The systemic lupus erythematosus Tri-Nation study: Cumulative indirect costs




We previously reported that patients with systemic lupus erythematosus (SLE) in the US incurred ∼19% and 12% higher direct medical costs than patients in Canada and the UK, respectively, without experiencing superior outcomes expressed as disease damage or quality of life. In the present study, we compared cumulative indirect costs over 4 years in these patients.


A total of 715 patients with SLE (269 US, 231 Canada, 215 UK) were surveyed semiannually for 4 years on employment status and time lost from labor and nonlabor market activities. Cross-country comparisons of indirect costs were performed.


In the US, Canada, and the UK, mean 4-year cumulative indirect costs (95% confidence interval [95% CI]) due to diminished labor market activity were $56,745 ($49,919, $63,571), $38,642 ($32,785, $44,500), and $42,213 ($35,859, $48,567), respectively, and cumulative indirect costs due to diminished nonlabor market activity were $5,249 ($2,766, $7,732), $5,455 ($3,290, $7,620), and $8,572 ($5,626, $11,518), respectively. Regression results showed that cumulative indirect costs (95% CI) due to diminished labor market activity in the US were $6,750 ($580, $12,910) greater than in Canada and $10,430 ($4,050, $16,800) greater than in the UK. Indirect costs due to diminished nonlabor market activity in the US were $280 (−$2,950, $3,520) less than in Canada and $2,010 (−$1,490, $5,510) less than in the UK, both results insignificant due to wide CIs.


Despite American patients incurring greater direct medical costs than Canadian and British patients, they do not experience superior health outcomes in terms of less productivity loss in either labor market or nonlabor market activities.


The aging of the population, escalating costs of health care, and rapid emergence of effective but expensive new therapies have led to increasing interest in determining the economic burden of rheumatic diseases. By weighing the costs and benefits associated with a given therapy, policy makers can determine how best to allocate resources. Cost of illness is often expressed in terms of direct and indirect costs: direct costs are the value of resources used in the prevention, diagnosis, treatment, and rehabilitation of a specified disease, and indirect costs represent the value of economic productivity lost because of disease-related work disability or premature mortality and may arise from both diminished labor and nonlabor market activity. Systemic lupus erythematosus (SLE) is a chronic multisystem disease that primarily affects women in the prime of life, and consequently indirect costs represent a significant proportion of the overall costs (1, 2). Diseases that affect women will particularly impact the nonlabor market activities such as housework and childcare (3). Therefore, estimating the total value of both labor and nonlabor market activities is crucial for comprehensive cost estimations.

In this study, our goal was to estimate cumulative indirect costs over a 4-year period in patients with SLE from the US, Canada, and the UK. We have previously demonstrated that patients with SLE in Canada and the UK incurred 19% and 12% lower direct costs than those in the US, respectively, but did not experience worse outcomes reflected as either more disease damage or poorer quality of life (4). In this study, we compared the indirect costs accrued by these same patients during this period.



Between July 1995 and February 1998, consecutive patients with SLE fulfilling at least 4 American College of Rheumatology (ACR) revised criteria for SLE (5, 6) were invited to participate in a comparative study on health expenditure and outcomes in SLE. Findings related to direct costs, damage accumulation, and quality of life have been described (4, 7); in this study, we focused on cumulative indirect costs. Patients were recruited from Johns Hopkins University School of Medicine (Baltimore) and the University of Pittsburgh in the US, the Montreal General Hospital and Hôpital Notre-Dame (Montreal) in Canada, and University College Hospital (London) and the Queen Elizabeth Hospital (Birmingham) in the UK. Approval was obtained from each center's institutional review board and informed consent was obtained from each participant.


At study entry and semiannually for a maximum of 4 years (final followup assessments occurred between May 1999 and October 2001), participants completed questionnaires on health resource utilization, employment status, days lost from labor and nonlabor market activities, and time lost by their caregivers in delivering health care to the patient or aiding the patient in obtaining health care or performing household tasks. Also at study entry and annually, patients reported on quality of life, social support, and satisfaction with health care. At study entry and conclusion, patients' treating physicians completed disease activity and damage measures.

Study instruments.

Health resource utilization and lost productivity were measured using a modified version of the economic portion of the Stanford Health Assessment Questionnaire (8). Quality of life was assessed by the Medical Outcomes Study Short Form 36 (SF-36) (9) and a visual analog scale (VAS) (10), social support was assessed using the Interpersonal Support Evaluation List (ISEL) (11), and patient satisfaction was assessed using the Medical Outcomes Study Patient Satisfaction Questionnaire (version IV) (12). Disease activity was assessed by the revised Systemic Lupus Activity Measure (SLAM-R) (13) and a VAS, and disease damage was assessed using the Systemic Lupus International Collaborating Clinics/ACR Damage Index (SDI) (14).

Calculating direct costs.

Given that prices differ across countries, we used a constant price per health service so that any observed cost differences would reflect differences in the quantity of resources used (7).

Calculating indirect costs.

Indirect costs were estimated according to the human capital approach, which attempts to value an individual's contribution to the economy and measures indirect costs in terms of time lost from work due to illness, either in the work force or unpaid work at home (2). Time lost from labor market activity represented 2 components: the additional weekly hours participants reported they would be working if they were not ill, and actual days patients reported missing from their current work schedule because of illness. Average sex- and age-matched wages as published by Statistics Canada (15) were then multiplied by the time lost to value losses in productivity. By applying Canadian wages to value lost productivity, we remained consistent with our previous analysis where Canadian prices were used to value health resources and allowed for comparison of indirect costs with the previously calculated direct costs.

Time lost from unpaid labor represented the difference between the number of hours patients devoted to domestic and volunteer work and the number of hours they would have devoted if they were not ill, and was computed in 3 steps. First, Statistics Canada data (15) were used to predict the expected time participants would devote to unpaid labor if they were not ill, based on their age, sex, marital status, and self report of time spent in paid labor. Second, similar estimates were also made according to patients' current work schedule. Third, to account for the effect of SLE on time actually devoted to unpaid labor, the prediction of current unpaid labor was decreased by a proportion equivalent to days missed from actual paid work for employed participants, or equivalent to days of limited activities for nonemployed participants. Days of self-reported limited activities were also accounted for in 2 ways, as either a 50% or 100% decrease in predicted current daily activity. Because the purpose of this study was to examine differences between the 3 countries while holding as many factors constant as possible, we chose to use survey data from a single country. Using data from American and British surveys, which have not necessarily been conducted in the same manner as those in Canada, would add an additional potential bias.

Time lost in nonlabor market activity was calculated by 2 methods: the replacement cost and the opportunity cost (2). In the replacement cost method, estimates of the value of lost time in household work are made based on expected earnings of service workers. However, these expected earnings may underestimate the indirect costs incurred by the patients in our study, who tend to have near average education. The opportunity cost method values household work as equivalent to the amount the person could have earned in the labor market. By using average sex- and age-matched wage rates, this method likely assigns a more realistic, higher value.

The time that the caregiver spent helping the patient receive health care services and the time the caregiver spent doing housework in excess of the patient's own losses in unpaid work were also included.

To demonstrate the influence on the economic approach chosen, indirect costs were also estimated using the friction cost method. This method only values days missed from current employment and productivity losses up to a maximum of 90 days following the termination of employment. Assuming that there is an ample pool of unemployed individuals with which the sick individual can be replaced, the costs incurred will only be those associated with lost productivity during the time of identifying and training the previously unemployed individual who is replacing the sick worker: the friction period. Thereafter, there are no further economic consequences of lost productivity to society. Furthermore, according to this method, the economic value of nonmarket work is effectively zero. In SLE, the zero value of nonmarket work and the limited period of market work losses may particularly undervalue the indirect costs of chronically ill women.

Although it would be optimal, we were unable to capture only SLE-related productivity loss because SLE causes such a broad spectrum of physical and mental problems that the patient is unable to determine what is attributable to SLE. Indirect costs were expressed in 2002 Canadian dollars.

Statistical analyses.

Demographics, disease characteristics, and costs were expressed across countries using means, SDs, medians, interquartile ranges, and proportions as appropriate. For patients who provided incomplete data (i.e., those who withdrew, were lost to followup, died, or provided data at study entry and conclusion but completed <4 productivity questionnaires), missing data were managed through multiple imputation using regression models with all available data from all patients as potential covariates. Potential covariates included age, sex, ethnicity (white versus nonwhite), education (both as years and categorical as <12 or ≥12 years), marital status (married versus unmarried), disease duration, health status (individual SF-36 subscales, summary scores, and patient-reported VAS), social support (ISEL total score), patient satisfaction with health care (individual subscales), direct and indirect costs, disease activity (both the SLAM-R and physician-reported VAS of current activity and activity over the past year), and disease damage. Consistent with our previous analysis (7), imputations were performed up to 4 years after entry for patients who died during the 4-year study. Alternative modeling strategies, such as omitting deceased patients or including them without performing imputations, would either create a selection bias or make it appear as if death were cost saving.

Cross-country comparisons of cumulative indirect costs were then performed using simultaneous regressions with indicator variables for the country where the patient was receiving care, with the US as the reference. Only study entry values of the abovementioned covariates were considered. These regressions also included cumulative direct costs, damage accumulation, and change in quality of life over the 4-year study period as outcomes. For all regressions, model selection was based on Bayes factor as approximated by the Bayesian information criteria.


A total of 715 patients were enrolled: 269 in the US, 231 in Canada, and 215 in the UK. A total of 157 (58%) patients in the US, 162 (70%) in Canada, and 166 (77%) in the UK completed the SDI at study entry and conclusion and at least 4 of 7 productivity questionnaires. Eighteen (7%) patients in the US, 13 (6%) patients in Canada, and 10 (5%) patients in the UK died. A total of 94 (35%) patients in the US, 56 (24%) in Canada, and 39 (18%) in the UK provided incomplete data. Baseline demographics and disease features by country are presented in Table 1.

Table 1. Baseline characteristics of study participants*
CharacteristicUS (n = 269)Canada (n = 231)UK (n = 215)
  • *

    SLAM-R = revised Systemic Lupus Activity Measure; IQR = interquartile range; SDI = Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; ISEL = Interpersonal Support Evaluation List.

  • Direct and indirect costs were based on patient self report over the 6 months preceding baseline and were annualized; these costs were not imputed. To compare the overall value of resource utilization across countries, diverse resources were collapsed into a single measure by assigning costs to these resources. By applying a constant price across countries (2002 Canadian $), any observed cost differences can be attributed to differences in pattern or frequency of resource use.

Age, mean ± SD years39.0 ± 11.943.2 ± 13.740.7 ± 12.1
Female, %95.193.594.8
White, %67.484.877.7
Completed secondary education, %85.062.368.3
Married, %52.247.863.3
Disease duration, mean ± SD years8.6 ± 6.210.0 ± 7.510.0 ± 7.1
SLAM-R (0 = no activity; 84 = maximum activity)   
 Mean ± SD4.1 ± 3.57.3 ± 4.96.3 ± 3.9
 Median (IQR)3.1 (1.0–6.0)6.2 (3.1–10.3)5.5 (4.0–8.6)
SDI (0 = no damage; 46 = maximum damage)   
 Mean ± SD1.7 ± 1.91.8 ± 2.41.3 ± 1.7
 Median (IQR)1.0 (0.0–3.0)1.0 (0.0–3.0)1.0 (0.0–2.0)
ISEL, total (0 = least support, 120 = most support)   
 Mean ± SD93.5 ± 19.996.1 ± 18.089.9 ± 19.7
Annual total direct medical costs (2002 Canadian $)   
 Mean ± SD5,055 ± 7,1944,968 ± 8,6464,763 ± 7,568
 Median (IQR)2,651 (1,301–6,314)2,085 (1,118–4,421)2,526 (1,263–5,009)
Annual total indirect costs (2002 Canadian $)   
 Mean ± SD16,345 ± 18,54611,101 ± 14,22212,925 ± 14,822
 Median (IQR)9,329 (775–28,712)4,678 (0–19,420)6,230 (133–24,940)

Within each country, there were no clinically meaningful differences in demographics, disease characteristics, and baseline direct and indirect costs between patients completing at least 4 productivity questionnaires and those who provided incomplete data (except deceased patients). In all countries, patients who completed at least 4 productivity questionnaires differed from deceased patients, with the difference being greatest in Canada. Deceased Canadian patients were older, had more disease activity and damage, and incurred higher medical expenditure, but did not differ in baseline indirect costs.

A breakdown of the cumulative number of hours of lost productivity is shown in Table 2. The only significant difference between the 3 countries was in the component of paid work consisting of the additional hours patients indicated they would work if they were not ill. The mean number of hours patients indicated they would work if they were not ill was 3,198 hours (95% confidence interval [95% CI] 2,742, 3,654) in the US, 2,041 hours (1,692,2,390,) in Canada, and 2,221 hours (1,814,2,627) in the UK. It is also of interest that unpaid labor may actually increase for some patients with SLE. This reflects the fact that patients reporting missing time from paid labor are expected to devote more time in unpaid labor, and that this substitution effect may compensate for some or all of the limiting effect of SLE.

Table 2. Time components (in hours of lost productivity)*
 US (n = 269)Canada (n = 231)UK (n = 215)
  • *

    Estimates based on unimputed data. Values are the mean (95% confidence interval).

  • Hours that nonemployed participants would be working if not ill, up to a maximum of 90 days, and given that they were employed in the preceding period. This estimate corresponds to the friction cost.

Paid work   
 Hours missed from current work schedule167 (107, 227)218 (148, 287)194 (147, 241)
 Additional hours would work if not ill3,198 (2,742, 3,654)2,041 (1,692, 2,390)2,221 (1,814, 2,627)
 Potential short-term disability99 (48, 150)156 (95, 218)93 (52, 133)
Unpaid work   
 Activities limited by 50%−290 (−410, −171)77 (−47, 200)35 (−86, 156)
 Activities limited by 100%56 (−101, 213)437 (252, 622)440 (259, 621)
Outside help   
 With patient activities limited by 50%605 (415, 796)303 (201, 405)508 (306, 711)
 With patient activities limited by 100%564 (382, 746)250 (161, 338)439 (257, 620)

When all patients were included using multiple imputation for those who provided incomplete data, the cumulative indirect costs due to diminished labor market activity were $56,745 (95% CI $49,919, $63,571), $38,642 (95% CI $32,785, $44,500), and $42,213 (95% CI $35,859, $48,567) in the US, Canada, and the UK, respectively (Table 3). Cumulative indirect costs due to diminished nonlabor market activity, calculated using the replacement cost method and by estimating a 50% decrease in activity on days that patients indicated they had limited their activities, were $5,249 (95% CI $2,766, $7,732), $5,455 (95% CI $3,290, $7,620), and $8,572 (95% CI $5,626, $11,518) in the US, Canada, and the UK, respectively. Less conservative estimates and estimates based on the friction cost method are also presented.

Table 3. Cumulative direct and indirect costs at 4 years in patients with systemic lupus erythematosus*
 US (n = 269)Canada (n = 231)UK (n = 215)
  • *

    Values are the mean (95% confidence interval) unless otherwise indicated. Missing data were imputed as described in the Patients and Methods section.

  • Provided a Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) at study entry and conclusion and a minimum of 4 resource questionnaires; 380 patients were missing only 1 resource questionnaire; 73% of patients provided an SDI at study entry and conclusion; 75% of the expected total number of resource questionnaires were obtained.

  • Indirect costs due to diminished nonlabor market activity were calculated using the replacement cost method, with an estimated 50% decrease in activity for days patients indicated they limited their activities.

  • §

    Total cumulative indirect cost is the sum of the cumulative indirect cost due to diminished labor market activity and cumulative indirect cost due to nonlabor market activity (replacement cost, 50%).

Data at entry and conclusion and ≥4 resource questionnaires, no. (%)157 (58)162 (70)166 (77)
Employed at baseline, %45.348.752.6
Cumulative direct costs (7) (2002 Canadian $)20,244 (17,764, 22,724)15,845 (13,509, 18,182)17,647 (15,557, 19,737)
Cumulative indirect costs (2002 Canadian $)   
 Labor market activity56,745 (49,919, 63,571)38,642 (32,785, 44,500)42,213 (35,859, 48,567)
 Nonlabor market activity   
  Replacement cost, 50%5,249 (2,766, 7,732)5,455 (3,290, 7,620)8,572 (5,626, 11,518)
  Replacement cost, 100%9,833 (7,230, 12,435)9,976 (7,363, 12,589)13,565 (10,232, 16,898)
  Opportunity cost, 50%6,764 (3,467, 10,062)7,275 (4,419, 10,131)11,317 (7,428, 15,206)
  Opportunity cost, 100%12,559 (9,107, 16,011)13,065 (9,635, 16,495)17,803 (13,407, 22,200)
 Total cumulative indirect cost§61,994 (55,798, 68,190)44,097 (38,333, 49,861)50,785 (44,224, 57,346)
 Total cumulative indirect cost (unimputed data)§62,595 (55,353, 69,836)39,042 (32,853, 45,232)48,300 (41,141, 55,459)
 Friction cost estimate4,638 (3,226, 6,050)6,529 (4,561, 8,497)4,993 (3,827, 6,159)

Regression results showed that cumulative indirect costs due to diminished labor market activity in the US were $6,750 (95% CI $580, $12,910) greater than in Canada and $10,430 (95% CI $4,050, $16,800) greater than in the UK (Table 4). Indirect costs due to diminished nonlabor market activity in the US were $280 (95% CI −$2,950, $3,520) less than in Canada and $2,010 (95% CI −$1,490, $5,510) less than in the UK, both results inconclusive due to wide 95% CIs.

Table 4. Regression model for cumulative indirect costs*
PredictorsCoefficient95% CI
  • *

    The other outcomes in the simultaneous regressions, cumulative direct costs, damage accumulation, and change in quality of life changed only trivially from the original analyses (4, 7). 95% CI = 95% confidence interval; SF-36 = Medical Outcomes Study Short Form 36.

  • Regression calculated using our most conservative assumptions (i.e., replacement cost estimates with a 50% decrease in activity on days patients indicated that they limited their activities).

Cumulative indirect costs due to diminished labor market activity  
 Canada−6.75−12.91, −0.58
 UK−10.43−16.80, −4.05
 Baseline indirect cost (labor market)4.594.21, 4.96
 Baseline SF-36 physical component summary score−0.64−0.86, −0.41
Cumulative indirect costs due to diminished nonlabor market activity  
 Canada0.28−2.95, 3.52
 UK2.01−1.49, 5.51
 Baseline indirect costs (labor market)−0.88−1.12, −0.64
 Baseline indirect cost (nonlabor market)0.980.67, 1.29
 Baseline physical functioning (SF-36 subscale)−0.13−0.19, −0.07
 Baseline role physical (SF-36 subscale)−0.07−0.11, −0.03
 Married5.643.09, 8.19
 Years of education−0.55−1.01, −0.08
 Baseline general satisfaction (patient satisfaction questionnaire subscale)−0.17−0.28, −0.05
 Baseline technical competence (patient satisfaction questionnaire subscale)0.150.02, 0.28


Few studies have examined the full economic burden of SLE (1, 2, 7). We have demonstrated that patients with SLE in the US incurred higher direct costs than those in Canada and the UK, but did not experience less disease damage (7) or a superior quality of life (4). In this study, the first transnational comparison of indirect costs in SLE, we found that the higher direct costs of American patients are also not associated with improved productivity either in the labor or nonlabor market.

Indirect costs are often not included in cost of illness studies, perhaps because they are more difficult to measure than direct costs. Nonetheless, they represent a large proportion of the total costs of illness in chronic diseases. We found that across countries, cumulative indirect costs represented ∼74% of the total costs. Even though indirect costs constituted a similar proportion of total costs in all countries, the absolute indirect costs were considerably higher in the American patients.

The higher indirect costs seen in the American patients were entirely due to additional hours patients indicated they would work if they were not ill. A higher level of education, as observed in patients from the US (85.0% completing high school in the US versus 62.3% in Canada and 68.3% in the UK), is likely associated with a higher earning capacity and, as a result, greater expectations in the workforce. This may account for why, when compared with patients from the other countries, the American patients would prefer to work additional hours if they were not ill. However, even after controlling for education, our regression analysis still showed that American patients incurred greater indirect costs. Other country-specific factors may also play an important role in producing the differences observed. Various social, political, and economic factors specific to each of the 3 countries may influence the indirect costs of any given illness. For example, differences in available social and welfare programs and unemployment rates may impact indirect cost estimates. Although the friction cost and human capital approaches yielded considerably different estimates, the conclusions of our study remain the same. US patients do not fare better; indirect costs estimated using the friction cost approach were also not lower in these patients.

There have been no other cost of illness studies in SLE that have provided estimates of indirect costs. Three studies have characterized the work disability experienced by patients with SLE without calculating costs. Partridge et al (16) demonstrated that 3.4 years after diagnosis, 40% of patients who had been employed at some time since diagnosis had become unemployed because of their illness. Zink et al (17) and Mau et al (18) described the employment rates in SLE and compared them with other rheumatic diseases (17) and the general population (18). After matching, an equal proportion of patients with SLE and rheumatoid arthritis (46%) under 65 years of age remained employed (17). When compared with the general population, the standardized employment ratios (SER) for patients with SLE with a disease duration <6 years did not differ from the general population; however, for patients with a disease duration 6–10 years and >10 years, the SERs were 0.80 (95% CI 0.74, 0.87) and 0.68 (95% CI 0.63, 0.73), respectively (18).

Even though cost assessment often relies on the use of patient questionnaires, there is little information on the validity and reliability of these cost questionnaires, particularly with respect to indirect costs. In a recent study (19), patient-reported economic questionnaires, administered every 3 months, yielded results comparable with payer data. Although information on administration of questionnaires every 6 months is not yet available, the results do suggest that there is validity in self report for calculation of indirect cost. Furthermore, although we attempted to be comprehensive in our indirect cost estimations, our methodology for calculating productivity loss in unpaid work may have been strengthened by asking participants to report their anticipated productivity if they were not ill rather than making assumptions based on time spent in paid labor. Nevertheless, these assumptions were likely more realistic than assuming that these patients had no losses at all in unpaid labor time. We also acknowledge that the 2 centers may not be representative of the entire country, and inclusion of more sites in future studies would ensure more representative populations.

Estimating the economic burden of chronic illness is essential for health policy makers to make informed decisions. In addition to the direct costs of medical treatment, comprehensive cost estimates must also incorporate the indirect costs attributed to lost productivity. Our study demonstrates that indirect costs represent a considerable proportion of the total cost of SLE. Furthermore, the higher direct costs incurred by American patients are associated with neither improved productivity nor enhanced health status.


Dr. Clarke had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Drs. Panopalis, Manzi, Gordon, Penrod, Joseph, Fortin, Sutcliffe, Esdaile, and Clarke.

Acquisition of data. Drs. Petri, Manzi, Isenberg, Gordon, Senécal, Fortin, Sutcliffe, Goulet, Choquette, Grodzicky, and Clarke.

Analysis and interpretation of data. Drs. Panopalis, Gordon, Penrod, Joseph, St. Pierre, Pineau, Choquette, and Clarke.

Manuscript preparation. Drs. Panopalis, Manzi, Isenberg, Gordon, Senécal, Joseph, Choquette, Esdaile, and Clarke.

Statistical analysis. Drs. Panopalis, Penrod, Joseph, St. Pierre, and Clarke.

Manuscript revision. Dr. Grodzicky.


The authors would like to thank the Tri-Nation staff: Tina Panaritis, BA, Popi Panaritis, Kathy Margonis, Maura Trifero, RN, Diane Ferland, RN, Carolyn Neville, BA, RN, and Mary Orsini-Dudin (Montreal General Hospital); Joan Rairie, RN, BSN (University of Pittsburgh); and Stephanie Heaton, RGN, Lupus UK Specialist Research Nurse (University of Birmingham). The authors would also like to thank all the participating physicians (University of Birmingham: Margaret Allen, PhD, MD; Simon Bowman, PhD, MD), Jennifer Gardner for her expert technical assistance, and the patients whose contribution made this study possible.