Complex Patients and Quality of Care in Medicare Advantage

Abstract

BACKGROUND:

New federal policies aim to focus Medicare Advantage (MA) plans on the needs of individuals with complex health conditions.

OBJECTIVES:

Examine enrollment patterns of MA beneficiaries with complex needs and the association of enrollment patterns with MA plan performance.

DESIGN:

Cross-sectional study

SETTING:

2015 Medicare Health Outcome Survey baseline survey

PARTICIPANTS:

273,336 MA beneficiaries in 467 MA plans who lived in the community

MEASUREMENTS:

Complex patients included individuals 65 and older with multiple self-reported chronic conditions and functional limitations and all under 65 disabled. Outcomes included 27 performance measures reported under the 5-Star Part C Star Rating. Linear probability regression was used to examine the association of concentration of complex patients and performance measures.

RESULTS:

Most complex patients were enrolled in general MA plans. Concentration of complex patients ranged from 25.9% in MA contracts in the lowest quintile to 68.9% in the top quintile. MA contract performance scores generally decreased as the concentration of complex patients increased. After adjusting for contract and enrollee characteristics, MA contracts with more complex patients performed less well on half of the Part C performance measures, including patient experience, preventive care, and chronic care measures.

CONCLUSION:

MA contracts with high concentration of complex patients have lower performance scores on more than half of Part C measures. Further study is needed to understand whether these performance measures are capturing the delivery of poor care, deficiencies in the health plan’s care systems, or whether some measures may not be appropriate for complex patients.

INTRODUCTION

There is growing policy interest in changing the delivery system to meet the needs of beneficiaries with complex health care needs.^1–3 The Affordable Care Act supported the supported the development of multiple models to address the needs of this population^3–5 and this focus has continued with the passage of the CHRONIC Care Act in 2018 which allows Medicare Advantage plans to specially target services for beneficiaries with complex health needs.⁶ In 2017, the National Academy of Medicine (NAM) released a publication focused on high-needs populations along with program and policy approaches to improve care and outcomes.³ The NAM report noted that measuring quality of care for high-needs populations requires moving beyond disease-specific quality measures to considering care coordination, quality of life, and independence. Furthermore, Boyd et al concluded that some common clinical quality measures are not appropriate for patients with complex needs.⁷

Most research on Medicare beneficiaries with complex needs has focused on the providers in fee-for-service Medicare program. These studies have found that the providers caring for this population often have lower quality ratings, higher readmission rates, and are less likely to offer preventive services.^8–12 We know less about the health plans that serve beneficiaries with complex health needs in the Medicare Advantage (MA) program. Understanding the characteristics and performance of MA plans that enroll beneficiaries with complex health needs is important since MA accounts for more than one-third of Medicare beneficiaries.¹³

We are not aware of any studies examining the quality experienced by health plans serving MA beneficiaries with complex health needs.¹⁴ A 2015 RAND report to the Centers for Medicare & Medicaid Services and a 2016 report to Congress by the Assistant Secretary for Planning and Evaluation (ASPE) found that social risk factors such as dual eligibility, race, and disability were associated with lower quality of care on some measures in MA,^15–17 but these studies did not focus on the characteristics of beneficiaries with complex needs. We therefore aim to show whether current measures of plan performance demonstrate that the needs of these individuals are being adequately met. Such information could assist MA plans in evaluating their care and assist CMS in developing policies to support MA plans in better assessing and meeting the needs of this population.

METHODS

Data Source

We used the 2015 Medicare Health Outcomes Survey (HOS) baseline cohort to determine the concentration of complex patients by MA plan. The HOS is a sample of randomly selected MA enrollees in MA contracts with more than 1,200 enrollees and all beneficiaries in MA contracts with 500 to 1,200 enrollees. MA contracts with less than 500 enrollees and employer-only contracts are not included in HOS.

We merged the HOS by MA contract identifier to publicly-available data files on Medicare Advantage Star Rating data published in 2017 (which largely draws on data collected in 2015). It is important to note that Medicare beneficiaries enroll in a specific MA plan (e.g., general plan, employer plan, or special needs plan), but for purposes of performance measurement and administration, MA plans are aggregated to the contract-level.

The University of Wisconsin and University of Maryland Institutional Review Boards determined that this study was exempt from review because study subjects cannot be identified.

Study Sample

The 2015 baseline HOS (Cohort 18) identified 567,241 beneficiaries for the survey of which 286,850 (50.57%) responded. These respondents represented 488 MA contracts. We excluded 21 plans and limited our analytic sample to Health Maintenance Organizations (HMOs) and Preferred Provider Organizations (PPOs), the most common plan types. This yielded a final analytic sample of 273,336 respondents, representing 16,168,341 individuals enrolled in 467 contracts.

Key Measures

Complex Patients.

Drawing on previous work defining complex patients,^{1, 3, 18} we classify MA enrollees as complex patients if they are disabled and under 65 or 65 and older, have 3 or more self-reported chronic conditions (hypertension, coronary artery disease, congestive heart failure, myocardial infarction, other heart conditions, stroke, chronic obstructive pulmonary disease, inflammatory bowel disease, arthritis of the hip or knee, osteoporosis, sciatica, diabetes, depression, colon cancer, lung cancer, breast cancer, prostate cancer, and other cancer), and at least one self-reported difficulty with a basic or instrumental activity of daily living (bathing, dressing, toileting, transferring, eating, walking, preparing meals, managing money, taking medication as prescribed, and doing errands). Previous studies have found patient-reported chronic conditions to be a valid approach that avoids reliance on claims-based measures that can be subject to bias due to varying financial incentives.²¹

Concentration of Complex Patients by Contract.

We determine the proportion of individuals with complex needs in each MA contract using survey weights to account for the probability of selection and nonresponse. We classify MA contracts by quintile of complex patient concentration.

Medicare Advantage Contract Characteristics.

We collected tax status, plan enrollment size, type of plan (HMO or PPO), and type of plan (Employer, Special Needs Plan, and general) from the HOS. The proportion of individuals enrolled in the low-income subsidy program was collected from the Medicare low-income subsidy public use file. Drawing on the NAM framework to identify social risk factors related to health¹⁴, we calculated the proportion of individuals in each MA contract living in a disadvantaged neighborhood defined using the Area Deprivation Index^{22, 23} to capture socioeconomic context. Using HOS, we also calculated the proportion of enrollees married and the proportion of enrollees living alone; proportion female; proportion Black and Hispanic; and proportion with no college education and annual incomes less than $20,000.

Dependent Variables

Quality of Care.

We examine 27 performance measures reported under the 5-Star Part C Rating in 2015 across all five domains: Domain 1: Staying Healthy, Domain 2: Managing Chronic Conditions; Domain 3: Member Experience with Health Plan; Domain 4: Member Complaints and Changes in the Health Plan’s Performance; and, Domain 5: Health Plan Customer Service (Appendix Table 1). We exclude four quality of care measures (SNP Care Management, Care for Older Adults – Medication Review, Care for Older Adults – Functional Status Assessment, Care for Older Adults – Pain Assessment) that were only reported by SNPs and one measure, Quality Improvement, because contract-level scores were not publicly reported.

Statistical Analysis

To generate estimates by MA contract, we use survey weights to account for the respondent’s probability of selection and probability of response. We describe the variation in the enrollment of complex MA beneficiaries overall and by plan type.

We examine the relationship between the concentration of beneficiaries with complex health needs and performance measures using bivariate analyses and multivariate linear regression models. To address the concern that differences in performance are due to differences in social risk factors and not differences in complexity, we account for a robust set of social risk factors including neighborhood disadvantage, income, education level, marital status, and living arrangements. We also control for differences in health plan characteristics (for-profit vs not-for-profit status, and HMO vs PPO), which could explain differences in care management that may affect quality of care. For the regression analysis, we report p-values that assess the difference of each level of concentration compared to the reference group (the lowest quintile), and we test the importance of accounting for the concentration of complex patients in the model using the Wald test.

RESULTS

In 2015, 488 Medicare Advantage contracts participated in the HOS baseline. There were 467 HMO and PPO contracts in our sample. Among these contracts, less than 30% were PPOs and not-for-profit. Contracts that only included SNP plans accounted for 14.6% of the sample. The average low-income subsidy rate was 35%. In HOS, 38% of respondents had complex needs. Overall, 66% of complex patients were enrolled in a general MA plan, 23% in SNP, and 12% in an employer plan. Reflecting a higher rate of complex patients in smaller plans, the average proportion of complex patients at the contract level was 43% (Table 1).

Table 1.

Medicare Advantage Contract Characteristics by Concentration of Complex Patients, Medicare Health Outcomes Survey, 2015

	Overall	Quantile of Complex Patients Concentration					P
		1st (Low)	2nd	3rd	4th	5th (High)
n	467	94	93	93	93	94
SNP Only (%)	14.6	0	0	3.2	16.1	53.2	<0.001
PPO (%)	26.3	40.4	35.5	30.1	19.4	6.4	<0.001
Not for Profit (%)	27.4	44.7	31.2	12.9	17.2	30.9	<0.001
Plan Size (%)							<0.001
<10,000	47.1	39.4	31.2	48.4	47.3	69.1
10,000 – 39,999	30.4	36.2	34.4	30.1	32.3	19.1
40000+	22.5	24.5	34.4	21.5	20.4	11.7
Average Percent Complex Patients	43.30	25.90	33.04	39.70	49.03	68.76	<0.001
Average Percent Black or African American	13.87	4.75	8.64	12.71	19.16	24.09	<0.001
Average Percent Hispanic	14.52	6.05	7.16	12.14	21.51	25.70	<0.001
Average Percent of Living in the Most Disadvantaged Neighborhood	14.23	4.90	9.78	12.43	19.80	24.22	<0.001
Average Percent Receiving Medicaid	31.77	7.23	9.93	19.36	39.92	82.13	<0.001
Average Percent Not Married	49.11	37.14	41.01	45.66	54.28	67.38	<0.001
Average Percent Living Alone	26.53	22.94	24.78	25.62	27.86	31.44	<0.001
Average Percent Female	56.89	56.15	56.08	56.06	58.12	58.01	<0.001
Average Percent Reporting Income<$20,000	29.74	13.80	19.05	26.37	37.68	51.73	<0.001
Average Percent Reporting Less than a College Education	47.59	34.38	40.50	47.76	55.88	59.46	<0.001

Table 1 presents the contract-level characteristics overall and by concentration of complex patients. The concentration of complex patients ranged from an average 25.9% in the 1^st quintile to 68.8% in the 5^th quintile. MA contracts in the top quintile of complex patients were more likely have less than 4 stars (Figure 1), HMO, for-profit, and more likely to be small plans (<10,000 enrolled) compared to contracts in the lowest quintile. MA contracts in the top quintile compared to the lowest quintile were more likely to enroll individuals living in the most disadvantaged neighborhoods, and having social risk factors such as not being married, living alone, and low incomes. Between the highest and lowest quintiles, there were statistically significant (p<0.05) differences in enrollment among both Black or African Americans (4.7% vs. 24.0%) and Hispanics (6.1% vs. 25.7%).

Figure 1.

Proportion of Medicare Advantage Plans by Star Rating and Quintile of Complex Pateint Concentration

We used bivariate analyses to examine whether contract-level performance on Part C performance measures varied by the concentration of complex patients. The results, presented in Table 2, show that performance scores generally worsened as the concentration of complex patients increased. We further examined whether these differences held after controlling for contract-level characteristics including tax status (non-profit vs. for-profit), plan type (HMO vs. PPO), contract enrollment, and enrollee characteristics. We present the coefficients and 95% confidence interval for the main independent variable: quantile of complex patients where the 1^st quantile (the lowest level) is the reference group in Figure 2 (Appendix Table 2 presents the results of the multivariate linear regression models). The coefficients can be interpreted as the percentage point differences compared to the 1^st quantile. For example, coefficients for the breast cancer screening model (the bottom row of Figure 2, Panel A) can be interpreted as follows: compared to plans with the lowest concentration of complex patients: contracts in the 2^nd quintile on average performed 3.91 percentage points lower; contracts in the 3rd quintile on average performed 5.93 percentage points lower; contracts in the 4th quintile on average performed 9.27 percentage points lower; and, contracts in the 5^th quintile on average performed 14.45 percentage points lower.

Table 2.

Average Contract Performance on Individual Star Rating Measures by Concentration of Complex Patients

		Quantile of Complex Patients Concentration
	N	1st (Low)	2nd	3rd	4th	5th (High)	Wald Test P-Value
Domain 1: Staying Healthy: Screenings, Tests and Vaccines
Breast Cancer Screening	336	78.05 (6.20)	74.46 (6.05)	72.31 (8.69)	71.07 (9.64)	66.75 (11.80)	<0.001
Colon Cancer Screening	362	72.35 (8.38)	69.53 (8.81)	68.86 (9.74)	64.07 (11.74)	63.81 (11.16)	<0.001
Annual Flu Vaccine	338	76.30 (6.61)	73.00 (7.23)	72.58 (6.08)	69.94 (7.11)	67.08 (9.87)	<0.001
Improving or Maintaining Physical Health	296	68.90 (3.32)	67.65 (3.06)	67.32 (3.55)	66.09 (3.43)	68.69 (5.87)	0.001
Improving or Maintaining Mental Health	296	85.63 (2.78)	84.70 (2.51)	84.86 (2.77)	83.87 (2.91)	84.46 (4.42)	0.036
Monitoring Physical Activity	360	52.58 (5.03)	50.68 (5.61)	49.37 (5.53)	49.14 (6.38)	50.12 (6.14)	0.002
Adult BMI Assessment	362	92.71 (7.11)	92.70 (9.01)	94.14 (8.23)	93.43 (6.55)	93.61 (6.04)	0.744
Domain 2: Managing Chronic Conditions
Osteoporosis Management	248	39.17 (16.09)	36.79 (15.42)	39.91 (18.42)	42.55 (21.56)	38.83 (18.63)	0.600
Diabetes Care – Eye Exam	362	74.51 (8.36)	69.70 (9.12)	68.89 (11.52)	68.97 (10.67)	69.99 (8.93)	0.001
Diabetes Care – Kidney Disease Monitoring	346	95.44 (1.86)	95.43 (2.41)	96.12 (1.72)	96.21 (1.77)	95.77 (2.34)	0.064
Diabetes Care – Blood Sugar Controlled	362	80.45 (9.80)	78.54 (10.04)	77.53 (10.75)	71.83 (13.39)	68.57 (13.12)	<0.001
Controlling Blood Pressure	359	73.42 (11.71)	70.74 (14.43)	72.65 (11.17)	65.59 (15.59)	63.36 (12.43)	<0.001
Rheumatoid Arthritis Management	293	81.18 (5.91)	79.56 (7.75)	76.58 (7.65)	76.32 (6.07)	73.29 (9.68)	<0.001
Reducing the Risk of Falling	337	53.16 (5.07)	53.32 (6.34)	56.31 (5.81)	60.07 (9.38)	67.82 (7.58)	<0.001
Plan All-Cause Readmissions#	344	10.79 (2.44)	10.39 (1.66)	10.27 (2.55)	11.15 (2.36)	11.53 (2.57)	0.011
Domain 3: Member Experiences with Health Plan
Getting Needed Care	317	84.61 (2.45)	83.73 (2.71)	82.35 (3.56)	81.59 (3.73)	81.24 (3.43)	<0.001
Getting Appointments and Care Quickly	338	78.01 (3.48)	76.77 (3.29)	74.98 (4.10)	73.85 (3.47)	73.46 (3.39)	<0.001
Customer Service	315	89.62 (2.46)	88.03 (3.00)	87.08 (2.95)	86.30 (3.39)	86.33 (2.75)	<0.001
Rating of Health Care Quality	318	86.62 (1.89)	86.14 (1.65)	85.65 (1.95)	84.94 (2.42)	84.89 (2.07)	<0.001
Rating of Health Plan	338	85.31 (3.30)	84.01 (3.28)	83.86 (3.07)	83.64 (3.82)	84.75 (3.07)	0.016
Care Coordination	329	86.33 (2.26)	85.73 (1.93)	84.61 (3.06)	84.53 (2.56)	83.83 (2.77)	<0.001
Domain 4: Member Complaints and Changes in the Health Plan’s Performance
Complaints about the Health Plan	353	0.17 (0.20)	0.22 (0.27)	0.25 (0.23)	0.30 (0.29)	0.20 (0.18)	0.017
Members Choosing to Leave the Plan*	346	7.81 (5.19)	10.63 (6.77)	11.50 (6.23)	14.33 (10.75)	12.31 (6.67)	<0.001
Beneficiary Access and Performance Problems	366	84.76 (18.27)	80.53 (25.71)	70.62 (25.19)	84.00 (15.83)	87.22 (17.86)	<0.001
Domain 5: Health Plan Customer Service
Plan Makes Timely Decisions about Appeals	232	95.62 (5.80)	93.98 (5.95)	91.20 (9.80)	95.32 (5.74)	94.46 (6.53)	0.022
Reviewing Appeal Decisions	232	90.83 (6.40)	88.64 (9.30)	87.32 (11.51)	90.84 (8.28)	88.54 (9.64)	0.254
Call Center – Foreign Language Interpreter and TTY Availability	346	95.98 (4.94)	95.24 (5.41)	94.48 (5.69)	92.75 (7.39)	90.35 (9.48)	<0.001

Note:

^#= Reverse coded (Lower values mean better care)

Figure 2.

Adjusted Difference in Part C Performance Measures by Level of Complex Patient Concentration Controlling for Plan Characteristics and Social Risk Factors

Notes. Panel A presents Domain 1: Staying Healthy. Panel B presents Domain 2: Managing Chronic Conditions. Panel C presents Domain 3: Member Experience with Health Plan; Panel D presents Domain 4: Member Complaints and Changes in the Health Plan’s Performance; and, Domain 5: Health Plan Customer Service

For breast cancer screening, we find a dose response in performance that is as the concentration of complex patients increases, performance decreases. We find similar patterns for eight measures: monitoring physical activity, diabetes: eye exam, diabetes: blood sugar controlled, getting needed care, getting appointments and care quickly, customer service, care coordination, and interpreter and TTY availabilities at call centers. Using the Wald test, we find the concentration of complex patients did not significantly contribute to a model explaining MA contract performance in slightly less than half of the measures (13 of 27 measures): 2 of 7 measures in Domain 1: Staying Healthy, 5 of 8 measures in Domain 2: Managing Chronic Conditions, 3 of 6 measures in Domain 3: Member Experiences with Health Plan, 1 of 3 measures in Domain 4: Member Complaints, and 2 of 3 in Domain 5: Health Plan Customer Service.

DISCUSSION

We find that most complex patients in MA were enrolled in general Medicare Advantage plans, not employer or Special Needs Plans. We find that MA contracts with high concentrations of complex patients are more likely to be for-profit, HMO, and smaller plans. MA contracts with high concentrations of complex patients are also more likely to perform less well on 22 of 27 Part C performance measures collected by the Medicare program, and these differences persisted after controlling for social risk factors on 14 measures.

Previous studies have found that the health risk differences between traditional Medicare and MA are narrowing.^{24, 25} Research by CMS and the ASPE has reported differences between MA contracts with respect to low-income subsidy and/or disability status.^{15, 16} In this study, we extend this literature finding that the distribution of complex patients is not equal between MA contracts, and that MA contracts with more complex patients tend to perform less well on clinical care performance measures after adjusting for social risk factors. However, we cannot rule out that one reason that these MA contracts may perform less well is that complex patients are more likely to have social risk factors that may make typical clinical pathways more difficult to implement.^{7, 26} It is also important to consider that complex patients are also more likely to have comorbid conditions, which may make typical clinical pathways inappropriate. Given the high proportion of complex patients in MA it is important to develop measures that reflect the care important to these patients.

These results are generally consistent with previous studies of the relationship between high-risk populations and quality. Similar to the RAND study of quality of care differences between Medicare Advantage contracts by low-income subsidy and disability status, we find quality of care ratings vary based on individual characteristics on a subset of measures.¹⁶ We did not find differences on 13 performance measures, which may be because any differences were already accounted for through adjustment (e.g., Improving or Maintaining Physical Health) or that there was no meaningful difference in contract performance by concentration of complex patients. Our study expands upon this analysis by including self-reported measures of chronic conditions and functional limitations to assess complexity. In addition, while a recent study finds high need individuals are more likely to disenroll from their MA plan,²⁷ we find that MA plans with higher concentrations of complex patients had higher disenrollment in the bivariate analysis, but not in the regression results.

These findings should be considered in the context of the study limitations. While there were over 400 MA contracts in 2015, our analytic sample did not include all contracts because the HOS sample does not include small contracts with enrollment less than 500 or employer-only plans. Our analysis excluded 21 MA plans that were not non-employer HMOs or PPOs that enrolled somewhat healthier populations and higher performers, which may bias our results away from the null. In MA, benefit design—including premium, cost-sharing, and supplemental benefits—can vary within MA contracts, and we cannot account for these differences in our analysis. We examined performance measures reported at the contract-level and because of data constraints cannot examine how performance may vary by patient subgroup. It is possible that adherence rates in many cases may be higher in complex patients because they are more likely to have multiple contacts with the health care system.²⁸ This study uses an ecological analytic approach to examine the correlation of contract enrollee characteristics and performance. As such, it is important to note that this approach cannot control for spurious relationships, and may lead to inflated p-values. Lastly, we define complex patients using self-reported chronic conditions and functional limitations whereas CMS uses claims-based diagnosis codes to classify health conditions. Applying our definition of complexity in claims data may result in a weaker relationship because claims are more likely to identify individuals as having multiple chronic conditions and functional limitations.

Despite these limitations, these findings have important implications. Many of the existing performance measures are not specifically designed to assess the quality of care for patients with complex conditions. For complex patients, a different set of measure may be necessary such as non-disease specific measures because not all measures apply equally well across the full spectrum of patients.^{7, 29} We find that MA plans with the highest concentration of complex patients do just as well as plans with the lowest concentration on two non-disease specific clinical quality measures (reducing fall risk and all-cause readmissions). It is important to note that the readmissions measure is adjusted for patient age but, fall risk is not adjusted for any patient factors. On the remaining 6 chronic condition-specific measures that are not risk-adjusted, MA contracts with highest concentration of complex patients perform significantly lower.

It is troubling that we find contracts with the highest concentrations of complex patients performed significantly less well on half of the patient experience measures (Getting Appointments and Care Quickly, Customer Service, and Rating of Health Plan) collected by Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey. Given that the CAHPS performance measures are case-mix adjusted, we expected to see no differences by concentration of complexity. These findings raise concerns that there may be meaningful differences in the care experiences of complex patients.

Among experts in the field, there is meaningful disagreement about whether to adjust for clinical and social risk factors in performance measurement.³⁰ Some argue that it is necessary to adjust for the clinical characteristics of enrolled individuals as well as the social risk factors that may affect the measure’s outcomes to appropriately compare providers.³¹ Others consider adjustment a way of institutionalizing different quality levels undermining the goal of improving quality of care for everyone. Similar arguments could be made on behalf of adjusting for patient complexity. Patient health status can make clinical treatment more complex and difficult, which might suggest that adjustment is appropriate. However, it is important to assure that health care providers should not be held to a lower standard for complex patients. A first step would be to select performance measures appropriate for complex patients.

The National Quality Forum’s framework recommends performance measures should be adjusted for risk factors when there is a conceptual justification and empirical evidence. In addition, we recommend that CMS review whether the included disease-specific clinical measures are appropriate for clinically complex populations. Medicare administrators could consider expanding report performance measures focused on the care for older adults (such as those measures SNPs are required to report) or rolling out new measures used in other programs specific to Medicare enrollees with complex needs such as examining hospital readmission rates for those with multiple chronic conditions.

In sum, we find substantial variations in MA contract characteristics and performance by concentration of complex patients. These differences in the underlying populations served by MA contracts raise important questions about whether contracts serving large concentration of complex patients are disadvantaged on clinical performance measures. Further study is needed to understand whether these performance measures are capturing the delivery of poor care, deficiencies in the health plan’s care systems, or whether the measures may not be appropriate for complex patients. Additional work is also needed to identify performance measures specific to the complex populations served by Medicare.