Total costs and predictors of costs in individuals with early inflammatory polyarthritis: a community-based prospective study

Transcription

1 Rheumatology 2002;41: Total costs and predictors of costs in individuals with early inflammatory polyarthritis: a community-based prospective study N.J.Cooper,M.Mugford 1,D.P.M.Symmons 2,E.M.Barrett 3 andd.g.i.scott 1 Department of Epidemiology and Public Health, University of Leicester, Leicester, 1 School of Health Policy and Practice, University of East Anglia, Norwich, 2 ARC Epidemiology Unit, University of Manchester, Manchester and 3 Norfolk Arthritis Register, St Michael s Hospital, Aylsham, UK Abstract Objective. To estimate the health service, non-health service and total costs and predictors of costs in individuals with early inflammatory polyarthritis (IP). Methods. We conducted a prospective longitudinal study over a 6-month period. The participants were a random sample of 133 individuals who had enrolled with the community-based Norfolk Arthritis Register (NOAR) database between 1994 and The main outcome measures were the mean (per person) 6-month health service cost, non-health-service cost and total cost associated with IP. Results. One hundred and fifteen of the 133 individuals who were recruited into the study completed 6 months of follow-up. The mean 6-month total cost was estimated to be 2800 per person, of which 14% was health service costs and the remainder non-health-service costs. Higher total costs were associated with lower health status and rheumatoid factor positivity. Conclusions. Early IP has a considerable impact on both the health-care system and, more importantly, society. Non-health-service costs (i.e. costs incurred by the individual with the disease, their family and friends) account for a substantial proportion (86%) of the total costs associated with early IP. KEY WORDS: Inflammatory polyarthritis, Cost of illness, Economics, Health-care costs, Non-health-care costs, Costs. The rapid increase in health-care expenditure in all countries has increased interest in the economic impact of individual diseases or disease categories. The economic burden of rheumatoid arthritis (RA) is thought to be substantial for both the person with the disease and the health services. Evidence from US-based studies has shown that people with RA face three times the cost of medical care, twice the rate of hospitalization and four times the number of ambulatory visits to physicians compared with an age- and sex-matched non-arthritic population w1x. Findings suggest that the total and per capita lifetime economic costs of RA are similar to those incurred by stroke and coronary heart disease patients in the USA w2x. The majority of studies published to date have been US-based and focused on long-term (established) RA Submitted 25 July 2001; revised version accepted 17 February Correspondence to: N. Cooper, Department of Epidemiology and Public Health, University of Leicester, Princess Road West, Leicester LE1 6TP, UK. in a hospital (clinic) setting w3x. There have been few studies (practical or theoretical) on the costs of RA based in Europe w4 7x and there is also a lack of studies reporting the costs of early RA, which are increasing in importance since the change in clinical practice to prescribe more aggressive treatment regimes earlier in the disease process w8x. Also, community-based studies are likely to be more representative of the wider disease population. In 1989 the community-based Norfolk Arthritis Register (NOAR) was set up with the primary aim of identifying and studying all new cases of inflammatory polyarthritis (IP), of which RA is a subset, in the former Norwich Health Authority by notification from general practitioners and hospital clinics w9, 10x. Since the register was established, approximately 250 people each year have agreed to take part in the project. From NOAR data and national population statistics, it can be estimated that there are approximately new cases of IP each year in the UK w9x. This paper reports the results of a 6-month longitudinal study designed to estimate the health 767 ß 2002 British Society for Rheumatology

2 768 N. J. Cooper et al. service, non-health-service and total costs of early IP, using data for a cohort of people from NOAR (i.e. a representative population sample). It provides informative data about how the burden of disease is distributed among not only the health-care system and the individual with the disease, but also other parts of the public sector, such as the individuals family and friends, and society as a whole. This study aimed to assist the decision-making process at policy and planning levels by documenting the extent of resource use in current practice and identifying where the major burden of cost lies in the treatment and care of these people. It also intended to identify the subgroups of people who incur the highest costs. It was not a primary aim of this study to inform choices about which treatment or therapy is the most cost-effective option for people with IP and RA, although we expect that data collected during the study will prove useful for such exercises. Methods Study design This was a prospective, prevalence-based, longitudinal study conducted from a societal perspective and designed to capture the 6-month costs associated with early IP. Study population The study population consisted of 133 people with early IP (i.e. with disease duration of 5 yr or less) who were enrolled in the community-based NOAR between 1994 and The study was conducted between June 1999 and May NOAR inclusion criteria are as follows: (i) age over 16 yr; (ii) swelling of two or more joints; (iii) symptom duration more than 4 weeks; and (iv) symptom onset after January Full details of the NOAR study method are published elsewhere w9, 10x. Sources of resource-use data Resource-use data were collected from participants at three different time points baseline, 3 months and 6 months from the time of recruitment into the study by the use of self-completion postal questionnaires including the EuroQol w11x, which measures health status on a scale from 0=worst imaginable health state to 1=best imaginable health state. The content and structure of the data collection instruments were informed by data extracted from the transcripts of two focus groups conducted on individuals with IP to obtain a better understanding of the types of costs people with IP and their families have to bear as a direct consequence of their illness. Prior to this study, the questionnaires were pilot-tested on two cohorts of individuals from different geographical settings (rural and urban) as a test of external validity, and the response data were validated against an alternative data source, where available, as a test of internal validity. To minimize recall bias, study participants were provided with a memory aid in the form of a diary in which to record their resource use and expenditure over the 3-month follow-up intervals. Completion of the diary was voluntary, and the diary was for the participants personal use only (i.e. it was not collected from them). The questionnaires were designed to obtain information regarding the utilization of health-care services and resources, out-of-pocket expenses, time lost from usual activities and household help from family and friends. The NOAR database provided sociodemographic data (age, sex, smoking status, social class and ethnic group) wsocial class was assigned using the Registrar General s system w41x, based on the occupation of the individualx and self-reported functional disability data measured annually using the Health Assessment Questionnaire (HAQ) (scored on a scale from 0=without difficulty to 3=unable to perform) w12x. Sources of unit cost data Unit costs to supplement the resource-use data were obtained from a number of different sources (Table 1). All costs were expressed in 1998u1999 UK prices. Cost analysis The costing methods used are presented in Table 2. The total 6-month cost per person was estimated by combining resource-use data with unit costs using the following TABLE 1. Sources of unit costs Type of cost Health professional visits and in-patient stays Alternative practitioner visits, aids and adaptations Travel costs Time costs Unpaid help Paid help Prescription costs Source of unit cost Out-patient visits (by profession), in-patient stay, day-unit visit w16x Respondent Cost per mile w15x Cost per hour of non-working (leisure) time w18x Cost per hour of unpaid (housework) work time w16x Cost per hour of working time w33x Cost per hour of help w16x Respondent National prescription charge Prepayment certificate

3 Costs of early inflammatory arthritis 769 TABLE 2. Costing methods Cost Forgone paid work Forgone unpaid (house)work Forgone leisure timeuactivities Informal care Health service Method Human capital w34, 35x Housekeeper replacement w36 38x Constant value per hour w39x Replacement of care w35x Generic costs w16x TABLE 3. Characteristics of study participants at baseline Number of people in study cohort 133 Female 71% RA classified a 50% Mean age wyr (S.D.)x 57 (13.9) Mean disease duration wmonths (S.D.)x 47 (81.6) Median EuroQol score (IQR) 0.62 ( ) Median HAQ score (IQR) 0.63 ( ) general costing formula: Total cost ¼ Xn X m (frequency) ij (unit cost) j i ¼ 1 j ¼ 1 where i is the ith individual (i = 1,... n) and j is the jth service received or resource used (j = 1,... m). The mean 6-month costs to the health service, the individual, their family and friends, and employer were also calculated. Statistical analysis The main descriptive statistics presented are the arithmetic mean, standard deviation (S.D.), median and interquartile range (IQR) w13x. Exploratory data analysis was used to study the distribution of costs. The study population was categorized into three age groups (16 39, and 60 yr and over) and their HAQ and EuroQol scores were categorized into tertiles. A multiple linear regression model was fitted to the log-transformed cost data and parameters were estimated using forward stepwise regression to identify which sociodemographic and clinical characteristics influenced cost. Sensitivity analysis The sensitivity of the costs estimated in the main analysis to changes in the following variables were investigated. Health service costs. In-patient care costs were varied by "40% and out-patient care costs by "15% to make allowances for variation in provider costs across the UK, as illustrated in the NHS reference costs w14x. Non-health service costs. The cost per mile travelled was varied from 45 to 185 pence w15x, the cost per hour of paid help was increased from 770 pence (weekday rate) to 1200 pence (weekend rate) w16x, and the cost of non-workinguleisure time was varied from 24.7% of working time cost for the main analysis w17, 18x to 40% (600 pence per hour) w17x. Results Out of the 133 people who enrolled in the study at baseline, 115 (86%) completed both the 3-month and the 6-month follow-up questionnaires. From Table 3 it can be observed that approximately 70% of the study population were female, the mean age was 57 yr and mean disease duration 47 months. It can also be observed that a Percentage of subjects classified as having RA using the ACR criteria cumulatively. TABLE 4. Mean 6-month cost (n = 115) ( sterling, 1998u9) Costs to Mean cost (95% CI a ) Median cost (IQR) % Health service 385 ( ) 154 (32 343) 14% Individual 1297 ( ) 265 ( ) 46% Relatives and 983 ( ) 44 (0 1192) 35% friends Employer 126 (70 188) 0 (0 60) 5% Total 2791 ( ) 869 ( ) 100% a Bootstrap 95% percentile confidence interval w40x. 50% of the study population were classified as having RA using the ACR criteria cumulatively w19, 20x. Cost analysis Over the 6-month study period, the mean total cost for the cohort of individuals under observation was 2791 per person w95% confidence interval (CI) 2080 to 3713x (Table 4). Only 14% of the total cost was incurred by the health-care service in the form of health-professional visits, in-patient stays and medication. The remaining 86% of total costs were non-health costs incurred by other agents wi.e. participant (46%), family and friends (35%) and employers (5%)x. Figure 1 presents a histogram of the 6-month costs for all 115 participants who completed both follow-up questionnaires. As is often the case with cost data, the distribution is heavily skewed due to a small number of individuals incurring high costs while the majority incur low costs. Health service and non-health service 6-month costs were found to be positively associated (Spearman s rank correlation coefficient r S = 0.177; P = 0.06). To investigate whether costs changed over time, the cohort was categorized into three subgroups depending on the number of years a subject had been registered with NOAR (0 to <2, 2 to <4, 04 yr). The boxplot in Fig. 2 shows that there was no statistically significant cost difference between subgroups. The largest proportion of costs incurred by the health service was for in-patient stays and day-unit visits. Despite there being only three reported in-patient stays (mean length 21 days) and nine day-unit visits, these accounted for 33% of the health service costs incurred by the whole cohort (Fig. 3). Visits by other health professionals (consultant, GP, nurse, therapists) accounted for 37% of the total cost to the health service. Only

4 770 N. J. Cooper et al. FIG. 1. A histogram of 6-month cost (UK ). FIG. 3. Pie chart of costs incurred by the health service. FIG. 2. Box plot of total 6-month costs vs disease duration. Thick lines represent the median, boxes 25 75% of values (IQR), thin lines 10 90% of values and asterisks and 0 the extreme values. 16 out of the 327 (5%) out-patient visits were private consultations paid for by the participant. Prescribed medication accounted for 30% of the total health-service costs. Almost two-thirds of the study participants were taking non-steroidal medications and approximately half were taking second-line medications. Non-health-service costs were incurred by the participants, their family and friends, and their employers (where applicable). The largest proportion of estimated cost was for household help (34%) (Fig. 4). This was mainly unpaid help from family and friends. Only 2.5% of all help was paid for by the participant. The mean number of hours of help per week required by participants was approximately 4.5. A large proportion of non-health-service costs were accounted for by time lost by participants from their usual activities either work (24%) or leisure (25%). The majority of time lost to work was unpaid work such as housework. This was to be expected, given the large proportion of the study cohort who were either female anduor retired. To enable cost comparisons with the results of earlier studies, the cost estimates above may also be expressed using the broader definitions of direct (health-care and non-health-care) and indirect (lost productivity) costs (a categorization previously favoured by health economists). Lost productivity costs are defined here as costs associated with production loss and replacement costs due to illness, disability and death of productive persons, both paid and unpaid w21x. Thus the indirect cost category includes forgone work (paid and unpaid), forgone leisure activities and the replacement cost of household (informal) care, whilst direct costs comprise the remainder. The mean direct and indirect costs were 794 and 1997 respectively. Statistical analysis To determine which sociodemographic and clinical characteristics influenced cost, a multiple linear regression model was fitted to the log-transformed cost data. Note that the cost data were transformed to obtain a more symmetrical distribution (Fig. 5). The sociodemographic and clinical variables considered for inclusion in the model are listed in Table 5. Note that, for the purposes of this analysis, the EuroQol and HAQ scores were categorized into tertiles. It can be observed from Table 5 that the mean cost appeared to increase with HAQ score and decrease with EuroQol score, as would be expected. There also appeared to be a difference between the mean costs for individuals who were positive and negative for rheumatoid factor. Note that the EuroQol score was not included in the regression model as it was found to be highly correlated with the HAQ score (r S = , P < 0.005). The final model included three of the seven variables, all of which were simultaneously significantly associated with cost

5 Costs of early inflammatory arthritis 771 FIG. 4. Pie chart of non-health-service costs. TABLE 5. Mean (S.D.) 6-month costs split by sociodemographic and clinical factors n Mean cost (S.D.) Median cost (IQR) Female (4470) 1237 ( ) Male (3716) 444 ( ) HAQ score a < (2518) 294 (88 838) (4471) 1130 ( ) > (4964) 3007 ( ) Age (yr) a yr (2548) 2920 ( ) yr (4790) 978 ( ) 060 yr (3984) 508 ( ) Social class a I IIIN (2948) 933 ( ) IIIM V (5145) 752 ( ) Rheumatoid factor b Positive (5359) 2251 ( ) Negative (3610) 518 ( ) RA classified a,c Yes (4777) 1000 ( ) No (3601) 482 ( ) EuroQol score a < (5591) 3382 ( ) (3248) 872 ( ) > (628) 221 (87 658) a At recruitment to NOAR; b at recruitment to economic study; c percentage of subjects classified as having RA using the ACR criteria cumulatively. TABLE 6. Linear regression model with log-transformed cost data b S.E.(b) P Constant < if ve Rheumatoid factor = < if þve HAQ score 1 = 0 otherwise if 0:36 to 1:00 HAQ score 2 = 0 otherwise < if >1:00 Age 1 = 0 otherwise if 060 yr n = 105; adjusted r 2 = 29 %. FIG. 5. Histogram of the log-transformed 6-month cost (UK ). (P < 0.05): rheumatoid factor, HAQ score (level 1 and 2) and age (Table 6). The adjusted r 2 value for this model was 29%, indicating that the variability in cost is only partly explained by the predictor variables in the model. Sensitivity analysis The uncertain parameters in the analysis of health service costs were the secondary care costs, which were based on generic unit costs for out-patient visits and in-patient stays. Such costs vary greatly from hospital to hospital across the UK and therefore the health service costs were reworked. The lower extreme for the mean 6-month cost (i.e. reducing out-patient costs by 15% and in-patient and day-unit costs by 40%) was 330 and the upper extreme (i.e. increasing out-patient costs by 15% and in-patient and day-unit costs by 40%) was 449. This was a change in health service costs of "15%. Similarly, the uncertain parameters in the analysis of non-health-service costs were unit cost of household help, cost per mile and cost per hour of non-workingu leisure time. The mean 6-month cost for the main analysis was When all of the above parameters

6 772 N. J. Cooper et al. were varied within the ranges defined in the Methods section, the lower extreme was 2176 and the upper extreme Discussion In the UK, most health-care is free at the point of use and funded from general taxation. However, much of the costs of illness still fall on other agents, such as patients and their families. Such costs are often not studied explicitly in health services research despite their importance in representing the part of total resources society devotes to health-care. Considering health service costs in isolation would have explained the costs associated with early IP (14%) only partially, and may lead to misallocation of resources if it places extra costs on other agents. For example, centralizing clinics may result in patients having to travel further to visit health professionals. As with the effects of disease course and treatment, costs also show vast variability between individuals with IP. Whilst the aggregated results of such a heterogeneous group are of limited value for decision-making, the mean cost (for the average individual) may be used to inform budget-setting for a new cohort of individuals. As can be observed from Fig. 1, a relatively small number of individuals incur high costs, which skews the distribution of cost data to the right and inflates the mean value above the median value. Although total costs cannot be derived from it, the median value does provide useful information on the most typical cost for an individual subject. Due to random variation, the number of high-cost individuals may vary considerably between years and rheumatology departments. Clearly, such variability may have substantive effects on tight budgets. In this cohort, 10% of the individuals incurred 50% of the total 6-month costs. The main component of health-service costs was in-patient stays and day care followed by prescribed medication, which accounted for 33 and 30% of total health-service costs respectively. The minority of study individuals who incurred an in-patient stay or day-care visit (6%) accounted for a large proportion of the total 6-month health-service costs (42%). These findings were similar to those for asthma w22, 23x and multiple sclerosis w24x. Overall, average (per person) health service costs were considerably lower than previous study findings. Reasons for this include differences in (i) the population under surveillance (e.g. early IP vs established disease); (ii) the classification of the disease (IP vs RA); (iii) the availability of a day-care unit; (iv) the base year of the study (i.e. change in clinical practice over time); (v) the availability of routine cost data for health-care; and (vi) study setting (community vs clinic). Only the studies by van Jaarsveld et al. w6x from The Netherlands and Newhall-Perry et al. w25x from the USA were concerned with the costs of early disease. The former study investigated the costs of RA over the first 6 yr after onset whereas the latter study concentrated on the first year after onset. In both studies the participants were, on average, more functionally disabled (mean HAQ 1.25) than in the present study (mean HAQ 0.89). The health-care costs that were estimated by van Jaarsveld et al. w6x greatly exceeded those reported in the present paper, which reflects the clinic setting of the study of van Jaarsveld et al. and the case mix of the population of interest. However, the study of Newhall-Perry et al. w25x was community-based, like this study, and therefore provided similar cost estimates. It is important to remind the reader that this was an evaluation of early IP and therefore costs to the health service are likely to increase still further as the disease becomes more progressive (e.g. surgery costs, especially total knee or total hip arthroplasty, combined with postoperative infections and other surgical complications) w26, 27x. The costs incurred by individuals as a result of their IP were found to be substantial w46% of total 6-month cost per person, equivalent to 17% of average income (f 1300 per person)x. In this study, only a small percentage of all household help was paid for (between 0.9 and 3.7%), which agrees with the findings of Medeiros et al. w28x. The cost of informal (unpaid) care is likely to vary depending on household and social capital available; for example, structures of society that provide help (e.g. self-help groups) and public resources, which may increase the potential of society (e.g. social services). Medeiros et al. w28x reported that low socioeconomic level and lack of social support resulted in reduced quality of life of caregivers. Therefore, it is important for future research to investigate and, if necessary, refine social support provisions to ensure greater quality of life for caregivers. Costs incurred by friends and family as a result of an individual s IP are often omitted from studies investigating the economic impact of IPuRA. This is a sizeable omission as friends and family costs accounted for a substantial proportion (35%) of total 6-month cost per person ( 983, 95% CI ). The majority of this cost was attributable to informal care around the home (30% of total cost), a result also found in multiple sclerosis research (26% of total cost) w24x. In later, more established disease, one of the major components of cost is reported to be long-term care in residential or nursing home accommodation, which is linked to deterioration in functional disability and quality of life w29x. In this study, time lost from work as a consequence of an individual s IP, which was covered by a statutory sick-pay scheme, was considered to be a cost to the employer and government. However, in some cases, especially short-time sickness absence, the employer may not incur the cost if it is possible for the work to be made up on an individual s return to work or absorbed by other workers at no extra cost. Other agents who incurred costs as a result of an individual s IP were social services anduor voluntary organizations, in the form of household help and providers of aids and home adaptations. Although such

7 Costs of early inflammatory arthritis 773 costs were minimal (0.03% of total 6-month cost per person) in this economic study of early IP, it is reasonable to assume that the role of social services and voluntary organizations will increase as an individual s IP progresses over time, especially for individuals without informal social support. Therefore, there is a need for future studies of more established IP to measure the extent of these costs to assist the planning of services for individuals without informal social support. Overall, there was no statistically significant association between total 6-month cost and the number of years registered with NOAR. Individuals with high health-service costs also tended to encounter higher non-health-service costs. The results from the multiple linear regression model showed cost to have a positive association with rheumatoid factor and HAQ score and a negative association with age. These results agreed with the findings of van Jaarsveld et al. w6x. The change in the treatment regime, together with the development of a new generation of therapies wfor example, cyclo-oxygenase-2 (COX-2) inhibitors w30, 31x and tumour necrosis factor a drugsx will have major effects on the cost of treating RA. Therefore, it is important that the efficacy and cost-effectiveness of these new therapies are investigated from the perspective both of the health provider and of society. It is important that future cost-effectiveness analyses are based on good-quality primary research on costs and cost behaviour of the type presented above. Such detailed cost data can then be used to inform the cost side of decision analysis models designed to assess the costeffectiveness of new treatment regimes at an early stage, in order to assess the desirability of further research. In addition to the above, new expensive therapies also require considerable clinical judgement to enable them to be used effectively w32x. Acknowledgements We thank the NOAR team and Nicola Wiles (University of Manchester), Shirley Pearce (University of East Anglia) and Julia Dossor (Norfolk and Norwich Healthcare Trust) for all their help and support. We also thank all the study participants, without whom this research would not have been possible. This research was funded by an Arthritis Research Campaign PhD Studentship awarded to NJC. References 1. Meenan RF, Yelin EH, Henke CJ, Curtis DL, Epstein WV. The costs of rheumatoid arthritis. A patient-oriented study of chronic disease costs. Arthritis Rheum 1978; 21: Stone CE. The lifetime economic costs of rheumatoid arthritis. J Rheumatol 1984;11: Cooper NJ. Economic burden of rheumatoid arthritis: a systematic review. Rheumatology 1999;39: Jonsson B, Rehnberg C, Borgquist L, Larsson SE. Locomotion status and costs in destructive rheumatoid arthritis a comprehensive study of 82 patients from a population of 13,000. Acta Orthopaed Scand 1992; 63: Mcintosh E. The cost of rheumatoid arthritis. Br J Rheumatol 1996;35: van Jaarsveld CHM, Jacobs JWG, Schrijvers AJP, Heurkens AHM, Haanen HCM, Bijlsma JWJ. Direct cost of rheumatoid arthritis during the first six years: a cost-of-illness study. Br J Rheumatol 1998;37: Cooper NJ, Mugford M, Scott DGI, Barrett EM, Symmons DPM. Secondary health service care and second line drug costs of inflammatory polyarthritis in Norfolk. J Rheumatol 2000;27: Newman S, Fitzpatrick R, Revenson TA, Skevington S, Williams G. Understanding rheumatoid arthritis. Routledge: London, Symmons DPM, Barrett EM, Bankhead CR, Scott DGI, Silman AJ. The incidence of rheumatoid arthritis in the United Kingdom: results from the Norfolk arthritis register. Br J Rheumatol 1994;33: Symmons D, Harrison B. Early inflammatory polyarthritis: results from the Norfolk Arthritis Register with a review of the literature. I. Risk factors for the development of inflammatory polyarthritis and rheumatoid arthritis. Rheumatology 2000;39: EuroQol Group. EuroQol a new facility for the measurement of health-related quality of life. Health Policy 1990;16: Fries JF, Spitz PW, Young DY. The dimensions of health outcomes: the health assessment questionnaire, disability and pain scales. J Rheumatol 1982;9: Barber JA, Thompson SG. Analysis and interpretation of cost data in randomised controlled trials: review of published studies. BMJ 1998;317: Department of Health. NHS reference costs, Motorweb News, phtm 16. Netten A, Dennett J, Knight J. Unit costs of health and social care. Canterbury: Personal Social Services Research Unit, University of Kent, Smith K, Wright K. Informal care and economic appraisal. A discussion of possible methodological approaches. Health Econ 1994;3: Department of Transport. Values of time and vehicle operating costs. Highways Economics Note No. 2. London: Department of Transport, Arnett FC, Edworthy SM, Bloch DA et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1987;31: Wiles N, Symmons D, Harrison B, Barrett E, Scott DGI, Silman AJ. Estimating the incidence of rheumatoid arthritis. Arthritis Rheum 1999;42: Brouwer WBF, Koopmanschap MA, Rutten FF. Patient and informal caregiver time in cost-effectiveness analysis: a response to the recommendations of the Washington Panel. Int J Technol Assess Health Care 1998;143: Krahn MD, Berka C, Langlois P, Detsky AS. Direct and indirect costs of asthma in Canada, Can Med Assoc J 1996;154: Smith DH, Malone DC, Lawson KA, Okamoto LJ, Battista C, Saunders WB. A national estimate of the

8 774 N. J. Cooper et al. economic costs of asthma. Am J Respir Crit Care Med 1997;156: Holmes J, Madgwick T, Bates D. The cost of multiple sclerosis. Br J Med Econ 1995;8: Newhall-Perry K, Law NJ, Ramos B et al. Direct and indirect costs associated with the onset of seropositive rheumatoid arthritis. J Rheumatol 2000;27: Blocka KL. Changing trends in the hospitalization of patients with rheumatoid arthritis and the future of the inpatient rheumatic disease unit. J Rheumatol 1994;21: Jonsson B, Larsson SE. Functional improvement and costs of hip and knee arthroplasty in destructive rheumatoidarthritis. Scand J Rheumatol 1991;20: Medeiros MMDC, Ferraz MB, Quaresma MR. The effect of rheumatoid arthritis on the quality of life of primary caregivers. J Rheumatol 2000;27: Pugner KM, Scott DL, Holmes JW, Hieke K. The costs of rheumatoid arthritis: an international long-term view. Semin Arthritis Rheum 2000;29: Lipsky PE, Abramson SB, Breedveld FC et al. Analysis of the effect of COX-2 specific inhibitors and recommendations for their use in clinical practice. J Rheumatol 2000;27: Freemantle N. Cost-effectiveness of non-steroidal antiinflammatory drugs (NSAIDs) What makes a NSAID good value for money? Rheumatology 2000;39: Kushner I, Ballou SP. Treatment of rheumatoid arthritis we re getting closer. J Rheumatol 1999;26: Office of National Statistics. Living in Britain. London: The Stationery Office, Liljas B. How to calculate indirect costs in economic evaluations. Pharmacoeconomics 1998;13: Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in health and medicine. New York: Oxford University Press, Smith J. The value of housework in Australian National Product. Treasury Seminar Series, Australian Bureau of Statistics. Measuring unpaid household work: issues and experimental estimates. Commonwealth of Australia, 1990; Brouwer WBF. The relationship between productivity and quality of life: some reflections and implications for measurement. ESR-QoL Workshop, Leicester, Sculpher MJ, Buxton MJ. The private costs incurred when patients visit screening clinics: the case of screening for breast cancer and for diabetic retinopathy. Health Economics Research Group, Brunel University, Efron B, Tibshirani RJ. The jacknife. In: An introduction to the bootstrap. New York: Chapman and Hall, 1993: Office of Population Censuses and Surveys. Standard occupational classification, Vol. 3. Social classifications and coding methodology. London: HMSO, 1991.