Main findings: the burden of patients with UPS
The objective of our study was to assess QoL and costs in a general sample of patients with UPS using generic measures. Patients with UPS reported a poor QoL. Their QoL was decreased the most by limitations in functioning due to physical health, and the least by limitations in functioning due to emotional problems.
The healthcare-related cost was estimated to be €3,122.93 (SD = €2,952.25) PPPY, which was based on the number of visits to healthcare services and the medication use. While the eight medical specialist visits per year in our study were comparable with the seven visits found by Barsky et al. [
10], the 15 primary care visits per year in our study were much higher than the four found by Barsky et al. [
10]. Our findings on primary care visits were more comparable with studies of high-utilizing patients with UPS [
8,
32], in which the mean number of visits per year ranged from 13.6 [
32] to 15.9 [
8]. In contrast to the high number of primary care visits, the mean number of 0.24 hospital days PPPY in our study was extremely low. Smith et al. [
9] found a mean of 1.9 hospital days per patient per three months. This difference might be related to the fact that our sample was drawn from a non-institutionalized population, which thus underrepresented hospitalizations.
The work-related cost due to absenteeism was estimated to be €2,403.92 (SD = €8,442.89) PPPY. The work-related cost due to absenteeism was based on the estimate of 67 disability days per employed patient per year. This number of disability days found in our study was comparable with the results of an earlier study [
6], which found 18.2 disability days per three months in primary care patients with UPS.
The work-related cost due to presenteeism was estimated to be €855.79 (SD = €3,635.31) PPPY and that due to paid substitution of domestic tasks was €433.27 (SD = €1,478.16) PPPY.
The mean total cost per patient with UPS per year was estimated to be €6,815.91.
Our main findings in relation to the literature: the burden of patients with UPS in comparison with other patient groups
As the allocation of healthcare resources to patient groups is partly based on who has ‘the highest burden’ , the burden of patients with UPS was compared with other patient groups and the general population whose QoL or costs had been measured in earlier studies using comparable outcome measures.
The SF-36 subscale means of patients with UPS were compared with those found in patients with major depression, in patients with cancer, and in the general population [
22,
33], resulting in Additional file
1. According to Osoba [
34], the Minimally Important Difference (MID) for SF-36 subscale scores is 10. MID is defined as ‘the smallest difference in score in the domain of interest which patients perceive as beneficial and which would mandate, in the absence of troublesome side effects and excessive cost, a change in the patient’s management’ [
35]. In comparison to other patient groups, patients with UPS had generally lower SF-36 subscale means, meaning a poorer QoL. However, they did not differ a MID from both other patient groups on the following subscales ‘Vitality’ , ‘Role functioning emotional’ and ‘Mental health’. On the ‘Vitality’ subscale, patients with UPS and those with major depression had comparable means that differed more than the MID from patients with cancer who reported higher vitality. On the ‘Role functioning emotional’ subscale, patients with UPS and those with cancer had comparable means that differed more than the MID from patients with major depression who reported lower emotional functioning. On the ‘Mental Health’ subscale, patients with UPS and those with cancer had comparable means that differed more than the MID from patients with major depression who reported lower mental health. Earlier studies that compared QoL of patients with UPS with that of patients with major depression [
4‐
7] also found that patients with UPS reported a poorer QoL in the physical domain and a relatively better QoL in the mental domain than patients with major depression. Earlier studies that compared QoL of patients with UPS with that of patients with a medical diagnosis found that patients with UPS reported a poorer QoL in all domains than those with a medical diagnosis [
8,
9].
The SF-6D median of patients with UPS was compared with those found in patients with mental disorders and chronic physical conditions and in the general population [
36‐
39], resulting in Additional file
2. According to Walters and Brazier [
40], the MID for SF-6D scores is 0.04. In comparison with other patient groups, patients with UPS had generally a lower utility, meaning a poorer overall QoL. The utility of patients with UPS was comparable to that of patients with the following disorders: panic disorder, dysthymia, social phobia, any mood disorder, and cancer. The utility of patients with UPS was more than the MID higher than that of patients with major depression.
The healthcare-related cost of patients with UPS was compared with the findings of the Cost of Illness study in the Netherlands 2007 [
41]. For this comparison, the cost had to be expressed as a percentage of total Dutch annual healthcare expenditure as shown in Additional file
3. The proportion of total Dutch annual healthcare expenditure associated with UPS was estimated to be over four percent. This healthcare expenditure was compared with those of all main disease categories and with those of patients with specific diseases and the general population [
41], resulting in Additional files
4 and
5. In comparison with the healthcare expenditure of all main disease categories, the expenditure of patients with UPS was comparable to that of the category ‘blastomas; cancer and benign tumors’ (see Additional file
4). In comparison with the healthcare expenditure of patients with specific diseases and the general population, only patients with dementia and those with a mental retardation had higher healthcare expenditure than patients with UPS (see Additional file
5).
The work-related cost due to absenteeism in paid work of patients with UPS was compared with that found in earlier studies in the general population, the healthy workforce, and the workforce with chronic illness [
42‐
45], using the mean number of disability days and the mean percentage of absenteeism per patient per year. The percentage of absenteeism is the number of lost working days due to absenteeism divided by the number of working days according to labor contract and expressed as a percentage. As Additional file
6 shows, patients with UPS had the highest number of disability days relative to that of other reference populations, even if patients who had passed the end of the friction period were excluded. Also, patients with UPS had the largest percentage of absenteeism compared with that of other reference populations.
The work-related cost due to presenteeism in paid work of patients with UPS was compared with that found in the general population and the workforce with chronic illness [
43,
44,
46]. As Additional file
7 shows, the mean cost of patients with UPS was higher than that of patients with psychiatric disorders but lower than that of patients with rheumatic arthritis.
The work-related cost due to productivity loss in unpaid work was calculated only if the unpaid work of patients with UPS was substituted by paid professionals. Because other studies summed up the costs for both unpaid and paid substitution [
44,
47], the resulting cost for substitution of unpaid work of patients with UPS could not be compared with that of different reference populations.
Limitations and strengths in the study
A number of limitations merits attention. The data of our cross-sectional study were collected from a randomized controlled trial on the effectiveness of a cognitive-behavioral group training and not from an epidemiological study. As patients had to be referred to this trial by a healthcare provider and patients had to agree to take part in this treatment trial, our sample might have been a selective group. Selection during the referral could have both increased and decreased QoL and costs; possible selection biases could have been only referring patients with more severe symptoms and a high healthcare utilization for reasons such as being those most recognized by physicians as having UPS, or only referring patients with mild symptoms for reasons such as being best treatable in a relative short group training. Also, selection during the acquirement of patients’ informed consent might have both increased and decreased QoL and costs; possible selection biases could have been only getting informed consent from patients with more severe symptoms for reasons such as being the most burdensome and eager to try treatment, or, alternatively, only getting informed consent from patients with mild symptoms for reason such as being the most vital to show up at their first appointment. When looking at the characteristics of our patient group, our patient group seems to be rather a general than a selective sample. The characteristics of our sample can be described as mainly female, average age of 45 years, of whom 41% had a comorbid DSM-IV Axis I disorder and 29% had a comorbid DSM-IV Axis II disorder. These characteristics are in line with those found in other studies showing that UPS is more prevalent in women in their forties [
4,
5,
7,
9,
48], and that 26 to 58% of the patients with UPS have a comorbid DSM-IV axis I disorder and 37 to 88,6% have a comorbid personality disorder [
49‐
58]. Therefore, we believe that our patient group is a representative group for adult patients with UPS as defined in our study.
UPS in our study was defined as physical symptoms that fulfilled the DSM-IV criteria for an undifferentiated somatoform disorder or for a chronic pain disorder. This definition is stricter than the general used definition of UPS which is physical symptoms that cannot be explained on the basis of a known medical condition. The prevalence of both undifferentiated somatoform disorder and chronic pain disorder in general practices totals 14.6% [
7], while the prevalence of UPS in primary care ranges from 20 to 74% depending on the definition and methods used to classify UPS [
1]. As costs in our study were calculated using a prevalence of only 14.6%, the real total costs associated with UPS might be higher, and the mean cost per patient might be lower.
The QoL of patients with UPS was not adjusted for sociodemographic characteristics such as age, gender, education and living situation. Also, the QoL of the patient groups used as reference was not adjusted for sociodemographic characteristics. Not adjusting for sociodemographic characteristics might have affected our findings, as, in general, patients who are older or have a low education report a poorer QoL in the physical domain and patients who are female or not living with a partner report a poorer QoL in the mental domain [
59]. Adjusting for the effects of sociodemographic characteristics seems to be artificial, as some illnesses, such as UPS and breast cancer, have different prevalences in various sociodemographic groups. By eliminating the effects of sociodemographic characteristics, results will visualize the burden of the illness itself and not the burden of patients, whereas treatments are indicated on the latter condition.
QoL and costs were not adjusted for comorbid mental and somatic disorders. As comorbidity reduces QoL [
7,
59] and increases costs [
10], the QoL might have been lower and costs might have been higher due to the comorbidity prevalent in our study. However, isolating these effects would also be artificial, as patients with UPS suffer from many comorbid mental [
54,
60], and somatic disorders [
8,
10].
Another potential limitation is that QoL and costs were measured using self-report questionnaires. Potentially, self-report is subject to errors caused by recall difficulties. Recall is easier: 1) if the time-period between the event and the recall is shorter, a two-week interval having been suggested as best; and 2) if events fluctuate dramatically [
61]. As the recall period of the questionnaire used on QoL and on the healthcare-related costs was four weeks, and as work-related events are likely to have fluctuated only slightly or moderately, errors in the recall may have followed.
The self-report data on the volume of healthcare consumption and production loss were extrapolated to estimate their volume at a one-year interval, an approach that would be appropriate only if the short intervals were chosen at random and were thereby generalizable. If this assumption was not met, the sample would have produced an inaccurate estimation of the volume of healthcare consumption and production loss.
Costs were estimated by multiplying the volume of healthcare consumption with reference prices and by multiplying the volume of production loss with the average productivity costs. Reference prices and productivity costs are average national cost prices and not actually paid prices. The advantage of the use of general cost prices is that they are a valuable contribution to decision-making at national level. However, the drawback is that they do not result in the actually paid costs and are less valuable for regional decision-making.
Work-related costs were based on a small sample of patients that resulted partly from our use of the friction-cost method, which reduced the number of employed patients in the cost estimations. The small sample of employed patients might also have been caused partly by the long duration of UPS in our study group. As our patients had endured their symptoms for a long time, their risk of losing paid employment due to symptoms had been increased.
The standard deviations around the cost estimates were large. This might be due to our relatively small sample size. However, despite a large population-based sample, Smit et al. [
47] also found broad confidence intervals around the cost estimates in their study of the cost of mental disorders. Because cost-data have large standard errors, they concluded that this is a common finding in health-economic studies.
For the cost estimations, both conservative and less conservative methods were used. A less conservative method used in the cost estimation was to carry out the recommendation of the TiC-P manual to assess costs attributable to health problems in general, instead of those only attributable to the target disease to overcome difficulties for patients to distinguish between them. This might have led to an overestimation of costs. On the other hand, costs might also have been underrated by other more conservative methods used. A conservative method was the use of the friction-cost method, which led us to count only the cost of an absence from work of less than 23 weeks [
24]. If we had counted absence from work irrespective of its duration (the so-called human-capital method), the number of patients with absenteeism would have doubled and the estimation of costs would have been higher. Also, the work-related cost due to presenteeism was estimated with a conservative method. The TiC-P manual provides two methods to calculate production loss due to presenteeism; the HLQ method and the Osterhaus method. In the HLQ method, the number of hours needed to catch up with work as reported by patients is used. In the Osterhaus method, the number of days of hindrance due to health problems and efficiency on these days as reported by patients is used. The study of Osterhaus et al. [
62] showed that the HLQ method was more conservative and led to a lower cost estimation for presenteeism than the Osterhaus method. Finally, the work-related cost caused by substitution of domestic tasks had been estimated in a conservative manner as only the hours substituted by paid professionals were used in the estimation of costs. If the volume of total lost household productivity was used, then the number of patients who used substitution of domestic tasks would have been over fourfold and the estimated costs would have been higher.
Another strength of our study is that we examined QoL and costs in patients with UPS who were referred both by primary and secondary services. The resulting heterogeneous population made our results more generalizable than those of most studies that explored only the burden of UPS in primary care [
4‐
10].
QoL and costs were assessed using generic measures. By using a generic measure for QoL which has the additional advantage of representing the different domains of QoL in one weighted score ‘utility’ , our study made it possible to compare patients with UPS with other patient groups to evaluate ‘who is in greatest need’. By using a generic measure for costs which assesses not only a broad spectrum of healthcare consumption but also production loss in paid and unpaid work, the study made it possible to compare patients with UPS with other patient groups to evaluate ‘who consumed which societal resources the most’.
The broad spectrum of cost-related outcomes, such as the number of healthcare visits, disability days, working hours lost to presenteeism in contrast to costs solely, have all the additional advantage to be internationally comparable, i.e. across countries which have different healthcare systems, healthcare services and labor markets. Presenting information on QoL and costs for different patient groups in quantified, unequivocal and comparable way helps politics and policymakers in their decision-making about allocation of healthcare resources.