Landon BE, Schneider EC, Normand ST, Scholle SH, Pawlson LG, Epstein AM. Quality of Care in Medicaid Managed Care and Commercial Health Plans. JAMA. 2007;298(14):1674-1681. doi:10.1001/jama.298.14.1674
Author Affiliations: Department of Health Care Policy, Harvard Medical School (Drs Landon and Normand); Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center (Dr Landon); Department of Health Policy and Management (Drs Schneider and Epstein), Department of Biostatistics, Harvard School of Public Health (Dr Normand); and the Section of Health Services and Policy Research, the Division of General Medicine, Brigham and Women's Hospital (Drs Schneider and Epstein), Boston, Massachusetts; National Committee for Quality Assurance, Washington, DC (Drs Scholle and Pawlson).
Context In contrast to the commercially insured population, the proportion of Medicaid beneficiaries enrolling in health maintenance organizations continues to increase.
Objective To compare quality of care within and between the Medicaid and commercial populations in 3 types of managed care plans: Medicaid-only plans (serving predominantly Medicaid enrollees), commercial-only plans (serving predominantly commercial enrollees), and Medicaid/commercial plans (serving substantial numbers of both types of enrollees).
Design, Setting, and Participants All 383 health plans that reported quality-of-care data to the National Committee for Quality Assurance for 2002 and 2003, including 204 commercial-only plans, 142 Medicaid/commercial plans (plans reported data for the Medicaid and commercial populations separately); and 37 Medicaid-only plans.
Main Outcome Measures Eleven quality indicators from the Healthcare Effectiveness Data and Information Set (HEDIS) applicable to the Medicaid population.
Results Among Medicaid enrollees, performance on the 11 measures observed in this study were comparable for Medicaid-only plans and Medicaid/commercial plans. Similarly, among commercial enrollees, there was virtually no difference in performance between health plans that served only the commercial population and those that also served the Medicaid population. Overall across all health plan types, the performance for the commercial population exceeded the performance for the Medicaid population on all measures except 1, ranging from a difference of 4.9% for controlling hypertension (58.4% for commercial vs 53.5% for Medicaid; P = .002) to 24.5% for rates of appropriate postpartum care (77.2% for commercial vs 52.7% for Medicaid; P = .001). Differences of similar magnitude were observed for commercial and Medicaid populations treated within the same health plan.
Conclusions Medicaid managed care enrollees receive lower-quality care than that received by commercial managed care enrollees. There were no differences in quality of care for the Medicaid population between Medicaid-only plans and commercial plans that also served the Medicaid population.
Although enrollment of commercially insured individuals in health maintenance organizations (HMOs) has decreased in recent years, HMOs continue to provide care for an increasing proportion of the Medicaid population.1 Between 1994 and 2004, enrollment in Medicaid managed care tripled from 7.9 million beneficiaries to more than 27 million beneficiaries. The proportion of Medicaid beneficiaries in managed care increased from 23% to more than 60% during the same time period.2
The transition to managed care, within both the Medicaid and commercial populations, has been driven in part by the potential cost savings believed to be obtainable by integrating care. The impact of managed care on quality of care for the Medicaid population has been controversial. HMOs may incorporate prevention and routine care to prevent serious and costly downstream complications and use population-management techniques to improve the delivery of services to their enrollees.3- 5 These techniques may be especially helpful to Medicaid recipients. State Medicaid programs also have adopted a variety of value-based purchasing techniques that require Medicaid managed care plans to measure and report on performance on core quality indicators and to undertake efforts to improve performance.6- 8 Nevertheless, health plans may institute programs or procedures that limit access to necessary medical services. The poorly educated, low-income, and immigrant populations often served by Medicaid health plans likely have less ability to negotiate the sometimes complex requirements of managed care systems.6,7,9
Within managed care, the type of health plan may make an important difference in quality of care. By specializing in the care of the Medicaid population, Medicaid-only plans may be able to provide superior care compared with commercial plans that also cover Medicaid populations but do not solely tailor services to this population. At the same time, concerns have been raised about the quality of care delivered in Medicaid-only plans.10 Many Medicaid health plans that focus predominantly on the Medicaid population are regional plans, and there is concern that these organizations may be undercapitalized, inexperienced in quality management, or rely on networks of lower-quality providers.
Despite intense interest from state and national policy makers and advocates, there is almost no information in the peer-reviewed literature on the quality of care delivered within health plans to Medicaid enrollees, how that quality compares with that received by their commercial enrollee counterparts, and how type of plan (eg, Medicaid only vs Medicaid/commercial) makes a difference.11 In this study, we examined performance on Healthcare Effectiveness Data and Information Set (formerly the Healthplan Employer Data and Information Set) (HEDIS) quality indicators in 3 types of managed care plans: Medicaid-only plans, commercial-only plans (health plans serving predominantly commercial enrollees), and Medicaid/commercial plans (health plans serving substantial numbers of both types of enrollees). We compared quality performance for the Medicaid and commercial populations by comparing between plans that focused exclusively on one or the other population with plans that provided care for both populations (eg, performance for Medicaid patients in Medicaid-only plans vs performance for Medicaid patients in plans that serve both commercial and Medicaid populations).
We obtained HEDIS data reported to the National Committee for Quality Assurance (NCQA) for all participating commercial and Medicaid health plans for the reporting years 2002 and 2003. The performance data reflect clinical care by each health plan for the 12-month period preceding the year of submission. The NCQA publishes detailed specifications for each measure and defines acceptable data sources, sampling methodologies, and data collection procedures.12 Acceptable data are obtained from administrative claims, medical record reviews of a random sample of eligible patients, or a combination of both. Health plans with both a commercial product and a Medicaid product submit HEDIS data separately for Medicaid and commercial populations. The data for this analysis include both health plans that allow their data to be released publicly through the NCQA Quality Compass database and those that submit their data confidentially but do not allow public release. Omission of the nonpublicly reported data could lead to biased estimates of quality.13,14
We selected measures of health plan quality that we thought would be most applicable to the care of enrollees in Medicaid programs (Table 1). These included measures relevant to children, women of childbearing age, and patients with chronic medical conditions: the principal targeted populations of the Medicaid program. For most of the outcome measures, less than 10% of the health plans were missing data, although rates of missing data were higher for measures that require chart review for larger proportions of those measured, such as controlling high blood pressure and asthma.
Using data maintained by the Centers for Medicare & Medicaid Services, we categorized health plans as Medicaid only if they predominantly served the Medicaid population (ie, > 75% of enrollment drawn from Medicaid); Medicaid/commercial if they served both the Medicaid and commercial populations and had less than 75% of enrollment drawn from Medicaid; and commercial only if they did not serve the Medicaid population. We obtained additional health plan characteristics from the Interstudy Competitive Edge 8.2, 2002 release (HealthLeaders InterStudy, Nashville, Tennessee) that reflected data from 2001 and contained annually updated information on HMOs operating in the United States.15 In some cases in which health plans had merged or reorganized recently preceding this study, we substituted health plan data from the 2004 release (reflecting data from 2003) to obtain a description of the appropriate organization. We considered variables in the Interstudy database that we hypothesized might be related to health plan performance. These included the number of years the plan had been in operation; model type: group/staff (the health plan employs the physicians or contracts with a single group), independent practice associations (the health plan contracts with independent practice associations that in turn represent individual physicians), mixed (usually a group/staff model health plan that has contracted with additional independent practice associations), or network (a health plan that contracts with multiple groups); profit status; national affiliation (defined by Interstudy as health plans operating in 2 or more states and with more than 10 000 total enrollees); health plan size; and census region of the country. Health plan affiliations were classified as national managed care firms, local independent plans, or local Blue Cross/Blue Shield affiliates. Each health plan was assigned to the US Census Bureau division in which it had the largest enrollment. We also used information on the NCQA accreditation status of health plans available in the Interstudy database. Health plans were categorized as excellent, commendable, or unaccredited.
After evaluating the distribution of total enrollment, we defined small plans as those with fewer than 15 000 total members (in all products) and large plans as those with greater than 75 000 total members based on natural cut points in the data.
We used a multistep algorithm to match plans in the Interstudy database with those in the NCQA database that began with a previously established linking database.16- 19 For unlinked plans, we either called the health plans directly to confirm alternative names or confirmed a match using information from Centers for Medicare & Medicaid Services or other sources. Because plan and market definitions are not identical between NCQA and Interstudy, multiple associations sometimes occurred. For instance, some Interstudy plans were linked with several submissions to NCQA. In those cases, we assigned the plan level characteristics to each NCQA submission. To resolve these discrepancies, our priority was to maintain the lowest level of aggregation possible for the performance data from NCQA, meaning that we chose to keep each submission to NCQA independent. In the few cases (approximately 30 health plans) in which a single entry in Interstudy was linked with 2 or more entries in the NCQA data, we were able to find an improved match by examining the 2004 Interstudy database, which usually had the same reporting unit as NCQA.
In some cases (n = 39), plans could not be linked to the Interstudy database. In most cases, these were smaller Medicaid plans that lacked commercial enrollment or local units of larger health plans for which no Interstudy data were available. For these health plans, all key variables of interest (eg, tax status, region, NCQA accreditation, enrollment) were determined through review of Internet information and telephone calls. We were not able to determine plan age using these methods.
We conducted 4 separate comparisons of plan performance: (1) for the Medicaid population, performance of Medicaid-only plans was compared with the performance of Medicaid/commercial plans; (2) for the commercial population, the performance of commercial-only plans was compared with the performance of Medicaid/commercial plans; (3) all plans serving the Medicaid population and reporting Medicaid performance were compared with all plans serving commercial populations and reporting commercial performance; and (4) within Medicaid/commercial plans, performance for the Medicaid population was compared with the performance for the commercial population. For these analyses, we report results for data submitted for reporting year 2003. We also examined the results for 2002 and the results were essentially identical.
We first tested unadjusted differences using t tests. We then estimated random-effect regressions models to examine the previous comparisons while controlling for clustering within plans and potential confounders including tax status, region, model type (staff/group, independent practice associations, etc), accreditation status, affiliation, and total plan enrollment. The outcome variable in each case was the number of eligible members in the plan having the particular indicator met. We assumed the probability of meeting an indicator was binomially distributed and that the log-odds of this probability varied randomly by plan and systematically by the plan characteristics. The plan-specific random effects were assumed to arise from a normal distribution. This strategy permitted us to allow some plans to have higher adjusted rates and some to have lower adjusted rates, as well as accounting for within-plan correlation. All confounders were included in the final regression models, although in some cases selected confounders were omitted because of lack of variability that prevented the models from converging. Finally, for chlamydia screening, we report only unadjusted rates because adjusted models did not converge.
Tests of association between plan type and the log-odds of having an indicator met were conducted by examining the significance of the coefficient of the plan-type indicator in the regression model. All analyses were performed using SAS software version 9.1 (SAS Institute Inc, Cary, North Carolina); the SAS procedure NLMIXED was used to estimate the random-effect regression models. We adjusted for multiple comparisons within a plan-type comparison using a Bonferroni adjustment so that within each of our 4 comparisons, the overall type I error rate was .05.
Among the 383 health plans included in the study, 326 health plans contributed commercial data and 137 health plans contributed Medicaid data (Table 2). Slightly more than 80% of health plans submitted data for both years. The commercial-only population was served by 204 health plans, 37 served only the Medicaid population, and 142 served both commercial and Medicaid populations.
Most health plans serving only the commercial market were for-profit (77.9%), whereas 51.4% of the health plans serving only the Medicaid market were for-profit (P < .001). The health plans were well distributed around the country. Few health plans (2.4% overall) were group or staff model plans. A higher proportion of Medicaid and Medicaid/commercial plans (62.2% and 52.8%, respectively) were independent local health plans, as compared with 27.5% of commercial-only health plans (both P < .001). Most plans serving the commercial population (including those also serving the Medicaid population) had been in operation for at least 10 years. Most Medicaid-only plans (52% of those with data), however, were less than 5 years old (P < .001 for both comparisons).
Unadjusted comparative health plan and population performance for 2003 is presented in Table 3. The results for 2002 were substantively the same. After restricting our analyses to the Medicaid population, the performance of Medicaid-only plans and of Medicaid/commercial plans for their Medicaid population was comparable across the 11 measures, although for cervical cancer screening, the performance of Medicaid/commercial plans was higher (64% vs 56%; P value statistically nonsignificant). Similarly, after restricting our analyses to the commercial population, there was no consistent difference in performance between health plans that only served the commercial population and those that also served both commercial and Medicaid populations (Table 3).
When both Medicaid and commercial populations were considered together, overall commercial population performance exceeded overall Medicaid population performance in all instances except 1 (Table 3), ranging from a difference of 4.9% for controlling hypertension (58.4% for commercial vs 53.5% for Medicaid, P = .002) to 24.5% for rates of appropriate postpartum care (77.2% for commercial vs 52.7% for Medicaid, P < .001). The 1 exception was chlamydia screening, for which Medicaid performance exceeded commercial performance (41.8% vs 25.3%, P < .001). When comparing performance for Medicaid and commercial populations within the same health plan, performance for the commercial population was uniformly better across all measures (Table 3). In multivariate models that controlled for health plan characteristics and clustering within health plans (Table 4), the findings were consistent with those in Table 3.
In this study, we compared the quality of care delivered by health plans serving the Medicaid population with those serving the commercial population. We found little difference in the quality of care provided to the Medicaid population served by Medicaid-only plans compared with the quality of care provided to the Medicaid population served by commercial plans that also served Medicaid enrollees. Similarly, there was very little difference in quality provided to commercial populations served by commercial-only plans compared with that provided to commercial populations served by Medicaid/commercial plans. In contrast, compared with the commercial population, the quality of care was substantially lower for the Medicaid population, regardless of plan type (with the exception of chlamydia screening). These findings suggest that the type of health plan enrolling the population (commercial, Medicaid/commercial, or Medicaid-only) is a less important determinant of the quality of care than differences in the characteristics of the population being served, the local provider networks in which they receive care, access to care, patterns of care seeking, and adherence to treatment recommendations.
The differences we observed between the quality performance for commercial and Medicaid enrollees are both statistically and clinically significant. For instance, cervical cancer screening rates and rates of diabetes control were more than 15 percentage points higher for commercial populations Analyses completed by NCQA suggest that differences in hemoglobin A1C control of an even smaller magnitude within commercial health plans could result in up to 15 000 fewer deaths per year nationally.20 Thus there is a clear opportunity for improving the health care and health of individuals cared for in Medicaid managed care.
Interestingly, chlamydia screening is the only measure showing higher performance in the Medicaid population vs the commercial population. Chlamydia is a sexually transmitted disease that is generally asymptomatic in women but can lead to serious sequelae.21,22 Factors affecting performance on this measure include physicians' perceptions about the prevalence of infection in their patient populations and outreach efforts to teenaged and young adults who may be covered under the insurance policies of their parents. Higher performance in Medicaid programs may be a result of Medicaid recipients being more likely to be treated in clinics where sexually transmitted disease screening is routinely implemented or where clinicians have more accurate understanding of the prevalence in that particular population.
These differences in quality performance also underscore the challenge of delivering high-quality care to the Medicaid population.23 Patients enrolled in Medicaid are socioeconomically disadvantaged and may face additional competing needs that make adhering to treatment recommendations difficult. Our findings suggest that mainstreaming Medicaid beneficiaries by enrolling them in health plans that also offer commercial insurance products does not appear better or worse than enrolling them in Medicaid-only health plans. While we found that performance in Medicaid/commercial plans was marginally higher than that for Medicaid-only plans on 10 of the 11 measures, these results were not statistically significant. In our previous research we found that health plans that focus predominantly on the Medicaid population have additional outreach services aimed at the special needs of the Medicaid population.9 Our data here suggest that these additional services might not be sufficient to bring the quality of care for the Medicaid population up to the level received by commercial populations for routine quality indicators such as those in HEDIS, although we were not able to assess the specific outreach programs of the plans in our study.
While HEDIS performance is similar across different types of health plans, the patterns we observed might still be explained in some part by the settings in which these patients obtain care. Disadvantaged populations covered through the Medicaid program might receive care from physicians or hospitals of generally lower quality. Presumably, these physicians would be included in both Medicaid-only plans and those that serve the Medicaid and commercial population, since performance is similar across those types of plans. Thus, even to the extent that the delivery systems overlap, enrollees may still see different clinicians based on segregation of residence by socioeconomic factors.24 In addition, many physicians refuse to participate in health plans that serve the Medicaid population so the delivery networks might not always overlap.
Prior studies of this important issue and related investigations are limited. One prior peer reviewed study compared HEDIS performance in Medicaid vs commercial HMOs for the pediatric population in 1999. Thompson and colleagues found higher performance for commercial health plans much as we did,11 and also that these differences persisted in health plans that served both the Medicaid and commercial populations. The NCQA has also released data comparing HEDIS performance for Medicaid and commercial populations, again showing lower performance for Medicaid enrollees.20 The NCQA data do not control for plan characteristics and location in their analysis. Neither source analyzed HEDIS performance in relation to the type of HMO (eg, Medicaid-only vs Medicaid and commercial). In the mid 1990s we (B.E.L. and A.M.E.) surveyed health plans serving the Medicaid market and found that Medicaid-only plans and plans that served both the Medicaid and commercial populations were similar in most aspects of quality management such as collection of performance data and initiation of quality improvement projects; but Medicaid plans were more likely to have programs to address the particular needs of the Medicaid population such as inadequate transportation and illiteracy.9 We did not collect data on HEDIS performance. In related work, Zaslavsky and colleagues examined patient-level commercial health plan HEDIS performance for individuals of differing socioeconomic status and found that scores tended to be lower for populations that were more disadvantaged.25,26
Our study is subject to several limitations. First, sociodemographic characteristics of enrolled populations are known to be associated with the quality of care and therefore could be an important confounder since commercial and Medicaid health plans enroll populations that differ on these characteristics.26,27 At the least, our results show that these alternate types of managed care plans do not make a difference in the quality of care for Medicaid enrollees, who seem to be at risk for receiving lower-quality care than the commercial populations.
Second, not all health plans in the country reported their data to NCQA, although the sample of plans in this study includes the vast majority of health plan enrollees nationally.
Third, not all health plans could be linked immediately to the Interstudy data. We therefore took additional steps including Internet searches and telephone calls to assure that we linked all of the health plans included in the study. However, we were not able to obtain all relevant variables (eg, plan age) for each plan.
Fourth, although we included quality indicators from several domains of care, our data include a limited number of process measures of quality and may not represent many of the dimensions of quality of care provided by health plans. We previously showed that Medicaid health plans have additional services aimed at certain specific needs of the Medicaid population. Therefore, Medicaid-only plans might have advantages in Medicaid-specific services (eg, use of translators) that were not covered in this article.
Finally, while these analyses document performance problems in Medicaid managed care, no similar information is available on care provided through traditional fee-for-service Medicaid. For example, it is not known whether enrollees served by Medicaid health plans are receiving better or worse care than Medicaid beneficiaries not enrolled in managed care.
In summary, our findings suggest that the Medicaid population receives lower-quality care than that received by commercial managed care enrollees and that this is true in both Medicaid-only plans and in commercial plans that also serve Medicaid beneficiaries. Neither mainstreaming of Medicaid beneficiaries in commercial plans nor relying on Medicaid-only plans seems to raise the quality of care to the level experienced by commercial populations. If reducing disparities in care nationally is an important goal of the US health care system, managed care is not a panacea. Additional resources will need to be devoted to designing and implementing specific interventions to improve the quality of care for Medicaid beneficiaries enrolled in managed care.
Corresponding Author: Bruce E. Landon, MD, MBA, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (Landon@hcp.med.harvard.edu).
Author Contributions: Dr Landon had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Landon, Schneider, Normand, Scholle, Epstein.
Acquisition of data: Landon, Scholle, Pawlson, Epstein.
Analysis and interpretation of data: Landon, Schneider, Normand, Scholle, Epstein.
Drafting of the manuscript: Landon, Schneider, Normand, Scholle.
Critical revision of the manuscript for important intellectual content: Schneider, Normand, Pawlson, Epstein.
Statistical analysis: Normand.
Obtained funding: Landon, Schneider, Normand, Epstein.
Administrative, technical, or material support: Normand, Epstein.
Study supervision: Epstein.
Financial Disclosures: None reported.
Funding/Support: This work was supported by grant 043782 from the Robert Wood Johnson Foundation (Dr Epstein).
Role of the Sponsor: The Robert Wood Johnson Foundation had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Disclaimer: Drs Pawlson and Scholle are employees of the National Committee for Quality Assurance.
Additional Contributions: We are indebted to Charlene Hsuan, BS, Department of Health Policy and Management, Harvard School of Public Health, for research assistance; Lin Ding, PhD, for assistance with expert statistical programming, and Eva Tomczyk, BA, Department of Health Care Policy, Harvard Medical School, for assistance with manuscript preparation. Ms Hsuan is now enrolled in the New York University School of Law and Ms Tomczyk is now with the Department of Social Medicine. The individuals named in this acknowledgment were paid employees of their respective institutions at the time of this study.