Kleinman, Nathan L. PhD; Cifaldi, Mary A. PhD, MSHA, RPh; Smeeding, James E. MBA, RPh; Shaw, James W. PhD, PharmD, MPH; Brook, Richard A. MS, MBA
Rheumatoid arthritis (RA) is the most common type of inflammatory arthritis, characterized by chronic destructive synovitis.1–4 Although the etiology is uncertain, it is likely that both genetic and environmental factors play a role in its cause.4 The physical effects, symptoms, and comorbidities associated with RA can impact a person's ability to work.5–7 Rheumatoid arthritis is also associated with shortened life expectancy and potentially reduced career length.8
The US National Arthritis Data Workgroup review of RA and other rheumatic conditions in the United States found that despite differences in methodologies, the RA prevalence estimates in developed countries were consistently 0.5% to 1% of the adult population.1 On the basis of data from Rochester, MN,4 and the Census Bureau, the workgroup estimated that in 2005, 1.3 million American adults (aged 18 years or older) or 0.6% of the population had RA.1
Cost-of-illness estimates have been derived, using self-reported information9 and a variety of US databases.4,10–13 The annual incremental costs in these studies ranged from $2379 to $10,716 per subject.4,10–13 This study aimed to assess the incremental direct and indirect costs and absence days associated with RA and project these incremental costs to the US employed population and the US population with RA.
This retrospective analysis used claims data from January 1, 2001, through June 30, 2010. Data were extracted from adjudicated health insurance claims within the Human Capital Management Services research reference database (RRDb). This database pools anonymous data from more than 20 US employers (including manufacturing, insurance, retail, transportation, telecommunication, health care, and pharmaceutical). The data are geographically dispersed through the United States and across a variety of health care plans, contain claims information from more than 900,000 employees, and have been used in other disease studies.14–19 The Human Capital Management Services database was similar to the US civilian noninstitutionalized population of employees in terms of sex (US women = 47.2%; RRDb women = 47.3%) and age (US average = 41.9 years; RRDb average = 39.4 years) in 2010.20 The data contain hours missed from work and absence payments made to employees (from payroll records and disability claims), compensation, racial information, and a variety of other job-related and demographic variables. Employees are classified as nonexempt (generally hourly) or exempt (generally salaried and not eligible to receive overtime pay). The database also contains information on spouses and eligible dependents. Confidentiality and anonymity of person-level data were maintained in accordance with the guidelines of the Health Insurance Portability and Accountability Act of 1996.
The RA cohort was identified from employees in the database with any primary, secondary, or tertiary International Classification of Diseases, Ninth Revision, codes for RA (714.xx = RA and other inflammatory polyarthropathies). Employees without any International Classification of Diseases, Ninth Revision, codes for RA were classified as potential controls.
Each subject was assigned an index date. For the RA cohort, the index date was the date of the first RA diagnosis found in the claims data. For employees without RA (controls), the index date was the average index date by company from the group of employees with RA. The study focused on the 12 months after each subject's index date. The employees were required to have medical and prescription drug coverage for the entire 12-month study period.
Several distinct categories of health-related absences were included. The analysis of each category was limited to the sub-cohort of employees eligible for that particular type of absence benefit.
Outcomes included direct costs, such as medical costs (based on plan-paid costs identified in the employee's medical claims) and prescription costs (based on plan-paid costs identified in the employee's prescription drug claims data). Components included health-related work absences (absenteeism or lost time measured in days) due to sick leave (SL), short-term disability (STD), long-term disability (LTD), workers' compensation (WC), and absence costs in the form of salary-replacement payments made to employees during health-related work absences due to SL, STD, LTD, and WC. Payments for WC also included direct medical payments made under the WC benefit. Unlike many studies that proxy absence at the half-day or daily level,9,10,12,21 the absence data in the Human Capital Management Services RRDb record the number of hours absent, allowing for fractional hours, and are therefore more continuous than discrete.
Sick leave is typically provided for brief illnesses lasting fewer than 2 weeks. Although on SL, employees generally receive 100% of their salary, the number of days of SL per year offered by employers varies by employer, and often by employee level and tenure.22 Short-term disability includes illnesses that last between 1 or 2 weeks and 6 months (generally). Employees usually receive 60% to 100% of their salary while on STD leaves. If the illness lasts longer than 6 months, the employee begins LTD and usually receives 50% to 70% of salary.23 Employees injured on the job may file a WC claim, which includes payments for medical services to treat the injury and salary-replacement payments (66% to 80% of salary, depending on location).24 Although this research focuses on four different types of absences, the categories are not substitutable and attempts to redistribute LTD absence to STD, SL, or another benefit, for example, would be difficult due to the eligibility requirements for these conditions.
The data spanned several years, and all cost variables were expressed as June 2010 US dollars, using non–seasonally adjusted consumer price indices for medical services, prescription drugs, and all consumer goods.25
To estimate the effects of RA on the overall employed US civilian noninstitutionalized population, the study's prevalence rate for RA was calculated. This rate was calculated as the number of eligible employees with RA divided by the number of eligible employees (employees with and without RA). The study's prevalence rate, incremental costs, and incremental absences were then applied to published estimates of the number of employees in the US civilian labor force (139 million persons).20
Differences in descriptive characteristics between the RA and control cohorts were compared using t tests for continuous variables and chi-squared tests for discrete variables. Characteristics included age, tenure (the number of years the employee has worked for his or her current employer), sex, marital status, race, exempt or nonexempt status, full-time or part-time status, annual salary, modified Charlson Comorbidity Index score,26 and region (grouped by first digit of employee's zip code).
Separate two-part regression models were run on SL-, STD-, LTD-, WC-, health care–, and prescription drug–dependent cost variables and on the SL, STD, LTD, and WC absence time variables. For example, in the cost or absence time models, logistic regression was first used to predict the likelihood of subjects having any costs or absence time (part 1). Generalized linear models were used with a gamma (γ) distribution and a log link function to model costs or absence time for subjects with more than zero costs or absences in the second part of the model. Results from the two parts were combined to produce estimates for each cohort. These regression methods account for the nonnormal distributions of the outcomes variables better than one-part linear models,27–29 and the two-part logistic and gamma distribution generalized linear models have been used successfully for both costs and absence time in prior studies.14–19
In each case, the multivariate regression models controlled for the impact of confounding factors, such as age, sex, marital status, race, exempt or nonexempt status, full-time or part-time status, salary, region, and the modified Charlson Comorbidity Index score. Only employees eligible for a specific work absence benefit were included in regression models for that benefit.
All models and statistics were generated via version 9.1 of the SAS System for Windows (SAS Institute, Inc, Cary, NC), and results were considered statistically significant when P ≤ 0.05.
A total of 340,740 employees were eligible for inclusion over the selected time frame (Table 1). Overall, the employees with and without RA were different in almost every descriptive measure reported. The RA cohort was significantly older, more likely to be women, more likely to work full-time, and had a higher modified Charlson Comorbidity Index score. The cohorts were also different in terms of marital status and race.
Direct and Indirect Costs
The regression-adjusted direct and indirect annual costs per employee for employees with and without RA are shown in Table 2. Cost comparisons between groups were significant for all cost categories (P < 0.05) except for LTD (P = 0.0505). Total annual costs were approximately $5212 higher for employees with RA than those without RA ($8707 vs $3495). The annual incremental direct costs and indirect costs associated with RA were $4687 (90% of total incremental cost) and $525 (10%), respectively.
Workers with RA had 3.58 more annual health-related absence days than the controls, missing an average of 7.92 workdays per year, whereas control employees missed 4.34 days per year (Table 3). Employees in the RA cohort had more lost time for SL, STD, and LTD and less lost time for WC than controls. The differences for SL (1.20 more days absent per year) and STD (1.91 days) were both highly significant (P < 0.0001).
Projections to the US Population
Using the study's prevalence rate of 0.79% and the $4687 incrementally higher “direct component” cost for employees with RA versus those without RA (medical and pharmaceutical claims), as well as indirect component (SL, STD, LTD, and WC) incremental costs of $525 and absence days of 3.58 per employee, projections can be made for the 2010 US civilian labor force.20 On the basis of these estimates, the total annual incremental burden of RA to the employer for the civilian labor force (Table 4) translates into $5.8 billion ($5.2 billion direct and $579 million indirect) and represents 4.0 million incremental lost days.
Despite an estimated annual incremental impact of $19.3 billion on patients, their employers, their family members, and the government,10 RA has arguably not received the public health attention that it deserves.
This is the first study to present objective data on the direct and indirect costs related to RA compared with non-RA controls in the United States, and this study does not use imputed, proxy, or self-reported data categorized by types of absences. The multivariate regression methodology used in this study has advantages (compared with matched case-control analysis and analysis of variance) in quantifying the impact of each confounding factor on the dependent variable being modeled and in being more appropriate for data that are not normally distributed. The two-part regression process examines the adjusted probabilities of having more than zero costs or absences and subsequently models those employees who had more than zero costs or absences. The incremental differences between the RA cohort and controls reported in this study represent the average employees—not the extremes. Finally, the database used for this study contains a broader set of categorized indirect cost and absence data types than those used in previous RA studies.9,12,21
Comparisons between this study and other studies need to be interpreted on the basis of their differences. We eliminated articles that were not specific to RA.12 Some other studies were limited by the lack of accurate data on costs and absences or absence types (in both costs and time). Some employer studies9,12,21 used older data that might not reflect the current diagnostic and treatment criteria around RA;21 estimated the costs of RA alone or in comparison with other conditions by multiplying the days absent by a standard wage;9,10,13 included payments for disability days but imputed costs for medically related absenteeism;10 did not use any control groups; or used self-reported information, which is subject to recall bias, to ascertain both the diagnosis of RA and the productivity and absenteeism measurements.9,12,13
This study found that employees with a diagnosis of RA have higher total medical costs than employees without RA. In particular, the costs for employees with RA were about 1.5 to 4 times higher than for controls for all cost metrics, consistent with the 2:1 ratio found elsewhere.21 Incremental direct costs were responsible for 90% of the incremental costs in this study, higher than the 65.7%12 and 83.4%21 reported in studies with different indirect cost components12,21 Using studies from the same database with similar methodologies, the significant difference of 1.20 annual SL days in this study can be compared with significant differences in other diseases, which ranged from 0.71 days to 2.79 days (Fig. 1).14–19 A study using another data set12 estimated SL days about 4 to 6 times higher than the present SL estimates and 1.5 to 2.8 times the total absence days from this study; however, the data might include vacation time.
A study of claims from commercially insured patients found that RA was not associated with either increased absenteeism or decreased productivity9; however, these employees reported investing significantly more effort in their work than did an age-sex matched cohort of employees without RA. The same publication also reported the opposite findings; in other words, RA was significantly associated with elevated absenteeism, but RA employees did not have a lower ratio of productivity to effort than other workers. Interestingly, when Kessler et al9 adjusted for comorbid conditions, their findings became not significant.9 Although this study did not adjust for the conditions used in Kessler, the models did control for the modified Charlson Comorbidity Index.
This study's estimate of incremental burden may be conservative. The estimate is representative of the year after identification in the database, which for some may be their first year of having RA but not for others with a diagnosis prior to entering the database. Although four indirect expenses to the employer (SL, STD, LTD, and WC) were quantified, the study did not require subjects to have all four benefit types for inclusion in either the RA or control cohorts, similar to methods in prior studies.10,14–19,21 Other potential employer expenses such as worker training and replacement costs and costs of workplace adaptations (eg, special keyboards, transportation devices, and other tools) were not included.10 In addition, this study's focus was on absenteeism, not presenteeism, which could be substantial and difficult to consistently measure9 and often is assessed by self-reported surveys.30
It is important to keep in mind that our point estimates for costs and absence days are based on the mean values for the cohort and do not reflect the wide range of outcomes among employees with RA (ie, not all are costly). Future analyses demonstrating the range of costs associated with RA should be considered. Also, the incremental costs associated with RA shown herein are representative of costs during the first year after the identification of their RA in the database. For some employees, this may be their first year of having RA, but this may not be the case for employees with a diagnosis prior to entering the database. Also, despite examining the costs of prescription therapies purchased by these employees, this study did not assess the impact of treatment.
The study was limited to employees with employer-sponsored insurance and 1 year of postdiagnosis eligibility, which may not be reflective of the overall RA population in which severe disability might impact the ability to maintain employment. The differences between employees with and without RA with regard to LTD may have been more pronounced if patients with more severe disease were included. The population was further restricted to individuals with International Classification of Diseases, Ninth Revision, diagnosis codes specific to RA, and the findings may not be representative of persons with RA who are not diagnosed, who are misdiagnosed, or who do not have a diagnosis in their medical records. Missing values for certain elements and treatment of missing data as a proper category in statistical analyses are a potential limitation of the research. Furthermore, although regression models controlled for differences between cohorts, using a broad set of available variables, other unavailable confounding factors may exist that influence the differences in outcomes between the RA and non-RA cohorts. The economic burden of undiagnosed and untreated RA may be greater than among those who carry a formal diagnosis and receive an appropriate intervention. Also, claims data do not indicate the severity of the RA. On the basis of the aforementioned facts, our prevalence rate and hence the projections to the US population may be considered conservative.
Estimated costs among workers with RA were consistently higher in every direct and indirect cost and absence category than among workers without RA—with 90% of the incremental costs related to direct medical care. The data emphasize the need for effective management strategies that can reduce the burden of illness and economic losses incurred.
The study concept and design was a collaborative effort of Mr Brook, Mr Smeeding, Dr Kleinman, and Dr Cifaldi. Mr Brook and Dr Cifaldi were responsible for study funding. Dr Kleinman conducted the primary analysis and Mr Brook, Mr Smeeding, Dr Cifaldi, Dr Kleinman, and Dr Shaw were responsible for the interpretation of the results. Mr Brook and Dr Kleinman drafted the original manuscript, and all authors contributed to revisions of the content. All authors take final responsibility for the final manuscript. In addition, the authors would like to acknowledge the contributions of Sanjoy Roy, MS, formerly an employee of Abbott Laboratories, Abbott Park, IL, and currently employed by Ethicon, Somerville, NJ, for his contributions to the study design, assistance in securing funding for this research, and initial reviews of the data. The authors would also like to acknowledge Stephanie E. Kirbach, PhD, Abbott Laboratories, Abbott Park, IL, for her contributions to the interpretation of the data and comments on drafts of this manuscript.
1. Helmick CG, Felson DT, Lawrence RC, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part I. Arthritis Rheum. 2008;58:15–25.
2. O'Dell JR. Rheumatoid arthritis. In: Goldman L, Ausiello D, eds. Cecil Medicine. 23rd ed. Philadelphia, PA: Saunders Elsevier; 2007. Available at: http://www.mdconsult.com/das/book/body/156141423-3/0/1492/1022.html#4-u1.0-B978-1-4160–2805-5..50290-1_12843
. Accessed June 22, 2011.
3. Lee DM, Weinblatt ME. Rheumatoid arthritis. Lancet. 2001;358:903–911.
4. Gabriel SE, Crowson CS, O'Fallon WM. The epidemiology of rheumatoid arthritis in Rochester, Minnesota, 1955–1985. Arthritis Rheum. 1999;42:415–420.
5. Treharne GJ, Hale ED, Lyons AC, et al. Cardiovascular disease and psychological morbidity among rheumatoid arthritis patients. Rheumatology. 2005;44:241–246.
6. Gabriel SE. Cardiovascular morbidity and mortality in rheumatoid arthritis. Am J Med. 2008;121(suppl 1):S9–S14.
7. Prior P, Symmons DPM, Hawkins CF, Scott DL, Brown R. Cancer morbidity in rheumatoid arthritis. Ann Rheum Dis. 1984;43:128–131.
8. Gabriel SE, Crowson CS, Kremers HM, et al. Survival in rheumatoid arthritis. Arthritis Rheum. 2003;48:54–58.
9. Kessler RC, Maclean JR, Petukhova M, et al. The effects of rheumatoid arthritis on labor force participation, work performance, and healthcare costs in two workplace samples. J Occup Environ Med. 2008;50:88–98.
10. Birnbaum H, Pike C, Kaufman R, et al. Societal cost of rheumatoid arthritis patients in the U.S. Curr Med Res Opin. 2010;26:77–90.
11. Khanna R, Smith MJ. Utilization and costs of medical services and prescription medications for rheumatoid arthritis among recipients covered by a state Medicaid program: a retrospective, cross-sectional, descriptive, database analysis. Clin Ther. 2007;29:2456–2467.
12. Ozminkowski RJ, Burton WN, Goetzel RZ, Maclean R, Wang S. The impact of rheumatoid arthritis on medical expenditures, absenteeism, and short-term disability benefits. J Occup Environ Med. 2006;48:135–148.
13. Silverman S, Dukes EM, Johnston SS, et al. The economic burden of fibromyalgia: comparative analysis with rheumatoid arthritis. Curr Med Res Opin. 2009;25:829–840.
14. Brook RA, Kleinman NL, Patel PA, et al. The economic burden of gout on an employed population. Curr Med Res Opin. 2006;22:1381–1389.
15. Kleinman NL, Brook RA, Melkonian AK, Doan JF, Baran RW. Health benefit costs and absenteeism due to insomnia from the employer's perspective. J Clin Psychiatry. 2009;70:1098–1104.
16. Brook RA, Rajagopalan K, Kleinman NL, Melkonian AK. Absenteeism and health benefit costs among employees with multiple sclerosis. Curr Med Res Opin. 2009;25:1469–1476.
17. Su J, Brook RA, Kleinman NL, Corey-Lisle P. The impact of hepatitis-C viral (HCV) infection on work absence, productivity, and healthcare benefit costs. Hepatology. 2010;52:436–442.
18. Brook RA, Kleinman NL, Choung RS, Melkonian AK, Smeeding JE, Talley NJ. Functional dyspepsia impacts absenteeism and direct and indirect costs. Clin Gastroenterol Hepatol. 2010;8:498–503.
19. Brook RA, Kleinman NL, Su J, Corey-Lisle PK, Iloeje UH. Absenteeism and productivity among employees being treated for hepatitis C. Am J Manag Care. 2011;17:657–664.
21. Birnbaum HG, Barton M, Greenberg PE, et al. Direct and indirect costs of rheumatoid arthritis to an employer. J Occup Environ Med. 2000;42:588–596.
24. Hunt HA. Is compensation for workplace injuries adequate? Employ Res. 2002;9:1–3.
25. Bureau of Labor Statistics. Consumer price indices for medical services, prescription drugs, and all consumer goods. Available at: http://www.bls.gov/cpi/data.htm
. Accessed April 1, 2011.
26. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383.
27. Blough D, Madden C, Hornbrook M. Modeling risk using generalized linear models. J Health Econ. 1999;18:153–171.
28. Manning W, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20:461–494.
29. Barber J, Thompson S. Multiple regression of cost data: use of generalised linear models. J Health Serv Res Policy. 2004;9:197–204.
30. Boonen A, Severens JL. The burden of illness of rheumatoid arthritis. Clin Rheumatol. 2011;30(suppl 1):S3–S8.
©2013The American College of Occupational and Environmental Medicine