Treating patients with fibromyalgia in primary care settings under routine medical practice: a claim database cost and burden of illness study

A cross-sectional and multicenter study was conducted from a retrospective review of medical outpatient records. The study population consisted of men and women from five renovated PC centers (Apenins-Montigalà, Morera-Pomar, Montgat-Tiana, Nova Lloreda, and La Riera) that are managed by a health management organization (Badalona Serveis Assistencials S.A. [BSA], Barcelona, Spain) and that cover a population of 110,440 inhabitants, 16.5% of whom are over 64 years of age. The assigned population is primarily urban. The organization is public with a private services supply and is managed according to a business model. The corporation staff, training policy, organization model, and services portfolio are similar to those of most PC centers in Catalonia (Spain), with a decentralized management model and with integrated structural services. This study is based on data obtained from an administrative medical database; consequently, it was not necessary to be approved by an ethics committee.

We analyzed men and women (older than 18 years of age) who were included in the database (n = 63,526). FMS patients, with a code in accordance with ICD-10v10 (International Statistical Classification of Diseases and Related Health Problems, 10th revision) (code M79.7) criteria for this disorder, were included in the analysis if they had at least one claim for FMS between 1 May 2006 and 30 April 2007. Subjects who did not have a claim for FMS were included in a reference group. Subjects referred to other PC centers, belonging to other geographic areas, visiting integrated specialists, or having serious psychiatric illnesses were excluded from the study. The clinical diagnosis of FMS was based on the presence of chronic and generalized bone and muscular pain, according to the ACR classification criteria established in 1990 [2]. FMS is defined by a history of at least 3 months of generalized and continuous pain on both sides of the body, above and below the waist, and axial skeleton, cervical, or front chest pain. Furthermore, there should be pain to touch in at least 11 out of the following 18 symmetric points: occipital, low cervical, trapezium, supraspinal, second intercostal space in the chondrocostal joint, epicondyle, gluteal, greater trochanter, and knee.

The study was performed in two phases. In the first, information on the primary sociodemographic and clinical (comorbidity) variables and the use of health care and non-health care resources (days of leave from work, subjects with permanent disability) related to the FMS was obtained from the database, omitting any data that could identify the patient. Information on resources used by patients as a consequence of their illness which were not financed by the National Health System such as special diets or non-pharmacological treatments (such as massages or acupuncture) was not gathered as it was not included in the database. In the second phase, telephone interviews were conducted of a sample of FMS patients selected at random from the database to evaluate self-perceived health and well-being. Interviews were conducted by members of the research team. The surveys averaged 40 minutes in duration and were conducted in the months of May and June of 2007, immediately after the data on the use of health care and non-health care resources were collected from the database. Prior to the interview, the selected patients were contacted to inform them about the study, obtain their consent for participation, and provide a verbal guarantee of confidentiality. Interviewers were previously subjected to training on the instruments of self-perceived health which had to be administered (see 'Patient-reported outcome instruments used in the study'). Selection of individuals was made using probability techniques (simple random sample stratified by age and gender), and sample size was calculated by adopting the following parameters: confidence level of 95%, bilateral test, infinite populations, precision of 2%, and anticipated prevalence of the illness of 2.2%. The total number of subjects to be interviewed was 212. Persons with a physical or psychiatric disability that limited them from responding to a telephone questionnaire, persons with incorrect phone numbers, subjects who were not located after three calls made on different days, and those who declined to participate were considered losses to the study.

The primary recorded data were age (continuous and by ranges), gender, and personal history (comorbidities) obtained from the PC International Classification (CIAP-2) [17], 7th component of diseases and health problems, consisting of blood hypertension (K86, K87), dyslipidemia (T93), diabetes mellitus (T90, all types), active smoking (P17), alcoholism (P15), obesity (T82), ischemic cardiopathy (K74, cardiac ischemia with angina; K75, acute myocardial infarction; K76, coronary ischemia), cerebrovascular events (including stroke and transient ischemic attack), presence of a cardiovascular event, chronic obstructive pulmonary disease (R95, chronic obstruction of airflow), bronchial asthma (R96), depressive syndrome (P70), failure of all types (heart, liver, and renal), neuropathy, malignancy, and dementia. Morbidity burden (patient's severity) and the number of health problems attended per patient per year were assessed using the Charlson index [18] and the resource utilization band (RUB) method [19].

Health resource utilization obtained from center records consisted of visits or appointments conducted at the PC center, referrals to reference specialists, requests for complementary support tests, emergency room visits, hospitalizations, and drug prescriptions financed by the National Health Service. Non-health care resource utilization consisted of workdays lost in the active population and early retirement (<65 years old) because of permanent disability due to FMS. The cost system design was defined by taking into account the characteristics of the organization and the level of development of available information systems. The unit of assistance used as the base for final calculation was the cost per patient assisted during the study period. Fixed costs (with imputation criteria) and variable costs were considered in accordance with their dependence on the volume of activity. Costs related to staff (salaries), consumer goods, and a set of expenses related to external services, in accordance with the General Accountability Plan for Health Care Centers, were considered fixed costs (structural), and those associated with diagnostic, therapeutic, or referral requests performed by staff at the center were considered variable costs. Economic assessments were generated for (a) complementary tests, including laboratory tests (mean expenses per application), conventional radiology (fee per each test requested), and support tests (fee per each test requested); (b) ordinary or urgent referrals to reference specialist doctors or to hospitals (referral adapted fee); (c) prescriptions (acute, chronic, or requested medical prescriptions; market price per container), and (d) workdays lost (professional average salary) and information regarding early retirement (before 65 years) because of permanent disability due to FMS. The fees used were obtained from analytical accountability studies conducted by the organization or from CatSalut established prices [20]. The mean cost per visit was obtained from average fixed costs, and a final direct distribution was made for each patient assisted during the study period.

Indirect costs were calculated according to human capital methodology [21, 22]. Two main components of these costs were calculated. First, workdays lost due to sick leave in the active population, as a result of FMS, were calculated as the sum of the yearly number of workdays lost (recorded in the BSA database during the study period) multiplied by daily average salary in active subjects. Second, we added the cost to society for those patients with early retirement prior to 65 years of age (permanent disability from usual working activity) due to FMS. These costs were calculated as the sum of the average salary for the calendar year (which is considered to be an opportunity cost) plus the pension received from Social Security because of the permanent disability from performing usual tasks, resulting in early retirement prior to 65 years of age. The annual average professional salary in the year 2006 was €18,714 and the pension received because of a permanent disability from Social Security was €1,906 per year per subject. Thus, the total cost per patient (Cp) was Cp = (mean cost per visit × number of visits [average fixed costs]) + variable costs + indirect costs. All costs were expressed in euros for the year 2007 and are shown as mean cost per patient per year.

Self-perceived health included self-assessment of pain intensity and impact on various facets of the patient's life, health condition, and quality of life as it relates to health. Validated Spanish versions of the following health instruments were administered: (a) Fibromyalgia Impact Questionnaire (FIQ) [23, 24], (b) the modified Brief Pain Inventory (BPI)-Short Form [25, 26], and (c) the five-item questionnaire on European quality of life (EQ-5D) [27, 28]. The FIQ was developed in an attempt to capture the total spectrum of problems related to FM and responses to therapy. The FIQ is a self-administered instrument that takes approximately 3 to 5 minutes to complete. It is scored in such a way that a higher score indicates a greater impact of the syndrome on the person. Each of the 10 items has a maximum possible score of 10. Thus, the maximum possible score is 100. The first item consists of 11 questions that make up a physical function scale. The 11 questions are scored and added to yield one physical impairment score. Each item is rated on a 4-point Likert-type scale. Items 2 and 3 have a 0 to 7 score range and then require standardization to a 0 to 10 scale. Items 4 through 10 are scored in a 0 to 10 score range and they do not require standardization. Finally, the FIQ total store is calculated by adding the scores of all 10 items [24]. The BPI is an instrument developed for use in epidemiological studies and clinical trials to evaluate the effectiveness of pain treatment. It consists of two dimensions: pain intensity (four items) and interference with activities (seven items). Each one of the items is scored using a numeric rating scale from 0 (absence of pain/interference with daily life) to 10 (worst pain imaginable/maximum impact on daily life). Scores for each subscale are obtained using the sums of the partial scores from the corresponding items divided by the number of items in each subscale, although a direct interpretation of each one of the items individually can be made [26]. The pain severity subscale can be interpreted in a categorized manner in three levels of intensity: mild (<4), moderate (≥ 4 to <7), and severe (≥ 7) [29]. A score greater than or equal to 5 on the interference subscale is considered to indicate the existence of pain interference in the patient's daily activities (this interpretation is also valid for each one of the items on the subscale) [30]. The EQ-5D was designed to assess the patient's perceived health status. This is a five-item generic measure of health status to assess the level of abnormality on five dimensions: movement, self-care, daily life activities, pain/discomfort, and anxiety/depression. Scores of these five items may be used to calculate a utility index, ranging from -0.6 to 1.0, with higher scores representing better health status. The instrument also includes a 20-cm visual analogue scale (EQ-5D VAS) ranging from 0 (the worst imaginable health status) to 100 (the best imaginable health status) [27, 28].

As a step prior to analysis, in particular to the source of information pertaining to computerized clinical records (Oficina Médica Informatizada de Atención Primaria Windows version, STACKS, Barcelona, Spain), data were carefully reviewed to study the distribution of frequencies and to check possible recording or codifying errors. The quality of computer-obtained data was considered adequate, and legal confidentiality requirements for recording were maintained as previously mentioned. A descriptive univariate statistical analysis was conducted, and the Kolmogorov-Smirnov test was used to check normality of the distribution. For the bivariate analysis, Student t tests and chi-square tests were used. A bivariate logistic regression analysis and analysis of covariance (ANCOVA) were conducted to adjust FMS findings (comorbidities, premature retirements, type of treatment, and so on) by age and gender. Costs between reference group and FMS patients were compared, in accordance with the recommendations of Thompson and Barber [31], using an ANCOVA with gender and age as covariates (with Bonferroni corrections for multiple-pair comparisons).

From the 110,440 subjects assigned to the five centers initially selected, 80,775 attended PC settings during the study period, with an intensity of use of 73.1% and a frequency of 4.7 visits per 100 inhabitants per year. Ultimately, 63,526 were recruited for the study. During the study period, 304,075 health problems and 494,122 PC visits were recorded; on average, 0.61 health problems were attended per visit per year. One point seven percent (n = 1,081; 95% confidence interval 0.6% to 2.2%) of patients had a diagnosis of FMS according to ICD-10 criteria. General characteristics, sociodemographics, and principal comorbidities of patients studied are presented in Table 1. The mean age of FMS subjects was slightly but significantly higher compared with the reference group (54.2 [10.1] versus 49.1 [17.9], P < 0.001), whereas the percentage of women was, as expected, significantly higher (96.7% versus 53.5%, P < 0.001). FMS patients showed significant and higher prevalence of a variety of common comorbidities such as psychiatric disorders (major depression, anxiety, and so on), neurological diseases (migraine), pain, and digestive tract diseases (gasroesophageal reflux disease [GERD], gastritis, and so on), with significantly higher values of morbidity burden (RUB and Charlson index). Of particular interest were migraines, major depressive disorders, irritable bowel syndrome, GERD, and pain that had gender- and age-adjusted odds ratios above 2 compared with the reference population. FMS patients also showed significantly higher values of total cholesterol, low-density lipoprotein cholesterol, and triglycerides compared with subjects in the control group (Table 2).

As a consequence of the higher morbidity burden and prevalence of comorbidities, the FMS group was associated with higher use of pain-related medications such as antidepressants, long-acting opioids, analgesics, and muscle relaxants (Table 3). Fifty-two point four percent of patients with FMS used benzodizepines versus 19.6% in the reference group (P < 0.001), 74.7% of subjects used nonsteroidal anti-inflammatory drugs versus 39.3% (P < 0.001), 25.3% used muscle relaxants versus 6.5% (P < 0.001), 30.5% used long-acting opioids versus 3.8% (P < 0.001), and 63.5% used other types of analgesics versus 33.3% (P < 0.001) in the general population. On average, after adjustments for age and gender, subjects with FMS were treated, in one year, with 1.96 more drugs than the reference population (Table 3). In addition, patients with FMS used significantly more health care resources (medical visits, referrals to specialists, and supplementary tests), except for hospital stays, than the reference group (Table 4). It is particularly interesting to note that patients with FMS made, on average, six more yearly medical visits than the reference population: five to the PC physician and one other visit to a specialist or to emergency services (P < 0.001 in all cases) (Table 4). Greater use of health care resources was accompanied by a higher average of workdays missed (20.9 versus 8.0 days, P < 0.001) and a greater number of subjects who receive a pension from Social Security due to permanent disability before the theoretical retirement age (29.9% versus 9.5%, P < 0.001) (Table 4).

Age was positively and significantly correlated to health care costs (r = 0.280, P < 0.001), whereas a weaker correlation, though significant, was shown with indirect costs (r = 0.072, P = 0.018) and total costs (r = 0.112, P < 0.001). This is explained by the fact that costs of lost earnings are not calculated after the theoretical retirement age, only health care costs.

The greater use of health care resources and absenteeism/employment disability in FMS patients is accompanied by significantly higher costs, in both the direct (except for hospitalization costs) and the indirect (sick leave and early retirement) costs. Once adjustments for age and gender were made, FMS subjects incurred €614 more in average annual health care costs (P < 0.001) (Table 4) and €4,397 more in the component of indirect costs (P < 0.001) in comparison with the reference group, totaling an extra annual average cost per patient of €5,010 (P < 0.001) (Table 4).

A second analysis assessed the relationship of total costs, direct and indirect, to the impact of FMS on self-perceived health, as determined using the FIQ, BPI, and EQ-5D scales mentioned above. The measures were administered to a subsample of 200 patients selected at random from among the 1,081 subjects with FMS included in the database. This patient population did not differ from the overall patient sample with FMS in the proportion of women (97.5% versus 96.5%, respectively; chi-square = 0.26, P = 0.612), pensioners (38.0% versus 43.9%, respectively; chi-square = 2.06, P = 0.152), or comorbidities according to the Charlson index (chi-square = 2.16, P = 0.142). Slightly significant differences in the mean age were noted (53.0 [8.5] versus 54.5 [10.4] years old, respectively; t = 1.92, P = 0.06), and as such, the sample to whom the measures were administered was considered to be comparable to and representative of the population of FMS patients in the BSA database. Persons with a physical or psychiatric disability that limited them from responding to a telephone questionnaire (n = 2), those with incorrect phone numbers (n = 3), subjects who were not located after three calls made on different days (n = 5), and those who declined to participate were considered losses to this analysis. In the end, information from the health questionnaires was obtained from a total of 200 patients. The mean score (standard deviation) in the FIQ for this subsample was 71.7 (16.9) points, with 6.9 (1.7) and 6.8 (2.1), respectively, as mean scores on the subscales of pain severity and interference on the BPI questionnaire. More than three quarters of the patients had pain interference with their daily activities (score of at least 5 on the pain interference subscale of the BPI), and 57% stated that the condition of their health had worsened with respect to the previous year (EQ-5D).

Total costs correlated significantly and moderately with the degree of impact of the illness on the patient according to the FIQ (r = 0.282, P < 0.001) due to the fact that both indirect and health costs also were significantly correlated, though moderately, with the FIQ (r = 0.265 and 0.202, respectively; P < 0.001 and P = 0.004, respectively). However, when patients were analyzed with the responses to the FIQ grouped by deciles, a significant linear relationship (P < 0.001) of moderate intensity of the association was observed, both with total costs (R² = 0.55) and with indirect and health care costs (R² = 0.51 and 0.67, respectively) (Figure 1).

The intensity of the pain present in the patient, according to the BPI pain severity subscale, also correlated in a significant and moderate way to the total costs (r = 0.270, P < 0.001) and indirect costs and less with the direct health care costs (r = 0.180, P = 0.011), showing statistically significant differences (P < 0.05) when patients were grouped according to pain intensity (Figure 2). Direct health care costs and pain intensity due to FM were driven largely by pharmacological costs (Figure 3). However, pain interference in general activities and daily lives of the patients measured with the BPI interference subscale correlated in a significant way with total (r = 0.340, P < 0.001), indirect (r = 0.323, P < 0.001), and health care (r = 0.217, P = 0.002) costs. Figure 4 more clearly demonstrates the very significant linear relationship (P < 0.001) with R² association coefficients greater than or equal to 0.7 between the total, indirect, and health care costs and the degree of interference in daily activities caused by the pain as measured by the 11-point numeric rating scale. Finally, when patients were classified according to item 5 of the EQ-5D questionnaire (presence of symptoms of anxiety and/or depression) based on no problems, some problems, and extreme problems, statistically significant increases (P < 0.05) were observed in total and indirect costs, but not in health care costs with greater responses of problems (Figure 5).