INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Enrollment in managed care continues to increase in both the public and private sectors (1). In 1995, 75% of privately insured individuals were in managed care (2). Medicaid programs have enrolled the Aid to Families with Dependent Children population in 48 states and the Supplemental Security Income population in 17 states (3, 4). Until recently, Medicare beneficiaries have been reluctant to join managed care plans. This is beginning to change, however. In 1996, approximately 10% of Medicare beneficiaries were enrolled in risk-contract managed care plans with over 100,000 Medicare beneficiaries enrolling in managed care every month. The Congressional Budget Office estimates that more than 29% of the aged Medicare population will be enrolled in capitated managed care plans by the year 2000 (5).

As enrollment in managed care continues to grow, more individuals with chronic illnesses are likely to enroll. While there is disagreement over how to define chronic illness, one recent study calculated that the 99 million Americans with one or more chronic illnesses accounted for 76% of total medical care spending in 1995 (6). The same study also showed that the prevalence of chronic illness increased with age with nearly 40% of the elderly not living in institutions experiencing limitations from one or more chronic illnesses. Given the rapid growth of Medicare and Medicaid managed care, the high prevalence of chronic illness among the elderly, and the financial incentives given by managed care plans for the elderly to enroll, enrollment of chronically ill Medicare beneficiaries in managed care can be expected to significantly increase in the coming years (7).

Under capitated arrangements, managed care organizations are paid a fixed amount of money for each enrollee, regardless of the amount of services actually provided to that specific individual. As a result, capitated managed care organizations assume financial risk for providing medical care to their enrollees. Increasingly, managed care organizations are passing on some or all of this risk to their physicians by paying them on a fully or partially capitated basis (8). Physicians who accept full capitation agree to provide or arrange for all physician, inpatient, and ancillary services in return for fixed payment per enrollee. More commonly, physicians are paid on a partial capitation basis where physicians accept a smaller capitation payment, but are at risk for a narrower range of services, such as physician services only (9).

Three significant clinical and financial questions emerge as more chronically ill individuals enroll in managed care and capitation becomes an important source of revenue for managed care plans and physicians. First, how should capitation payments be adjusted to reflect the higher health care costs associated with individuals with chronic illness? Unless capitation rates are adjusted to reflect the higher expected costs associated with chronic illness, plans or capitated providers have a financial incentive to avoid, disenroll, or undertreat chronically ill individuals. Second, what level of risk is appropriate for capitated physicians to assume and what methods are available to limit their risk? Physicians whose practices are oriented toward the chronically ill may face substantial levels of financial risk under capitation. Third, how can access to essential services and providers be assured for individuals with chronic illness? Most managed care plans have limited experience with this population and may not be familiar with the mix of services they require. The data presented here may assist managed care plans to develop their networks and assess whether individuals with COPD have access to appropriate providers. In addition, states attempting to regulate managed care will also be interested in how to monitor whether appropriate care is being provided to individuals with chronic illnesses.

We examined one chronic illness, chronic obstructive pulmonary disease (COPD), to begin to address these questions. COPD is an example of a chronic illness that is expensive to treat (10) and is a major contributor to mortality and morbidity among the elderly (11). Although COPD is the fourth leading cause of death from chronic illness in the United States (11), little information is available on the extent and spectrum of resource utilization for the care of patients suffering from this disease. In this study, we analyzed utilization and expenditure patterns for a nationally representative sample of fee-for-service aged Medicare beneficiaries and report some of the financial risks facing managed care plans and physicians, including pulmonologists and others, who accept capitated payments for these individuals. Data on utilization and expenditure patterns on a nationally representative sample of aged Medicare beneficiaries enrolled in managed care are not collected. For this reason we chose to analyze fee-for-service care. In conducting this analysis, we do not intend to imply that fee-for-service care is the "gold standard." Instead, the data are used simply as a point of reference. In addition to the analysis of Medicare data, we consulted published guidelines and clinical literature in an attempt to establish the extent of services required in treating a chronic illness such as COPD.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design and Population

This is a national, cross-sectional study of utilization and expenditures for aged Medicare beneficiaries with COPD. To be eligible for our analysis, persons had to be Medicare beneficiaries older than 65 yr of age living in the U.S., with fee-for-service coverage, who had both Medicare Part A and Part B coverage in fiscal year (FY) 1992. Encounter and expenditure data for beneficiaries enrolled in HMOs were not available.

Chronic obstructive pulmonary disease is the name given to a broad spectrum of obstructive pulmonary diseases. For purposes of this study we selected the ICD-9 diagnosis codes 491.2 (chronic obstructive bronchitis with or without mention of acute exacerbation) and 492.x (emphysema) (12). These codes closely match the American Thoracic Society's most recent definition of COPD which is that disease state characterized by "presence of airflow obstruction due to chronic bronchitis or emphysema, progressive in nature, which may be accompanied by airway hyperreactivity and may be partially reversible" (11, 13). Patients with asthmatic bronchitis were excluded because this population is generally considered to have different pathophysiological characteristics (11).

Data Construction

Data were obtained from a 5% nationally random sample of paid claims for Medicare beneficiaries in FY 1992. The file, maintained by the Health Care Financing Administration, is commonly used for research purposes. The enrollment file contains demographic information on each Medicare beneficiary and the claims files contain expenditure and utilization data at the individual beneficiary level for all Medicare covered services. We combined the enrollment and claims files to develop individual records for each Medicare beneficiary in our sample.

Skilled nursing home, hospice, and home health care expenditure and utilization data were not included, since they are frequently "carved out" of standard benefit packages offered by managed care plans. Diagnostic and procedure codes for all other covered services were reviewed, and when a code of 491.2 or 492.x or both was found in any utilization record, that individual was identified as having COPD. These individuals formed the analysis file.

Expenditures were grouped into three categories: hospital inpatient, hospital outpatient, and physician. Hospital inpatient and hospital outpatient include short-term, psychiatric, rehabilitation, long-term, and other hospitals. Physician expenditures could be grouped by specialty and also included the following services: laboratory, ambulance, and medical supplies.

Expenditures reflect actual payments made by the Medicare program in a fee-for-service setting in FY 1992. Total expenditures for Medicare beneficiaries with COPD represent expenditures for all Medicare covered services provided to a beneficiary with COPD during the year, including services unrelated to COPD. We compared expenditures for beneficiaries with COPD with all Medicare beneficiaries in the database, and examined the distribution of expenditures for services received by beneficiaries with COPD compared with all Medicare beneficiaries. Expenditure information was not available on deductibles and coinsurance paid by Medicare beneficiaries for Medicare covered services. The expenditures in this study also do not include services that may have been provided to beneficiaries, but were not covered by Medicare.

Comorbidities

In order to determine if certain comorbidities were associated with higher expenditures and different utilization patterns among Medicare beneficiaries with COPD, the claims data were further stratified by specific comorbid diagnoses and procedures, as well as a comorbidity index.

Diagnoses were selected on the basis of a previous study establishing certain diagnoses as being associated with the majority of hospitalizations for acute COPD exacerbations (13). Subgroups by additional diagnosis included patients who had at least one ICD-9 coding in Part A or Part B claims for the following diseases: upper respiratory infection (460, 462, 464, 465, 466, 487), lower respiratory tract infection (480, 481, 482, or 483), septicemia (038.x), and heart failure (428).

We also examined several procedures and therapies with well-established criteria for usage which may serve as markers of resource intensity and disease severity among individuals with COPD. One therapy examined was long-term oxygen therapy (LTOT) (14). The LTOT subgroup contained all patients with at least one code for an oxygen delivery device as identified by the 1992 Health Care Financing Administration Common Procedure Code System (HCPCS): E1401 through E1404 (15). Additional subgroups, by procedure, were identified by the 1992 HCPCS Current Procedure Terminology (CPT) coding system and included bronchoscopy (31615, 31622, 31625, 31628, 31629, 31645, 31646), intubation (31500), and tracheostomy (31600, 31610) (16).

The Deyo-adapted Charlson index (D-CI), a measure of comorbidity based on ICD-9 coding for 17 diagnosis indicators, was used to assess a broad range of comorbid conditions and their impact on expenditures for Medicare beneficiaries with COPD (17). Each comorbid condition is assigned a weight from 1 to 6 based on adjusted relative risk for this indicator giving the index a possible range of 0 to 33, with the D-CI calculated as the sum of the weights of the indicated comorbidities (18). This index was applied to the list of Medicare beneficiaries with COPD to provide a D-CI score for each individual.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characteristics of Study Population

From the 1992 Medicare 5% sample which contained 1,221,615 aged Medicare beneficiaries, 42,472 individuals met our criteria for COPD. The average Medicare expenditure per beneficiary with COPD was $8,482 or 2.4 times the average expenditure for all Medicare beneficiaries (Table 1). The percentage of Medicare beneficiaries, total per capita expenditures, and the ratio of per capita expenditures for Medicare beneficiaries with COPD compared with all Medicare beneficiaries, stratified by age and gender, are presented in Table 1. Per capita expenditures for the three age groups < 75, 75-84, and  85 were similar. Beneficiaries with COPD were slightly more likely to be male than female; however, per capita expenditures were similar for men and women.

A relatively small percentage of the individuals accounted for a large percentage of the total expenditures for Medicare beneficiaries with COPD (Figure 1). The 10% with highest expenditures accounted for nearly half of total Medicare expenditures for this population.

View larger version (23K): [in this window] [in a new window]   Figure 1.   Percentage of payments versus percentage of population.

Expenditures by Comorbidities and Procedures

In general, comorbidities and/or specific procedures were associated with higher per capita expenditures among beneficiaries with COPD (Table 2). A diagnosis of COPD and upper respiratory tract infection increased per capita expenditures by only 10% compared with all Medicare beneficiaries with COPD. Those beneficiaries with a diagnosis of pneumonia, however, had average expenditures that were over 2 times the average for all beneficiaries with COPD. Comorbidities of heart failure and septicemia were associated with expenditures 1.6 and 3.0 times the average for all beneficiaries with COPD.

Oxygen-dependent COPD beneficiaries had per capita total expenditures 1.6 times that of the average Medicare beneficiary with COPD. Beneficiaries with COPD and a bronchoscopy had per capita expenditures that were 2.5 times those of all Medicare beneficiaries with COPD. Intubation and tracheostomy were associated with per capita expenditures of 3.7 and 7.8 times the average Medicare beneficiary with COPD, respectively. Patients with procedure coding for tracheostomy represented less than 1% of the sample COPD population although they account for greater than 15% of total expenditures.

Higher comorbidity, as measured by the D-CI, was roughly linearly associated with total expenditures up to an index of 5 (Table 2). Nearly 40% of the COPD patients had a D-CI score of zero. These patients had total expenditures similar to those of the sample of all Medicare beneficiaries (expenditure ratio 1.0 compared with all Medicare beneficiaries). The average expenditures of the COPD population corresponded to a D-CI between 1 and 2. Higher D-CIs were associated with the minority of patients with highest expenditures. For example, only 10% of patients had a D-CI of 4 or greater, however, this group accounted for nearly one-quarter of total expenditures.

Distribution of Expenditures

The distribution of expenditures for Medicare beneficiaries with COPD was compared with all Medicare beneficiaries in the database (Table 3). Hospital inpatient expenditures for the COPD population accounted for 64% of total per capita expenditures compared with 57% for all Medicare beneficiaries. The average hospital inpatient expenditures for a Medicare beneficiary with COPD were 2.7 times the average inpatient expenditures for all Medicare beneficiaries. Outpatient and Part B expenditures for a beneficiary with COPD accounted for only 5% and 31% of total per capita expenditures compared with 9% and 34% for the average Medicare beneficiary. However, average expenditures for outpatient and Part B services for a beneficiary with COPD were 1.5 and 2.2 times the average expenditures for all Medicare beneficiaries in the sample.

Expenditures by Specialty

When categorized by physician specialty, per capita expenditures were typically higher among beneficiaries with COPD (Table 4). Pulmonology was the only specialty that had much greater expenditures among the Medicare beneficiaries with COPD. Expenditures per capita for pulmonologists for beneficiaries with COPD were 7.5 times the expenditures per capita for all Medicare beneficiaries. The higher expenditures for pulmonologists may be attributable to a sicker patient population using these providers. However, it was not possible to adjust for differences in severity of illness using the claims data. It is also not possible to determine if the care provided by pulmonologists resulted in lower expenditures in other areas or better outcomes. Total per capita expenditures for supply companies among COPD patients were 5.8 times those of all Medicare beneficiaries. Further analysis suggests that most of the difference is attributable to oxygen equipment.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Under most capitated systems, payments are often not adjusted for the health status of the enrolled population. Medicare's capitation payment, for example, is adjusted for demographic characteristics, institutional status, and welfare status, but includes no direct adjustment for the health status of enrollees (20). In this study we have shown that total per capita expenditures for Medicare beneficiaries with COPD were 2.4 times the mean for all beneficiaries. Without some method for limiting financial risk under capitation, health plans and capitated providers may have a financial incentive to avoid individuals with COPD, or to encourage the individual to disenroll if he or she develops COPD.

One approach to limit risk under capitation would be to designate COPD as a "carve out" condition. Under a "carve out" system, separate payments are developed for certain high-cost conditions and usually payment levels are based on the average cost of that condition. For example, a managed care plan or a fully capitated physician group enrolling a Medicare beneficiary with COPD might be paid a capitation rate that is approximately 2.4 times the rate of the average Medicare beneficiary. This method may encourage capitated physicians and plans to accept Medicare beneficiaries with COPD.

However, the high variability in expenditures for Medicare beneficiaries with COPD may make a carve out based solely on a diagnosis of COPD difficult. Figure 1 shows that spending on Medicare beneficiaries with COPD is highly skeweda small portion of beneficiaries with COPD accounts for a large portion of total expenditures for this group. In this study, approximately 70% of Medicare beneficiaries with COPD had expenditures below $8,482, the average expenditures for a beneficiary with COPD. A capitation rate based on the average expenditures for enrollees with COPD could encourage providers and plans to actively seek those individuals with COPD whose costs are expected to be lower than average, while discouraging the enrollment of enrollees expected to incur higher costs.

For this reason it may be important to identify specific demographic or clinical factors that identify enrollees with higher expected costs. The results of this study suggest that among the COPD population, age and gender explain very little of the variation in expenditures for beneficiaries with COPD. These results are consistent with literature showing that demographic features are not highly predictive of present or future expenditures at the person level (21). We also examined whether certain clinical indicators available in the claims data were associated with higher expenditures. In addition to being predictive of higher expenditures, useful prospective risk adjusters should be resistant to incentives for upcoding and inefficient management (22). Connors and coworkers have reported that patients with acute COPD exacerbations have 6-mo hospital readmission rates greater than 50%, thus adding substantially to the cost of caring for these patients (13). Infection, mainly respiratory, and congestive heart failure (CHF) have been shown to be associated with the majority of acute exacerbations of COPD (13). This analysis shows that beneficiaries with COPD and an additional diagnosis of pneumonia, CHF, or sepsis have higher annual per capita expenditures. A possible prospective risk adjustment system could adjust capitation rates based on such associated diagnoses. One potential problem with using these comorbid diagnoses for payment is that they are based partly on subjective clinical criteria, because available objective tests lack sufficient diagnostic value. A payment system increasing reimbursement based on such diagnoses potentially creates incentives for the inclusion of borderline cases for financial gain (22).

Prospective risk adjustments based on more objective criteria are also possible. Long-term oxygen therapy, for example, is common among COPD patients with high expenditures, is likely predictive of future resource utilization, and has the added benefit of relying on existing objective Medicare payment guidelines (14). In this study, tracheostomy and intubation are found to be associated with the minority of COPD patients with the highest expenditures. COPD patients undergoing these procedures have markedly elevated yearly expenditures. Given the seriousness of these procedures, it is much less likely that they will be undertaken for financial gain.

Finally, comorbidity indexes may potentially be used to establish a prospective risk adjustment system. These allow aggregate assessment of risk using a range of clinical data to adjust prospective payment rates based on claims data for a large population. In this study when the D-CI was applied to the COPD population, a roughly linear association between yearly expenditures and increasing Charlson index was found, suggesting such indexes may be useful for prospectively adjusting capitation rate. Additional advantages include automatic adjustment for nonpulmonary comorbidities, low administrative burden when applied to large population databases, and reduction of clinical subjectivity given the multiple diagnoses combined to make up the index.

No prospective risk adjustment system can be expected to be perfectly predictive of future expenditures (23). Health plans and physician groups will always remain vulnerable to unpredictable high-cost cases. From this study the top 1% of aged Medicare beneficiaries with COPD, on average, have 26 times the expenditure as the average beneficiary. Some retrospective risk adjusters, most commonly reinsurance, may be desirable to protect against random variations. When physicians purchase individual stop-loss reinsurance, the insurance company pays when expenditures for an individual Medicare beneficiary exceed a specified amount. Thus, in effect, the physician is paying not to accept risk above a certain expenditure threshold. Some level of copayment above the threshold may be included to provide incentive for continued efficient management.

There are several potential limitations to this study. First, fee-for-service data from 1992 may not reflect current medical practice. Second, spending and utilization patterns in a fee-for-service setting may not reflect spending and utilization patterns in a managed care setting. Third, using 1 yr of data does not allow one to determine if expenditures are consistently higher over multiple years. Finally, there are limitations with claims data. Many of these limitations have been described previously, including accuracy of coding; a lack of the usual markers for mortality, morbidity, or intensity of resource use for COPD patients; and a lack of outcome measures (21, 24, 25). As a result, conclusions cannot be made regarding the appropriateness of services.

Despite these limitations, this study demonstrates the need for further analytical studies to understand the wide variation of expenditures for individuals with COPD. The importance of characterizing these expenditure patterns and their association with clinical risk factors will allow implementation of appropriate risk adjustment strategies to maintain access to care. Moreover, such analyses will allow targeted intervention to those most severely affected by COPD and who are at higher risk for costly medical complications.