University of Michigan Medical School, 1135 Catherine St, Ann Arbor, MI 48109 (bcher@umich.edu).

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

    Concept and design: Cher, Ryan, Sheetz.

    Acquisition, analysis, or interpretation of data: Cher, Hoffman, Sheetz.

    Drafting of the manuscript: Cher, Hoffman, Sheetz.

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

    Statistical analysis: Cher, Ryan, Sheetz.

    Obtained funding: Ryan.

    Administrative, technical, or material support: Cher, Ryan.

    Supervision: Hoffman.

    Conflict of Interest Disclosures: Dr Ryan reported receiving grants from the University of Michigan during the conduct of the study. Dr Hoffman reported receiving grants from the Agency for Healthcare Research and Quality outside the submitted work. No other disclosures were reported.

    Funding/Support: This work was supported by the M-Cubed program at the University of Michigan.

    Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Additional Information: Mr Cher is a medical student at the University of Michigan Medical School in Ann Arbor.

    References
    1.
    Joynt Maddox  KE, Reidhead  M, Qi  AC, Nerenz  DR.  Association of stratification by dual enrollment status with financial penalties in the Hospital Readmissions Reduction Program.   JAMA Intern Med. 2019;179(6):769-776. doi:10.1001/jamainternmed.2019.0117 PubMedGoogle ScholarCrossref
    2.
    Barnett  ML, Hsu  J, McWilliams  JM.  Patient characteristics and differences in hospital readmission rates.   JAMA Intern Med. 2015;175(11):1803-1812. doi:10.1001/jamainternmed.2015.4660 PubMedGoogle ScholarCrossref
    3.
    Gaskin  DJ, Zare  H, Vazin  R, Love  D, Steinwachs  D.  Racial and ethnic composition of hospitals’ service areas and the likelihood of being penalized for excess readmissions by the Medicare program.   Med Care. 2018;56(11):934-943. doi:10.1097/MLR.0000000000000988 PubMedGoogle ScholarCrossref
    4.
    Sills  MR, Hall  M, Cutler  GJ,  et al.  Adding social determinant data changes children’s hospitals’ readmissions performance.   J Pediatr. 2017;186:150-157. doi:10.1016/j.jpeds.2017.03.056 PubMedGoogle ScholarCrossref
    5.
    Sills  MR, Hall  M, Colvin  JD,  et al.  Association of social determinants with children’s hospitals’ preventable readmissions performance.   JAMA Pediatr. 2016;170(4):350-358. doi:10.1001/jamapediatrics.2015.4440 PubMedGoogle ScholarCrossref
    6.
    Roberts  ET, Zaslavsky  AM, Barnett  ML, Landon  BE, Ding  L, McWilliams  JM.  Assessment of the effect of adjustment for patient characteristics on hospital readmission rates: implications for pay for performance.   JAMA Intern Med. 2018;178(11):1498-1507. doi:10.1001/jamainternmed.2018.4481 PubMedGoogle ScholarCrossref
    7.
    Meddings  J, Reichert  H, Smith  SN,  et 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. doi:10.1007/s11606-016-3869-x PubMedGoogle ScholarCrossref
    8.
    Hoffman  GJ, Hsuan  C, Braun  T, Ponce  N.  Health equity and hospital readmissions: does inclusion of patient functional and social complexity improve predictiveness?   J Gen Intern Med. 2019;34(1):26-28. doi:10.1007/s11606-018-4635-z PubMedGoogle ScholarCrossref
    9.
    Blum  AB, Egorova  NN, Sosunov  EA,  et al.  Impact of socioeconomic status measures on hospital profiling in New York City.   Circ Cardiovasc Qual Outcomes. 2014;7(3):391-397. doi:10.1161/CIRCOUTCOMES.113.000520 PubMedGoogle ScholarCrossref
    10.
    McCarthy  CP, Vaduganathan  M, Patel  KV,  et al.  Association of the new peer group–stratified method with the reclassification of penalty status in the Hospital Readmission Reduction Program.   JAMA Netw Open. 2019;2(4):e192987. doi:10.1001/jamanetworkopen.2019.2987 PubMedGoogle Scholar
    11.
    Joynt Maddox  KE, Reidhead  M, Hu  J,  et al.  Adjusting for social risk factors impacts performance and penalties in the Hospital Readmissions Reduction Program.   Health Serv Res. 2019;54(2):327-336. doi:10.1111/1475-6773.13133 PubMedGoogle ScholarCrossref
    12.
    Centers for Medicare & Medicaid Services. Hospital Readmissions Reduction Program (HRRP). Updated February 11, 2020. Accessed July 14, 2019. https://www.cms.gov/medicare/medicare-fee-for-service-payment/acuteinpatientpps/readmissions-reduction-program.html
    13.
    Ibrahim  AM, Nathan  H, Thumma  JR, Dimick  JB.  Impact of the Hospital Readmission Reduction Program on surgical readmissions among Medicare beneficiaries.   Ann Surg. 2017;266(4):617-624. doi:10.1097/SLA.0000000000002368 PubMedGoogle ScholarCrossref
    14.
    Glance  LG, Kellermann  AL, Osler  TM, Li  Y, Li  W, Dick  AW.  Impact of risk adjustment for socioeconomic status on risk-adjusted surgical readmission rates.   Ann Surg. 2016;263(4):698-704. doi:10.1097/SLA.0000000000001363 PubMedGoogle ScholarCrossref
    15.
    Anderson  JE, Li  Z, Romano  PS, Parker  J, Chang  DC.  Should risk adjustment for surgical outcomes reporting include sociodemographic status? a study of coronary artery bypass grafting in California.   J Am Coll Surg. 2016;223(2):221-230. doi:10.1016/j.jamcollsurg.2016.05.008 PubMedGoogle ScholarCrossref
    16.
    Martin  JR, Wang  TY, Loriaux  D,  et al.  Race as a predictor of postoperative hospital readmission after spine surgery.   J Clin Neurosci. 2017;46:21-25. doi:10.1016/j.jocn.2017.08.015PubMedGoogle ScholarCrossref
    17.
    Strombom  P, Widmar  M, Keskin  M,  et al.  Assessment of the value of comorbidity indices for risk adjustment in colorectal surgery patients.   Ann Surg Oncol. 2019;26(9):2797-2804. doi:10.1245/s10434-019-07502-9 PubMedGoogle ScholarCrossref
    18.
    Campbell  DA  Jr, Englesbe  MJ, Kubus  JJ,  et al.  Accelerating the pace of surgical quality improvement: the power of hospital collaboration.   Arch Surg. 2010;145(10):985-991. doi:10.1001/archsurg.2010.220PubMedGoogle ScholarCrossref
    19.
    Englesbe  MJ, Brooks  L, Kubus  J,  et al.  A statewide assessment of surgical site infection following colectomy: the role of oral antibiotics.   Ann Surg. 2010;252(3):514-519. doi:10.1097/SLA.0b013e3181f244f8 PubMedGoogle Scholar
    20.
    Portier  K, Ida  KK.  The ASA Physical Status Classification: what is the evidence for recommending its use in veterinary anesthesia?—a systematic review.   Front Vet Sci. 2018;5:204. doi:10.3389/fvets.2018.00204 PubMedGoogle ScholarCrossref
    21.
    Hu  J, Gonsahn  MD, Nerenz  DR.  Socioeconomic status and readmissions: evidence from an urban teaching hospital.   Health Aff (Millwood). 2014;33(5):778-785. doi:10.1377/hlthaff.2013.0816 PubMedGoogle ScholarCrossref
    22.
    Talutis  SD, Chen  Q, Wang  N, Rosen  AK.  Comparison of risk-standardized readmission rates of surgical patients at safety-net and non–safety-net hospitals using Agency for Healthcare Research and Quality and American Hospital Association data.   JAMA Surg. 2019;154(5):391-400. doi:10.1001/jamasurg.2018.5242 PubMedGoogle ScholarCrossref
    23.
    Gu  Q, Koenig  L, Faerberg  J, Steinberg  CR, Vaz  C, Wheatley  MP.  The Medicare Hospital Readmissions Reduction Program: potential unintended consequences for hospitals serving vulnerable populations.   Health Serv Res. 2014;49(3):818-837. doi:10.1111/1475-6773.12150 PubMedGoogle ScholarCrossref
    24.
    Bynum  J, Lewis  V.  Value-based payments and inaccurate risk adjustment—who is harmed?   JAMA Intern Med. 2018;178(11):1507-1508. doi:10.1001/jamainternmed.2018.4142 PubMedGoogle ScholarCrossref
    25.
    National Academies of Sciences, Engineering, and Medicine 2017. Kwan LY, Stratton K, Steinwach DM, eds.  Accounting for Social Risk Factors in Medicare Payment: Identifying Social Risk Factors. National Academies Press; 2016.
    26.
    Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation. 2013 Measures updates and specifications: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized mortality measure (version 7.0). Published March 2013. Accessed March 13, 2020. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Mortality_AMI-HF-PN_Measures_Updates_Report_FINAL_06-13-2013.pdf
    27.
    Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation. Hospital-level 30-day all-cause unplanned readmission following coronary artery bypass graft surgery. Published February 1, 2014. Accessed March 13, 2020. https://www.qualitynet.org/inpatient/measures/readmission
    ×