Health Care Utilization in Chronic Obstructive Pulmonary Disease: A Case-Control Study in a Health Maintenance Organization

Background  Information about health care utilization and costs among patients with chronic obstructive pulmonary disease (COPD) is needed to improve care and for appropriate allocation of resources for patients with COPD (COPD patients or cases) in managed care organizations.

Methods  Analysis of all inpatient, outpatient, and pharmacy utilization of 1522 COPD patients continuously enrolled during 1997 in a 172,484-member health maintenance organization. Each COPD case was matched with 3 controls (n = 4566) by age (±5 years) and sex. Information on tobacco use and comorbidities was obtained by chart review of 200 patients from each group.

Results  Patients with COPD were 2.3 times more likely to be admitted to the hospital at least once during the year, and those admitted had longer average lengths of stay (4.7 vs 3.9 days; P<.001). Mean costs per case and control were $5093 vs $2026 for inpatient services, $5042 vs $3050 for outpatient services, and $1545 vs $739 for outpatient pharmacy services, respectively (P<.001 for all differences). Patients with COPD had a longer smoking history (49.5 vs 34.9 pack-years; P = .002) and a higher prevalence of smoking-related comorbid conditions and were more likely to use cigarettes during the study period (46.0% vs 13.5%; P<.001).

Conclusions  Health care utilization among COPD patients is approximately twice that of age- and sex-matched controls, with much of the difference attributable to smoking-related diseases. In this health maintenance organization, inpatient costs were similar to and outpatient costs were much higher than national averages for COPD patients covered by Medicare.

CHRONIC OBSTRUCTIVE pulmonary disease (COPD), including emphysema, chronic bronchitis, and chronic obstructive asthma, is the fourth leading cause of death in the United States, and mortality rates for COPD are expected to increase during the next decade, particularly among women.^1,2 The prevalence of COPD in the general population is difficult to determine, but estimates derived from the National Health Interview Survey suggest that chronic bronchitis and emphysema affect 1 in 8 men and 1 in 10 women older than 65 years.³ Chronic obstructive pulmonary disease is a major cause of chronic disability and a leading reason for visits to office-based physicians.⁴

Despite the high prevalence, morbidity, and mortality of COPD, remarkably little is known about its impact on health care costs and utilization of services. Most available data are based on patients with severe end-stage disease^5,6 or special populations such as Medicare recipients.^7,8 These studies suggest that utilization of health care services among patients with COPD (COPD patients or cases) is dramatically increased. For example, per capita inpatient costs for COPD patients who are covered by Medicare were more than double those of average Medicare recipients ($5409 vs $2001 in 1992), with nonrespiratory care services accounting for much of the difference.⁷ Unfortunately, details about outpatient and pharmacy utilization in these studies are limited or not available. Furthermore, information about smoking-related comorbid conditions and their contribution to overall health care utilization is limited. These studies also underrepresent COPD patients younger than 65 years and those otherwise not eligible for Medicare.

Data on health care utilization in COPD are needed so that health care plans and providers in capitated systems can be compensated more fairly for treating COPD patients. During the past decade, state and national Medicare and Medicaid systems have encouraged their patients to enroll in managed care organizations (MCOs) as a way of controlling costs while maintaining access to services.^9,10 However, because of the increased costs and risks associated with treating chronic illnesses such as COPD, MCOs are reluctant to enroll these patients in capitated programs and tend to disenroll those who become acutely ill.¹¹ To address this problem, the US Congress mandated that the Health Care Financing Administration (HCFA) implement a risk-adjusted payment system for Medicare health maintenance organization (HMO) plans by January 1, 2000.¹² Several states had already instituted risk-adjusted payment systems for their Medicaid programs.¹³ The risk adjustment method of the HCFA is likely to be based on Medicare fee-for-service inpatient data because of the lack of accurate and detailed outpatient data.¹⁴ Basing risk adjustment on inadequate data could have a number of unintended results. Health maintenance organizations with successful outpatient programs that reduce COPD hospitalizations may be penalized, causing them to withdraw from the Medicaid managed care program.^14,15 Alternatively, HMOs may be encouraged to increase hospital admissions or to develop other behaviors that would defeat the intention of the HCFA to control costs.¹⁶

A better understanding of utilization of health care services among COPD patients is also needed to assess the economic impact of interventions and improve quality of care.^17,18 For example, although smoking cessation improves lung function in COPD patients,¹⁹ MCOs often do not provide smoking cessation treatment, in part because of the lack of information on how this intervention would affect overall health care utilization.^17,20 New treatments for COPD will have to demonstrate a favorable impact on overall health care utilization in addition to improving more traditional outcome variables such as spirometry values and dyspnea scores. To improve understanding of health care utilization among persons with COPD, we conducted a case-control study examining inpatient, outpatient, and pharmacy utilization for all 1522 COPD patients enrolled in an HMO during 1997.

Lovelace Health Systems (LHS) is a staff and network model health care delivery system serving most of New Mexico. The Lovelace Health Plan (LHP), a division of LHS, is New Mexico's oldest HMO, serving more than 700 employer groups and 172,484 plan participants in 1997. Facilities owned and operated by LHS include an acute care hospital, 2 multispecialty clinics, and 9 primary care centers located in Albuquerque, Santa Fe, and Las Cruces. Lovelace Health Systems employs 285 staff physicians and contracts with an additional 37 hospitals and more than 2000 physicians.

The Managed Care Database is a multi-HMO relational database maintained by the Southwest Center for Managed Care Research, a division of the Lovelace Respiratory Research Institute. This longitudinal database contains more than 1 million patient-years of data from the LHP and other HMOs collected since 1988. Electronic data from outpatient, inpatient, pharmacy, laboratory, and enrollment records are collected quarterly from the LHS computer systems, normalized, subjected to quality assurance edits, and formatted into tables in a relational database. This database has been used in previous health care utilization and cost studies.^21,22

All patients continuously enrolled in the LHP from January 1, 1996, through December 31, 1997, with a primary or secondary diagnosis of COPD were identified. The 1996 calendar year served as the case-finding year and 1997 as the year for which utilization and cost records were reviewed. The continuous enrollment criterion ensured accrual of a full year of utilization data, avoided any biases that might be introduced by patients switching in and out of fee-for-service plans, and reduced bias introduced by services given at the end of life, which are known to be dramatically increased.²³ Chronic obstructive pulmonary disease was defined as International Classification of Diseases, Ninth Revision (ICD-9)²⁴ codes 491.x (chronic bronchitis), 492.x (emphysema), and 496 (COPD, unspecified), which are the codes that most closely fit the American Thoracic Society definition of COPD.²⁵ Patients younger than 30 years were excluded to avoid patients with bronchiectasis, cystic fibrosis, and reversible asthma. Three control patients were matched to each case by age (±5 years) and sex. Controls were selected from the health care plan population who had no COPD diagnosis codes in 1996 and who met the same continuous enrollment criterion as cases. Ethnicity was assigned using a computer program that identifies Hispanic and non-Hispanic ethnicity or race based on surname, with an accuracy of 90%.²⁶

Data for all 1997 health care encounters of cases and controls were extracted from the Managed Care Database. Data elements collected from enrollment records included age, sex, and dates of enrollment and disenrollment. Data elements extracted from inpatient and outpatient encounter and claims records included date of service, location of service, up to 3 ICD-9 diagnosis codes, and all codes from the Physicians' Current Procedural Terminology²⁷ associated with each visit. Data elements derived from pharmacy records included the date a prescription was filled, American Hospital Formulary Service class, and average wholesale price. Prescriptions filled by pharmacies outside the health care plan network were not available for inclusion.

Inpatient and outpatient costs were estimated using charges appearing on claims. The LHP generates claims for all health care services provided through its staff model HMO (internal claims) and processes claims submitted by external providers (external claims). For external claims, the amount requested by the submitting provider was used. Although this figure does not represent the final amount paid to the provider, it is likely to reflect market rates. For study patients receiving Medicare benefits, the costs used were the total costs and not just the Medicare benefit or any portion covered through supplementary medical insurance. Pharmacy costs were based on the average wholesale price of each prescription drug fill.

We reviewed the medical charts of 200 randomly selected COPD cases and 200 matched controls to obtain information on smoking history and prevalence of comorbid illnesses and to confirm the validity of the electronic case-finding method. The abstracting instrument was based on the American Thoracic Society Respiratory Symptoms Questionnaire modified to ascertain histories, examinations, and tests that support a COPD diagnosis.²⁸ Comorbid illnesses were categorized using a system developed to identify prognostically important chronic conditions.²⁹

Data from the chart abstraction were compared using paired t tests for continuous variables and the McNemar test for paired proportions. Differences in utilization and costs between case and control groups were tested using the Wilcoxon rank sum test.³⁰ Because many controls had zero costs for the study period, particularly for inpatient utilization, we also used Tobit models to estimate costs and develop 95% confidence intervals (CIs).³¹ Tobit analysis is a hybrid of probit analysis and multiple regression that takes into account the concentration of observations at the limiting value. Predictor variables considered in the Tobit models included ethnicity, case-control status, and a variable for the strata.

A total of 1522 COPD cases were identified who met enrollment criteria for the study. The mean age of cases was 67.5 years, with 72.3% aged 65 years or older; 48.9% were female. Patients with COPD were less likely than controls to be Hispanic (17.8% vs 30.8%), which is consistent with ethnic differences in smoking behavior found in this area.³² Cases and controls were relatively stable customers of the health care plan; the average duration of enrollment was more than 8 years for both groups, with only 18.3% of cases and 16.9% of controls enrolled for 3 years or less.

Among the 200 COPD cases and 200 controls whose medical charts were reviewed, cases were significantly more likely to be current or former cigarette smokers, and smokers in the case group had a higher average pack-year smoking history (Table 1). The most common primary pulmonary diagnosis in the case group was COPD (68.0%), followed by chronic bronchitis (7.5%), emphysema (3.0%), chronic obstructive asthma (12.0%), and miscellaneous other pulmonary diagnoses (9.5%). In the control group, 3.2% of former or current smokers had a documented history of chronic cough, dyspnea, or recurrent bronchitis, but did not have a clinical diagnosis of COPD.

Patients with COPD had a higher prevalence of chronic comorbid conditions, particularly nonpulmonary diseases related to smoking, including heart disease, cancer, neurologic injuries, and gastritis (Table 1). Cases had an average of 3.7 chronic medical conditions (including lung disease) vs 1.8 for controls (P<.001, paired t test). Only 6.0% of cases did not have another chronic medical condition.

During 1997, 25.0% of cases (n = 380) were admitted to the hospital at least once, compared with 10.6% of controls (n = 486) (Table 2). Of patients who were hospitalized, cases also were admitted more frequently (1.8 admissions per hospitalized case vs 1.6 admissions per hospitalized control; P = .01). Cases also had a longer average length of stay (4.7 vs 3.9 days; P<.001).

Respiratory system disorders were the principal discharge diagnosis in 31.2% of case hospitalizations, but cases also had a higher proportion of discharge diagnoses than controls in almost every major diagnosis category (Table 2). Admissions for cardiovascular diseases, the leading discharge diagnosis category in controls, were almost twice as common among cases.

Patients with COPD had significantly higher utilization in all outpatient service areas (Table 3). As expected, a large difference in utilization was seen in respiratory care services, with 27.9% of COPD cases using the pulmonary specialties clinic at least once during 1997, compared with 2.4% of controls. However, the greatest difference in overall utilization is seen in primary care and other nonpulmonary specialty services. Although the proportion of COPD cases seeing a primary care physician at least once during the year (89.2%) was only slightly greater than that of controls (84.1%), the average number of primary care visits for cases was 54% higher. Patients with COPD were also significantly more likely than controls to use outpatient cardiology services (30.1% vs 19.5%) and emergency department services (21.8% vs 13.1%) at least once during the year (for all comparisons, P<.001, Wilcoxon rank sum test).

Significant differences in outpatient pharmacy utilization between cases and controls are seen in all major drug categories (Table 3). The greatest difference is seen in respiratory medications, but COPD patients also had a substantially higher number of prescription fills for cardiovascular drugs, antibiotics, psychotherapeutic drugs, and pain medications.

Total health care costs for cases were twice those of controls (Table 4), with a marginal cost of COPD in this HMO of $5864 per case in 1997. The greatest proportional difference in costs was seen in inpatient care, with the average cost per case 2.5 times that per control. However, costs for outpatient services for cases were significantly increased in all service areas, and total outpatient costs (including pharmacy) were greater than inpatient costs for cases and controls. The costs of outpatient pulmonary services, home oxygen therapy, and respiratory drugs averaged $729 per case, but pulmonary-related utilization accounted for only 25% of the total difference in outpatient utilization between groups.

Tobit models were used to estimate cost differences between case and control groups adjusted for differences in ethnicity and in the proportion of patients with zero cost. The estimates were similar to those in the unadjusted analysis, with substantially higher costs found in the case group in each category. The adjusted differences between case and control groups were as follows: inpatient cost, $3074 (respiratory admissions, $1333 [95% CI, $1125-$1541]; nonrespiratory admissions, $1740 [95% CI, $1141-$2340]); outpatient services, $1788 (95% CI, $1515-$2061); outpatient pharmacy, $794 (95% CI, $732-$856); and total cost, $6039 (95% CI, $5221-$6858).

Health care utilization and costs of the 1522 COPD patients in this HMO were more than twice those of age- and sex-matched controls. Not surprisingly, utilization of respiratory services and medications was higher among cases by a factor of more than 12. However, utilization and costs in all service areas were significantly higher among cases, and most of the difference in utilization was not directly attributable to pulmonary services. Tobacco use and smoking-related nonpulmonary illnesses, especially cardiovascular diseases, were significantly more prevalent among cases, with corresponding increases in related utilization and costs.

Total inpatient costs for the COPD patients in this HMO were similar to and total outpatient costs much higher than those seen in recent Medicare fee-for-service data.⁷ In our study, mean inpatient costs were $5446 and mean outpatient costs were $5446 for COPD patients aged 65 years and older. Grasso and colleagues⁷ reported that the average 1992 Medicare costs per COPD patient were $5409 for inpatient services, $469 for outpatient services, and $2604 for Medicare part B services, which include outpatient procedures and physician fees. The outpatient costs for our controls aged 65 years and older were also much higher than the 1992 Medicare average. These figures are not directly comparable because our inpatient and outpatient cost estimates include physician services, because we excluded patients who died, and because these figures do not account for annual deductibles and copayment expenses, which are higher in the Medicare fee-for-service program. We also cannot adjust for differences in disease severity between our COPD cohort and the national Medicaid database, although the percentage of patients using home oxygen therapy was similar (20% and 16%, respectively).⁷

Our study has several important implications for health policy planning in COPD. First, the data suggest that enrolling Medicare COPD patients in MCOs will result in only modest, if any, reduction in health care costs. Utilization among COPD patients is very high regardless of the type of health care payment or delivery system, and improved access to outpatient services may increase overall utilization.³³ Second, any risk-adjusted payment system will be problematic in COPD because of the high prevalence of comorbidities that significantly affect health care utilization. In the study by Grasso et al,⁷ costs for COPD patients without other serious illnesses were only minimally elevated, but COPD patients with just 1 comorbid condition were more than double the national average. Third, any policy that increases payment for patients with a diagnosis of COPD is likely to result in a substantial number of newly diagnosed cases. In our chart review, 3.2% of controls had a history of smoking and chronic symptoms consistent with COPD, suggesting that at least 3000 patients in this HMO could be labeled immediately as having COPD, and probably many thousands more could receive a diagnosis of COPD with appropriate screening. If it becomes financially advantageous for patients to receive a diagnosis of a specific condition, they undoubtedly will, as was seen when the diagnosis-related group payment system was implemented.³⁴

Our study reveals an obvious intervention that could improve care and reduce costs in this population: smoking cessation. Approximately 46% of COPD cases were still using cigarettes during the survey period, compared with 13.5% of controls. Smoking cessation reduces rates of respiratory symptoms and infections³⁵ and reduces the overall decline in pulmonary function,¹⁹ compared with continued smoking. Smoking cessation also reduces the risk of coronary artery disease, respiratory and nonrespiratory tract cancers, and neurovascular disease.³⁵ Although the economic benefits of smoking cessation need to be examined in prospective studies, aggressive smoking intervention in COPD will promote better health and is likely to have cost utility comparable with that of other commonly accepted tertiary interventions.^17,20

Our study has limitations that limit its generalizability to other populations. Patients with COPD who died or disenrolled from the health care plan for other reasons were excluded, so our cost estimates are likely to underestimate the true costs for the entire population with diagnosed COPD. This study was conducted in a single HMO, and the results may not be directly comparable with those of COPD patients enrolled in other health care plans, those who have only Medicare coverage, or those without any health insurance. The proportion of cases who were female (48.9%) was higher than that seen in most COPD studies, which may represent a selection bias. Health care utilization for COPD varies by region,³⁶ and sex and ethnic differences in smoking behavior may also affect utilization. These limitations would not have affected the differences we observed between case and control groups, and therefore would not affect the overall results or conclusions.

The MCOs may have advantages over traditional fee-for-service systems that allow more cost-effective management in COPD, but a fair risk-adjustment system for this condition requires a better understanding of the factors affecting utilization. Clear diagnostic standards need to be developed before implementation of the program, and research is needed on the relationship between severity of airflow obstruction and health care utilization. Increased payment for COPD care must also be linked to specific goals and guidelines for management, including smoking cessation therapy, to improve the quality of care.

Accepted for publication March 3, 2000.

This study supported by National Research Service Award K08 HL03615 from the Division of Lung Diseases, National Heart, Lung, and Blood Institute, National Institutes of Health, Rockville, Md (Dr Mapel), and a grant from SmithKline Beecham, Collegeville, Pa.

We are grateful to Susan Paine, MPH, and Curtis Hunt, MS, for assistance with data analysis, to Ann Von Worley and Andrea Romero for abstraction of medical records, and to Rita Elliot, MA, for editorial review. We wish to give special thanks to Eva Lydick, PhD, for her ideas and support.

Reprints: Douglas W. Mapel, MD, MPH, Epidemiology and Cancer Control Program, University of New Mexico Health Sciences Center, 2325 Camino de Salud NE, Albuquerque, NM 87131-5306 (e-mail: dmapel@salud.unm.edu).