Medical care expenditures and earnings losses of persons with arthritis and other rheumatic conditions in the United States in 1997Total and incremental estimates.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 50, No. 7, July 2004, pp 2317–2326 DOI 10.1002/art.20298 © 2004, American College of Rheumatology Medical Care Expenditures and Earnings Losses of Persons With Arthritis and Other Rheumatic Conditions in the United States in 1997 Total and Incremental Estimates Edward Yelin,1 Miriam G. Cisternas,2 David J. Pasta,3 Laura Trupin,1 Louise Murphy,4 and Charles G. Helmick4 lion). The largest components of these expenditures were inpatient care (39%), ambulatory care (29%), and prescriptions (14%). The mean increment in medical care expenditures specifically attributable to AORC among those ages 18 years and older was $1,391 (total ⬃$51.1 billion). Persons with AORC ages 18–64 years earned $3,812 less on average than did other persons of these ages (total $82.4 billion). Of this average, $1,579 was attributable to the AORC (total $35.1 billion). Conclusion. In 1997, persons with AORC incurred direct and indirect costs of $269.3 billion, of which $86.2 billion was attributable to these conditions. Objective. To provide estimates of the total medical care expenditures and earnings losses associated with arthritis and other rheumatic conditions (AORC), as well as the increment in such costs specifically attributable to these conditions, in the US in 1997. Methods. The estimates were derived from the 1997 Medical Expenditures Panel Survey (MEPS), a national probability sample of 14,147 households including 34,551 persons, of whom 4,776 self-reported arthritis. After weighting, those who self-reported AORC represent 38.4 million persons. We tabulated all medical care expenditures of the adult MEPS respondents, stratified by arthritis and comorbidity status, and then used regression techniques to estimate the increment in health care expenditures attributable to AORC, after taking comorbidity, demographic characteristics, and insurance status into account. Using the same methods, we also estimated the magnitude of the earnings losses sustained by persons of working ages (18–64 years) who had AORC. Results. Persons with AORC incurred mean total medical care expenditures of $4,865 (total $186.9 bil- Information about the economic impact of illness has become central to health policy debates, affecting the allocation of research funding among conditions and the choice of therapies for specific conditions. Because of the growing importance of cost of illness studies in health policy, the literature concerning the cost of all forms of musculoskeletal conditions for the US as a whole (1–8) and for other nations (9–13), as well as the cost of specific musculoskeletal conditions, including various forms of arthritis and other rheumatic conditions (AORC) (14–37), has been expanding rapidly. In general, the studies of the economic impact of all forms of musculoskeletal conditions have been derived from population-based surveys, but not directly observed individual-level medical care expenditures or employment status and earnings. In contrast, studies that have tracked the actual costs or expenditures for individuals have used clinical-based samples and are, thus, not representative of the general population. In the present study, we used the results of the Medical Expen- Supported by grants from the Arthritis Foundation and from the Arthritis Program of the Health Care and Aging Studies Branch, Centers for Disease Control and Prevention. 1 Edward Yelin, PhD, Laura Trupin, MPH: University of California, San Francisco; 2Miriam G. Cisternas, MA: MGC Data Services, Carlsbad, California; 3David J. Pasta, MS: DMA Corporation, Palo Alto, California; 4Louise Murphy, BSc, Charles G. Helmick, MD: Centers for Disease Control and Prevention, Atlanta, Georgia. Address correspondence and reprint requests to Edward Yelin, PhD, Rosalind Russell Medical Research Center for Arthritis, UCSF, Box 0920, San Francisco, CA 94143-0920. E-mail: firstname.lastname@example.org. Submitted for publication August 21, 2003; accepted in revised form March 1, 2004. 2317 2318 YELIN ET AL ditures Panel Survey (MEPS), which melds the methods of the 2 kinds of studies by using a population-based sampling frame and then prospectively tracking actual expenditures and employment and earnings among respondents. The specific goals of the study were as follows: 1) to provide estimates of all medical care expenditures on behalf of all persons with any form of AORC in the US in 1997, estimates of the increment in expenditures that was specifically attributable to the AORC among persons ages 18 years and older, and estimates of the fraction of total medical care expenditures attributable to the AORC and 2) to provide estimates of lost earnings among persons with AORC ages 18–64 years and estimates of the increment in lost earnings attributable to the AORC among persons of these ages. A method developed by the investigators in an earlier study concerning all forms of musculoskeletal disease (7) was used to estimate the increment in medical care expenditures attributable to AORC. An expansion of that method was used to estimate the increment in earnings losses. SUBJECTS AND METHODS Data source. Data from the MEPS, a joint endeavor of the Agency for Healthcare Research and Quality and the National Center for Health Statistics, was used in the present study. MEPS is designed to provide data on health care use, medical care expenditures, sources of payment, and insurance coverage for a representative sample of the civilian noninstitutionalized population of all ages in the US. MEPS also tracks the employment status and earnings of that sample. The full MEPS data include survey responses from this sample of households (MEPS-H), information from their providers about their medical conditions, information from their healthcare plans about the specifics of these plans, and a separate sample of nursing home residents (38). The MEPS-H sample is derived from the previous year’s National Health Interview Survey respondents, who are, in turn, derived from a clustered, random sample of the civilian noninstitutionalized population, with an oversample of African Americans and Hispanics. Data for nursing home residents were not included in the analyses. In the present study, we used data from the 1997 MEPS-H, which included 14,147 households with 34,551 persons (39). MEPS-H data are collected through 6 rounds of interviews over a 2.5-year period. The first 3 interviews, which cover expenditures, employment, and earnings over an entire year, provided the data used in this analysis. The interviews are used to collect information on health and functional status, health care utilization and expenditures, employment and earnings, as well as basic demographic information. The health status section elicits data on the specific medical conditions each respondent has. The information from the self-report of the respondents is then used to code conditions to 3-digit levels using the International Classification of Diseases, Ninth Revision (ICD-9) system. The utilization and expenditures and the employment and earnings sections elicit information about the previous 4 months in order to improve the reliability of responses. In the MEPS, expenditure data derive from a combination of the MEPS-H interviews and information obtained from providers. Expenditures in MEPS are defined as the actual expenditures for the medical care services used, regardless of the source of payment (38–40). In contrast to studies of the costs of illness in which costs are tabulated even if uncompensated care is provided, on the assumption that resources are being consumed regardless of payment, in studies of medical care expenditures, the analyst is studying the actual exchange of money. Because MEPS is based on expenditures rather than costs, there are health care encounters enumerated in the survey for which no expenditures were made. In an entirely fee-for-service system, one could track all expenditures among respondents. However, in many forms of managed care, charges are not rendered when services are provided and, hence, there are no expenditures specific to the medical care encounters. Accordingly, in such instances, MEPS-H imputes expenditures based on the charges incurred within the fee-for-service sector for similar services provided to similar individuals. Analyses. Data partitions. We first partitioned the 1997 MEPS-H data into the following chronic condition groups, based on ICD-9 codes: persons with only AORC (ICD-9 codes 274, 354, 390, 391, 443, 446, 710–716, 719–721, and 725–729), persons with both AORC and non-AORC chronic conditions, persons with 1 non-AORC condition, persons with ⱖ2 nonAORC chronic conditions, and persons with no chronic conditions. The definition of AORC is based on a standard developed by the National Arthritis Data Workgroup (6). Because MEPS-H data used only 3-digit ICD-9 codes, we could only use the Workgroup’s 3-digit ICD-9 codes. For a few Workgroup ICD-9 codes for which the presence of a fourth digit is normally necessary, we used the corresponding 3-digit ICD-9 code if we thought it would be composed primarily of the Workgroup’s targeted condition. Chronic conditions in general are defined by the protocol devised by Hoffman et al (41), which was designed to provide a conservative estimate of prevalence. Of all 34,551 MEPS respondents, 4,776 met the criteria for any form of AORC. General considerations. Because MEPS-H is based on a 2-stage cluster sample rather than a simple random sample of the noninstitutionalized population, it is necessary to weight the data to make inferences for the US population. In MEPS-H, the sampling weights also take into account nonresponse in the households targeted for inclusion and attrition among respondents after completion of the first interview (42). We used SUDAAN software to account for the cluster sampling design in all analyses requiring the calculation of the standard error of estimates (43). Description of expenditures. We began by enumerating medical care expenditures on behalf of persons of all ages by condition group and by category of health services, as well as the distribution of total health care expenditures. In the foregoing analysis, we tabulated all expenditures among persons in the condition groups, regardless of whether the condi- MEDICAL CARE EXPENDITURES AND EARNINGS LOSSES IN ARTHRITIS, 1997 tion in question accounted for the expenditures. In the tables for these analyses, we indicate the estimates that had low statistical reliability, that is, a relative standard error ⬎30%. Analysis of increment in health care expenditures. In order to assess the incremental contribution of AORC to health care expenditures, we estimated a series of regressions separately for adults with and without AORC. We then simulated the expected level of expenditures for the AORC group as if they did not have the AORC by applying the parameter estimates derived from the group without the condition to the data for the group with the condition. The increment was then calculated as the difference between these 2 values (44). To make these calculations with respect to expenditures for ambulatory care, inpatient care, prescription drugs, and residual expenditures, we followed the 2-stage method outlined by Duan et al (45). This method accounts for the distribution of health expenditures in which many persons have no health expenditures or relatively low expenditures, while a small proportion has very high expenditures. In this method, one first uses a logistic regression to estimate the probability that an individual has any expenditures and then uses ordinary least-squares regression on the logarithm of the expenditure to estimate the level of expenditures among those with strictly positive expenditures. We estimated the incremental contribution of AORC to total expenditures by a 4-stage model, which has also been described by Duan et al (45). The first stage uses a logistic regression to predict the probability of any medical expenditures. The second stage uses a separate logistic regression to predict the probability of any hospital expenditures given the presence of medical expenditures. The third stage uses an ordinary least-squares regression to predict the logarithm of the level of total costs among persons without hospitalizations, and the fourth stage uses ordinary least-squares regression to predict the logarithm of total costs among persons with hospitalizations. Total expenditures include ambulatory and inpatient care, prescription drugs, and a residual category that comprises home healthcare, vision aids, dental visits, and medical devices. For both the 2-stage and 4-stage models, the log transformation is used in the ordinary least-squares regressions to account for the skewed distribution of expenditures. The resulting estimate, when transformed back to the original units by exponentiation, is biased downward (that is, the expected value of the estimate is lower than the expected value of the population mean). Duan (46) has shown that a simple adjustment called the smearing estimate can be used to adjust for this bias. The smearing coefficient is the mean of the exponentiated residuals from the regression on the logtransformed variable. All the regression models included independent variables for the presence or absence of AORC and the following 9 other high-cost chronic conditions: hypertension, other forms of heart disease, pulmonary disease, stroke, other neurologic conditions, diabetes, cancer, mental illness, and musculoskeletal conditions other than AORC. In addition to the indicator variables for conditions, we included variables for age categories (45–64 years and ⱖ65 years, with 18–44 years as the reference category), sex (female versus male), race (white versus nonwhite), ethnicity (Hispanic 2319 versus non-Hispanic), marital status (single, widowed, separated, and divorced, with currently married as the reference category), level of formal education (high school graduate, some college, college graduate, and some graduate school, with less than high school as the reference category), and health insurance status (only public insurance or no insurance, with those having any private insurance as the reference category). For the 2-stage model, the final estimated expenditure for each individual was obtained by exponentiating the predicted logarithm of expenditures, multiplying that value by the smearing coefficient, and multiplying that, in turn, by the predicted probability of a positive expenditure from the logistic regression. The calculations for the 4-stage model were similar. All of these analyses were limited to the 22,376 MEPS respondents who were 18 years of age or older and who had complete data for all covariates used in the multistage models, of whom 4,457 met the study criteria for any form of AORC. Calculation of attributable fractions. Attributable fractions, which are the proportion of all health care expenditures in the entire population that would not occur if no one had AORC, were calculated by dividing the increment in total medical care expenditures attributable to AORC by the total medical care expenditures for the entire population. We cross-validated the estimates of attributable fractions in each of 12 cells defined by age and sex. In the cross-validation studies, we estimated 10 different values for the attributable fraction within each cell. We estimated the models on ninetenths of the data and used that model to calculate the attributable fraction in the other one-tenth of the data. The largest range of estimates within a cell was 8 percentage points, but in most of the cells, the range was ⱕ3 percentage points. Estimation of employment and earnings losses. The estimations of earnings losses were limited to persons ages 18–64 years with a history of ever being employed as recorded in MEPS. Earnings losses were estimated only for persons who were ages 18–64 years, both for the substantive reason that labor force participation rates decline precipitously at age 65 years, as individuals reach the earliest age of eligibility for full Social Security retirement benefits, and for the methodologic reason that estimates among persons ages ⱖ65 years, based on a relatively small sample size, would not be reliable. To estimate earnings losses, we tabulated the actual employment status of persons with and without AORC who were between the ages of 18 years and 64 years, and, among those who were employed, the hours of work. The absolute level of lost wages was then calculated as the sum of 2 products. The first product was the cost of lost wages among those who were not working at all. This product is equal to the difference in the employment rates of persons with and without AORC times the number of persons with AORC times the mean wage among all employed persons with AORC. The second product was the cost of lost wages among those who continued to work. This product is equal to the difference in the hours of work per year by employed persons with and without AORC divided by the mean hours of work of all persons (yielding the percentage reduction in hours) times the number of persons with AORC times the mean wage among all employed persons with such conditions. Analysis of increment in earnings losses. We then estimated the increment of earnings losses in a manner analogous 2320 YELIN ET AL Table 1. Noninstitutionalized population of all ages, by arthritic condition status, US, 1997 Chronic condition status No. of persons (millions) % of total population % with arthritic conditions All arthritic conditions Arthritic conditions only Arthritic and nonarthritic conditions All nonarthritic conditions 1 nonarthritic condition only ⱖ2 nonarthritic conditions No chronic/comorbid conditions 38.423 5.635 32.788 131.749 58.609 73.14 101.106 14.2 2.1 12.1 48.6 21.6 27.0 37.3 100.0 14.7 85.3 to the estimate of the increment in expenditures. Specifically, we began by estimating separate logistic regressions of the probability of employment among persons with and without AORC and then estimated separate ordinary least-squares regressions of the level of annual earnings among the members of the 2 groups with earnings. The regressions included the following independent variables: age (35–44 years, 45–54 years, and 55–64 years, with 18–34 years as a reference), sex, race (white versus nonwhite), ethnicity (Hispanic versus nonHispanic), marital status (single, widowed, separated, and divorced, with married as the reference category), education level (high school graduate, some college, college graduate, and some graduate school, with less than high school as the reference category), and indicator variables for the same comorbid conditions as described in the direct cost analyses above. After the regressions were completed, we applied the parameter estimates derived from regressions on persons without AORC to the data from those with such conditions to simulate the expected level of earnings among such persons if they did not have AORC. The increment in earnings losses was then calculated as the difference between the AORC values and simulated non-AORC values. In the ordinary leastsquares regressions, the natural logarithm of earnings was used; the results were then transformed back to absolute dollar amounts using separate smearing estimates calculated among persons with and without AORC in a manner analogous to the estimations of medical care expenditures described above. Sensitivity analyses. In order to assess the extent to which estimates of the prevalence and economic impact of AORC were sensitive to the inclusion of potentially problematic ICD-9 codes within the AORC rubric, we eliminated the following ICD-9 codes from the rubric: codes 354 (mononeuritis of the upper extremity; 85% of 255 sample cases were carpal tunnel syndrome in later analyses), 390 and 391 (acute rheumatic fever), and 443 (other peripheral vascular disease; 39% of 25 sample cases were Raynaud’s syndrome in later analyses). The elimination of these conditions from the AORC rubric reduced the prevalence of AORC by 3.1% (from 38.4 to 37.2 million) and reduced aggregate medical expenditures by 2.4% (from $186.9 billion to $182.4 billion), while increasing the average expenditure level by 0.8% (from $4,865 to $4,902). Because the results were relatively insensitive to the elimination of these codes and because relying on the definition of AORC from the National Arthritis Data Workgroup uses a standard that permits comparison with other studies (6,47), we report below the estimates based only on that definition. We also calculated the sensitivity of the estimate of the increment in direct costs to the inclusion of different forms of several of the independent variables, including age (inclusion of age and age-squared), education (truncation of education at “college graduate”), and comorbidity (inclusion of count of all chronic conditions, rather than disease-specific indicator variables). In no instance did the estimate of the increment in direct costs vary by more than 7.1%. In a similar manner, we calculated the sensitivity of the estimate of the increment in direct costs to the form of the comorbidity variable. Replacement of the chronic condition indicator variables with the count of chronic conditions increased the estimate of the increment in indirect costs by 14.6%. RESULTS Medical care expenditures. Distribution of conditions. In 1997, 38.423 million persons (14.2% of the civilian noninstitutionalized population) reported any form of AORC. Of those, 32.788 million (12.1% of the entire population and 85.3% of all persons with AORC) also reported having 1 or more non-AORC conditions (Table 1). In that same year, 131.749 million persons (48.6%) reported non-AORC conditions in the absence of AORC and another 101.106 million (37.3%) reported no chronic conditions whatsoever (all told, there were 232.855 million persons without AORC in 1997). A higher proportion of persons with AORC than without AORC were women and white, while a slightly smaller percentage were of Hispanic origin. Such persons were also ⬎2 decades older on average, and reflecting this, they were more likely to be widowed, separated, or divorced and less likely to never have been married (data on characteristics of persons with and without AORC not presented in tables). Distribution of expenditures. The 38.423 million persons with any form of AORC had average total medical care expenditures in 1997 of $4,865 ($1,074 among the 5.635 million without non-AORC, $5,516 among the 32.788 million with such conditions). This is much higher than the average of $2,397 among persons with non-AORC conditions or than the $500 among those without chronic conditions (Table 2). The 3 largest MEDICAL CARE EXPENDITURES AND EARNINGS LOSSES IN ARTHRITIS, 1997 2321 Table 2. Health care expenditures (by type) of the noninstitutionalized population of all ages, by arthritic condition status, US, 1997 Office-based and hospital outpatient, mean dollars (row %) Condition status Arthritic conditions Arthritic conditions only Arthritic and nonarthritic conditions Nonarthritic conditions 1 nonarthritic condition ⱖ2 arthritic conditions No chronic conditions All Persons ER and inpatient, mean dollars (row %) Inpatient¶ Home healthcare Prescriptions filled Other# Total expenditures, mean dollars 103 (2) 35 (3) 115 (2) 1,905 (39) 320 (30)** 2,177 (39) 398 (8) 15 (1)** 464 (8) 687 (14) 81 (8) 792 (14) 377 (8) 210 (20) 406 (7) 4,865 1,074 5,516 82 (3) 59 (5) 101 (3) 30 (6) 66 (3) 951 (40) 379 (33) 1,409 (41) 149 (30) 787 (39) 99 (4) 29 (3) 155 (5) 9 (2)** 108 (5) 330 (14) 121 (11) 498 (15) 23 (5) 266 (13) 248 (10) 188 (16) 296 (9) 137 (27) 225 (11) 2,397 1,142 3,403 500 2,039 Physicianrelated* Other† Total‡ ER§ 967 (20) 290 (27) 1,083 (20) 428 (9) 123 (11) 480 (9) 1,394 (29) 413 (38) 1,563 (28) 687 (29) 365 (32) 944 (28) 151 (30) 587 (29) 488 (20) 269 (24) 664 (20) 115 (23) 417 (20) 198 (8) 96 (8) 280 (8) 36 (7) 170 (8) Other, mean dollars (row %) * Includes office-based visits to physicians, physician outpatient expenses billed separately, and associated outpatient facility expenses. † Includes all nonphysician office-based and outpatient visits and associated facility expenses and unknown office-based and outpatient expenses. ‡ Includes office-based visits to physicians, physician outpatient expenses billed separately, associated outpatient facility expenses, all nonphysician office-based and outpatient visits and associated facility expenses, and unknown office-based and outpatient expenses. § Emergency room (ER) includes facility and separately billed provider expenses. ¶ Includes facility and separately billed provider expenses. # Includes dental visits and other medical supplies and equipment. ** Estimate has low statistical reliability (relative standard error ⬎30%). components were inpatient costs (39%), ambulatory care (29%), and prescription drugs (14%). Overall, in 1997, persons with any form of AORC accounted for $186.9 billion in expenditures, the equivalent of ⬃2.3% of the gross domestic product for the US in that year, with all but $6.1 billion accounted for by persons reporting AORC as well as non-AORC (Table 3). The 38.423 million persons with any form of AORC represent 14.2% of the US population but account for 34% of all medical care expenditures in the MEPS. Increment in expenditures and attributable fractions. The $186.9 billion in expenditures among persons with AORC are only partially attributable to the AORC; a significant proportion is due to the other chronic conditions such persons have, as well as to their acute care and wellness care. In Table 4, we present the results of our analysis of the proportion of various kinds of expenditures attributable to AORC among persons ages ⱖ18 years. Overall, we estimated that all adults with AORC had an increment in total expenditures of $1,391 per person beyond what would be expected of similar persons without such conditions. This is equivalent to a population attributable fraction of 10%, meaning that eliminating AORC would reduce total health care expenditures by 10%. The attributable fraction rose with age, from 6% among persons ages 18–44 years, to 12% among those ages 45–64 years, and to 18% among those ages ⱖ65 years, as befits a set of conditions associated with aging. The attributable fraction associated with AORC was larger for men of all ages (11%) than for women of all ages (9%). The attributable fraction among men ages Table 3. Distribution of total health expenditures of the noninstitutionalized population of all ages, by arthritic condition status, US, 1997 Condition status Minimum 5% 25% Median 75% 95% Maximum Mean Total, billion dollars (column %) Arthritic conditions Arthritic conditions only Arthritic and nonmusculoskeletal conditions Nonarthritic conditions 1 nonarthritic condition ⱖ2 arthritic conditions No chronic conditions All persons 0 0 0 60 0 154 575 98 799 1,644 349 2,004 4,306 910 5,114 20,041 3,703 22,332 191,297 65,316 191,297 4,865 1,074 5,516 186.9 (34) 6.1 (1) 180.9 (33) 0 0 0 0 0 0 0 78 0 0 223 108 424 0 83 635 337 1,068 89 394 1,847 854 2,870 319 1,415 9,644 4,494 13,735 1,958 8,478 351,156 337,262 351,156 136,109 351,156 2,397 1,142 3,403 500 2,039 315.8 (57) 66.9 (12) 248.9 (45) 50.5 (9) 553.2 (100) Dollars 2,783 722 1,044 1,017 1,666 471 673 522 12,187 6,482 3,464 2,241 10,248 5,717 2,991 1,540 94,303 53,934 26,308 14,062 102,026 55,063 28,207 18,756 196,329 108,997 54,515 32,817 Total Inpatient Prescription drugs Attributable fractions by expenditure type Residual Total 14,897 4,599 5,644 4,654 21,862 6,075 7,813 7,973 36,759 10,674 13,458 12,628 514 391 599 539 649 491 757 680 598 453 698 627 629 543 665 715 882 771 932 993 768 668 812 867 13 6 19 25 16 7 23 29 15 7 21 28 795 734 535 1,066 150 196 ⫺22 310 401 407 195 605 1,050 899 927 1,248 1,045 902 942 1,248 1,042 895 930 1,241 12 7 12 28 3 2 ⫺1 11 7 4 5 19 121 87 186 108 107 86 194 74 112 86 190 85 369 236 388 456 522 341 557 654 456 295 483 567 5 3 10 8 4 3 10 5 5 3 10 6 165 140 172 168 181 154 192 182 176 149 185 177 387 323 433 462 444 373 496 530 418 350 467 498 7 4 9 12 9 5 11 15 8 4 10 14 1,762 1,788 1,695 1,962 1,163 1,268 1,182 1,278 1,391 1,467 1,377 1,539 2,465 2,051 2,474 2,931 2,886 2,405 2,949 3,456 2,689 2,238 2,728 3,209 11 7 15 22 9 6 11 16 10 6 12 18 IncreIncreIncreIncreIncreWith ment, Adjusted Attributable ment, Adjusted Attributable ment, Adjusted Attributable ment, Adjusted Attributable ment, Adjusted Attributable arthritis $† cost, $‡ fraction, % $† cost, $‡ fraction, % $† cost, $‡ fraction, % $† cost, $‡ fraction, % $† cost, $‡ fraction, % Outpatient * MEPS ⫽ Medical Expenditures Panel Survey. † Average increment per person with the condition. ‡ Adjusted cost per person whether or not they have the condition. 4,449 1,193 1,717 1,539 22,435 12,199 6,455 3,781 Total With arthritis MEPS respondents* Estimated population (thousands) Increment in expenditures and attributable fractions for arthritic conditions for the US population ages ⱖ18 years, 1997, by age and sex group and by expenditure type Both sexes All ages 18–44 years 45–64 years ⱖ65 years Females All ages 18–44 years 45–64 years ⱖ65 years Males All ages 18–44 years 45–64 years ⱖ65 years Sex, age Table 4. MEDICAL CARE EXPENDITURES AND EARNINGS LOSSES IN ARTHRITIS, 1997 Table 5. Individual and aggregate employment rates of persons with arthritic conditions ages 18–64 years, raw differences and increments attributable to the arthritic conditions, US, 1997 2323 Table 6. Individual and aggregate earnings losses of persons with arthritic conditions ages 18–64 years, raw differences and increments attributable to the arthritic conditions, US, 1997 Employment Value Earnings Dollars Employment rate of persons without arthritic conditions, raw, % Employment rate of persons with arthritic conditions, raw, % Individual employment rate gap, raw, % Aggregate employment gap, raw, no. of persons* Individual incremental employment rate gap attributable to arthritic conditions, % Aggregate incremental employment gap attributable to arthritic conditions, no. of persons* 91.0 Aggregate lost earnings due to lower employment rates among persons with arthritic conditions, raw, billions Per capita earnings gap due to lower employment rates among persons with arthritic conditions, raw Aggregate lost earnings among employed persons with arthritic conditions, raw, billions Per capita lost earnings among employed persons with arthritic conditions, raw Net aggregate earnings gap, raw, billions Net per capita earnings gap, raw Aggregate incremental net earnings gap attributable to arthritic conditions, billions Per capita incremental net earnings gap attributable to arthritic conditions 73.2 79.6 11.4 2,542,603 4.0 889,948 * Represents the number of affected individuals (⬃22.2 million) times individual employment gap. ⱖ65 years was particularly pronounced (22%), suggesting that the presence of AORC among men of these ages adds dramatically to their overall medical care expenditures. Among all persons with AORC, the attributable fraction associated with this group of conditions was highest for outpatient services (15%) and lowest for prescription medications (5%). The attributable fractions associated with inpatient services and the residual category were 7% and 8%, respectively. Although the attributable fraction for inpatient services was only 7% overall, it was much higher for men, regardless of age (12%), particularly men who were ⱖ65 years old (28%). Presumably, the relatively high attributable fraction among men these ages reflects the growing prevalence of total joint replacement surgery. Another way to visualize the magnitude of the costs attributable to AORC is by estimating the aggregate impact of the mean increment in expenditures. Thus, the average increment of $1,391 among each of the 36.759 million persons ages ⱖ18 years with AORC aggregated to $51.1 billion, or ⬃0.6% of the gross domestic product for 1997. Employment/earnings losses. As a result of their markedly lower employment rates (91.0% versus 79.6%; difference 11.4%), ⬃2.5 million persons with AORC who were not then employed would have been working if they had had the same employment rate as persons without such conditions (Table 5). However, after controlling for differences in demographic characteristics and comorbidity between the 2 groups, the incremental employment gap was 4.0%, or ⬃0.9 million persons. 3,293 9.2 519 82.4 3,812 35.1 1,579 The raw wage loss costs associated with the forgone earnings of the ⬃2.5 million persons with AORC who would have been working but were not was $73.2 billion ($3,293 per working-age person) (Table 6). The wage losses of the then currently employed persons with AORC that were associated with their lower earnings as compared with persons without such conditions was $9.2 billion ($519 per working-age person with AORC), leaving an overall net loss associated with Table 7. Factors associated with employment of persons with arthritic conditions ages 18–64 years, controlling for demographic characteristics and 9 comorbid conditions, US, 1997 Characteristic Female Age 18–34 years 35–44 years 45–54 years 55–64 years Race Nonwhite White Ethnicity Non-Hispanic Hispanic Education Less than high school High school graduate Some college College graduate Some graduate school Marital status Married Widowed, separated, divorced Never married Odds ratio 95% confidence interval for odds ratio 0.58 0.44–0.75 Reference 0.55 0.37 0.12 0.33–0.94 0.22–0.64 0.07–0.20 Reference 1.65 1.17–2.32 Reference 0.87 0.58–1.29 Reference 1.9 2.85 3.72 5.5 1.41–2.56 1.94–4.18 2.29–6.04 3.09–9.82 Reference 1.05 1.07 0.79–1.39 0.66–1.72 2324 YELIN ET AL AORC of $82.4 billion ($3,812 per working-age person with AORC). After controlling for demographic characteristics and comorbidity, the increment in the net individual earnings gap attributable to AORC averaged $1,579, for a total of $35.1 billion. Demographic factors associated with employment status. Because almost all of the earnings losses associated with AORC occurred among those who were not working, stopping work loss is essential to containing the earnings losses associated with this group of conditions. Table 7 shows the demographic factors associated with employment status among persons ages 18–64 years with AORC and a work history. The characteristics associated with a significantly reduced odds of employment included female sex (odds ratio 0.58; 95% confidence interval 0.44–0.75) and increments of age, while increasing increments of education and being white were associated with an increased odds of employment. DISCUSSION We performed 2 kinds of estimates of both the medical care expenditures and earnings losses associated with arthritis and other rheumatic conditions. In the first estimate of medical care expenditures, we recorded the magnitude and distribution of all medical care expenditures among the 38.423 million persons with AORC, and we found that such persons experienced mean total expenditures of $4,865, or $186.9 billion overall. Hospital admissions accounted for 39% of these expenditures; ambulatory care (29%) and prescription drugs (14%) were the next 2 largest components. Although average expenditures of $4,865 among persons with AORC were substantial, most individuals avoided such high levels of expenditures. Among persons with AORC alone, the average expenditures were only $1,074 and, even at the 95th percentile, were relatively small ($3,703). Among persons with both AORC and non-AORC chronic conditions, median expenditures were only $2,004 and only rose to $5,114 at the 75th percentile. Nevertheless, expenditures incurred on behalf of persons with AORC represent a substantial drain on the nation’s economy. In 1995, the National Arthritis Data Workgroup reported that both direct and indirect costs of all forms of musculoskeletal conditions, including AORC, for 1988 was 2.5% of the gross national product of the US (6). In the present study, however, we found that total medical expenditures for AORC alone in 1997 were equivalent to ⬃2.3% of the gross domestic product for that year (48), and would suggest that both the accounting methods used in the MEPS and the aging of the population may have contributed to the increase in the estimated impact of these conditions. In the second set of analyses of medical care expenditures, we estimated the proportion of total expenditures attributable to AORC among persons ages ⱖ18 years. Overall, AORC was associated with an attributable fraction of 10% for medical expenditures, or $1,391 per person with AORC, which aggregates to $51.1 billion, the equivalent of about 0.6% of the gross domestic product in 1997. Since in a recession, by definition, the economy retrenches by ⱖ1% of the gross domestic product for at least 2 consecutive quarters, the increment in medical care expenditures attributable to AORC has an impact slightly less than that of a small recession, but unlike a recession, it occurs in perpetuity. In the unadjusted analysis, persons with AORC ages 18–64 years who were not working were estimated to be responsible for $73.2 billion in earnings losses, while those who were employed were responsible for losses of $9.2 billion, resulting in an estimate of net earnings losses of $82.4 billion. The finding that those who were not employed were responsible for the bulk of the lost income is consistent with clinical studies in specific rheumatic conditions that have shown a total loss of employment to be more common than a reduction in hours or a change in work activities (49). After controlling for demographic characteristics and comorbidity, the incremental value of the aggregate net earnings gap attributable to AORC was ⬃$35.1 billion, or $1,579 per person with AORC ages 18–64 years. In a recent estimate of the economic burden of musculoskeletal conditions, Rice and colleagues (1) calculated that the direct costs of AORC in 1995 were $21.7 billion, or about $22.8 billion in 1997 dollars. In the present study, we estimated that the increment in medical care expenditures attributable to AORC was $51.1 billion in 1997 dollars. Thus, the direct accounting of medical care expenditures made possible by the development of the MEPS may have resulted in a substantial increase in the estimate of the direct costs attributable to AORC. In the same study by Rice, it was estimated that the indirect costs of AORC among persons ages 18–64 years were $60.8 billion in 1995, or about $64.0 billion in 1997 dollars. Thus, the estimate of the earnings losses attributable to AORC in the current study ($35.1 billion) is considerably smaller than the estimate made by Rice and colleagues. The ability to control for differences in characteristics other than the AORC between persons with and without AORC may MEDICAL CARE EXPENDITURES AND EARNINGS LOSSES IN ARTHRITIS, 1997 account for the lower estimate in the present study. The estimate of earnings losses, however, is within the range of estimates reported by Dunlop and colleagues (8). Some AORC conditions were excluded from the analysis because 4-digit ICD-9 data that would identify these conditions were not available. This underestimation of cost was likely balanced by the presence of some non-AORC conditions that were included in the analysis because 3-digit ICD-9 codes are less specific. There are 4 reasons why medical care expenditures may have been underestimated in this study. First, the study sample was limited to the US civilian noninstitutionalized population and, thus, expenditures by institutionalized and military populations were not included. Second, the MEPS measures only medical expenditures and does not capture services for which there is no exchange of money. Third, we did not include in this analysis expenditures for complementary and alternative medicine, which is becoming an increasingly common treatment modality, especially among individuals with AORC (50,51). Fourth, the data from the current study precede the introduction of biologics for the treatment of rheumatoid arthritis, more-aggressive testing for the presence and treatment of osteoporosis, and the growth of the coxibs and/or the use of protonpump inhibitors with traditional nonsteroidal antiinflammatory drugs. However, it should be pointed out that the relatively low prevalence of RA means that the introduction of biologics would add, at most, several billion dollars to the estimate of the aggregate increment of ⬃$51 billion that was attributable to all forms of AORC. Moreover, although it is becoming increasingly common to prescribe bone densitometry for women at risk of developing osteoporosis, the added costs of such testing would be balanced by the reduction in the number of women receiving estrogen replacement therapy. Nevertheless, all of these developments have undoubtedly increased the incremental costs associated with AORC since 1997, making the estimates of the present study relatively conservative. There are some limitations that would apply to the estimates of the increment in direct and indirect costs. The MEPS survey provides no information with which to estimate whether persons with AORC differ from the rest of the population in their level of perseverance in the face of medical symptoms with respect to the kinds of medical care services procured or their willingness to stay at work. These 1997 cost estimates, and the 1996 expenditures presented by Dunlop et al (8), are the first 2325 national population–based AORC cost estimates based on directly observed individual-level expenditure and employment data. This study is the first to provide attributable fractions for medical care expenditures due to AORC. These expenditure estimates are an essential component of the ongoing effort by the Centers for Disease Control and Prevention to characterize the burden of AORC in the US for a single year, 1997. Thus far, the impact of AORC on hospitals (52) and ambulatory care (47,53) has been reported, and these data will soon extend to characterizing the burden of AORC in nursing homes and assessing mortality rates among persons with AORC. The results reported here indicate that a larger proportion of the indirect costs (41% of all $82.4 billion in indirect costs, or $35.1 billion) than of the direct costs (29% of all $186.9 billion, or $51.5 billion) of AORC is attributable to the AORC. This finding from a population-based study of persons with all forms of AORC is consistent with studies of persons with discrete rheumatic conditions (19) and indicates that the prevention of work loss and the resultant earnings losses should be central to public health policy aimed at reducing the impact of AORC, even as the cost of medical care continues to garner a disproportionate amount of the public’s attention. REFERENCES 1. Rice D. The economic burden of musculoskeletal conditions, US, 1995. In: Praemer A, Furner S, Rice D, editors. Musculoskeletal conditions in the US. Rosemont (IL): American Academy of Orthopedic Surgeons; 1999. p. 139–62. 2. Rice D. Estimating the cost of illness. Hyattsville (MD): National Center for Health Statistics; 1966. Health economic series. No. 6. 3. Cooper B, Rice D. The economic cost of illness revisited. Soc Secur Bull 1979;39:21–35. 4. Rice D, Hodgson T, Kopstein A. The economic costs of illness: a replication and update. Health Care Finance Rev 1985;7:61–80. 5. Wigle D, Mao Y, Wong T, Lane R. Economic burden of illness in Canada, 1986. Chronic Dis Can 1991;12:1–27. 6. Yelin E, Callahan L. The economic cost and social and psychological impact of musculoskeletal conditions. Arthritis Rheum 1995;38:1351–62. 7. Yelin E, Herrndorf A, Trupin L, Sonneborn D. A national study of medical care expenditures for musculoskeletal conditions: the impact of health insurance and managed care. Arthritis Rheum 2001;44:1160–9. 8. Dunlop D, Manheim L, Yelin E, Song J, Chang R. The costs of arthritis. Arthritis Rheum 2003;49:101–13. 9. Badley E. The economic burden of musculoskeletal disorders in Canada is similar to that for cancer, and may be higher. J Rheumatol 1995;22:204–6. 10. Access Economics. Cost of arthritis to the Australian community. The Arthritis Foundation Submission to the Industry Commission Inquiry into Charitable Organizations. Canberra: Arthritis Foundation of Australia; 1994. p. 14–22. 2326 11. Freedman D. Arthritis: the painful challenge. In: Searle Social Research Fellowship Report. Chicago: GD Searle; 1989. 12. March L, Bachmeier C. Economics of osteoarthritis: a global perspective. Baillieres Clin Rheumatol 1997;11:817–34. 13. Levy E, Ferme A, Perocheau D, Bono I. Les couts socioeconomiques de l’arthrose en France. Rev Rhum 1993;60:63s–7s. 14. Jonsson B, Rehnberg C, Borgquist L, Larsson S. Locomotion status and costs in destructive rheumatoid arthritis. Acta Orthop Scand 1992;63:207–12. 15. Eberhardt K, Larsson B, Nived K. Early rheumatoid arthritis: some social, economical, and psychological aspects. Scand J Rheumatol 1993;22:119–23. 16. Allaire S, Prashker M, Meenan R. The costs of rheumatoid arthritis. Pharmacoeconomics 1994;6:513–22. 17. Prashker M, Meenan R. The total costs of drug therapy for rheumatoid arthritis: a model based on costs of drug, monitoring, and toxicity. Arthritis Rheum 1995;38:318–25. 18. Maclean K, Knight K, Shekelle P, Paulus H, Brook R. Costs attributable to arthritis: implications of costs [abstract]. Arthritis Rheum 1996;39 Suppl 9:S317. 19. Yelin E. The costs of rheumatoid arthritis: absolute, incremental, and marginal estimates. J Rheumatol 1996;23 Suppl 44:47–51. 20. Lanes S, Lanza L, Radensky P, Youd R, Meenan R, Walker A, et al. Resource utilization and cost of care for rheumatoid arthritis and osteoarthritis in a managed care setting. Arthritis Rheum 1997;40:1475–81. 21. Clarke A, Zowall H, Levinton C, Assimakopoulos H, Sibley JT, Saga M, et al. Direct and indirect medical costs incurred by Canadian patients with rheumatoid arthritis: a 12 year study. J Rheumatol 1997;24:1051–60. 22. Sangha O, Stucki G. Economic impact of rheumatologic disorders. Curr Opin Rheumatol 1997;9:102–5. 23. Jaarsveld C, Jacobs J, Schrijvers A, Heurkens A, Haanen H, Bijlsma J. Direct cost of rheumatoid arthritis during the first six years: a cost of illness study. Br J Rheumatol 1998;37:837–47. 24. Clarke A, Levinton C, Joseph L, Penrod J, Zowall H, Sibley J, et al. Predicting the short-term direct medical costs incurred by patients with rheumatoid arthritis. J Rheumatol 1999;26:1068–75. 25. Yelin E, Wanke L. An assessment of the annual and long-term direct costs of rheumatoid arthritis: the impact of poor function and functional decline. Arthritis Rheum 1999;42:1209–18. 26. Liang M, Larson M, Thompson M, Eaton H, McNamera E, Katz R, et al. Costs and outcomes in rheumatoid arthritis and osteoarthritis. Arthritis Rheum 1984;27:522–9. 27. Thompson M, Read J, Liang M. Feasibility of willingness to pay measurements in chronic arthritis. Med Decis Making 1984;4: 195–212. 28. Gabriel S, Crowson C, Campion M, O’Fallon W. Direct medical costs unique to people with arthritis. J Rheumatol 1996;24:719–25. 29. Rothfuss J, Mau W, Zeidler H, Brenner M. Socioeconomic evaluation of rheumatoid arthritis and osteoarthritis: a literature review. Semin Arthritis Rheum 1997;26:771–9. 30. Allaire S, DeNardo B, Szer I, Meenan R, Schaller J. The economic impacts of juvenile rheumatoid arthritis. J Rheumatol 1992;19: 952–5. 31. Lubeck D, Brown B, Holman H. Chronic disease and health system performance: care of osteoarthritis across three health services. Med Care 1985;23:266–77. 32. Gabriel S, Crowson C, O’Fallon W. Costs of osteoarthritis: estimates from a geographically defined population. J Rheumatol 1995;22 Suppl 43:23–5. 33. Yelin E. The economic impact of osteoarthritis. In: Baker J, Brandt K, editors. Reappraisal of the management of patients with osteoarthritis. Springfield (NJ): Scientific Therapeutics Information; 1993. p. 10–3. YELIN ET AL 34. Clarke A, Esdaile J, Bloch D, Lacaille D, Danoff D, Fries J. A Canadian study of the total medical costs for patients with systemic lupus erythematosus and the predictors of costs. Arthritis Rheum 1993;36:1548–59. 35. Lacaille D, Clarke A, Bloch D, Danoff D, Esdaile J. The impact of disease activity, treatment, and disease severity on short term costs of systemic lupus erythematosus. J Rheumatol 1994;21:448–53. 36. Gironimi G, Clarke A, Hamilton V, Danoff D, Bloch D, Fries J, et al. Why health care costs more in the US: comparing health care expenditures between systemic lupus erythematosus patients in Stanford and Montreal. Arthritis Rheum 1996;39:979–87. 37. Webster B, Snook S. The cost of compensable low back pain. J Occup Med 1990;32:13–5. 38. Cohen S. Sample design of the 1996 Medical Expenditure Panel Survey household component. Rockville, MD: Agency for Health Care Policy and Research; 1997. MEPS Methodology Report No. 2. AHCPR Publication No. 97-0027. 39. Cohen S. Sample design of the 1997 Medical Expenditure Panel Survey household component. Rockville, MD: Agency for Healthcare Research and Quality; 2000. MEPS Methodology Report No. 11. AHRQ Publication No. 01-0001. 40. Agency for Healthcare Quality and Research. MEPS household component 1996 full year use and expenditure data: public use documentation. Publication No. C6-C11. Rockville, MD: Agency for Healthcare Research and Quality; 1999. 41. Hoffman C, Rice D, Sung H. Persons with chronic conditions: their prevalence and costs. JAMA 1996;276:1473–9. 42. Cohen S, DiGaetano R, Goksel H. Estimation procedures in the 1996 Medical Expenditure Panel Survey household component. Rockville, MD: Agency for Health Care Policy and Research; 1999. MEPS Methodology Report No. 5. AHCPR Publication No. 99-0027. 43. Shah B, Barnwell B, Bieler G. Technical manual: statistical methods and algorithms used in SUDAAN release 7.0. Research Triangle Park (NC): Research Triangle Institute, 1996. 44. National Institute on Drug Abuse. The economic costs of alcohol and drug abuse in the United States, 1992. Appendix B. Bethesda, MD: National Institutes of Health; 1998. NIH Publication No. 98-4327. 45. Duan N, Manning W, Morris C, Newhouse J. A comparison of alternative models for the demand for medical care. J Bus Econ Stat 1983;1:115–26. 46. Duan N. Smearing estimate: a nonparametric retransformation method. J Am Stat Assoc 1983;78:605–10. 47. Impact of arthritis and other rheumatic conditions on the healthcare system—United States, 1997. MMWR 1999;48:349–53. 48. US Bureau of the Census. Statistical Abstract of the US, 2000. Washington (DC): Dept. of Commerce (US); 2000. p. 451. 49. Yelin E, Henke C, Epstein W. The work dynamics of the person with rheumatoid arthritis. Arthritis Rheum 1987;30:507–12. 50. Astin J, Pelletier K, Marie A, Haskell W. Complementary and alternative medicine use among elderly persons: one-year analysis of a Blue Shield Medicare supplement. J Gerontol A Biol Sci Med Sci 2000;55:M4–9. 51. Resch K, Hill S, Ernst E. Use of complementary therapies by individuals with ’arthritis’. Clin Rheumatol 1997;16:391–5. 52. Lethbridge-Cejku M, Helmick C, Popovic J. Hospitalizations for arthritis and other rheumatic conditions: data from the 1997 National Hospital Discharge Survey. Med Care 2003;41:1367–73. 53. Hootman J, Helmick C, Schappert S. Magnitude and characteristics of arthritis and other rheumatic conditions on ambulatory Medicare care visits, United States, 1997. Arthritis Rheum 2002; 47:571–81.