Healthcare Utilization and Costs for Insured Patients With Fibromyalgia
Fibromyalgia was related to a pattern of initial increase in utilization and costs, followed by a decrease 7 to 12 months after diagnosis.
Fibromyalgia syndrome is a chronic disorder defined by the presence of pain and tenderness at a minimum of 11 of 18 defined tender points.1 It is commonly associated with a number of other symptoms including fatigue, abnormal sleep patterns, functional bowel disturbances, cognitive dysfunction, anxiety, depression, genitourinary symptoms, and others. Patients with fibromyalgia
experience disability and reduced health-related quality of life.2 The prevalence of fibromyalgia in the United States is approximately 2%,3,4 and the disorder occurs more commonly among females, with a female-to-male ratio of 9:1.5 In recent years, a number of possible pathophysiologic pathways have been described6-10; however, there is still no cure for the syndrome and there is conflicting evidence regarding different types of therapies, making the management of fibromyalgia challenging.11
In order to standardize and improve the approach to the patient with fibromyalgia syndrome, 3 medical societies have developed evidence-based treatment guidelines.12-14 Although the 3 sets of recommendations have similarities, there are some variations due to the different search strategies and inclusion criteria used in selecting pertinent evidence among the 3 societies.11
The guidelines from the American Pain Society (2004)13 and from the association of the scientific medical societies in Germany (2008)14 assigned the highest level of recommendation to aerobic exercise, cognitive behavioral therapy, amitriptyline, and multicomponent treatment, while the recommendations from the European League Against Rheumatism (2007)12 assigned the highest level of recommendation based on available evidence to the use of amitriptyline, tramadol, selective serotonin reuptake inhibitors (SSRIs), and anticonvulsants. The American Pain Society guidelines recommend a multidimensional approach that includes nonpharmacologic and pharmacologic components. 13 The first step according to the American Pain Society guidelines should include educating the patient regarding the diagnosis and the treatment of comorbid illnesses; the second step should include aerobic exercise, cognitive therapy, and tricyclic antidepressants or cyclobenzaprine; and the third step should include
single or combined use of serotonin-norepinephrine reuptake inhibitors (SNRIs), SSRIs,15 tramadol, or anticonvulsants, 13 which have been found to have a beneficial effect. Another recently published meta-analysis by Häuser et al (2009),16 which analyzed 18 randomized placebo-controlled trials with tricyclic and tetracyclic antidepressants (TCAs), SSRIs, SNRIs, and monoamine oxidase inhibitors (MAOIs), found that TCAs had a large effect size for pain reduction in fibromyalgia versus a medium effect size for MAOIs and a small effect size for SSRIs and SNRIs.
The challenges of diagnosing and treating fibromyalgia leave many patients suffering with chronic symptoms that generate high direct healthcare costs17,18 and indirect costs related to loss of revenue and absenteeism.19-21 A prospective multicenter study showed that the mean total healthcare cost over 12 months for fibromyalgia patients was $9573 in 2005 dollars, which was about 3 times higher than the cost for their nonfibromyalgia counterparts. However, the authors recognized that this mean was skewed by high utilizers.22 Penrod et al described fibromyalgia disability and number of comorbidities as important determinants of direct costs; however, the patients in that study were only followed for 6 months.20 A more detailed understanding of healthcare costs and their determinants among patients with fibromyalgia is lacking, particularly during the time that the fibromyalgia diagnosis
is made. Evidence that current fibromyalgia treatment recommendations are followed in real-world practice is also lacking.
We hypothesized that making the diagnosis of fibromyalgia would impact the patterns of utilization by decreasing the number of office visits and the number of tests and procedures performed after the diagnosis, and by increasing the number of rehabilitation visits and the use of medications recommended for the treatment of this condition (eg, antidepressants, muscle relaxants,
anticonvulsants). This study followed subjects for 24 months (12 months before and 12 months after fibromyalgia diagnosis) divided into four 6-month time periods to evaluate how the diagnosis of fibromyalgia modified utilization and costs in real-world clinical practice and to determine whether the treatment patterns reflected current recommendations.
A retrospective cohort analysis using medical and pharmacy claims data for members in Humana’s commercial and Medicare Advantage (with eligibility, medical and pharmacy benefits) population was conducted. Three electronic databases were merged: (1) member file containing demographic information for each member (age, sex, type of insurance, and geographical region); (2) medical file for each member containing up to 9 recorded International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, number of billed units and allowed charges, Current Procedural Terminology codes for procedures, and J codes for in-office injections; and (3) a pharmacy file containing all Medi-Span Generic Product Identifier numbers of pharmacy-dispensed medications, quantity dispensed, and allowed charges.
The study protocol and relevant supporting information were submitted and approved by the University of Miami Institutional Review Board. A limited data set was used in this study in compliance with the Health Insurance Portability and Accountability Act of 1996.
We identified members 18 years and older with at least 2 medical claims (other than tests and procedures) for diagnosis ICD-9-CM code 7291 (myalgia and myositis unspecified)21-23 between June 1, 2002, and March 1, 2007 (identification period), and with at least 24 months of continuous enrollment (at least 12 months prior to and at least 12 months after the diagnosis). The index date was defined as the first medical claim for fibromyalgia.
We included members enrolled in commercial HMO, preferred provider organization (PPO), or Medicare Advantage
plans. To increase the likelihood of capturing incident cases, we required that subjects had no prior diagnosis of unspecified myalgia and myositis in the 12 months preceding the first fibromyalgia diagnosis (index date; Figure 1).
The study period was divided into a 12-month prediagnosis period and a 12-month postdiagnosis period. In order to better observe patterns of change in utilization and costs before and after the diagnosis was made (which would have been less evident if a 12-month prepost analysis had been conducted), the prediagnosis and postdiagnosis periods were subdivided into four 6-month time periods (
): time period −2 (7-12 months before index date); time period −1 (1-6 months before index date); time period +1 (1-6 months after index date); and time period +2 (7-12 months after index date).
Definition of Pain-Related Medications
We defined medications as being “pain-related” based on 2 criteria: (1) if they belonged to medication classes recommended and/or used in clinical practice for the treatment of fibromyalgia22,24 and (2) if they were designated as analgesics or adjuvant medication in the World Health Organization “analgesic ladder.”25 We included in this category opioids, nonsteroidal anti-inflammatory drugs (NSAIDs), antimigraine agents, antidepressants, anticonvulsants, benzodiazepines, muscle relaxants, steroids, and sedatives/hypnotics. We also further divided the opioids, NSAIDs, anticonvulsants, and antidepressant classes. Because tramadol is considered narcoticlike, we reported tramadol as a specific subset of the opioid class.
Baseline characteristics assessed during the 12-month prediagnosis period included demographics (sex, age, and type of health insurance [commercial HMO and PPO, and Medicare Advantage]), comorbidities, and the Deyo-Charlson26 modified comorbidity index. The Deyo-Charlson modified index is an extensively studied and validated comorbidity index, which includes 17 diseases
that have been selected and weighted on the basis of the strength of their association with mortality. It has been adapted for use with ICD-9-CM databases. Comorbidities were determined using ICD-9-CM codes present in all 9 positions of the ICD-9-CM diagnosis codes during the 12-month prediagnosis period.
We analyzed healthcare utilization and costs by 4 broad service categories: outpatient, emergency department (ED), inpatient, and pharmacy. The outpatient category was subdivided into office visits (which included primary care and specialist visits, but excluded psychiatry visits), tests and procedures, rehabilitation and durable equipment, and psychiatric visits. The inpatient category was subdivided into hospitalization and long-term facility. The pharmacy category was subdivided into pain-related and
non—pain-related medications. The pain-related subcategory included the medications described above, and the non–pain-related category included all other medications.
Service categories were defined based on claim type codes, revenue codes, bill type codes, provider specialty codes, and Berenson- Eggers Type of Service codes (details are available from the authors by request). Since more than 1 subcategory might be included in 1 single claim (eg, office visit and test or procedure in 1 single claim), the total number of claims for the outpatient service category might be lower than the sum of the number of claims for each of the subcategories within it (ie, office visits, tests and procedures, rehabilitation and durable equipment, and psychiatric utilization).
Healthcare and medication utilization was measured as the mean number of claims per patient and the percentage of subjects utilizing the service or medication (overall and by medication class). Further, medication utilization was measured as the number and percentage of subjects who filled a prescription for at least 1 medication in each medication category and subcategory. We
also calculated the overall mean number of prescription claims per person during each of the 4 time periods. We used Generic Product Identifier numbers to ascertain medication use. The complete list of Generic Product Identifier codes and names used is available upon request from the corresponding author. We did not include the mean number of prescription claims per person by
medication categories and subcategories because of the very low frequencies of utilization in most of the medication subcategories. A table that depicts these data is also available upon request to the corresponding author, and reveals the same trends shown in the tables included in this report.
Healthcare costs included net amount paid by Humana plus the amount paid by the member including copayments and deductibles. Total allowed reimbursed amounts were used in all analyses of healthcare costs. Each outcome was measured and reported for each of the 4 time periods. That allowed us to evaluate patterns of change in utilization of services and medications, as well as costs during the year that preceded and the year that followed the diagnosis of fibromyalgia.
Descriptive statistics were produced for each set of the study measures. We reported means and standard deviations for continuous variables and frequency counts for categorical variables. Statistical significance of differences in the prediagnosis and postdiagnosis periods was calculated using the paired t test for normally distributed continuous measures; otherwise, a Wilcoxon signed-rank test was used. McNemar’s test was used to determine the statistical significance of differences in categorical measures. All comparisons were made between consecutive time periods. Each table reports 3 distinct P values, which represent the following comparisons: time period −2 to time period −1, time period −1 to time period +1, and time period +1 to time period +2. The purpose of this analysis was to determine the statistical significance of temporal changes in the specified outcomes. SAS version 9.1 (SAS Institute, Cary, North Carolina) was used for all statistical analyses. Statistical significance was defined as P <.05.
reports the baseline characteristics of 11,232 subjects identified with fibromyalgia. Overall, our population was more likely to be female and to be enrolled in a commercial plan. The mean Deyo-Charlson score was 0.73, which reveals that this cohort had a low burden of costly diseases. The 5 most frequent diagnoses among fibromyalgia patients were spondylosis (45.4%), essential hypertension (35.4%), other connective tissue disease (34.7%), other nontraumatic joint disorders (34.5%), and
other respiratory conditions (18.2%).
Utilization of Pain-Related Medications
reports the number and percentage of utilization of pain-related medications. In the prediagnosis time periods, the most commonly used pain medications were opioids, for which 28.1% and 33.0% of the cohort had at least 1 claim in time periods −2 and −1, respectively. NSAIDs were used by 19.8% and 23.1% of the cohort during time period −2 and time period −1, respectively. The percentage of utilization in all major categories of painrelated medications, with the exception of antimigraine medications, increased significantly following diagnosis (time period −1 to time period +1; P <.001). The largest absolute increase from time period −1 to time period +1 was for muscle relaxants, which increased from 15.1% to 20.9% (P <.001), followed by opioids and NSAIDs, which increased from 33.0% to 38.0% and from 23.1% to 27.5%, respectively. However, the largest relative increases for
major categories of pain-related medications from time period −1 to time period +1 were 38.0% for muscle relaxants (absolute change from 15.1% to 20.9%) and 31.0% for anticonvulsants (from 8.9% to 11.7%), driven by an increase of 36.8% for second-generation anticonvulsant medications (from 6.0% to 8.2%). Utilization of antidepressants increased 15.8% (from 22.4% to 25.9%) during those same consecutive periods (−1 to +1). The mean number of prescriptions for pain-related medications filled per member in the 6 months before and after diagnosis of fibromyalgia is shown in the eAppendix, available at www.ajpblive.com.
Among other pain-related medication subcategories, we observed an increase of 46.0% for tramadol (absolute change from 4.8% to 6.9%), 42.2% for TCAs (from 5.9% to 8.3%), 38.1% for SNRIs (from 2.8% to 3.8%), and 37.5% for non-tramadol long-acting opioids (from 2.9% to 4.0%). The use of tramadol and non-tramadol short-acting opioids, NSAIDs, corticosteroids, second generation anticonvulsants, TCAs, and muscle relaxants showed a significant declining utilization pattern from time period +1
to time period +2, while the percentages of utilization of first-generation anticonvulsants, non-tramadol long-acting opioids, sedatives and hypnotics, and SSRIs and SNRIs did not undergo a significant change during this period.
In the outpatient setting, office visits, use of tests and procedures, use of rehabilitation services, and psychiatric utilization showed an increasing pattern from time period −2 through time period +1, with a subsequent declining trend during time period +2. Emergency department visits showed a similar pattern, which was evident from the percentage of members utilizing services and also from the mean number of distinct clinical encounters per member (Table 3). All changes were statistically significant and especially relevant for office visits and for tests and procedures, which had higher mean numbers of encounters up to time period +1. Although the same patterns were seen in the use of inpatient care, the changes in utilization were only significant when comparing time period −1 with time period +1.
When examining changes in healthcare costs related to outpatient visits before and after the fibromyalgia diagnosis by time period, we found that during the prediagnosis year the per patient per month (PPPM) costs increased from time period −2 to time period −1 and reached a peak during time period +1. However, during the postdiagnosis year the opposite pattern occurred, showing a decrease in PPPM costs from time period +1 to time period +2 (
). Per patient per month costs for office visits increased 18.7% ($117.63 to $139.56; P <.001) from time period −2 to time period −1, continued to increase through time period +1 (46.4%; $139.56 to $204.37), and later decreased by 14.4% ($204.37 to $174.31) during time period +2. A similar pattern of change in costs for tests and procedures was observed from time period −2 to time period −1 with an increase of 17.5% ($102.75 to $120.71) followed by an increase of 42.2% during time period +1 ($120.71 to $171.62) and a later decrease of 23.3% ($171.62 to $131.67) during time period +2. Similar patterns of change in costs were observed for rehabilitation, psychiatric, and ED visits. All changes were statistically significant. Costs related to the use of medications (pain-related and not pain-related) had significant increases during all 4 time periods of the study in spite of a reduction in utilization during time period +2 for most of the subcategories analyzed.
We stratified the costs according to type of insurance and observed the same trends and significant findings. However, we did observe some differences between the 2 types of insurance. Most importantly, the PPPM costs tended to be slightly lower among Medicare beneficiaries. The other difference was that the office visits category had smaller variations than those observed in commercial enrollees. The other categories had similar magnitudes of change.
This retrospective observational study of patients with a new ICD-9-CM claim for fibromyalgia evaluated healthcare utilization and costs during 1 year preceding and 1 year following the diagnosis of fibromyalgia, divided into 4 time periods, using a large health benefits claims database. Our cohort had a slightly lower representation of females (70%) than other previously published studies, which included 75% to 81% females. The age distribution of our cohort was similar to that in previous reports, with a higher proportion of diagnosis between the ages of 45 and 54 years.22,26,27
The patterns observed in pain medication utilization in this study showed high emphasis on short-acting opioids and NSAIDs, both before and after the fibromyalgia diagnosis. Although tramadol has been recommended for the treatment of pain in fibromyalgia,12 NSAIDs and more potent opioids are not recommended unless other therapies have failed, and they should be used in conjunction with other medications.12 Several antidepressants, muscle relaxants, and anticonvulsant medications have proved to be effective in the pharmacologic treatment of this condition.11 Moreover, their use is supported by several sets of published recommendations.12-14 A recent meta-analysis (Häuser et al, 2009) also showed that antidepressant medications are associated with improvements in pain, depression, fatigue, sleep disturbances, and health-related quality of life in patients with fibromyalgia.15 Another meta- analysis of 18 randomized controlled trials evaluating the effect of antidepressants on pain reduction showed that while all subcategories have a beneficial effect, TCAs had a large effect size in pain reduction among subjects with fibromyalgia, MAOIs had a medium effect size, and SSRIs and SNRIs had a small effect size.16
However, in spite of the treatment guidelines and results from meta-analyses, only around 11% of patients in this study used second-generation anticonvulsants (eg, pregabalin) and only 25% of patients used antidepressant medications during the year that followed the fibromyalgia diagnosis. Moreover, we observed that TCAs, which are considered the first line of therapy and have a considerable amount of evidence demonstrating their beneficial effects, were used only in a small proportion of patients and had a utilization reduction during the last 6-month time period of the study.
It is encouraging to observe that use of SSRIs, SNRIs, and anticonvulsants increased from the prediagnosis period to the postdiagnosis period; nevertheless, we expected physicians to favor them over other medications such as TCAs, tramadol, and muscle relaxants over the use of short-acting opioids and NSAIDs, particularly during the first year after diagnosis. We acknowledge that a potential explanation for the relatively low use of these classes may be that the first set of recommendations was published in 200413 and this cohort was followed from mid 2002 until 2007 (thus, many practitioners may not have been aware that these medication classes were indicated for fibromyalgia). However, the fact that in our population the largest relative increase was for medications recommended in the guidelines (eg, tramadol, anticonvulsants, muscle relaxants) supports our
hypothesis that making the fibromyalgia diagnosis may lead to therapy that is more reflective of current practice guidelines. Another cohort study previously reported that subjects started on pregabalin had a decrease in the use of NSAIDs and narcotics.27 Further studies should be done to evaluate whether more recently diagnosed cohorts have experienced a larger absolute increase in the use of medications considered first-line therapy and to assess predictors of utilization from the provider and patient perspectives.
In terms of healthcare utilization, the study findings showed that use of all subcategories of outpatient services progressively increased from time period −2 through time period +1, and later showed a significant reduction during time period +2. A potential explanation is that during the time periods that immediately precede and follow the first ICD-9-CM claim related to fibromyalgia (time periods −1and +1), the subject is undergoing the necessary workup to exclude other conditions and that this increased utilization plateaus during the months after the diagnosis of fibromyalgia is made. While our initial hypothesis of decreasing patterns of utilization was not readily apparent, our study supports a decreasing trend for tests and procedures during time period +2. The cost trends mirror the utilization variation before and after the fibromyalgia diagnosis. The increase in costs that occurred in the time periods before and after the diagnosis of fibromyalgia (time period −1 and time period +1) and the significant decrease in costs during time period +2 were mostly related to a decrease in outpatient services, particularly office visits and tests and procedures. Pharmacy costs continued to increase during all time periods; however, the driver in this sector was primarily non—pain-related medications.
These data both negate and affirm our initial hypothesis. Making the fibromyalgia diagnosis was associated with an overall increase in utilization in the short term (which contradicts our hypothesis), but this utilization showed a clear decreasing trend in months 7-12 after the diagnosis. Longer follow-up is needed to determine whether this trend continues beyond 12 months postdiagnosis and overall utilization returns to prediagnosis rates.
A recent study also supports these findings, as it reported that healthcare costs were higher among those recently diagnosed with fibromyalgia compared with those with established fibromyalgia.28 Annemans et al29 previously used costs before diagnosis in a trend analysis to predict later costs assuming the diagnosis had never been made, and these predicted costs were compared with the observed costs after diagnosis. They estimated significant cost savings related to making the fibromyalgia diagnosis
in the Great Britain population, particularly due to a decrease in diagnostic tests.
Future studies should explore whether making the fibromyalgia diagnosis also improves the quality of life of subjects suffering from this condition.
Several limitations deserve mention. First, our study reported all-cause utilization and not fibromyalgia-specific utilization. To compare the utilization of the fibromyalgia subjects with that of the general population, we have selected a propensity score matched control group and will report that methodology and analyses in a separate communication. Second, our selected follow-up time proved to be too short to adequately evaluate our hypotheses. We followed the cohort for only 12 months postdiagnosis and could not observe whether costs continued to decrease beyond 1 year. In order to document whether utilization
continues to decrease or stabilizes over a longer period of time, we are extending the follow-up period and will report these results separately. Third, the nature of claims data analyses does not allow us to determine causality and could be associated with regression to the mean due to the presence of high utilizers. Fourth, the guidelines and pharmaceuticals approved for fibromyalgia changed over time, and these changes were not considered for our analyses. Lastly, this analysis did not take into consideration the effect of copayment differences within the pharmacy benefit structure. While copayment may alter the choice of medication and ultimately the pharmacy cost, the contribution to overall healthcare spending is negligible.
Our study has several strengths. These include a large sample size, the ability to identify very detailed categories of utilization of services and medications, the availability of actual cost data, and a relatively long follow-up period.
After the fibromyalgia diagnosis was made, we observed a pattern of increase in utilization and costs up to time period +1 followed by a decrease in time period +2, particularly for office visits and tests and procedures. Longer follow-up is needed to determine whether this decrease continues beyond 12 months after the fibromyalgia diagnosis and whether utilization returns to prediagnosis levels or is further reduced. The utilization of recommended therapies such as antidepressants, tramadol opioids, and anticonvulsants increased after the diagnosis was made; however, the percentage of fibromyalgia members using those medications in period +2 was lower than the percentages using other nonrecommended medications such as non-tramadol opioids, suggesting that further dissemination of treatment guidelines is needed.