Cost Differences for COPD With and Without Physician-Diagnosis

Abstract

Previous studies have presented divergent estimates of the cost of illness of COPD due to differences in methodology. The objective of this study was to examine differences between register-based estimates versus population-based estimates on the burden of COPD. This study therefore examined differences in costs of COPD among physician-diagnosed and un-diagnosed subjects. During a one-year period, four telephone interviews were made with 212 randomly selected subjects with COPD derived from the Obstructive Lung Disease in Northern Sweden (OLIN) studies. Health care resource utilization and productivity losses were measured, and the costs were also transformed with the estimated COPD prevalence in Sweden. Average annual costs were SEK 18,252 (USD 2,207, EUR 2,072), and SEK 9,327 (USD 1,128, EUR 1,059) for subjects with and without a physician-diagnosis, respectively. Although lower per individual, the costs of undiagnosed subjects accounted for approximately 40% of the total costs in Sweden, since the majority of subjects with COPD in Sweden lack a physician-diagnosed disease. In conclusion, we found that the costs due to COPD differed considerably between those with and without physician-diagnosed disease. This study indicates that register-based studies result in underestimated costs of COPD.

Key words: :

• COPD
• Cost of illness
• Diagnosis
• Underestimation

INTRODUCTION

Chronic Obstructive Pulmonary Disease (COPD) is a common, disabling, and costly disease, and is a major health problem [[1], [2]]. In most cases disease progress is slow [[3], [4]]. Subjects with COPD remain undiagnosed for many years, despite having multiple respiratory symptoms, and often subjects are not correctly identified until late in the disease process [[5], [6], [7], [8]]. Thus, COPD is a hidden disease, since a majority of subjects suffering from the disease have not been identified by the health care system [[2], [6][7], [8]]. A European estimate is that only 25% of the COPD cases in the general population are diagnosed [[9]].

The large proportion of unidentified cases complicates the estimation of the burden of the disease. Despite the fact that COPD causes high costs, few health economics studies of COPD have been performed [[10]]. Most of them have been register studies based on identified cases and health statistics without correction for the large underdiagnosis [[11], [12], [13], [14]]. To the best of our knowledge, no information is available describing the societal costs of COPD taking into account the effects of subjects with COPD but without a relevant diagnosis.

We have previously estimated the economic consequences of COPD in Sweden using a well-defined and representative cohort of subjects with COPD derived from large studies of the general population [[15]]. The total costs were approximately three times higher than the costs estimated by a Swedish health register based study [[13]]. Our hypothesis is that the main reason for the discrepancies between the two studies is due to the fact that the majority of subjects with COPD have not been identified by the healthcare system, and thus are not available for studies of health economics based on registers. The aim of this study was to examine the differences in costs among diagnosed and un-diagnosed subjects with COPD, and to evaluate how the societal costs of COPD can be identified by using different methods.

MATERIAL AND METHODS

In this study, we chose to use the prevalence approach, since considerable changes in treatment patterns of COPD are expected to occur in the future. Data were collected by telephone interviews with randomly selected subjects with COPD. The Ethics Committee at the University Hospital of Northern Sweden in Umeå approved the study.

Study population

The study population and the selection procedure have been described previously in detail [[15], [16]]. Briefly, the sample was generated from large-scale population studies of the epidemiology of obstructive airways diseases and type-1 allergy in northern Sweden, the OLIN studies [[17], [18]]. Longitudinal studies of a number of cohorts have been underway since 1985, including a total of approximately 40,000 children, adults and elderly. The study cohort in the current study was derived from two surveys performed from 1993 to 1998, in which pulmonary function tests were performed on 4,600 subjects [[2], [5]]. The subjects having COPD were identified in these epidemiological studies.

The diagnosis of COPD was made according to the British Thoracic Society's (BTS) criteria [[19]]: FEV₁/VC < 70% and FEV₁ < 80% of predicted values. According to BTS guidelines, COPD was divided into mild, moderate, or severe disease (Table 1). In addition, persons with FEV₁/VC < 70% and FEV₁ ≥ 80% of the predicted value, which corresponds to the Global Initiative for Obstructive Lung Disease (GOLD) criteria for mild COPD, GOLD stage 1 [[20]], were also included in the study. Subjects with other diseases, which explained their impaired pulmonary function, were excluded.

Table 1. The participants by age, gender, and severity of disease

		Age (years)
FEV1 categories(% predicted)	25–50		51–70		> 70		Total (M/F)
	D	U-D	D	U-D	D	U-D	D	U-D
≥80%	2	13	0	8	1	9	3	30
60–79%	8	16	8	24	4	19	20	59
40–59%	3	1	12	16	20	20	35	37
< 40%	0	2	12	4	6	4	18	10
Total	13	32	32	52	31	52	76 (41/35)	136 (79/57)

D: Subjects with physician-diagnosed COPD.

U-D: Subjects without physician-diagnosed COPD.

In total, approximately 800 subjects with COPD were identified. Due to limitations in resources, not all could be invited to the current study, instead a sample size allowing estimations of reliable and representative costs was calculated. The study sample was thus stratified by age and severity of COPD. All subjects with FEV₁ < 60% of predicted values were included in the study cohort because of low numbers. Of those with FEV₁ ≥ 60% of the predicted value, a random sample was drawn from each stratum, and a study cohort of a sufficient size to allow measures of significant differences between the different groups was obtained.

Study sample

The sampling procedure resulted in an invitation of 261 subjects, who fulfilled the criteria for COPD. A number of 212 subjects (43% women) participated, while 38 persons could not be traced, and 11 individuals refused to participate in the study. Demographics of the study sample in terms of age, gender, FEV₁ of predicted and smoking habits are shown in Table 2. Subjects were defined as physician-diagnosed by answering, “yes” to the following question: ‘Have you been diagnosed by a physician as having COPD, chronic bronchitis, or emphysema?’ Of the non-diagnosed subjects a large majority had also consulted health care due to respiratory complains, but had not been accurately diagnosed.

Table 2. Demographics of the study sample

	Physician-diagnosis (n = 76)	No physician-diagnosis (n = 136)	p-value
Mean age (SD)	65.9 (11.1)	63.6 (12.9)	n.s
Gender (Female %)	46.1	41.9	n.s
FEV1% of predicted, mean (SD)	49.9 (18.7)	68.7 (16.9)	< 0.0001
Smoking habits			n.s
Non-smokers n (%)	4 (5.3%)	31 (22.8%)
Ex-smokers n (%)	38 (50.0%)	59 (43.4%)
Smokers n (%)	27 (35.5%)	42 (30.9%)
Missing n (%)	7 (9.2%)	4 (2.9%)

Age and FEV₁% were tested with t-tests (p = 0.17, 1.3e-9, respectively); gender: proportions test (p = 0.66); smoking habits: chi-square test (p = 0.21), n.s = no significant difference.

Methods

A letter informing about the study was sent to the subjects before the interviews were conducted. During a one-year period, the 212 subjects were interviewed over telephone. A specially designed and pilot-tested questionnaire was used to record resource use including both respiratory-related healthcare utilization and healthcare utilization due to other reasons. However, only costs related to respiratory diseases were included in the estimations. Each subject was interviewed four times during 1999 (March, June, September, and December), and the questions covered resource use for the preceding 3-month period (Appendix and Table 3). All interviews were carried out by the same person (SAJ), and each interview lasted 10–45 minutes. Three attempts were made to reach each respondent, and a total of 201 persons took part in all four interviews, while five died during the study and the remaining six subjects did not complete all four interviews. Costs were estimated as previously described [[15]], using the patient-year approach to deal with discontinued subjects [[21]]. Although the methodology for estimating costs is under development [[22]], we do not choose to employ new methods as it might disturb the study of the effect of presence/absence of a physician-diagnosis. FEV₁ was also measured during the study to try to detect any change in the severity classification among the subjects compared to the measurement performed before the study period.

Table 3. Unit costs for healthcare contacts (SEK, in year 1999)

Contact	Unit costs
Visit to primary care physician	1 170
Visit to other primary care personnel	340
Visit to hospital-based physician	1 170
Visit to other hospital-based personnel	340
Hospitalisation 24 hours, dept. of pulmonary medicine	2 810
Hospitalisation 24 hours, dept. of medicine	3 650
Hospitalisation 24 hours, intensive care	5 440
Hospitalisation 24 hours, primary care centre	3 030
Oxygen therapy per month	1 109

Source: County Council in Norrbotten, Sweden.

Costs

The total costs that were recorded reflected both direct and indirect expenses. The direct costs were divided into hospitalisations, drugs, planned and acute healthcare visits and contacts, oxygen therapy, and equipment used as aids, such as wheelchairs. The indirect costs were divided into loss of productivity due to disability pensions and absence from work due to sick leave. Only costs related to respiratory diseases were included. The exchange rates were USD 1 = SEK 8.27, and EUR 1 = 8.81, the mean value for 1999.

Unit costs

Unit costs for year 1999 for healthcare resources, including physician salaries, capital and overhead costs were obtained from the county council in Norrbotten, Sweden, where all patients lived (Table 3). One day of intensive care in hospital was estimated at SEK 5,440 [[23]]. Drug prices were obtained from the 1999 Swedish Pharmacopoeia [[24]], and unit costs for oxygen therapy were estimated as the total monthly costs for an oxygen concentrator prescribed by Sunderby Central Hospital of Norrbotten, Luleå.

The value of the indirect costs was based on the human capital approach, which is the one most commonly used today [[25]]. The cost of each day absent from work was calculated as the average salary in Sweden in 1999, including payroll taxes and social security contributions.

Analysis

Since a societal perspective has been used, both direct and indirect costs were estimated. The resource use for each subject with a physician-diagnosed COPD and each subject without a physician-diagnosed disease was multiplied by the 1999 unit costs (Table 3). Multiple regression was used to determine the influence of COPD severity class (one intercept for each four classes), the presence of a physician diagnosis, age and smoking status as independent variables on total cost, and on direct cost, respectively. Due to the skewness of the data, we tried different transforms. A log transformation of the costs was finally used. The models have not been used to predict costs, and no predicted costs are presented. Sensitivity analyses were also performed, first by including the subjects referring themselves as asthmatics in the group with a diagnosis, i.e., either COPD or self-reported asthma, then by excluding all subjects with FEV₁ ≥ 80% of predicted.

The average cost per patient estimates in the severity groups were weighted using prevalence weights by severity group, and estimated proportions with and without a physician-diagnosed disease by severity group, in order to get the societal costs for COPD in Sweden among those with and without a physician-diagnosed disease. We used previous estimates of the prevalence, broken down by disease severity and official population data, in order to get prevalence-weight [[15]]. The estimated total number of subjects with COPD in Sweden in 1999 was about 678,000 individuals [[2], [5], [26]].

RESULTS

Among the interviewed subjects, 76 persons had physician-diagnosed COPD, while 136 persons did not. Of the undiagnosed subjects, 56 persons referred themselves as having asthma. The majority of those without a physician-diagnosis had a mild disease (65.5%), and only 7.4% had a severe COPD in the study sample (Table 1). Among those with physician-diagnosed disease, 30.3% had a mild disease and 23.7% had a severe COPD.

Resource use

Subjects with a physician-diagnosed COPD consumed considerably more healthcare resources and drugs, and a greater proportion of them had disability pensions and absence from work due to respiratory diseases. The average number of days in hospital during one year due to respiratory diseases amounted to 4.1 days for subjects with a physician-diagnosed disease compared to 0.5 days for those without, and the average number of outpatient visits amounted to 2.8 visits and 0.8 visits, respectively. Among those with physician- diagnosed COPD, there were 22.4% subjects with a disability pension due to respiratory diseases, while only 5.1% of those without physician-diagnosed disease had a disability pension.

Total costs

The total costs for the Swedish society were twice as high for subjects with a physician-diagnosed COPD, SEK 18,252 (USD 2,207), compared to subjects without, SEK 9,327 (USD 1,128) (Table 4 and Table 5). The costs were higher for subjects with a physician-diagnosed COPD regardless of disease severity (p = 0.0017). In particular, there was a substantial difference in costs among subjects with severe COPD. For both diagnosed and undiagnosed COPD, the indirect costs were higher than the direct costs, 58% and 55%, respectively.

Table 4. The annual societal costs per patient for those with physician-diagnosed COPD (SEK, 1999)

FEV1 categories %
	< 40%	40–59%	60–79%	≥80%	Prevalence-weighted
	N = 18	N = 35	N = 20	N = 3	mean
Direct costs
Drugs	13,672	6,614	4,561	1,683	3,661
Hospitalisation	38,778	6,249	0	0	2,258
Outpatient care	3,776	1,214	1,530	1,560	1,590
Equipment aids	403	88	0	0	26
Oxygen therapy	2,957	0	0	0	111
Total direct costs	59,586	14,165	6,091	3,243	7,646
95%CI	8,653–93,794	4,311–21,062	2,800–9,023	−1,997–6,486	4,218–10,378
% of group total	(42.1)	(27.2)	(41.4)	(99.3)	(41.7)
Indirect costs
Disability pensions	77,344	36,860	7,449	0	9,930
Absence from work	4,461	1,031	1,169	24	676
Total indirect costs	81,805	37,891	8,618	24	10,606
95%CI	30,288–128,036	5,371–63,535	−7,167–16,953	−24–47	3,466–15,795
%of group total	(57.9)	(72.8)	(58.6)	(0.7)	(58.3)
Total costs	141,391	52,056	14,709	3,267	18,252
95%CI	51,746–218,793	18,458–79,541	−2,068– 25,350	−1,950–6,533	10,212–24,775

*Total costs (according to Table 4 and 5) were higher in more severe patients (p < 0.0001) and higher in patients with a physician diagnosis than in patients without one (p = 0.0017) according to the regression analysis (Adjusted R-square = 0.74). Regression analysis of the direct costs (Table 4 and 5) gave similar results, p < 0.0001 and p = 0.0027, respectively (Adjusted R-square = 0.71).

Table 5. The annual societal costs per patient for those without physician-diagnosed COPD (SEK, 1999)

	FEV1 categories
	< 40%	40–59%	60–79%	≥80%	Prevalence-weighted
	N = 10	N = 37	N = 59	N = 30	mean
Direct costs
Drugs	13,023	3,263	1,439	700	1,721
Hospitalisation	2,022	7,921	835	0	1,344
Outpatient care	2,140	967	505	958	863
Equipment aids	104	136	521	0	184
Oxygen therapy	0	674	0	0	86
Total direct costs	17,289	12,961	3,300	1,658	4,198
95%CI	8,971–24,549	−180–21,852	930–5,026	377–2,626	2,406–5,704
%of group total	(36.0)	(37.9)	(54.1)	(68.0)	(44.9)
Indirect costs
Disability pensions	29,796	20,596	2,525	0	4,531
Absence from work	790	597	271	779	598
Total indirect costs	30,586	21,193	2,796	779	5,129
95%CI	−13,318–61,171	−3,026–41,191	−2,479–5,513	−779–1,559	1,004–8,377
%of group total	(64.0)	(62.1)	(45.9)	(32.0)	(55.1)
Total costs	47,875	34,154	6,096	2,437	9,327
95%CI	3,308–81,980	7,056–56,918	−691–10,181	−2–4,159	5,162–13,127

*Total costs (according to Table 4 and Table 5) were higher in more severe patients (p < 0.0001) and higher in patients with a physician diagnosis than in patients without one (p = 0.0017) according to the regression analysis (Adjusted R-square = 0.74). Regression analysis of the direct costs (Table 4 and Table 5) gave similar results, p < 0.0001 and p = 0.0027, respectively (Adjusted R-square = 0.71).

The subjects without a diagnosis of COPD who referred themselves as having asthma had lower costs (SEK 14,084) than those with a diagnosis of COPD, but higher costs than those without any diagnosis (SEK 7,343). Also after including these subjects in the group with a diagnosis, the total costs were higher in subjects with severe disease (p = 0.003) and in subjects with a diagnosis than in those without (p < 0.0001) (Adjusted R-square = 0.79). The same relation was found when excluding subjects with FEV₁ ≥ 80% of predicted, (p < 0.0001), and (p = 0.002), respectively (Adjusted R-square = 0.76), and this exclusion decreased the total costs among the diagnosed subjects with SEK 1,708 (or 9%) and among the undiagnosed with SEK 1,274 (14%). Smoking status and age did not significantly influence the total costs in any of the regression models.

Direct costs

The annual direct costs were estimated at SEK 7,646 per person for those who had a physician-diagnosed COPD and SEK 4,198 for those who did not. The main cost driver was drugs in both groups, 48% and 41%, respectively. In mild disease, there was no difference between the groups, and costs for drugs and for healthcare visits or contacts were the main cost drivers. Moderate and severe disease, on the other hand, had different key cost drivers. Hospitalisation was the main cost driver among those with a physician- diagnosed COPD, while drugs were the main cost driver among those without a COPD- diagnosis. The costs were higher for subjects with diagnosed COPD than for those without for all COPD severity levels (p = 0.003). The direct costs were positively correlated to disease severity independent of presence or absence of physician-diagnosis (p < 0.0001).

The inclusion of the asthmatics in the group with a diagnosis resulted in a similar pattern as for total costs. Severe disease (p = 0.003) and a diagnosis (p < 0.0001) were related to higher costs (Adjusted R-square = 0.78). The relation was similar when subjects with FEV₁ ≥ 80% of predicted were excluded (Adjusted R-square = 0.74). As for total costs, age and smoking status had no significant influence on the direct costs.

Indirect costs

The annual indirect costs were estimated at SEK 10,606 per person for those who had physician-diagnosed COPD and SEK 5,129 for those who had not. The main cost driver was loss of productivity due to disability pension both among those with as well as without a physician-diagnosed COPD. The greatest difference was found in severe disease, where costs for disability pension amounted to SEK 77,344 and SEK 29,796, respectively. Also the indirect costs were positively correlated to disease severity in both groups.

Costs to society

As the number of subjects with COPD in Sweden was estimated at about 678,000 persons (95% CI 593,751-764,484) [[3], [6], [27]], the annual total cost to the Swedish society could be estimated at about SEK 5.4 billion for those who had a physician-diagnosed COPD and SEK 3.7 billion for those without a diagnosed COPD. Although the average total costs for subjects without physician-diagnosed COPD were lower, the costs accounted for approximately 40% of the total costs in Sweden, because the prevalence-weighted estimate of the proportion of persons in Sweden lacking a physician-diagnosed COPD was 79%.

However, the composition of the total costs for the society differed considerably between subjects with and without physician-diagnosed COPD (Figure 1). Those with a diagnosed COPD were fewer, but consisted to a greater degree of subjects who had severe and moderate COPD. Persons without physician-diagnosed COPD were more common, but, on the other hand, they had a milder and less costly disease. So, the non-diagnosed subjects accounted for the largest proportion of the total costs in the mild and moderate disease groups.

Figure 1 The distribution of total societal costs among diagnosed and undiagnosed subjects with COPD.

DISCUSSION

Reliable and valid knowledge about the size and types of costs that different diseases impose on the society is important in order to plan and implement health-policy interventions. COPD is a common, disabling and frequently undiagnosed disease [[2], [9]]. It increases considerably in prevalence by age and is strongly associated with smoking [[1], [2], [3], [4]]. Due to the great proportion of undiagnosed subjects it is needed to study the costs also for persons without a physician-diagnosed disease, in order to fully understand the total societal burden of COPD.

In this study we examined the cost differences between subjects with and without physician-diagnosed COPD. Costs were measured at an individual level and aggregated to the level of Sweden as a whole by the estimated prevalence of the disease. No previous study has analysed the cost differences between diagnosed and undiagnosed persons with COPD, and these results, thus, provide a more complete understanding of the cost of COPD. We found large cost differences between these two groups based on presence or absence of physician-diagnosis of COPD. The individual costs for those with physician-diagnosed COPD were twice as high as for those without diagnosed COPD. One possible interpretation of that observation could be that the presence of a physician-diagnosis identifies those patients who are heavily affected by their disease, which does not seem unlikely.

The societal costs for those without a physician-diagnosed COPD accounted, however, for 40% of the total costs of COPD, reflecting an underdiagnosis of COPD in society [[2], [5], [6], [7], [8], [9]]. Still after including those referring themselves as having asthma among those with a diagnosis, the pattern was similar. As the costs for subjects with FEV₁ ≥ 80% of predicted were limited, the exclusion of them did not significantly change the relation of costs between diagnosed and un-diagnosed subjects. By using a multiple regression model, we could demonstrate that there were differences in costs between the different groups of severity and between subjects with and without a physician-diagnosis also after correction for possible other determinants of costs. The analysis showed that severity of disease and having a diagnosis of COPD (both when including and excluding the subjects regarding themselves as asthmatics) were significantly related to both total and direct costs, while smoking status itself and age did not significantly influence the costs. Thus, the costs are related to the disease and not to the causes of the disease. Our data have been used together with data from other studies to create a regression model for estimating costs of COPD [[27]].

COPD is largely underdiagnosed [[6], [7], [8], [9]]. Studies in northern Sweden confirmed that the disease could go unrecognised until a severe and irreversible loss of FEV₁ has taken place. Only approximately 30% of subjects with airflow limitation defined as FEV₁/VC < 0.7 and FEV₁ < 80% of predicted values, which conforms with the BTS criteria for COPD, had been diagnosed as having either chronic bronchitis, emphysema, or different labels of COPD prior to the study [[2], [5]]. Similar proportions of undiagnosed COPD were found in the US NHANES III study [[6]], the Dutch DIMCA study [[7]], and the Spanish IBERPOC study [[8]].

Questionnaires are commonly used in respiratory epidemiological studies. Register-based data are useful when the diagnoses made by clinicians are clear, as can be the case for instance with cancer. For many diseases, it is difficult to distinguish a symptomatic subject from a subject with the disease under study. This is particularly true for COPD, as subjects often do not seek health care for their symptoms and adapt their lifestyle and activity levels to their disease. Subjects contacting health care for common colds, but also other respiratory complains are often not examined with spirometry and remain undiagnosed. Others prefer to label themselves as asthmatics. Furthermore, subjects do not always receive the correct diagnosis even when seeking health care. This is especially true for non-smokers, which is supported by the observation that 23% of the undiagnosed subjects in our study sample were non-smokers.

Also, subjects with FEV₁ < 40% of predicted values had not previously been identified by the health care system. However, the proportion of diagnosed subjects increased with increasing impairment of lung function. Earlier detection and diagnosis of COPD are needed to limit progression of the disease to more severe and costly stages. When estimating the societal costs for COPD, register-based studies do not provide information about the large number of undiagnosed subjects. Further, data collected in structured interview-based studies can be detailed, and the risk of selection bias and under estimated costs is reduced because it is possible to “coach” the patient.

Interview-based studies also have limitations. The questionnaire must contain clear questions that are easy to understand in order to minimize the risk of errors. In a COPD population, which mainly consists of elderly, there is an obvious risk for subjects misunderstanding questions. The questionnaire must be standardized, which means that the questions must be formulated in exactly the same way in all settings. In this study, all interviews were carried out by one person (SAJ) to eliminate inter-observer bias [[15]]. The same sample has previously shown high competence in completing five different health-related quality of life questionnaires [[28]].

Patients with severe COPD may be underrepresented in general population samples. They are mostly elderly, are often hospitalised, and may have difficulties to participate. This may have underestimated the costs somewhat in our study. However, the participation rate was particularly high among the elderly. Further, the proportion of subjects using oxygen therapy corresponded well with the reported proportion from the hospital, which is managing the oxygen therapy in the study area.

Several studies have shown that there is a risk of recall bias in patients' self-reports [[29], [30]]. In this study, structured interviews were performed every third month over 1 year in order to decrease that risk. Further, a patient diary was used in the periods between the interviews in order to minimize the risk even more. In order to obtain the correct doses and strengths of the medications, the patients were told to bring their medications to the telephone when interviewed.

There are difficulties in comparing results from studies with different study designs. One study previously performed in Sweden showed that the total costs for COPD amounted to SEK 2.8 billion (USD 339 million, EUR 318 million) by using register data [[13]]. The large number of un-diagnosed subjects with COPD largely explains the discrepancy in costs between this study and the previously performed register-based study. Presuming that our assumption about the prevalence is correct, underdiagnosis would result in an underestimation of costs of 3-4 billions SEK. In addition, we included cost components that the register-based study could not do (absence from work, equipment aids and oxygen therapy). On the other hand, some element of bias may have occurred in our cost estimates. The prevalence estimates from the northern part of Sweden may be somewhat high, although the prevalence of symptoms common in COPD in central and southern Sweden is similar to those in northern Sweden [[17], [31], [32]]. Thus, when transforming the costs to Sweden as a whole, the costs may have been overestimated. However, this possible overestimation is probably much less in magnitude than the underestimation in the register-based study.

The data in this study are valid, since they reflect a real-life setting. Though a limited study sample, the results can be considered as reliable, as representative samples with subjects fulfilling criteria for COPD generated from large-scale studies of the epidemiology of obstructive airways diseases were used, well reflecting the general population in northern Sweden.

Further, the study sample was stratified by age and severity of disease thus allowing prevalence-weighted estimates. However, there is a risk that the costs for the undiagnosed subjects could have been somewhat biased, as some subjects may have been diagnosed as having COPD by a physician without having been informed.

In conclusion, we found that the costs due to COPD differed considerably between those with and without a physician-diagnosed disease. The per patient costs for those with a diagnosis were on average twice as high as for those without a diagnosis, but when estimating the total costs for the society, the undiagnosed subjects accounted for 40%. The costs for the society caused by COPD are considerably higher than has been found in register studies [[13]]. Our data confirm the substantial underdiagnosis of COPD, which has been found in other studies [[2], [5], [6], [7], [8]].

GLOSSARY

• BTS = British Thoracic Society

• COPD = Chronic obstructive pulmonary disease

• FEV1 = Forced expired volume in 1 second

• GOLD = Global Initiative for Chronic Obstructive Lung Disease

• HE = Health economics

• HRQL = Health-related quality of life

• OLIN = The Obstructive Lung Disease in Northern Sweden

• VC = Vital capacity

APPENDIX

A specially designed and pilot-tested questionnaire was used to record resource use due to respiratory diseases. As the direct costs consist of costs for outpatient care, hospitalisations, medications, oxygen therapy, and equipment aids, we asked questions about every subject's resource use in order to be able to estimate the direct costs. The questionnaire was detailed, and every item consisted of a number of questions. When interviewing the subjects about outpatient care, we asked if it was an outpatient visit, home visit, or a telephone contact, and if the visit was to a specialized physician, primary care physician, nurse, or other specified health care provider. About hospitalisations due to respiratory diseases, a number of questions were asked including how many days and in which clinic the subjects had been hospitalised. Regarding use of medications, the subjects were asked which medications due to respiratory diseases they used, the number of doses, the strength, and if they took the medications every day or as needed.

As the indirect costs consist of loss of productivity due to disability pensions and absence from work due to sick leave, the questionnaire included questions about the work situation for the subjects in order to be able to estimate the indirect costs. The subjects were asked if they were working, had retired, or if they had a disability pension. When early retired, we asked for the cause, and if it was a full-time or a part time disability pension. Subjects, who were working, we asked for how many days they had been reported sick due to respiratory diseases.

ACKNOWLEDGMENTS

The authors thank SRN Ann-Christine Jonsson for help with the collection of data, and the statistician Elsy Jönsson for help with selecting the study sample.

Financial support from AstraZeneca R&D, Lund, Sweden, and the Swedish Heart-Lung Foundation is gratefully acknowledged.