Ronald J. Ozminkowski, Ph.D., Mark W. Smith, Ph.D., Rosanna M. Coffey, Ph.D., Tami L. Mark, Ph.D., Cheryl A. Neslusan, Ph.D., and John Drabek, Ph.D.
The MEDSTAT Group
Most people receive their health insurance through employer-sponsored plans. Yet little is known about the prevalence of chronic conditions and disabilities among plan enrollees. Similarly, employer-sponsored plans make extensive use of managed care, but little is known about the impact of managed care on those with severe chronic illness. This study was funded by the Office of Disability, Aging, and Long-Term Care Policy of the U.S. Department of Health and Human Services to better understand the role of private insurance plans in caring for people with chronic conditions and disabilities. It estimates the prevalence of chronic illness, analyzes the factors affecting the choice of indemnity plans versus managed care, and estimates the impact of managed care on service use and expenditures. The study also investigates the leading risk adjustment systems as a possible method for paying plans more appropriately in serving this population.
A. Background--Chronic Illness and Managed Care
Chronic illness and functional disability occur surprisingly frequently among the population of the United States. Roughly 14 percent of adults between the ages of 18 and 65 experience a disability that limits their functional activity level (Adler, 1995). In addition, as many as 31 percent of children have special health care needs due to chronic illness or functional limitations (Harris-Wehling and Ireys, 1995).
Chronically ill and disabled individuals often require a broad range of health and social services to maximize functional abilities and improve health status. Managed care has been touted as having great potential for those with disabling chronic illness because a single case manager can take responsibility for guiding patients through the maze of services and providers that may be necessary to treat chronic conditions. However, managed care also has been equated with “managed cost,” implying that more consideration is given to reducing costs than to patient health and functioning. Which view is more accurate?
Nearly 50 percent of disabled people in the United States are covered by managed care plans (Fama, Fox and White, 1994), yet little research has been done to determine whether managed care is more or less beneficial to them than traditional indemnity health insurance. This is especially true for those covered outside of Medicare or Medicaid by private sector, employer-based plans.
B. Study Questions
The Private Payers Study represents a major step in the development of knowledge concerning chronically ill individuals covered by private health plans--a knowledge base that will aid both government policymakers and private firms interested in reducing costs, expanding health care choices, and assuring appropriate care.
The study attempts to answer the following questions:
What is the prevalence of chronic illness and disability among the population enrolled in employer-based health insurance plans?
Are individuals with chronic illness more or less likely to choose managed care or indemnity plans, if given a choice?
How does the type of health plan selected by chronically ill and disabled enrollees affect their service use and cost, after accounting for differences in enrollee characteristics?
To what extent can risk-adjustment systems help employers and health plans predict expenditures of their chronically ill and disabled enrollees?
Further discussion of these questions is preceded by a description of the employers we studied and their health plans, how we identified individuals with potentially disabling chronic illness, and the statistical methods we used.
1. Selection of Employers and Health Plans
The Private Payers Study was based on administrative claims for reimbursement of health care services, submitted under private-sector insurance plans by two large firms. To preserve their confidentiality we will refer to them as Employer A and Employer B. Table ES-1 compares their health insurance arrangements.
Employer A is a large firm with offices in more than 30 cities across the country. In 1995, its health plans covered over 400,000 people. Employer A offered two health plans at each location. One was an indemnity plan. Indemnity plans, sometimes called fee-for-service (FFS) plans, do not have a predetermined network of providers; enrollees receive similar reimbursement for care by any provider. The other offering was a particular type of managed care plan called a point-of-service (POS) plan. It arranged a network of providers but also covered services outside the network, while penalizing individuals financially for seeking care outside the network. The POS assigned each member a primary care gatekeeper and required the gatekeeper’s approval for specialist visits.
Employer B is a large state government. In 1995, its health plans covered over 200,000 people. Employer B offered seven health maintenance organizations (HMOs), one indemnity plan, and one preferred provider organization (PPO). The HMOs did not reimburse expenses for health care by providers outside of their networks. The PPO, like the POS plan of Employer A, reimbursed enrollees for care outside its network, but at a higher coinsurance rate to employees. Unlike the POS, however, the PPO did not assign a primary gatekeeper.
|TABLE ES-1. Characteristics of Study Population, 1995|
|Sponsor|| Number of
| Network of
|Employer A||>400,000||Indemnity (1)||No||Yes|
|Point of Service (1)||Yes||Yes|
|Employer B||>200,000||Indemnity (1)||No||Yes|
|Health Maintenance Organization (7)||Yes||No|
|Preferred Provider Organization (1)||Yes||Yes, with high copayments|
2. Study Population
To study the experiences of chronically ill and disabled people in private health plans, we had to find which employees (and dependents) had potentially disabling chronic conditions. In the Private Payers Study we did this by classifying diagnoses from claims and encounter records in two ways. One was based on clinical judgment. The other relied on results of a study in the literature that used responses to a major national survey to identify disabling conditions. The two methods, described briefly below, are described in more detail in Chapter 4.
In an earlier report, researchers at The MEDSTAT Group (Crown et al., 1998b) developed a set of criteria for identifying potentially disabling chronic illness on the basis of diagnosis codes and other information available in medical claims databases. Potential disability was defined as any mental or physical problem that typically results in loss of normal functioning. Many diagnosis codes for physical and mental conditions are indicative of a potentially disabling chronic condition by themselves, while others are indicative of such impaired health status only at later stages of disease. Also, although there are broad areas of overlap between the physical and mental criteria for children and adults, some conditions are specific to each age group.
Potentially disabling chronic conditions were identified through several steps. First, a clinical coding specialist selected conditions thought to potentially result in partial or total disability. Any conditions known to be invariably terminal were excluded from consideration. The preliminary list of conditions was forwarded to a consulting physician for judgment regarding the appropriateness of inclusion. This list was then reviewed by staff at ASPE and further revised by the coding specialist and physician. The result of this process was a detailed list of over 300 ICD-9-CM diagnosis codes for adults and over 300 for children that were applied to the medical claims data.
There is an alternative definition of disabling chronic illness that focuses on activity limitations. Developed by LaPlante (1989) using the 1983-1986 waves of the National Health Interview Survey, it includes 37 conditions (20 for adults and 17 for children) that were found to be highly correlated with limitations in activities of daily living. As Appendix A shows, some but not all of the activity-limiting conditions are also per se disabling conditions as defined by Crown et al. (1998). In some of the analyses conducted during this project comparisons were made between the two definitions of disability.
D. Statistical Methods
Several different analyses were performed during this project, which focused primarily on differences between those choosing indemnity or managed care options. Comparisons of several measures of personal characteristics, service use, and expenditures were made for key non-elderly sub-populations--active employees, dependent children and spouses, and early retirees. Raw differences were tabulated for several measures, and standard tests of statistical significance were applied. Such comparisons get at the gross differences between groups.
Simple comparisons, however, cannot disentangle differences in the characteristics of enrollees among plans from differences in the effect of plans on utilization. To identify the effect of managed care on health care utilization and expenditures, differences in enrollee characteristics across plans need to be accounted for. Higher utilization and expenditures in one insurance type relative to another may reflect underlying differences in enrollees as well as differences in the plan themselves. For example, if older individuals tend to choose indemnity plans over managed care plans and are also less healthy, then a finding of higher utilization and payments in indemnity insurance may be solely attributable to the age variation among the plan types, not to the plan itself.
We then estimated the effect of plan type on utilization and expenditures controlling for two categories of confounding influences: patient characteristics available in our data and unmeasured factors systematically related to insurance choice. This second category is important to consider since a patient’s true health is not completely observable to insurers. “Adverse selection” occurs if people whose poor health is unknown to insurers choose more generous plans. In this case, premiums will not accurately reflect costs in the population. In response, insurance plans will have an incentive to raise premiums, which may price some individuals out of the market.
To some extent the greater burden of people with chronic illness on plans could be alleviated if payments to plans were adjusted for the population they serve. Some employers and insurers differentiate on the basis of certain characteristics, for instance having separate policies for families and individuals or for active employees and retirees. A number of risk-adjustment models have been proposed to pay plans more appropriately. In this study, four systems--Hierarchical Coexisting Conditions (HCCs) with employer-specific adjustment factors, HCCs with pre-determined adjustment factors, Adjusted Clinical Groups (ACGs), and Adjusted Diagnostic Groups (ADGs)-- were applied to the employer data. The implications for total health expenditures, and for mental health expenditures alone, were estimated.
1. Study Findings
This project used evidence from medical claims databases of two employers to answer the four questions raised above about how health plans managed the care of chronically ill enrollees. The main findings are noted briefly below. More in-depth findings and discussions are contained in the full report.
Question 1: What is the prevalence of chronic illness and disability among the population enrolled in employer-based health insurance plans?
Private health plans cover a significant number of people with chronic disabilities.
For the two employers combined, 19 percent of the people (employees and their dependents) had disabling chronic conditions. However, the insured populations of the two firms had slightly different rates of potentially disabling chronic conditions: 17 percent of the covered lives for Employer A, 21 percent for Employer B.
It is not clear why this difference between the two employed populations exists. Differences in the nature of the businesses, differences in employee benefits (especially disability retirement), and the competing health care benefits provided by other employers in the area surrounding Employer B may have meant that Employer B attracted more employees (or their dependents) with chronic illness.
Question 2: Are individuals with chronic illness more or less likely to choose managed care or indemnity plans, if given a choice?
People with potentially disabling chronic conditions are more likely to choose indemnity plans, but a substantial proportion choose managed care.
About 58% of service users with chronic illness chose Employer A’s indemnity plan, while 42% chose the POS plan. The indemnity plan had a higher proportion of chronically ill service users in every category: active employees, early retirees, children, and other dependents. The indemnity plans also had a higher proportion of people with activity-limiting conditions. Employer B had the same experience. Of the service users with potentially disabling chronic illness in Employer B’s plans in 1995, 40 percent chose the indemnity plan, versus 33 percent the PPO, and 27 percent the HMOs.
People with potentially disabling chronic conditions are not homogeneous with respect to the type of insurance they choose.
For each employer, we estimated the probability of enrollment into managed care versus indemnity plans, controlling for differences in patients’ demographic characteristics and health status. For those having at least one paid claim in 1995 for one of the potentially disabling chronic mental and physical conditions, we found that being male and younger increased the likelihood of choosing managed care as opposed to indemnity coverage in both firms. Early retirees having both a mental and physical condition, as opposed to just one or the other, were less likely to choose managed care. Living in a metropolitan statistical area (MSA) increased the probability of enrollment in the POS option in Employer A, while using more outpatient services in the past or having a child with a potentially disabling chronic condition lowered the probability of HMO enrollment.
Question 3: What is the impact of health-plan type on utilization and expenditures, taking into account differences in the populations that enroll in different types of insurance plan?
Utilization and expenditures were slightly greater in the indemnity plan than in the managed care plan, but these differences are strongly influenced by casemix differences among plans.
For Employer A, we found that utilization and expenditures were generally lower, on average, in the POS plan. After adjusting for population characteristics, however, fewer differences emerge, and those that remain are generally smaller. For example, on average, POS enrollees with potentially disabling chronic conditions had 15 prescriptions filled in 1995, whereas indemnity plan enrollees filled 20. After adjusting for population characteristics, this difference of five prescriptions was reduced to 3.6. For Employer B we also found that simple descriptive comparisons resulted in mean utilization and expenditures in 1995 that were generally lower in the HMO plan than in the indemnity plan. In contrast to Employer A, after adjusting for population characteristics, more differences remain (hospital admissions, outpatient visits, outpatient expenditures and total expenditures). As with Employer A, however, those that do remain are generally smaller in magnitude.
For both employers, those persons with chronic conditions who also had activity- limiting conditions used more services and had higher expenditures than those that did not.
Greater differences in use and expenditures between managed care and indemnity plans were observed for Employer B than for Employer A.
Among adults there was no significant difference in inpatient expenditures across plan types for Employer A once population differences were taken into account--slightly greater hospital use in the managed care plan was offset by lower cost per user. For Employer B, however, both hospital use and expenditures per user were significantly lower for adults in the HMO plan than for those in the indemnity. A similar pattern emerged for outpatient expenditures, but the differences for Employer B were not quite large enough to be statistically significant.
There is some evidence consistent with adverse selection among members of Employee A’s health plans, but not among those in Employer B’s plans.
For Employer A, we found some evidence consistent with adverse selection-those who were less-healthy tended to join the indemnity plan and have higher levels of use and payments, based on factors that were unobservable in the data. This situation may result in premium increases or service cutbacks in this plan type through time. In Employer B’s plans, contrary to expectations, we found some evidence that those having higher use and expenditures were also more likely to have joined the HMO plan. This is probably due to the relatively high price of the indemnity plan. Some people with potentially disabling chronic conditions (in particular, those who were inherently higher users of health care and those that were more expensive) may have viewed the indemnity plan as too expensive compared to the HMO options. If those people had joined the indemnity plan, then the utilization and expenditure differences between indemnity and managed care would have increased.
These results suggest that the apparent relative efficiency and cost savings of managed care versus indemnity may be significantly affected by underlying casemix differences. Furthermore, such differences may be difficult to predict without detailed data on plan enrollees, especially since casemix may depend critically on the relative prices of the plans. Employers and policymakers must closely investigate the relationship between health-plan type and cost savings in competing plans.
Question 4: To what extent can risk-adjustment systems help employers and health plans predict expenditures of their chronically ill and disabled enrollees?
To answer this question, we must explain the concept of risk adjustment and the systems we tested. Risk adjustment is a tool to achieve more precise methods of payment to health plans than has been traditionally used. It attempts to account for the higher-than-average cost of treating people who are expected to be high service users. By improving the match between payments and actual expenditures, a risk-adjustment system reduces the incentive of insurers to avoid potentially expensive users, including those with chronic illness, by offering less generous benefits. Employers can use risk adjustment to set capitated rates for plans, or if they are self-insured, to assist in judging plans’ efficiency.
Age, sex, and region are the categories used most often by insurers to set premiums for employees and employers. We investigated the ability of leading risk- adjustment systems to predict the expenditures of those having potentially disabling chronic conditions relative to what insurers normally use. The systems we studied were Hierarchical Coexisting Conditions (HCCs) with employer-specific adjustment factors, HCCs with pre-determined adjustment factors, Adjusted Clinical Groups (ACGs), and Adjusted Diagnostic Groups (ADGs).
To predict total health care expenditures, we used these systems to study 10 different groups of potentially disabling chronic conditions: arthritis, asthma, cancer, chronic obstructive pulmonary disease, diabetes, heart failure, psychiatric disorders, seizure disorders, stroke, and ulcerative colitis. Full details of the study and results are described in Chapter 8.
Risk adjustment provides a substantial improvement over current payment methods for this population.
Risk-adjusted projected payments were much closer to actual expenditures than were payments adjusted for demographics alone. Both of the risk-adjustment systems studied performed substantially better than simple adjustment based on age, sex, and area hospital wage index. Risk adjusted models based on age, sex, and regional prices under-predicted expenditures for the 10 chronic conditions by more than 50 percent, resulting in substantial financial losses to insurers.
Of the risk adjustment systems we studied, the one performing the best was the one based on the Hierarchical Coexisting Conditions (HCCs) approach using employer-specific adjustments.
The HCCs model reduced the prediction error to less than 15 percent for all chronic conditions except stroke, which had an error of 21 percent. Prediction errors were no more than 3 percent for arthritis, asthma, diabetes, psychiatric conditions and the activity-limiting conditions.
We did not find evidence that particular chronic conditions are significantly under- or over-estimated by the models across employers.
The HCC risk-adjusted model predicted Employer A’s stroke expenditures to be 21 percent higher than the actual expenditures, while the same model under- predicted Employer B’s stroke expenditures by 10 percent. This pattern is encouraging since it suggests that errors in the prediction of total payments are not systematic for a particular chronic condition.
Risk adjustment models were also applied to mental health expenditures alone, since many employers typically use different plans or payment methods for mental health services than for other health services. When choosing among health plans, employers have the option to “carve out” mental health care by assigning it to third-party plans that specialize in treating psychiatric illness. As discussed above, risk-adjustment systems offer employers a way to judge among competing health plans. However, these systems were designed to predict expenses across all types of illness, not mental illness alone. The purpose of our second risk- adjustment study was to assess how well common risk-adjustment systems can predict actual expenditures for mental health care.
For this study, we examined the viability of two risk adjustment schemes for setting capitation payments to Employer B’s carve-out plan, using 1994 and 1995 data. A single carve-out plan was responsible for all psychiatric health care delivered by Employer B’s indemnity and PPO plans. The specific risk-adjustment systems used were Ambulatory Care Groups (ACGs), Hierarchical Coexisting Conditions (HCCs) and Adjusted Diagnostic Groups. We predicted health care expenditures for members of the indemnity and PPO plans, comparing predictions that included the risk-adjustment factors to predictions based only on demographics-age, sex, and area hospital wage index. Complete study methods and results are reported in Chapter 8.
Risk adjustment may substantially improve payment methods for psychiatric conditions, but there is still room for improvement to make payments more equitable.
ACGs, HCCs and ADGs improved the ability to predict actual expenditures for mental health care, relative to predictions based on demographics alone. The difference between actual and predicted expenses for Employer B’s HMO was an average of $133 per user per year for the model with demographic variables, $129 for demographic variables plus ACGs, $116 for demographic variables plus ADGs, and $115 for demographic variables plus HCCs.
There is still room for improvement in risk-adjustment systems when forecasting mental health expenditures. Adding controls for the type of psychiatric disability (e.g., major depression or schizophrenia) increased the ability to match payments with expenditures even when HCCs or ADGs were used. ADGs and HCCs underpredicted actual expenses by 22-28 percent for individuals with psychiatric disabilities, but adding controls for the type of disability reduced the prediction error to 15-18 percent.
Mixed payment systems offer some of the advantages of capitating mental health services, but they also reduce some incentives for cost containment.
A mixed payment system combines traditional reimbursement with capitation. For example, in a 50/50 mixed payment system a health plan would receive 50 percent of a risk-adjusted capitation payment for its population plus 50 percent of actual costs. Partial capitation encourages health plans to reduce costs, but the incentive is weaker than under a fully capitated system. We compared the predicted profits or losses from several alternative payment systems for a sample of HMO enrollees. A 50/50 mixed payment system generated profits and losses roughly 50% lower than those predicted under full capitation, and even smaller gains and losses were found under other payment systems.
Our main conclusions for employers and other purchasers of health insurance include the following:
Firms seeking to save money by offering managed care plans to their employees need to consider the type of plans being offered.
Managed care plans often have lower health care utilization and expenditures for chronically ill individuals. Managed care plans may save money relative to indemnity plans, but the impact of these plans is not uniform. For example, the managed care plans offered by Employer B had a stronger impact on total costs than the POS plan offered by Employer A. The POS and HMO plans experienced differing abilities to control costs and utilization relative to the two indemnity plans, although the POS and HMO plans in this study were available at different firms, making direct comparisons problematic.
Risk adjustment helps to close the gap between payments and expenditures, but current risk-adjustment methods are imperfect and may still lead to substantial losses or profits.
Risk adjustment based on Adjusted Care Groups (ACGs), Ambulatory Diagnostic Groups (ADGs), or Hierarchical Coexisting Conditions (HCCs ) led to substantially more equitable payments than did age and sex adjustment alone. Thus, risk- adjusted payment systems have the potential to reduce, but perhaps not eliminate, incentives for plans to under-serve or to avoid enrolling people with chronic illness. Risk adjustment is a sensible method for health plan sponsors to ensure financial stability and for firms to evaluate competing plans. Its actual impact on both enrollees and plans must be monitored carefully, however.
E. Future Research
The four studies answer some basic questions about the experience of people with potentially disabling chronic conditions in private-sector health care plans. They augment our knowledge in several key areas: the prevalence of potentially disabling chronic conditions among privately insured individuals; relationships between managed care coverage and service utilization and expenditures; and better ways to pay plans for services covered. This information can be used by corporate leaders and government policymakers to provide incentives for health plans to recruit, accept, and appropriately care for those with chronic conditions.
At the same time, this research raises a number of questions that deserve further study. Below are topics for further research, each of which will aid our understanding of how private-sector health plans treat those with chronic illnesses and how corporate leaders and public policymakers can use this information.
What characteristics of health plans are most attractive to people with chronic illness?
Increasing health care quality and access and lowering costs for chronically ill people requires knowledge of their preferences. By exploring the characteristics of competing health plans we may learn additional ways to encourage chronically ill individuals to enroll in managed care plans and to find better ways to make sure they receive all of the services they need, in an efficient manner. Together these may increase individuals’ satisfaction while decreasing employers’ costs through higher productivity and lower absenteeism.
How can risk-adjustment systems be modified to better balance the competing objectives of access and cost management for chronically ill individuals?
New payment systems have been developed which share features of both capitated and traditional (fee-for-service) reimbursement methods. Further study is necessary to determine how these hybrid payment systems can be modified to allow health plans to earn reasonable returns while simultaneously ensuring appropriate health care for chronically ill individuals.
How does the quality of care in managed care and indemnity plans compare for people with potentially disabling chronic conditions?
The research we conducted shows that utilization and expenditure sometimes differ across plan types. The next logical step is to determine whether the appropriateness and quality of care differ too. This should be followed by research linking care patterns to outcomes in the different plan types, to see if there are ways to maximize and equalize the quality of care for people with potentially disabling chronic conditions.
How might better preventive care services benefit those with potentially disabling chronic conditions?
As an example of why this issue is important, it has been shown that better preventive care for people with diabetes may postpone the onset of disabling or life-threatening complications, such as blindness, amputation, and end-stage kidney failure (Diabetes Control and Complications Research Group, 1996). Thus, the more we can learn about differences in preventive services by plan type, the better able providers and policy makers will be to draft policies that assure the appropriate use of those services.
The answers to these questions could greatly enhance the existing knowledge base that corporate and public policymakers draw upon when considering methods for better meeting the needs of people with potentially disabling chronic conditions.