Hawkins, Kevin PhD; Wang, Sara PhD; Rupnow, Marcia F. T. PhD
* Summarize the demographic characteristics of migraine patients that may relate to the cost burden imposed by this disorder.
* Compare average annual indirect expenditures in the migraine patients and matched control subjects; translate these figures into total annual dollar costs; and identify the costliest component (or components) of total indirect expenses.
* Contrast the annual direct and indirect costs attributed to migraine.
Migraine is a common, chronic, and often debilitating disorder.1,2 It affects approximately 11% of the US population, with considerably higher prevalence in women (16% to 18%) than men (6% to 7%).1,2 Its related disability is substantial, as functional impairment accompanying headaches is seen in at least 90% of patients, and this can disrupt every aspect of day-to-day living, including work, school, and family and social relationships.2–5
Migraine is most prevalent in people between the ages of 25 and 55 years, the peak years of productivity; work attendance and productivity are adversely affected in up to 75% of patients.1,2,6–8 Employers of individuals with migraine bear an enormous burden due to the effect of disability on lost productivity and health care costs.6,9 Indeed, migraine and headache are among the top ten most costly diseases to employers.10 Studies have shown that the indirect costs of migraine, reflecting missed workdays (absenteeism) and diminished on-the-job performance (presenteeism), are significantly higher than direct costs (ie, the cost of medical care).9,11–13
In one early study, Osterhaus et al13 assessed costs related to lost labor (absenteeism and presenteeism) via a survey mailed to 940 patients with migraine. The survey received a 70% rate of response; all respondents met criteria for migraine as defined by the International Headache Society (IHS) and had been participants in at least one clinical trial evaluating migraine therapy. More than 50% of employed respondents reported missing at least 2 days of work per month due to migraine; symptoms of migraine adversely affected work performance in the majority (89%). These symptoms had an effect on work productivity for about 1 week per month. Results of this study estimated that in the US working population with migraine, the cost to employers for lost labor was between $5.6 and $17.2 billion per year. In a more recent publication, Hu et al9 estimated the indirect cost burden of migraine in the United States by combining self-reported data from an epidemiologic study (American Migraine Study), a Canadian population-based survey,14 and data from the US Bureau of Labor Statistics. These investigators estimated the average workdays with migraine per year to be 7.5 for men and 7.6 for women, and the missed workdays per year to be 3.8 and 8.3 for males and females, respectively. Migraine was estimated to cost employers approximately $13 billion ($16.5 billion in 2004 dollars) annually as a result of missed workdays or impaired function at work, of which $8 billion ($10.1 billion in 2004 dollars) was due to absenteeism. In contrast, direct medical costs related to migraine amounted to only about $1 billion ($1.3 billion in 2004 dollars) per year.
Recognition of the indirect cost burden of chronic disease has assumed greater importance, because it must be incorporated with direct cost estimates to gain an accurate view of the societal impact of the disease and cost-effectiveness of various treatments.13 Cost-effectiveness data are becoming mandatory in efforts to control health care costs. Employers will also benefit from knowledge of the impact of chronic disease in the workplace because they can implement affordable medical plans that ensure effective treatment and other accommodations to minimize disability and improve day-to-day functioning.9,15
To provide more current estimates of indirect cost burden associated with migraine, this retrospective analysis utilized Thomson-Medstat's MarketScan Health and Productivity Management (HPM) database (Thomson-Medstat, Cambridge, MA) to estimate the indirect burden of illness (BOI) of migraine and project this estimate to the US population. This study adds to the literature by using a more recent data set, which more accurately reflects current migraine treatment regimes. Also, this study uses current modeling techniques (propensity scoring), as well as providing migraine BOI estimates for short-term disability and workers' compensation, which have not been estimated before. The direct BOI of migraine, also estimated in this study, is presented elsewhere.16
Materials and Methods
To estimate the indirect BOI of migraine, we compared the average annual indirect expenditures of a group of employees with migraine with a matched group of employees without migraine (control group). Data were obtained from Thomson-Medstat's HPM database for the 2002 through 2003 calendar years. The HPM database contains information on more than 800,000 employees during 2002 and 2003. More specifically, it includes information regarding benefit plan enrollment, inpatient and outpatient health care services, pharmaceutical claims, workplace absence records, short-term disability (STD), and workers' compensation (WC) claims for employees at 10 large corporations in 43 states. Health care (eg, inpatient, outpatient, and prescription medications) and indirect data (absenteeism, STD, WC) were linked to employee enrollment data from these ten US employers.
The BOI of migraine was defined as the difference in average indirect expenditures per person between migraine and control cohorts. Indirect cost components included in this study are workplace absence, STD, and WC.
Study Sample and Inclusion/Exclusion Criteria
Patients were included in the migraine group if they had a primary diagnosis of migraine (ICD-9 346.xx) or menstrual migraine (635.4) at any time during 2002 or 2003, or had a primary diagnosis of tension headache (307.81) or headache (784.0x) and received a prescription for a migraine-specific abortive at any time during 2002 or 2003, or received a prescription for an abortive drug at any time during 2002 or 2003. Migraine-specific abortive drugs included serotonin (5-HT) receptor agonists (triptans), ergotamine preparations, and migraine combination agents.
Although the control patients met the exclusion criteria, they did not meet the inclusion criteria for migraine (ie, they did not have evidence of migraine). The control sample was randomly drawn and was 5 times larger than the migraine cohort. Controls were selected so that gender (80% female) and age distributions matched those in the migraine cohort.
Employees were excluded if they were not continuously enrolled in 2002 or 2003 or had missing data elements (age, gender, location, urban residence, plan type, diagnosis codes to determine health status).
Employees with migraine were included only once in the sample. If they met inclusion criteria in both 2002 and 2003, only 2002 data were used.
Variable Definitions and Data Analyses
Demographic variables included age, gender, year (2002 or 2003), residence derived from ZIP code (categorized rural or urban), US region of residency (Northeast, North Central, South, or West), and type of insurance. The latter variable was categorized as fee-for-service (FFS), preferred provider organization and point of service (PPO/POS), and capitated. As capitated claims generally have missing or underreported payment information relative to “true” resources involved with the service, capitated payments were imputed at the claim level. Capitated claim payments were assigned from noncapitated claims matched by procedure code, place of service (eg, inpatient, emergency room), and region.
The number of comorbid conditions in each patient was measured by the Charlson Comorbidity Index (CCI).17 Psychiatric Diagnostic Groups (PDG) measured the number of psychiatric conditions in each patient.18
Workplace absence expenditures were derived from payroll information provided by the employers contributing to the data set. Payroll periods were either biweekly or monthly records indicating number of absence hours for the employee. Records specified either the date(s) the employee was absent or that the absence occurred during the payroll period. Absence expenditures were estimated as hours absent from work multiplied by $30 per hour; this wage estimate is consistent with previously published data.10
The STD claims were provided by employers, and each indicated the start and end date of the STD episode. STD expenditures were estimated as number of hours missed from work on STD claims multiplied by $30 per hour, then multiplied by 70%, which represents percentage of wage replacement used by these employers.
WC expenditures were taken directly from the payment field on WC claims, which were provided by the employers. These expenditures represented replacement wages as well as health care costs while away from work. The WC claims provided the first and last date that the employee was receiving WC.
Only employees who could contribute work-absence, STD, and WC data were included in the study analysis. All payments were expressed in 2004 US dollars.
Descriptive analysis comparison of migraine and control groups was performed before and after matching. The χ2 and student t tests were used to test for differences in categorical and continuous variables, respectively.
The average indirect BOI was estimated by subtracting average annual indirect costs of the control cohort from the migraine cohort. Propensity score matching was used to equalize measurable differences between employees with migraine and the control group. Controls were matched to the migraine cohort according to the predicted probability of having a migraine; this probability was estimated for each patient on the basis of a logistic regression analysis of having a migraine that controlled for demographics (age, gender, region, location, year, and type of insurance) and overall comorbidities (based on the CCI and PDG). Patients in the control cohort were matched to migraine patients using a 1:1 nearest-neighbor-within-caliber approach. Patients with migraine and controls who were not matched were eliminated from the sample.19
A second-stage regression was used to estimate the indirect burden of migraine. Specifically, the second-stage regression used total indirect expenditures as the dependent variable and the same independent variables used in the propensity score matching, plus a dummy indicator to denote migraine patients. The second-stage regression controlled for any remaining differences between the cohorts after matching. As is common,16 a generalized linear model (GLM) was used in the second stage. All analyses were performed using SAS software version 9.1 (SAS Institute, Inc., Cary, NC).
National Indirect BOI
A national indirect BOI was estimated by projecting the results of this study using the national weights of the Medical Expenditure Panel Survey (MEPS).20 Specifically, BOI by age, gender, and indirect component was multiplied by the accompanying age and gender-specific MEPS national weight and then summed across the indirect cost components. More specifically, average BOIs (the mean expenditure for migraine patients minus the mean expenditure for the controls) for each indirect component (absence, WC, and STD) were estimated by age (10-year increments) and gender. The BOI in each cell (indirect component, age group, and gender) was then multiplied by the number of employed workers (by age and gender) in the United States in 2003 according to MEPS. The BOIs were summed across all the cells to produce a total.
The results of this study will vary with different wage rate assumptions. A sensitivity analysis was conducted by changing the wage rate from $30 per hour to $27.31 per hour. This new rate is from the recently released report on employer costs for employee compensation from the Bureau of Labor Statistics.* For the sensitivity analysis, the entire model was re-estimated, just as in the main study, but this new wage rate was utilized to estimate wages and STD payments.
Employee demographics before and after matching are shown in Table 1.
Before matching, inclusion criteria were met by 6622 employees with migraine and 32,866 employees without migraine. Age and gender were similar in each cohort, but all other variables differed significantly. Employees with migraine had higher comorbidity scores, were less likely to live in an urban area, were more likely to be in a capitated insurance plan, and were more likely to reside in the southern and western United States than were employees without migraine.
Almost all (98%) migraine employees were matched (6516/6622). All variables that were used in the matching process were not significantly different between cohorts after matching (Table 1).
Among migraine employees and matched controls, the mean age was 39.5 years, and women constituted 71% of the studied population. Approximately 95% resided in an urban area, and about 50% resided in the South, 20% in the Northeast, and 14% each in the North Central and West regions. Capitated insurance plan members made up 70% of the study population. The average CCI and PDG scores were about 0.28 and 0.25, respectively (Table 1).
As depicted in Fig. 1, average annual indirect expenditures were 2.75-fold higher in the migraine group compared with the control group ($4453 vs $1619), a statistically significant finding (P < 0.001). The difference in costs between the two groups is the estimated migraine indirect BOI ($2834 per patient). Most of this was attributed to absenteeism (75%), whereas STD and WC accounted for 21% and 4% of the indirect BOI, respectively. The dollar amounts of these latter expenditures and other indirect cost data are shown in Table 2.
Second-Stage Regression Results
The results of the GLM model were very similar to the post-match descriptive results and are not presented. Specifically, the indirect BOI of migraine from the GLM model was only 2% different than the post-match descriptive results described above. This implies that the propensity score matching was successful in equalizing the two cohorts.
National Indirect BOI
The national annual indirect BOI of migraine, excluding presenteeism, was estimated to be $12 billion. A breakdown of this amount by indirect components revealed the following (Fig. 2):
* Absenteeism, $9.66 billion (81%).
* STD, $1.55 billion (13%).
* WC, $0.78 billion (6%).
When the wage rate was lowered from $30 per hour to $27.31 per hour, the average annual BOI decreased from $2834 per patient to $2590. The national indirect BOI went from $12 billion to $11 billion. Thus, decreasing the wage rate by 9.4% decreased the BOI by about 9%. We would expect the BOI to fall at a rate slightly less than the wage, because the wage rate directly affects absence and STD payments, but not the WC payments (which only make up about 4% of the BOI).
The objective of this study was to update estimates of the indirect BOI of migraine. This was accomplished by comparing indirect expenditures in terms of absenteeism, STD, and WC claims for employees with migraine to those of a matched control cohort of employees without migraine. After matching and regression-based adjustments, it was shown that employees with migraine cost approximately $2834 more per year than employees without migraine. Based upon projection of study results to the US population, it was estimated that employees with migraine cost employers approximately $12 billion per year due to absenteeism, STD, and/or WC. Most of the indirect migraine expenditures were driven by absenteeism. In a separate component of this study (data not shown), the direct annual health care cost attributed to migraine was estimated to be about $11 billion, somewhat less than the indirect-cost component.
These data confirm that the overall burden of migraine on society is large. In a cross-study comparison, the estimated per-employee absenteeism plus STD expenditures for migraine in this study were shown to be greater than those of other common conditions such as chronic obstructive pulmonary disorder, heart disease, depression, chronic renal failure, diabetes, and asthma.
Compared with the study by Hu et al,8 which found absenteeism costs to be about $8 billion ($10.1 billion 2004 dollars), the estimate of absence cost from the current study ($9 billion), which used a different methodologic approach, was consistent. It should be noted that our total indirect BOI estimates cannot be directly compared with the estimate from Hu et al, as the two studies captured different sets of indirect cost components. Specifically, Hu et al did not include STD and WC. Our study, on the other hand, excluded presenteeism.
Further perspective on differences between the present study and that of Hu et al can be found in data from Goetzel et al study,9 which evaluated the impact of medical costs, absenteeism, STD, and presenteeism on overall costs associated with 10 medical conditions, including cancer, diabetes, heart disease, migraine/headache, and arthritis. For all conditions, presenteeism was the most important driver of overall costs, and the impact of presenteeism was greatest in patients with migraine/headache, accounting for 89% of the total cost burden for this condition. These data suggest that the $12 billion estimate for the national indirect BOI of migraine in the present study (based on absenteeism, STD, and WC) is likely an underestimate of the total indirect BOI. We did not attempt to estimate presenteeism based on modeling techniques; rather, we are presenting the components of the indirect BOI that could be estimated from real data captured in the HPM database. In the study of Hu et al,8 annual costs related to reduced productivity were approximately $5 billion ($6.3 billion in 2004 dollars). A recent study by Stewart et al21 employed the American Productivity Audit to estimate the loss of productivity in the workplace resulting from various health conditions. The study reported that compared with 22 other health conditions, “headache” was the most common pain condition resulting in lost productive time at work (5.4% of the workforce) and cost employers $15.4 billion in presenteeism costs ($16.1 billion in 2004 dollars).
In addition to the lack of presenteeism component, other limitations of this study should be acknowledged. This was a retrospective database study, and only information recorded in the database was available for analyses. The effects of unmeasured variables are unknown. In addition, the sample is representative of the 10 large corporations in the United States, which is only a subset of the US working population. It is not known whether the difference in indirect costs between the migraine and control groups would be similar in a working population outside of these 10 employers.
Despite these limitations, this study offers an important methodologic approach to estimating the indirect BOI based on productivity databases such as HPM. The exact medical reasons for absenteeism, STD, and WC are often not captured in such databases, and therefore, it is not possible to directly identify what claims were due to migraine. By using the propensity score matching technique, we can identify a matched control group with similar characteristics and estimate the incremental costs incurred by the migraine cohort compared with the control cohort.
Employers should consider the annual indirect cost of migraine when investigating the cost and value of various medical plans that relate to migraine benefits.17 Facilitating access for the employee experiencing migraine into the health care system and a subsequent treatment plan offering effective therapeutic intervention can reduce the chances of absenteeism and work productivity losses, thus reducing indirect costs.7,11,22
This study has shown that employees with migraine incur greater indirect costs for employers than employees without migraine. Specifically, the national annual indirect BOI of migraine (including absenteeism, STD, and WC) was estimated to be $12 billion, with almost $10 billion of this accounted for by absenteeism.
Migraine imposes a substantial indirect cost burden. Employers may benefit from the awareness of these indirect cost data. Increased knowledge and access to appropriate migraine management will likely reduce this indirect cost burden.
1. Lipton RB, Bigal ME. Migraine: epidemiology, impact, and risk factors for progression. Headache. 2005;45(Suppl 1): S3–S13.
2. Lipton RB, Stewart WF, Diamond S, Diamond ML, Reed M. Prevalence and burden of migraine in the United States: data from the American Migraine Study II. Headache. 2001;41:646–657.
3. Dueland AN, Leira R, Burke TA, Hilyer EV, Bolge S. The impact of migraine on work, family, and leisure among young women—a multinational study. Curr Med Res Opin. 2004;20:1595–1604.
4. Lipton RB, Stewart WF, von Korff M. Burden of migraine: societal costs and therapeutic opportunities. Neurology. 1997;48(Suppl 3):S4–S9.
5. Dahlof CG, Solomon GD. The burden of migraine to the individual sufferer: a review. Eur J Neurol. 1998;5:525–533.
6. Warshaw LJ, Burton WN. Cutting the costs of migraine: role of the employee health unit. J Occup Environ Med. 1998;40:943–953.
7. Kryst S, Scherl E. A population-based survey of the social and personal impact of headache. Headache. 1994;34:344–350.
8. Hu XH, Markson LE, Lipton RB, Stewart WF, Berger ML. Burden of migraine in the United States: disability and economic costs. Arch Intern Med. 1999;159:813–818.
9. Goetzel RZ, Long SR, Ozminkowski RJ, Hawkins K, Wang S, Lynch W. Health, absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting U.S. employers. J Occup Environ Med. 2004;46:398–412.
10. Solomon GD, Price KL. Burden of migraine. A review of its socioeconomic impact. Pharmacoeconomics. 1997;11(Suppl 1):1–10.
11. Edmeads J, Mackell JA. The economic impact of migraine: an analysis of direct and indirect costs. Headache. 2002;42:501–509.
12. Ferrari MD. The economic burden of migraine to society. Pharmacoeconomics. 1998;13:667–676.
13. Osterhaus JT, Gutterman DL, Plachetka JR. Healthcare resource and lost labour costs of migraine headache in the US. Pharmacoeconomics. 1992;2:67–76.
14. Edmeads J, Findlay H, Tugwell P, Pryse-Phillips W, Nelson RF, Murray TJ. Impact of migraine and tension-type headache on life-style, consulting behaviour, and medication use: a Canadian population survey. Can J Neurol Sci. 1993;20:131–137.
15. 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.
16. Hawkins K, Wang S, Rupnow M. Direct cost burden of migraine among members of US employers. Presented at the 48th Annual Scientific Meeting of the American Headache Society, June 22–25, 2006, Los Angeles, CA. Poster # F68.
17. D'Hoore W. Bouckaert A, Tilquin C. Practical considerations on the use of the Charlson comorbidity index with administrative data bases. J Clin Epidemiol. 1996;49:1429–1433.
18. Ashcraft ML, Fries BE, Nerenz DR, et al. A psychiatric patient classification system. An alternative to diagnosis-related groups. Med Care. 1989;27:543–557.
19. Ganguly R, Martin B, Dorfman J, Rizzo J. In search of an unbiased estimate of treatment effect using observational data: a comparison of propensity scoring and Heckman two stage sample selection models. ISPOR Connections. 2004;10:2–5.
20. US Department of Health and Human Services. Agency for Healthcare Research and Quality Medical Expenditure Panel Survey. Available at: http://www.mcps.ahrq.gov
. Accessed March 2006.
21. Stewart WF, Ricci JA, Chee E, Morganstein D, Lipton R. Lost productive time and cost due to common pain conditions in the US workforce. JAMA. 2003;290:2443–2454.
22. Rapoport AM, Adelman JU. Cost of migraine management: a pharmacoeconomic overview. Am J Manag Care. 1998;4:531–545.
*http://www.bls.gov/news.release/pdf/ecec.pdf, referenced January 15, 2007. Cited Here...