Sarah Hudson Scholle, DrPH

 To determine whether implementation of patient-centered oncology standards in 5 medical oncology practices improved patient experiences and quality and reduced emergency department (ED) and hospital use.

 Retrospective, pre-post study design with a concurrent nonrandomized control group.

 We used insurance claims to calculate all-cause hospitalizations and ED visits and primary care and specialist office visits (n = 28,826 eligible patients during baseline and 30,843 during follow-up) and identify patients for a care experiences survey (n = 715 preintervention and 437 postintervention respondents). For utilization and patient experience outcomes, we compared pilot practices’ performance with 18 comparison practices using difference-in-differences (DID) regression models accounting for practice-level clustering. We assessed pilot practice performance on 31 quality measures from the American Society of Clinical Oncology Quality Oncology Practice Initiative program.

 There were no statistically significant differences in hospital, ED, or primary care visits between the pilot and comparison groups over time, but there was a significant increase in specialty visits for the pilot group (adjusted DID of 0.07; 95% CI, 0.01-0.13; 

 = .03). For care experiences, pilot practices improved more on shared decision-making (4.03 DID composite score; 

 = .013), whereas the comparison group improved more on access (–6.36 DID composite score; 

 < .001) and exchanging information (–4.25 DID composite score; 

 = .013). On average, pilot practices improved performance on 65% of core quality measures from baseline to follow-up.

 This pilot of patient-centered oncology care showed improved quality but no impact on hospitalizations/ED use and mixed results for patient experiences. Findings are consistent with early evaluations of primary care patient-centered medical homes.

Practices implementing a patient-centered oncology care pilot demonstrated improvement on patient-centered process measures but no impact on utilization and mixed results for patient experiences relative to comparison practices.

The greatest process measure improvements occurred in symptom assessment and care planning.

There were no statistically significant differences in hospital, emergency department, and primary care utilization, but there was a significant increase in specialty provider visits for the pilot group.

Pilot practices improved in shared decision-making, whereas comparison practices improved in access and exchanging information.

Findings are consistent with early evaluations of primary care patient-centered medical homes.

Cancer is the second leading cause of death in the United States,

 and about 14.5 million individuals currently live with cancer.

 There is ample evidence of cancer care quality gaps, including non–evidence-based treatment and poor communication, care planning, and coordination.

 Patients with cancer report issues with receiving insufficient or inadequate information.

 The Institute of Medicine, now the National Academy of Medicine (NAM), noted that communication problems often contribute to poor outcomes, and other studies show that clinicians ask for patient preferences in medical decisions only about half the time.

 Often, patients undergoing treatment for advanced cancer do not understand that treatment is not aimed at curing the disease, do not have discussions about treatment preference, or have delays in palliative care discussions that result in active treatment being prolonged during the last weeks of life, contrary to patient preferences.

Patients with cancer often see multiple providers and undergo multiple treatment modalities. Although medical oncologists are typically considered the “captains” of the cancer treatment team, initial diagnosis and treatment may be decided during consultations with a primary care provider and surgeon and without the opportunity for a fully informed decision process or development of a comprehensive treatment plan.

 The lack of clarity in roles continues through treatment and survivorship.

 Current payment systems exacerbate these integration issues. Oncologists are unable to bill for shared decision-making, services to help patients navigate the health care system, or support for emotional problems.

The National Committee for Quality Assurance (NCQA) defined a new model for patient-centered oncology care based on multiple reports from the NAM

 that have called for greater attention to improving delivery of patient-centered care for patients with cancer. The standards for patient-centered oncology care are built on the chassis of NCQA’s successful Patient-Centered Medical Home (PCMH) program, in which primary care practices are responsible for coordinating accessible, continuous, and team-based care for patients, and the Patient-Centered Specialty Care program, which promotes interactions between PCMH practices and “neighbor” specialty practices.

 While supporting this concept of the neighbor, leading oncology societies have suggested that oncology practices may serve in both roles of the “home” for patients (particularly during active treatment) as well as the “neighbor” during periods of transition and survivorship.

 To develop the patient-centered oncology model, NCQA first convened a multistakeholder advisory panel to identify the components of the patient-centered specialty practice standards that should be built out for oncology, as well as items from the PCMH program that are relevant, and then reviewed the recommendations with other multistakeholder committees. The most recent version of the oncology medical home standards can be found on NCQA’s website.

In this study, we evaluated a pilot of the patient-centered oncology care model in oncology practices in southeastern Pennsylvania to determine whether the patient-centered oncology standards improve patient experiences and quality and reduce emergency department (ED) and hospital utilization.

We selected southeastern Pennsylvania as the demonstration location because Independence Blue Cross, an insurer with large market share, was willing to participate in the evaluation. Five practices consisting of 2 large academic medical centers, 2 private physician-owned practices, and 1 hospital-based outpatient department volunteered to pilot patient-centered oncology care. The pilot practices began implementing the standards in January 2014. Practices received implementation support, including monthly webinars and technical assistance, through December 2016.

Our evaluation consisted of a retrospective, pre-post study design with a concurrent nonrandomized control group of 18 local medical oncology practices for the utilization and patient experience outcomes (

). The comparison practices were similar in size and ownership, were located in the same community, and participated in the payer network. Compared with comparison practices, the population served by the pilot practices was younger, was more likely to be male and non-White, had higher average risk scores, was less likely to have common comorbid conditions, and had a somewhat different profile of cancer types (

This project was reviewed and approved by the Chesapeake Research Review Inc Institutional Review Board (IRB) and IRBs at 3 of the participating sites.

 We identified 4 project time periods for the utilization analysis. “Baseline” occurred before the project began, from August 2011 to July 2013; “start-up” occurred from August 2013 to June 2014, when the pilot practices prepared for and began implementation; “intervention” occurred from July 2014 to December 2015, during which the practices continued implementing change; and “follow-up” occurred post implementation, from January to July 2016. For each of these 4 time periods, we identified a cross-section of patients attributed to a pilot or comparison practice. Specifically, for each time period, we used National Provider Identifiers and Tax Identification Numbers to identify all patients in the payer network who had an evaluation and management (E&M) claim at any medical oncology provider during that time period and then excluded patients without a claim at a pilot or comparison practice. We attributed these patients to a pilot or comparison practice if they had an E&M claim for which the performing provider was in the practice and for whom the practice provided all services, a majority of services, or a plurality of services. Plurality was defined by count of E&M visits, with ties broken by the most recent E&M visit date of service relative to other oncology practices. Most patients were exclusive with the index practice (ie, had no E&M claims with any other oncology practice during the time period).

We calculated unadjusted rates of all-cause hospitalizations, all-cause ED visits, and primary care provider (PCP) and specialist office visits per patient per month in each of the 4 study time periods for the pilot and comparison groups. We used a difference-in-differences (DID) regression model with fixed effects for practices to estimate the effects of exposure to the intervention on utilization. Generalized estimating equations with robust standard errors were used to account for heteroscedasticity, autocorrelation, and clustering of patients within practices.

 The dependent variable in the model was the utilization rate in the time period of interest. Therefore, the effects of the intervention were represented by the coefficient estimates for project time period interacted with status (pilot or comparison). The models also controlled for practice fixed effects and a Cotiviti (formerly Verisk Health) DxCG risk score estimating the cost of underlying illness burden.

values < .05 significant. We conducted sensitivity analyses to assess whether results differed based on the population definition: all continuously enrolled patients, patients with a history of cancer since 2009, and patients in active treatment.

We surveyed patients pre- and post intervention who had recently received chemotherapy using version 2.0 of the Consumer Assessment of Healthcare Providers and Systems (CAHPS) Cancer Care Survey; it had 5 composites and an overall rating of drug therapy treatment team.

 For the preintervention survey sample, we attributed patients who had a claim for chemotherapy at pilot and comparison practices from January through June 2014 for the early intervention and for the postintervention time period of January through June 2016. We also added patients who were eligible for the preintervention survey to the postintervention survey sample.

Both surveys were in the field for 10 weeks, starting with an initial mailing, followed by reminder telephone calls and then a subsequent round of mailing and reminder phone calls. We removed ineligible patients from the sample frame after the mailing, including patients who were deceased, had an undeliverable address, or were no longer insured at the time the survey was administered (preintervention response base, 2304; postintervention response base, 1788). We received 715 preintervention responses and 437 postintervention responses for response rates of 31% and 24%, respectively. We removed patients who had skipped relevant survey items (12 patients each at preintervention and post intervention) and patients who indicated that they did not receive drug therapy for cancer at the attributed pilot or comparison practice during the previous 6 months (97 patients at preintervention and 165 patients at post intervention). After removing ineligible patients from the response base, there were 606 valid preintervention survey responses (175 pilot, 431 comparison) and 260 postintervention survey responses (81 pilot, 179 comparison).

We calculated descriptive data on the CAHPS survey demographic variables (Table 2) and composite scores on a 0 to 100 scale using proportional scoring and the summated rating method based on the CAHPS macro.

 This method calculates the mean responses to each item, after transforming each response to a 0 to 100 scale (100 representing the most positive response on any given item response scale; 0 representing the least positive). For example, on a yes/no response scale, if “yes” represents the most positive response, then yes is equal to 100 and no is equal to 0; on an always/usually/sometimes/never response scale, if “always” represents the most positive response, then always is equal to 100; “usually,” 67; “sometimes,” 33; and “never,” 0. A higher score means that practices were rated more positively for care on that item. To estimate the effect of exposure to the intervention, we used a DID model with fixed effects for practices. The dependent variables included the 5 survey composites and an overall rating. The effects of the intervention were represented by the coefficient estimates for survey time period (pre- or post) interacted with status (pilot or comparison). Covariates included standard CAHPS case-mix variables (age, education, self-reported health status, sex, race, ethnicity, and help with completing survey) and nonresponse weighting.

We used 31 measures from the American Society of Clinical Oncology’s Quality Oncology Practice Initiative (QOPI) to evaluate improvement in quality, including 25 core measures for patients receiving chemotherapy with new cancer diagnoses and 6 palliative care measures for patients with new advanced cancer diagnoses. The baseline sample included patients with a new cancer diagnosis from August 2011 through August 2013, and the follow-up sample included patients with a new cancer diagnosis from July 2015 through June 2016. For the core measures, pilot practices reported on a total sample of 308 patients at baseline and 349 patients at follow-up. For the palliative care measures, they reported on a total of 153 patients at baseline and 139 patients at follow-up. Demographic data for the quality measure patient sample are included in 

For the core and palliative care measures, we calculated mean, minimum, and maximum pilot practice performance rates. We calculated the percentage point difference in average performance rates across pilot practices, and we compared average pilot practice performance rates with national and regional benchmark data for all practices in the United States and all practices in HHS Region 3 (which includes Pennsylvania). There were no available benchmark data for the palliative care measures because these measures were not yet included in the regular QOPI reporting program.

There were no statistically significant (

 < .05) DIDs between the pilot and comparison groups on all-cause hospitalizations, all-cause ED visits, or PCP office visits (

). Among patients with active chemotherapy, the unadjusted rate of all-cause hospitalization was 0.05 per member per month (PMPM) for the pilot practices compared with 0.05 PMPM for the comparison group; at follow-up, the rates were 0.10 and 0.08, respectively, and the adjusted DID was –0.03 (95% CI, –0.27 to 0.21). The pilot group was associated with an increase in specialist visits from baseline to intervention period relative to comparison practices. The unadjusted rates of specialty visits were 0.93 PMPM for the pilot group and 0.98 for comparison group at baseline; at follow-up, the rates were 1.4 and 1.3, respectively. The adjusted DID of 0.12 (95% CI, 0.05-0.19) was significant (

 = .002). The sensitivity analysis results were fairly consistent and nonsignificant across all 3 patient groups: those in active treatment, continuously enrolled patients, and patients with a history of cancer since 2009.

Nonresponse analyses showed that respondents were significantly more likely to be older and White and also more likely to have more consistent insurance coverage, coronary artery disease, and a recent chemotherapy visit. In the postintervention group, respondents were also more likely to have lower DxCG risk scores. Respondents and nonrespondents did not differ with regard to sex, education, or presence of other comorbid conditions.

The DID analysis showed mixed results of the pilot on patient experiences (

). Pilot participation was significantly associated with greater improvement in scores on 1 composite, shared decision-making, where the pilot score improved from 76.6 to 83.2 vs a change from 77.9 to 79.7 in the comparison group, yielding a DID result of 4.03 (95% CI, 0.86-7.21). In contrast, the comparison group had greater improvement on 2 composites, access (–6.36 DID; 

 < .001) and exchanging information (–4.25 DID; 

 = .013). There were no statistically significant differences in affective communication, patient self-management, or the overall rating. Detailed results for items comprising the composites suggest items that contribute to the overall findings (not shown). Notably, the pilot practice score on “asked for patient opinion about treatment choices” increased from 79.4 to 90.1 for the pilot group compared with an increase from 80.7 to 83.2 for the comparison group. Further, the pilot practice score on “gave patient clear instruction how to contact them after hours” declined from 89.5 to 77.1 from pre- to post intervention in the pilot group, compared with an increase from 87.8 to 90.8 for the comparison practices.

On average, the pilot practices improved performance between baseline and follow-up on 65% of the 25 core measures, had no improvement on 8% of measures, and declined in performance on 27% of measures (

]). Measures that addressed care planning and assessment had the greatest improvements in performance, whereas performance declined on measures that assessed proper follow-up of identified problems. Mean pilot practice performance rates improved more than 10 percentage points and exceeded the national benchmark by follow-up on the care planning measures for documenting chemotherapy intent (80% at baseline vs 93% at follow-up) and discussing intent of chemotherapy with patients (73% vs 87%), discussing infertility risk (28% vs 63%) and fertility preservation options (53% vs 64%), and documenting chemotherapy treatment summary (41% vs 56%) and providing summary to patient (6% vs 37%).

Mean performance rates improved more than 10 percentage points for each of the symptom assessment palliative care measures for patients with advanced cancer. Similarly, the pilot practices improved on the measures for assessing pain and emotional well-being within the first 2 visits for patients with a new diagnosis of cancer; however, they had lower performance on the measures for addressing problems identified during pain and emotional well-being assessments.

Oncology practices participating in this patient-centered care pilot demonstrated an increase in specialist visits, no impact on hospitalizations and ED use, and mixed results for patient experiences, despite improvements in several patient-centered processes. These results are consistent with early studies of the primary care PCMH, which showed that financial incentives are needed to drive improvement in patient experiences and reductions in utilization and that time is required for outcomes to improve. The increase in specialty visits could indicate that the pilot practices were managing patients more closely, which is the intention of patient-centered care and better symptom management. However, more time may be needed to see a reduction in ED and hospital use. Also, practices may be affiliated with hospitals in which the incentive to reduce use of these services is dampened. Higher levels of care coordination require more staff and documentation resources, and neither this pilot project nor payers offered pilot practices increased reimbursement fees for implementing the patient-centered care model. With respect to the mixed results for patient experiences, one explanation could be that pilot practices had only partially demonstrated implementation of certain standards. For instance, comparison practices had greater improvement than pilot practices on the access composite, and we learned that at follow-up, the pilot practices had not implemented all standards related to care access, including providing timely clinical advice before and after office hours.

There are several other explanations for the mixed results. Throughout the pilot, many of the practices were undergoing upgrades to electronic health record (EHR) systems and changes in ownership or organizational structure, which key informants reported may have disrupted care and affected patient experience. In addition, the practices faced barriers in fully implementing the patient-centered oncology care model due to a lack of resources. This intervention did not include a financial payment strategy, and practices did not receive any payment changes from payers to help transform or manage patients. In contrast, recent studies have shown that where there has been greater investment in systems for population management or alternative payment strategies, there have been decreases in costs and utilization and increases in quality of care and patient satisfaction.

 Some of these studies did not evaluate changes in patient experience or quality but focused only on utilization and cost.

This study has several limitations. The intervention was not randomized, and practices that volunteered to participate differed somewhat from the comparison practices on utilization outcomes from baseline to the start-up period. We did not have information on the implementation of patient-centered standards or quality measure performance for the comparison sites. Thus, participating and comparison practices might have differed on unobserved characteristics that could bias our estimates of intervention effects (eg, volunteering practices might have more engaged leadership that could drive observed differences that our models would erroneously attribute to the intervention; changes in EHR systems in the pilot or the comparison sites may have affected observed differences). A second limitation is that we did not perform a test of parallel trends. With only 2 years of baseline data and a small sample size, we had concerns that a parallel trends test would be underpowered and thus uninformative. A third limitation involves concerns about ceiling effects on the patient experience survey and the ability of oncology practices to improve on the composite scores over time, although there is evidence demonstrating the use of CAHPS scales for assessing and improving quality.

Finally, the approach we used to attribute patients to pilot and comparison practices to establish which practices should be accountable for patients’ care is one that is commonly used, but we found that it was imperfect and the patterns of attribution changed over time. The percentage of survey respondents who denied receiving care at the attributed practice was higher at the second survey, and the number of patients attributed to the pilot practice decreased between baseline and follow-up. We found few claims for patients in 1 of the pilot practices, which was due to the practice billing chemotherapy through the affiliated hospital rather than through the medical oncology practice, and we were unable to separate the practice patient population claims from the hospital patient population claims. In addition, Independence Blue Cross made changes to the provider identification system between the baseline and follow-up period, which may have affected the number of patients attributed to practices in the follow-up utilization and survey samples. Any imperfections in attribution would constitute error in the main identification variable, which would bias any effect estimates of the intervention toward the null hypothesis (because the error term of the regression becomes correlated with the measurement error in the independent variable itself).

Oncology practices participating in this patient-centered care pilot demonstrated improvement in quality on several patient-centered processes, particularly symptom assessment and care planning, but patient experiences showed mixed results. Use of specialty care increased, but there was no impact on hospitalizations or ED use. These findings are consistent with early evaluations of the PCMH in primary care.

 This was the first such intervention to be tried and evaluated in oncology practices. Future research is needed to evaluate different oncology practice interventions and identify effective approaches for improving patient-centered outcomes in oncology and among subgroups of patients.

The authors thank Shelley Fuld Nasso, MPP, National Coalition for Cancer Survivorship; Ellen Stovall, National Coalition for Cancer Survivorship; John Sprandio, MD, Oncology Management Services; Susan Tofani, MS, Oncology Management Services; Patti Larkin, RN, MSN, American Society for Clinical Oncology; and Johann Chanin, RN, MSN, consultant, for their guidance throughout this study.


 National Committee for Quality Assurance (LMR, MT, TP, SHS), Washington, DC; RAND Inc (MF), Boston, MA; Independence Blue Cross (AS-M), Philadelphia, PA.

 Research reported in this paper was funded through a Patient-Centered Outcomes Research Institute (PCORI) Award (IH-12-11-4383). The statements in this paper are solely the responsibility of the authors and do not necessarily represent the views of PCORI, its Board of Governors, or its Methodology Committee.

 Ms Roth’s and Dr Scholle’s institution, the National Committee for Quality Assurance, recognizes practices as oncology medical homes. Since 2016, Dr Friedberg has received financial support for research from the Agency for Healthcare Research and Quality, American Board of Medical Specialties Research and Education Foundation, American Medical Association, Center for Medicare & Medicaid Innovation, Centers for Medicare & Medicaid Services, Cedars-Sinai Medical Center, Commonwealth Fund, Milbank Memorial Fund, National Institute on Aging, National Institute on Drug Abuse, National Institute of Diabetes and Digestive and Kidney Diseases, National Institute on Minority Health and Health Disparities, Patient-Centered Outcomes Research Institute, and Washington State Institute for Public Policy. Since 2016, Dr Friedberg has received payments from Consumer Reports for consulting services. Dr Friedberg also has a clinical practice in primary care at Brigham and Women’s Hospital and thus receives payment for clinical services, via the Brigham and Women’s Physician Organization, from dozens of commercial health plans and government payers, including but not limited to Medicare, Medicaid, Blue Cross and Blue Shield of Massachusetts, Tufts Health Plan, and Harvard Pilgrim Health Plan, which are the most prevalent payers in Massachusetts. Dr Friedberg also receives compensation from Harvard Medical School for tutoring medical students in health policy. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (LMR, MT, MF, AS-M, SHS); acquisition of data (LMR, TP, AS-M); analysis and interpretation of data (LMR, MT, TP, MF, AS-M); drafting of the manuscript (LMR, MT, MF, AS-M, SHS); critical revision of the manuscript for important intellectual content (LMR, MF, SHS); obtaining funding (SHS); administrative, technical, or logistic support (TP); and supervision (LMR, SHS).

 Lindsey M. Roth, MPP, National Committee for Quality Assurance, 1100 13th St NW, Third Floor, Washington, DC 20005. Email: roth@ncqa.org.

1. Hoyert DL, Xu J. Deaths: preliminary data for 2011. 

2. National Center for Health Statistics. Health, United States, 2015: with special feature on racial and ethnic health disparities. CDC. 2016. Accessed May 17, 2017. https://www.cdc.gov/nchs/data/hus/hus15.pdf

3. SEER cancer statistics review, 1975-2012. Surveillance, Epidemiology, and End Results Program. Updated November 18, 2015. Accessed May 17, 2017. http://seer.cancer.gov/csr/1975_2012/

Delivering High-Quality Cancer Care: Charting a New Course for a System in Crisis

5. Ayanian JZ, Zaslavsky AM, Guadagnoli E, et al. Patients’ perceptions of quality of care for colorectal cancer by race, ethnicity, and language. 

. 2005;23(27):6576-6586. doi:10.1200/JCO.2005.06.102

6. Ayanian JZ, Zaslavsky AM, Arora NK, et al. Patients’ experiences with care for lung cancer and colorectal cancer: findings from the Cancer Care Outcomes Research and Surveillance Consortium. 

. 2010;28(27):4154-4161. doi:10.1200/JCO.2009.27.3268

7. McInnes DK, Cleary PD, Stein KD, Ding L, Mehta CC, Ayanian JZ. Perceptions of cancer-related information among cancer survivors: a report from the American Cancer Society’s studies of cancer survivors. 

. 2008;113(6):1471-1479. doi:10.1002/cncr.23713

Cancer Care for the Whole Patient: Meeting Psychosocial Health Needs

Patient-Centered Cancer Treatment Planning: Improving the Quality of Oncology Care: Workshop Summary

10. Lee CN, Chang Y, Adimorah N, et al. Decision making about surgery for early-stage breast cancer. 

. 2012;214(1):1-10. doi:10.1016/j.jamcollsurg.2011.09.017

11. Zikmund-Fisher BJ, Couper MP, Singer E, et al. Deficits and variations in patients’ experience with making 9 common medical decisions: the DECISIONS survey.

. 2010;30(5 suppl):85S-95S. doi:10.1177/0272989X10380466

12. Dow LA, Matsuyama RK, Ramakrishnan V, et al. Paradoxes in advance care planning: the complex relationship of oncology patients, their physicians, and advance medical directives. 

. 2009;28(2):299-304. doi:10.1200/JCO.2009.24.6397

Articles

Practices implementing a patient-centered oncology care pilot had improved quality, but utilization and patient experiences did not differ from comparison practices.

Patient-Centered Oncology Care: Impact on Utilization, Patient Experiences, and Quality

 People with serious mental illnesses (SMI), including schizophrenia, bipolar disorder, and major depression, experience early mortality, partly due to comorbid physical health conditions such as diabetes and hypertension. This study examined the quality of diabetes and hypertension care for Medicaid and Medicare enrollees with SMI.

 We conducted a retrospective analysis of medical records and claims data from 3 health plans: a Medicaid plan for disabled adults, a Medicaid plan for low-income adults, and a Special Needs Plan for individuals dually enrolled in Medicaid and Medicare. The study population included 258 adults with SMI and diabetes and 241 adults with SMI and hypertension.

 Existing quality measures for diabetes and hypertension from the Healthcare Effectiveness Data and Information Set (HEDIS) were adapted and applied to the SMI population for the 2012 calendar year. The rates of diabetes care and hypertension control for people with SMI were compared with national averages for Medicaid and Medicare managed care plans to examine disparities in care.

 Adults with SMI receive poor-quality care for diabetes and hypertension. Depending on the health plan, performance on the diabetes care and hypertension control HEDIS measures was 14 to 49 percentage points lower among the SMI population than the general Medicaid and Medicare populations. 

 Findings highlight disparities in care for individuals with SMI compared with the general Medicaid and Medicare populations. Health plans demonstrated substantial room for improvement on almost all diabetes and hypertension HEDIS measures for the SMI population.

Health plans demonstrated substantial room for improvement on almost all Healthcare Effectiveness Data and Information Set (HEDIS) diabetes and hypertension measures for individuals with serious mental illness (SMI). 

		The SMI population contributes to disproportionately high healthcare costs. 

		Disparities in care for the SMI population is mostly driven by the lack of utilization of ambulatory medical care. 

		The diabetes and hypertension outcome measures developed for this study fill critical gaps in quality measurement for the SMI population. 

		Health plans and state agencies can use the measures to evaluate physical and mental health integration efforts and to monitor and improve the quality of care for the SMI population.

Individuals with a serious mental illness (SMI), including schizophrenia, bipolar disorder, and major depression, are at higher risk of chronic physical health conditions, such as diabetes and hypertension. This is partly due to poor diet, side effects of psychiatric medications, and lack of access to primary care.

 Study results suggest that roughly 15% of individuals with schizophrenia and 18% with major depression have diabetes compared with 9% of the general population.

 Nearly 40% of individuals with bipolar disorder and 48% of individuals with major depression experience hypertension compared with 30% of the general population.

 These conditions contribute to early mortality rates among individuals with SMI.

Hypertension and diabetes have often progressed by the time they are identified in individuals with SMI,

 who are less likely to receive high-quality care for these diseases.

 Data from clinical trials and national epidemiological studies suggest that at least 30% of individuals with schizophrenia or bipolar disorder receive poor medication treatment for diabetes and 62% receive poor medication treatment for hypertension.

Policy makers, health plans, and providers are learning that improving care quality and outcomes for the SMI population will require addressing chronic, comorbid physical health conditions.

 State Medicaid programs recognize that although only about 10% of Medicaid beneficiaries have SMI,

 their complex conditions make their care costly. Current programs and delivery system reforms promoting integration of primary and behavioral healthcare are promising. Many health plans are testing strategies for integrating and coordinating care for this population, with special attention on physical health.

However, there is limited published information on how well plans manage diabetes and hypertension care for individuals with SMI. As part of an effort to develop behavioral health quality measures for health plans, we pilot-tested measures focusing on diabetes and hypertension care for this population. This article presents measure performance for 2 Medicaid plans and a Medicaid and Medicare Dual-Eligible Special Needs Plan (Medicaid & Medicare D-SNP) that serves individuals enrolled in Medicaid and Medicare. We examined disparities in diabetes and hypertension care for those with SMI relative to the general Medicaid managed care population. Study findings are intended to provide both insight into the quality of care for the SMI population and a baseline for measuring progress. 

To examine variation in measure performance, we conducted outreach through health plan trade organizations and used a list of health plans maintained by the National Committee for Quality Assurance (NCQA) to recruit geographically diverse plans that serve different types of Medicaid populations. Three plans (1 each in the Northeast, Midwest, and West regions) participated: a Medicaid plan for disabled adults (Medicaid-Disabled Plan), a Medicaid plan for low-income adults (Medicaid-Adult Plan), and a Dual-Eligible Special Needs Plan (D-SNP) for individuals enrolled in Medicaid and Medicare (Medicaid & Medicare D-SNP). Plan enrollment ranged from 13,000 (Medicaid Adult, Medicaid & Medicare D-SNP plans) to 131,000 (Medicaid-Disabled Plan). All 3 plans were responsible for medical, pharmacy, and mental health benefits; had access to physical and behavioral health records; have been in business for at least 20 years, and offer lines of business other than the products in this study. Two of the 3 have NCQA accreditation.  

We adapted 2012 Healthcare Effectiveness Data and Information Set (HEDIS) measures for diabetes and hypertension to apply to the SMI population (see 

 for measure descriptions). We asked plans to report on member use of ambulatory care: the percentage of members with SMI who had an ambulatory physical healthcare visit in 2012.

Data were collected in 2013. Health plans followed standard procedures for gathering data to report HEDIS measures (all plans annually report diabetes and hypertension measures to NCQA). To identify a sample of members eligible for each measure, each plan used claims/encounter data from calendar year 2012 to identify members with SMI, diabetes (2011 claims were also used for diabetes), and hypertension. Plans also used claims data to report rates of ambulatory physical healthcare. 

Plan members were eligible if they: 1) were continuously enrolled from January 1 through December 31, 2012; 2) were 18 years or older during the same period; and 3) had at least 1 inpatient or 2 outpatient claims with a diagnosis of schizophrenia or bipolar disorder, or at least 1 inpatient claim with a diagnosis of major depression (previous studies included schizophrenia, bipolar, and major depression in the SMI population

; we required an inpatient stay for major depression to ensure that individuals included in the measure denominator had severe depression). Across plans, 884 members with SMI were identified (345 in Medicaid & Medicare D-SNP, 219 in Medicaid-Disabled, 320 in Medicaid-Adult). After identifying the entire SMI population, the plan assigned individuals into 2 denominator groups: 1) members who received at least 1 diabetes medication or had at least 1 claim for diabetes in 2011 or 2012 and 2) members diagnosed with hypertension in 2012. 

The final sample included 258 members with SMI and diabetes (135 in Medicaid & Medicare D-SNP, 40 in Medicaid-Disabled, 83 in Medicaid-Adult) and 241 members with SMI and hypertension (91 in Medicaid & Medicare D-SNP, 104 in Medicaid-Disabled, 46 in Medicaid-Adult). Members could qualify for both groups (64 members qualified for both). For both groups, the final sample included a fairly even distribution of individuals with schizophrenia, bipolar disorder, and major depression. 

Health plan staff manually reviewed members’ medical and behavioral health records (when available) for the same period. Two plans had access to electronic health records or other electronic systems and the third plan accessed hard copies of medical records from which they extracted information to calculate measures. Manual abstraction confirmed the hypertension diagnosis (to confirm eligibility for the measure denominator, per HEDIS specifications) and determined if the member received services consistent with the numerator requirement. Because behavioral health records provided little additional information, study results were based on medical records.

To examine interrater reliability, plan staff performed double abstraction for a subset of records (69 patients with diabetes, 67 patients with hypertension); interrater reliability (kappa =  0.65-1) was substantial or almost perfect.

 Plans provided de-identified data that did not require institutional review board approval. 

We used plan data to calculate the percentage of members with SMI and diabetes and/or hypertension who received care consistent with measure specifications. To analyze potential disparities, we compared the SMI population’s average rate with the general population’s average HEDIS Medicaid rate in 2012 for the 3 plans using data in Quality Compass 2013 with the permission of the NCQA. We also compared rates of diabetes care and hypertension control for members with SMI, with the HEDIS national averages for the same measures, which included roughly 180 Medicaid and 500 Medicare managed care plans.

Average health plan performance among the SMI population ranged from 13.2% for eye exams to 48.0% for glycated hemoglobin (A1C) testing (

). Variation was widest on medical attention for nephropathy (a 55-percentage-point difference across plans) and smallest for eye exams (a 15-percentage-point difference across plans). 

The plan-average HEDIS Medicaid rates for the general population, compared with the SMI population’s mean performance rates on the diabetes indicators, were 16 percentage points (A1C testing) to 45 percentage points (eye exam) lower. Compared with 2012 HEDIS national average Medicaid rates, mean performance rates on the diabetes indicators were 14 percentage points (A1C <8%) to 40 percentage points (eye exam) lower for the SMI population. When looking at 2012 HEDIS national average Medicare rates, the average rates across plans on the diabetes indicators were 19 percentage points (blood pressure control) to 49 percentage points (medical attention for diabetic nephropathy) lower for the SMI population.

Controlling Hypertension Measure Performance 

About 40% of members with hypertension had their blood pressure controlled by the end of 2012 (ranging from 12.5%-57.7% across plans) (Table 2). The average rate was 17 percentage points lower than plans’ HEDIS Medicaid rates for the general population. Compared with 2012 HEDIS national average Medicaid rates, performance on hypertension control was 15 percentage points lower among the SMI population. Compared with 2012 HEDIS national average Medicare rates, performance on hypertension control was 21 percentage points lower among the SMI population. 

Variation in the receipt of ambulatory care across plans contributed to poor measure performance. In the Medicaid & Medicare D-SNP and Medicaid-Disabled plans, members with at least 1 ambulatory visit in 2012 ranged from 88% to 99% for the diabetes and hypertension groups, respectively. In contrast, in the Medicaid-Adult plan (which had the worst performance on all measures), 19% of members with diabetes and 15% with hypertension had at least 1 ambulatory visit in 2012. The lower performance of this plan was driven in part by less utilization of ambulatory care and, therefore, fewer opportunities to monitor care.  

Health plans play an important role in delivering care for the SMI population. Physical and behavioral healthcare providers often deliver care in fragmented systems, limiting their ability to ensure that individuals with SMI receive comprehensive care. In contrast, health plans are positioned to use data to monitor delivery of care across providers and to intervene for members who are not receiving adequate care. The results of this pilot study suggest that individuals with SMI enrolled in Medicaid and Medicare plans receive poor care for diabetes and hypertension compared with the general Medicaid and Medicare managed care populations. We found wide variation in plan performance for the SMI population. With the exception of 1 measure (eye exam for diabetes), the Medicaid & Medicare D-SNP had the highest performance across measures. 

Although it is difficult to draw broad conclusions from only 3 health plans, performance variations may reflect differences in plan capacity to manage care for this population. Medicaid & Medicare D-SNPs are designed to coordinate care for dual-eligible populations with complex health conditions. At least 1 other study found that dual-eligible enrollees were more likely to receive diabetes testing than Medicaid-only enrollees.

 Performance differences could reflect that health plans serve different populations and operate in different state contexts, in terms of service systems and available community resources. Utilization of ambulatory care corresponded to measure performance, suggesting that getting people into care is key to managing conditions. Health plans may identify individuals with SMI who are at risk for poor outcomes simply by monitoring the use of ambulatory physical healthcare.  

Previous studies using claims data, clinical registries, and information collected at mental health clinics have found poor management of diabetes and hypertension among the SMI population.

 This study’s findings provide a snapshot of the quality of care for individuals with SMI who are enrolled in comprehensive health plans, using data collected from health records. This study also used common measures for diabetes and hypertension care for direct comparison with care quality for the general plan population. Although these plans were responsible for physical and behavioral health benefits, individuals with SMI had poor management of diabetes and hypertension. Given the side effects of psychiatric medications, providers should routinely monitor the comorbid medical conditions of SMI populations.  

This pilot study was limited to health plans, and the sample size for each plan was modest due to variation in the prevalence of members with SMI and diabetes or hypertension. As other plans use these or similar measures, it will be important to compare their performance with this study’s findings and to examine sources of performance variations. 

This study highlights disparities in care for diabetes and hypertension for individuals with SMI who are enrolled in Medicaid and Medicare managed care plans. Although health plans demonstrated variation in measure performance, the average on most indicators was much lower for the SMI population than for the general population. Measures developed for this study fill critical gaps in quality measurement for the SMI population. Following pilot testing, they were endorsed by the National Quality Forum and are now available for health plans, state agencies, and other organizations to monitor quality of care for the SMI population. They may also be useful for ongoing healthcare delivery system demonstrations and reforms toward improving care for this population.

 National Committee for Quality Assurance (JL, SM, SHS, RL, MP), Washington, DC; Mathematica Policy Research, Inc (JB), Washington, DC; Office of the Assistant Secretary for Planning and Evaluation (DEBP), Washington, DC; Substance Abuse and Mental Health Services Administration (LP), Rockville, MD. 

 The authors conducted this work under contract to the Office of the Assistant Secretary for Planning and Evaluation (ASPE), US Department of Health and Human Services (HHS) (HHSP23320100019WI, HHSP23337001T). Funding was provided by the Substance Abuse and Mental Health Services Administration (SAMHSA), HHS. Views and opinions expressed are those of the authors and do not necessarily reflect the views, opinions, or policies of ASPE, SAMHSA, or HHS.

 Drs Liu and Scholle, Ms Morton, Mr Patel, and Ms Lewis, work for National Committee for Quality Assurance, a nonprofit organization that develops and maintains quality measures. The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article. 

 Concept and design (JL, JB, MP, RL, DEBP, LP, SHS); acquisition of data (JL, JB, RL, DEBP, SHS); analysis and interpretation of data (JL, JB, SM, MP, RL, SHS); drafting of the manuscript (JL, JB, SM, MP, RL, SHS); critical revision of the manuscript for important intellectual content (JL, JB, SHS); statistical analysis (JB, SM); obtaining funding (JB, DEBP, LP, SHS); administrative, technical, or logistic support (MP, RL, DEBP, LP). 

 Junqing Liu, PhD, National Committee for Quality Assurance, 1100 13th St NW, Ste 10000, Washington, DC 20005. E-mail: liu@ncqa.org. 

This study highlights disparities in care for diabetes and hypertension for individuals with serious mental illness compared with the general Medicaid and Medicare populations.

Disparities in Diabetes and Hypertension Care for Individuals With Serious Mental Illness

 This study examined physical and mental health, health symptoms, sensory and functional limitations, risk factors, and multimorbidity among older Medicare managed care members to assess disparities associated with race/ethnicity.

 We used data on 236,289 older adults from 208 Medicare plans who completed the 2012 Medicare Health Outcomes Survey to compare 14 health indicators across non-Hispanic whites, blacks, American Indians/Alaskan Natives, Asians, Native Hawaiians/Pacific Islanders, multiracial individuals, and Hispanics. Logistic regression models that clustered on the plan estimated the risk of indicators of adverse health and functional status.

 Even after controlling for key patient sociodemographic factors, race/ethnicity was significantly associated with most adverse health indicators. Except for Asians, all racial/ethnic minority groups were significantly more likely than whites to report poor mental health status, presence of most health symptoms, sensory limitations, and activities-of-daily-living disability. Important differences were observed across racial and ethnic groups.

Despite some exceptions, elders of racial/ethnic minority background are generally at higher risk than non-Hispanic whites for a broad range of adverse health and functional outcomes that are not routinely assessed. Limitations include bias related to self-reported data and respondent recall. Future research should consider ethnic subgroup variations; employing newer techniques to improve estimates for smaller groups; and prioritizing and identifying opportunities for care improvement of diverse enrollee groups by considering specific needs. To improve the health status of the elderly, service delivery targeting the needs of specific population groups, coupled with culturally appropriate care for racial/ ethnic minorities, should also be considered.

Critical patient-reported outcomes of care are not routinely assessed among diverse elderly racial/ethnic groups in Medicare managed care.

		Certain non-white racial/ethnic groups—blacks, Hispanics, American Indians, and Native Hawaiians&mdash;had worse functional status, health symptoms, and risk factors than whites. Asians were an exception.

		As the proportion of older adults from non-white groups grows more rapidly than whites, it will be crucial to understand health outcomes in diverse enrollee populations.

		To improve the health status of the elderly and reduce disparities, service delivery targeting the needs of specific population groups should be considered.

The maintenance and improvement of health and functioning is a major goal in providing care to older adults in the United States.

 Improving health and eliminating disparities in quality of care are a primary objective of Healthy People 2020, a national initiative aimed at helping to improve health-related quality of life and other aspects of population health that have been identified as health system priorities by the Institute of Medicine.

Further, the Affordable Care Act (ACA) includes provisions to address health disparities, including enhanced understanding and data collection regarding care for specific racial and ethnic groups in federally supported health programs like Medicare managed care (MC).4 Moving forward, MC plans will have strong incentives to address healthcare disparities, given that payment will be tied to their ability to provide better clinical quality and patient experiences for all members, as measured by common performance metrics (eg, Healthcare Effectiveness Data and Information Set or Consumer Assessment of Healthcare Providers and Systems measures). Among older adults, those of racial and ethnic minority backgrounds have been found to receive lower quality of care than whites,

including documented disparities in the delivery of primary and preventive services,

 Increasingly, patient-reported outcomes, including functional health outcomes, are being used as quality indicators.

 However, little is known about how functional health outcomes vary across racial and ethnic groups. An understanding of patterns of health and functional status across racial and ethnic groups is needed to develop targeted interventions aimed at improving health outcomes.

Moreover, existing Medicare disparities research has focused primarily on blacks and whites—and, to a lesser extent, Hispanics&mdash;with much less analysis of the experiences of other racial/ethnic groups: American Indians/Alaskan Natives, Asians, Native Hawaiians/Pacific Islanders, or individuals of more than 1 race.

 The proportion of elderly people within these other groups is expected to increase more quickly than that of blacks or whites by more than double—or even triple&mdash;by midcentury. 

 While studies have documented disparities among these other groups, often they do not simultaneously compare groups within a single study, and draw conclusions based on varying studies and sources, wherein populations are not necessarily comparable because of differences among data sources or health measures used.

 Because of sample size limitations, many smaller racial/ethnic groups are also excluded from national comparative assessments (eg, Native Hawaiian/ Pacific Islander, American Indian/Alaskan Native, and Asian populations). When included, there is often insufficient data for a large proportion of health indicators used in such assessments. 

 Studies that do include more diverse racial/ethnic groups have revealed somewhat mixed findings: studies generally find lower-quality care among minority groups compared with whites, but there is evidence that Asians fare better than whites for some aspects of healthcare.

 Further, there is limited evidence for Native Hawaiians/Pacific Islanders because they are often included with Asians.

This study examined racial/ethnic differences in a large variety of health indicators and functional status measures among adults 65 years and older from a diversity of racial/ ethnic groups, using data from a nationally representative sample of Medicare beneficiaries enrolled in Medicare health plans. It sought to (1) assess racial/ethnic differences in the prevalence of 14 health status indicators across diverse racial/ ethnic groups within a single study; and (2) examine whether health status differs by race/ethnicity even after adjusting for patients’ age, sex, socioeconomic status, region, and length of enrollment in their Medicare plan. Unlike much earlier research, this study used data on critical end points of care that are not routinely assessed (eg, functional health status), rather than health services receipt. The study also examined estimates for smaller groups often excluded from Medicare comparisons because of insufficient data (eg, Native Hawaiian, American Indian, and Asian populations).

Data are from the 2012 baseline Medicare Health Outcomes Survey (HOS), an annual mail survey with telephone follow-up sponsored by the Centers for Medicare & Medicaid Services. The survey collected health and demographic information from a nationally representative sample of Medicare beneficiaries enrolled in Medicare health plans, and had an overall response rate of 47.5%. Race/ethnicity was self-reported. The core measure of health was the Veterans-RAND 12- Item Short Form Survey (VR-12), an instrument embedded within the HOS that allows calculation of 0-100 Physical Component Summary (PCS) and Mental Component Summary (MCS) scores, with 0 representing the worst health and 100 representing the best health for each.

 Other measures of health status in the survey assessed perceived health, presence of health symptoms, sensory limitations, risk factors, functional impairments, and chronic conditions.

The study sample included 236,289 elderly Medicare health plan members from 208 plans in 2012, who were 65 years and older, residing in the 50 US states or the District of Columbia, and who self-reported their race/ethnicity.

The main independent variable was Medicare beneficiaries’ self-reported race/ethnicity. Seven race/ethnicity categories were created based on respondents’ self-report of Latino or Hispanic descent and race (white, black, American Indian/ Alaskan Native, Asian, Native Hawaiian/Pacific Islander). A person identifying as Hispanic was coded “Hispanic” irrespective of race. Among those not self-identifying as Hispanic, a person reporting a single race was coded in the corresponding race group; a person reporting more than 1 race was coded as multiracial. The resulting race/ethnicity categories were Hispanic (of any race), non-Hispanic white, non-Hispanic black, non-Hispanic American Indian/Alaskan Native, non- Hispanic Asian, non-Hispanic Native Hawaiian/Pacific Islander, and non-Hispanic multiracial (individuals reporting more than 1 race).

The main dependent variables were 14 health status indicators from the HOS, assessing general health status, presence of health symptoms, sensory limitations, risk factors, functioning, and presence of chronic conditions.

Three measures of general health status and functioning were assessed: self-report of perceived health, VR-12 PCS score, and VR-12 MCS score. To focus on indicators of worsethan- average health, each measure was dichotomized as an adverse health indicator. Perceived health was dichotomized as fair or poor (vs excellent, very good, or good) health. The PCS measure was coded as poor physical health—scoring in the lowest quartile of the PCS score distribution (

30.11) versus higher. The MCS measure was similarly coded as lowest quartile (

45.05) versus higher. The choice to dichotomize the PCS and MCS health variables was motivated partly by the desire to understand racial/ethnic disparities for risk of being in the poorest health, focusing on the most vulnerable of the elderly, rather than capturing distinctions between average and better than average health.

Respondents reported the presence or absence of 5 health symptoms: chest pain symptoms in the last 4 weeks when exercising or resting (vs no symptoms); shortness of breath symptoms in the last 4 weeks (when lying flat, sitting, walking <1 block, or climbing 1 flight of stairs vs no symptoms); foot symptoms in the last 4 weeks (numbness, tingling, inability to feel hot/cold, or nonhealing sores on feet vs no symptoms); arthritis pain the last 4 weeks (severe or moderate pain vs mild, very mild, or no pain); and depressed mood for much of the last year (yes or no).

Respondents were asked to report vision problems (ability to see well enough to read newspapers, with glasses or contacts if needed) or hearing problems (ability to hear most things, with a hearing aid if needed). Respondents were classified as obese (body mass index [BMI] >30 kg/m2) versus not, based on self-reported weight and height. Current smoking status was dichotomized as smoking every day or some days versus not smoking at all.

Adverse indicators included activity-of-daily-living (ADL) disability and multimorbidity. Disability was assessed through a single dichotomous indicator of any (vs no) selfreported difficulty, or inability to perform any of 6 ADLs: bathing, dressing, eating, getting in or out of chairs, walking, and toileting. A multimorbidity indicator was defined as reporting 4 or more (vs 3 or fewer) of 14 chronic conditions: hypertension, angina pectoris or coronary artery disease, congestive heart failure, heart attack, other heart condition, stroke, respiratory conditions, inflammatory bowel disease, hip/knee arthritis, hand/wrist arthritis, osteoporosis, sciatica, diabetes, and any cancer.

Covariates were age (in years), gender, marital status (married vs not), education (some college or higher vs no college), Medicaid eligibility (based on administrative data), length of enrollment in Medicare health plan (in months), and census region (New England, Mid-Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain, or Pacific).

Analyses were performed using SAS version 9.1 (SAS Institute, Cary, North Carolina). The patient was the unit of analysis. The unadjusted prevalence of all 14 adverse health indicators was compared by race/ethnicity using the χ

 test, and the difference in prevalence of an indicator for each racial/ ethnic group, relative to non-Hispanic whites, was reported. To further examine the association of race/ethnicity with health, 14 multivariate logistic regressions were used to predict each health indicator from all racial/ethnic indicators other than non-Hispanic whites and the full set of covariates described above, accounting for clustering of members within plans using PROC GENMOD. Multivariate results were reported as adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Because this study examined 14 dependent variables for each individual, a Bonferroni correction was used to interpret 

 values. Thus, at the alpha testing level of .05, only 

 values <.003 (.05/14) were considered significant.

The 236,289 elderly Medicare health plan members in the study sample included 176,994 (74.9%) white; 20,553 (8.7%) Hispanic; 22,729 (9.6%) black; 1054 (0.5%) American Indian/ Alaskan Native; 8228 (3.5%) Asian; 729 (0.3%) Native Hawaiian/Pacific Islander; and 6002 (2.5%) multiracial (>1 race) survey respondents. Blacks, Hispanics, American Indians/Alaskan Natives, Native Hawaiians/Pacific Islanders, and multiracial respondents had lower levels of education and homeownership than did whites, and higher levels of Medicaid eligibility (

). Asians had higher levels of education than whites, but also reported lower levels of home ownership and higher levels of Medicaid eligibility.

Health and functional health status as measured by VR-12 PCS and MCS scores, number of ADL impairments, and number of chronic conditions were generally worse among those of racial/ethnic minority backgrounds (Table 1) than whites. Hispanics, blacks, American Indians/Alaskan Natives, and multiracial individuals had slightly lower mean PCS scores than other racial/ethnic groups. All minority groups had lower mean MCS scores than whites. The mean number of functional ADL limitations and chronic conditions was slightly higher among Hispanics, blacks, American Indians/Alaskan Natives, Native Hawaiians/Pacific Islanders, and multiracial individuals than other groups. With the exception of lower mean MCS scores, Asians had similar or better functional status on most measures than whites. Native Hawaiians/Pacific Islanders had functional status similar to whites.

The prevalence of 14 adverse health indicators by race/ ethnicity appears in 

. Adjusted ORs for all racial/ethnic minority groups relative to whites appear in 

. A higher proportion of those of racial/ethnic minority backgrounds had fair or poor perceived health compared with whites (Table 2). Blacks (41.9%) and Hispanics (45.8%) had the highest prevalence of fair or poor perceived health, while whites had the lowest prevalence (24.1%). The proportion of beneficiaries in the sickest quartile of PCS scores ranged from 33.7% and 32.5% for American Indians/Alaskan Natives and multiracial individuals respectively to 18.6% and 21.3% for Asians and Native Hawaiians/Pacific Islanders. A higher proportion of all racial/ethnic minority groups also had poorer mental health (sickest quartile of MCS scores) than did whites, with the most pronounced absolute difference of 19 and 15 percentage points among Hispanics and American Indians/Alaskan Natives.

Adjusting for sociodemographic factors and region eliminated some differences between whites and racial/ethnic minority groups, but did not greatly alter these patterns (Table 3). Racial/ethnic minority groups generally had significantly higher odds of fair or poor perceived health than whites, from an OR of 1.30 for Hispanics and American Indians/Alaskan Natives (95% CI, 1.24-1.37 and 1.13-1.49, respectively) to an OR of 1.40 (95% CI, 1.32-1.49) for multiracial individuals. After adjustment, differences in the proportion of individuals who were in the lowest quartile of PCS were no longer significant for most groups. Moreover, Asians and Native Hawaiians/ Pacific Islanders were significantly less likely than whites to have low PCS scores (OR, 0.52 [95% CI, 0.48-0.56] and OR 0.67 [95% CI, 0.55-0.81], respectively). All racial/ethnic minority groups had significantly higher adjusted odds of poor mental health than whites (from OR 1.22 [95% CI, 1.14-131] to 1.53 [95% CI, 1.32-1.78]).

. Racial/ethnic minority groups were generally more likely than whites to report chest pain; shortness of breath; and foot, arthritis, and depression symptoms (Table 2). Prevalence of symptoms was usually highest among blacks, Hispanics, American Indians/Alaskan Natives, Native Hawaiians/Pacific Islanders, and multiracial individuals. Conversely, Asians had the lowest prevalence of arthritis symptoms and were similar to whites in prevalence of shortness- of-breath symptoms.

Most, but not all, significant findings in bivariate comparisons remained significant after multivariate adjustment, but patterns were largely similar (Table 3). Hispanics, blacks, American Indians/Alaskan Natives, Native Hawaiians/Pacific Islanders, and multiracial individuals had significantly higher odds than whites of most symptoms. Conversely, Asians had significantly lower odds than whites for shortness of breath (OR, 0.85; 95% CI, 0.80-0.90) and arthritis symptoms (OR, 0.60; 95% CI, 0.56-0.65).

. Hispanics and Native Hawaiians/Pacific Islanders had the highest prevalence of vision and hearing problems. All racial/ethnic groups except Asians had a higher prevalence of obesity than whites. Blacks had the highest prevalence of obesity relative to whites (40.6% vs 28.6%), while Asians had a much lower prevalence than whites (8.1% vs 28.6%). Most groups had rates of smoking that were similar to, or lower than, those of whites.

Adjusting for sociodemographic factors did not greatly alter the patterns for vision problems (Table 3). Most racial/ ethnic minority groups had significantly higher odds of vision problems than whites (from OR 1.36 [95% CI, 1.27-1.45] to 1.67 [95% CI, 1.32-2.10]). Racial/ethnic minority groups also had significantly higher odds of hearing problems than whites (from OR 1.18 [95% CI, 1.08-1.30] to 1.68 [95% CI, 1.30- 2.17]). Reflecting the patterns observed in bivariate results, most groups except blacks and Asians had few significant differences from whites in obesity; blacks retained significantly higher odds of obesity than whites (OR 1. 43; 95% CI, 1.38-1.48), while Asians retained significantly lower odds (OR 0.22; 95% CI, 0.19-0.25). Except for multiracial individuals, most racial/ethnic groups also had either no significant difference or lower odds of being smokers than did whites.

. Blacks, American Indians/Alaskan Natives, Native Hawaiians/Pacific Islanders, and multiracial individuals had worse functioning relative to whites, while Asians had better functioning (Table 2). The prevalence of multimorbidity (more than 3 chronic conditions) was similar among most groups (within approximately 

5-7 percentage points of one another), with Asians reporting the lowest prevalence.

Adjustment did not greatly alter these patterns (Table 3). Blacks, American Indians/Alaskan Natives, and multiracial individuals had significantly higher odds of ADL impairment than whites (from OR 1.11 [95% CI, 1.07-1.15] for blacks to OR 1.37 [95% CI, 1.20-1.57] for American Indians/Alaskan Natives and multiracial individuals), while Asians had significantly lower odds than whites (OR 0.63; 95% CI, 0.59-0.67). Most racial/ethnic groups did not report significantly different odds for multimorbidity than whites, while Asians and Hispanics retained significantly lower odds than whites (OR 0.66 [95% CI, 0.61-0.69] and OR 0.90 [95%CI, 0.86-0.94], respectively).

Findings from this nationally representative survey of Medicare MC members 65 years and older provide new evidence of significant racial/ethnic disparities in health status indicators. To our knowledge, this is one of the first national assessments of health and functional status among racially and ethnically diverse Medicare beneficiaries. The results provide unique national estimates of prevalence for many of these health indicators among older adults and add evidence to the literature on health status—based disparities among older adults. Even after controlling for key sociodemographic factors, race/ethnicity was significantly associated with most adverse outcomes on health indicators. Except for Asians, all racial/ethnic minority groups were significantly more likely than whites to report poor mental health status, presence of most health symptoms, sensory limitations, and ADL disability.

Unlike much earlier research, this study assesses health indicators that are critical end points of care but not routinely assessed (eg, functional health status), and not just receipt of health services. These indicators reflect quality-of-life issues that are especially important for older adults. Moreover, while the findings are consistent with prior Medicare disparities studies, many prior studies were typically limited to blacks, whites, or Hispanics, while this study examined smaller groups (eg, Native Hawaiians, American Indians, Asians, and multiracial individuals—a growing and heterogeneous group) typically excluded from national comparisons because of small numbers or unreliable race/ethnicity data.

 This is critical, given the growing diversity of the aging, with proportions of elderly from many racial/ethnic subgroups expected to grow faster than those of blacks and whites.

By providing a snapshot of disparities and health quality across diverse indicators, these data are useful in highlighting areas of greatest need for improving quality and equitable care in older US adults, particularly mental health status and health symptoms, where differences between minorities and whites were the most pronounced. These symptoms typically result from common chronic conditions such as diabetes, heart failure, and arthritis that are often the target of quality measurement and improvement efforts. Older adults from minority racial/ethnic groups may have unique needs, including diverse health beliefs or attitudes,

 and may report or perceive health status differently.

Teasing out these dynamics and changing business practices to include culturally appropriate care could potentially improve care delivery and close gaps in quality of care.

 Examples of culturally appropriate care include improved language helpline support, availability of educational materials in multiple languages, and availability of non-English pharmacy labeling.

Promising analytic techniques that pool information across time may also increase the power to compare smaller racial/ethnic groups.

This study examines disparities in important patient-reported functional outcomes not routinely assessed among diverse racial/ethnic groups in Medicare managed care.

Sarah Hudson Scholle, DrPH

Articles

Patient-Centered Oncology Care: Impact on Utilization, Patient Experiences, and Quality

Disparities in Diabetes and Hypertension Care for Individuals With Serious Mental Illness

Beyond Black and White: Race/Ethnicity and Health Status Among Older Adults

Health Plan Resource Use Bringing Us Closer to Value-Based Decisions

Comparison of Administrative-only Versus Administrative Plus Chart Review Data for Reporting HEDIS Hybrid Measures

Practice Systems for Chronic Care: Frequency and Dependence on an Electronic Medical Record