Healthcare Service Use, Costs, and Treatment Patterns within a Cohort of Experienced and New Users of Preventive Migraine Medications

Background: Migraine is a debilitating disease associated with increased use of healthcare services. Pharmacological interventions include acute medications to reduce symptoms and restore patient functioning, and preventive migraine medications (PMM), to reduce frequency, duration, and intensity of migraine symptoms . This study examined treatment and associated healthcare service use and costs between PMM naïve and experienced patients. Methods: Migraine patients initiating treatment with a PMM from January 1, 2010-June 30,2014 were identified in the IBM MarketScan Commercial and Medicare Supplemental Databases. Migraine medication use, service utilization, and costs were examined over the 12 months following PMM initiation; outcomes were compared between patients experienced with and naïve to PMM treatment. Results: Adherence and persistence with PMMs was low, with only 24% of patients adherent to their index PMM. Rates of discontinuation were high, with 71.4% of the sample discontinuing their PMM over the 12-month follow up. Utilization of acute medications was common, as was PMM switching in experienced patients. Annual healthcare costs were $12,044 and $19,093 for the PMM naïve and experienced populations respectively. Migraine-specific service use accounted for approximately 20% of all-cause healthcare costs. The PMM experienced cohort consistently evidenced higher service use and costs than the PMM naïve cohort. Conclusion: Utilization of PMM remains suboptimal and is accompanied by both lack of PMM adherence and high use of acute medications. Rates of PMM switching and use of acute medications suggest that patients have unmet needs regarding migraine management. Improved treatment regimens that effectively manage migraine symptoms are needed to improve patient level of functioning while reducing healthcare costs

disability. [9] Pharmacological intervention, which includes preventive migraine medications (PMM) and acute migraine drugs, is the focus of migraine treatment; although due to their association with transformation to chronic migraine, overuse of acute medications should be avoided. Non-pharmacological approaches, including avoidance of triggers and alleviation of modifiable risk factors, are also important aspects of treatment. [4,9,10] Generically available PMMs are effective in increasing quality of life, reducing migraine frequency, and lowering healthcare resource use and costs; these medications can also help to limit overuse of acute medications. [11][12][13][14][15][16] Despite the benefit in migraine management afforded to patients by PMMs, studies have shown that there is poor adherence to these medications. [17][18][19] Further, PMMs have been found to be underutilized, with fewer than half of those patients eligible for PMMs receiving treatment. [2,5,16,20] Poor management of migraine symptoms is associated with increased healthcare costs. Medications that can effectively manage migraine symptoms stand to not only improve patient outcomes, but also reduce the burden of disease to both patients and society. [4,5,9] This study used the IBM MarketScan Commercial and Medicare Supplemental claims databases to examine real world treatment approaches within a population of PMM users that are treatment naïve and a population of PMM-experienced patients initiating treatment with a new PMM. Patterns of PMM and acute medication use were examined to evaluate the effectiveness of generically available treatments. Further, healthcare service utilization and cost outcomes were examined within the population to evaluate the burden associated with treatment. An implicit goal of this analysis is to provide a backdrop for new and emerging PMMs.

Data Source
This retrospective, observational study utilized administrative claims data contained in the IBM MarketScan ® Commercial and Medicare Supplemental Databases from January 1, 2008 through June 30, 2015. These databases comprise enrollment and demographic information as well as inpatient medical, outpatient medical, and outpatient pharmacy claims data collected from employees, dependents, retirees, and members of more than 200 large self-insured U.S. employers and health plans.

Study Design
The study sample included patients initiating a new course of treatment with an oral PMM.
Patients were required to meet the following criteria: Evidence of a migraine on the index date or during the prior 30 days, as indicated by meeting ≥1 of the following criteria: ≥1 inpatient (IP) claim with a diagnosis of migraine (ICD-9-CM diagnosis 346.xx) in any position (in the 30 days before or on the index date) ≥2 emergency room (ER) claims with a diagnosis of migraine in any position (≥1 claim must have occurred on the index date or during the prior 30 days, the second claim may have occurred at any time in the 12-month period prior to the index date) ≥2 outpatient (OP) claims with a diagnosis of migraine in any position (≥1 claim must have occurred on the index date or during the 30 days prior to index, the second claim had to have occurred between 7-180 days before that claim) ≥1 OP claim with a diagnosis of migraine in any position and ≥1 pharmacy claim for a triptan or ergotamine or topiramate (≥1 of these claims in the 30 days before or on the index date and the 2nd 7-180 days before that claim) ≥2 pharmacy claims for a triptan or ergotamine (≥1 claim must have occurred on index or during the 30 days prior to index, the second claim must have occurred between 7-180 days before that claim) No evidence of HIV or malignancy during the baseline or follow-up periods No evidence of onabotulinumtoxinA administration during the baseline or follow up periods Eligible patients were further classified as either PMM experienced, defined as having evidence of prior PMM use during the baseline period, or PMM naïve, defined as having no 6 claims for a PMM during the baseline period. PMM experienced patients were additionally required to have a diagnosis for migraine within the 90 days prior to their first PMM claim in the baseline period to ensure the medication was prescribed for migraine and not for other indicated conditions. The PMM experienced cohort was further stratified into a subgroup that used only one class of PMM during the baseline period and a subgroup that used two or more PMM classes during the baseline period.

Outcomes
Patient demographics and comorbidity burden were examined at study index, and over the last 12 months of the baseline period, respectively. All-cause and migraine-specific healthcare service utilization, including treatment with PMMs or acute migraine medications, and associated costs were assessed over the 12-month follow up period.
Migraine related service utilization and costs were classified as inpatient claims with a diagnosis for migraine in the primary position and outpatient claims with a diagnosis for migraine in any position. Migraine related pharmacy use was classified as claims for PMMs or acute migraine medications. PMMs included: angiotensin-converting enzyme inhibitors (ACE-I), alpha-agonists, angiotensin II receptor blockers (ARB), anticonvulsants, antihistamines, beta blockers, calcium channel blockers (CCB), serotonin-norepinephrine reuptake inhibitor (SNRI) antidepressants, and tricyclic antidepressants, while acute migraine medications included: barbiturates, ergots, muscle relaxants, neuroleptics, nonsteroidal anti-inflammatory drugs (NSAIDs), opioids, other analgesics, and triptans.
Adherence and persistence with PMM therapy was examined over the 12-month follow up period. Adherence was calculated as the proportion of days covered (PDC) defined as the number of days with a PMM on hand divided by the 12-month follow up (365 days); PDC was expressed as a percentage between 0 and 100%. Patients with a PDC value of 80% or greater were classified as being adherent to PMMs. Persistence was calculated as the 7 number of days from study index to PMM discontinuation, with discontinuation defined as a gap in therapy ≥ 60 days or the end of follow up. PMM utilization after discontinuation was examined over the remainder of the follow up period. Use of acute migraine medications during the follow up period was also assessed; specifically, utilization of acute medications over the 30 days prior to and 90 days following PMM discontinuation.

Analyses
For sample characteristics and outcomes, categorical variables were reported as frequency and percent, while continuous variables were reported as mean, median, and standard deviation. Differences between the PMM naïve and PMM experienced cohorts were examined using chi-square tests on categorical measures and t-tests on continuous measures. P-values <0.05 were a priori considered statistically significant. All analyses were conducted using SAS version 9.4 (SAS Inc., Cary, NC).

Patient Sample
A total of 74,533 patients met the study eligibility criteria. Of these patients, 29,919 had no evidence of PMM use in the baseline period and were classified as PMM naïve. The remaining 44,614 patients were classified as PMM experienced, as they had PMM prescription claims and a migraine diagnosis within 90 days of the first fill in the baseline period. Among the PMM experienced patients, 26,279 (58.9%) used only one class of PMM during the baseline period, and 18,335 (41.1%) used two or more PMM classes.
The sample was primarily composed of females of middle age with commercial insurance.
PMM experienced patients were slightly older and more likely to be female compared to PMM naïve patients (p <0.001; Table 1). Patients in the PMM experienced cohort also had a higher comorbidity burden than PMM naïve patients as (evidenced by a higher mean Deyo Charlson Comorbidity Index (DCI)); a significantly increased proportion of PMM 8 experienced patients exhibited diagnoses for multiple comorbidities including anxiety and depression compared to the PMM naïve cohort (Table 1). Further, the PMM experienced cohort was significantly more likely than the PMM naïve cohort to evidence a migraine or chronic migraine diagnosis during the baseline period (p <0.001, Table 1). Among PMM experienced patients, those with two PMM classes during baseline were slightly older (p <0.001) and had a significantly higher comorbid burden than those with one PMM class.

PMM Utilization
During the baseline period, among PMM experienced patients, 58.9% used only one PMM class, 30.0% used two PMM classes, and 11.1% used three or more classes of PMM (Table   2). Anticonvulsants were the most commonly used class of PMM at index in both the naïve and experienced cohorts; other commonly used classes included beta-blockers, tricyclics, and SNRIs. The same pattern of results was observed whether experienced patients used one or more than one PMM class during the baseline period. A significantly larger proportion of the PMM experienced cohort used each of the PMM medication classes at index compared to the PMM naïve cohort, except for anticonvulsants, which were utilized by a significantly larger proportion of the PMM naïve cohort (p <0.001, Table 2). PMM experienced patients were also more likely to exhibit use of multiple PMMs at index (polytherapy) compared to PMM naïve patients (12% v. 2%).
Overall adherence to index PMMs was low, with a mean ± SD PDC of 0.43 ± 0.34; further, only 24% of the population was adherent (PDC ≥80%) to the index PMM over the 12-month follow up period (data not shown in Table). Consistent with the low rates of adherence, rates of discontinuation were high. The majority (71.4%) of the sample discontinued their PMM; mean ± SD time to discontinuation was 162.0 ± 142.2 days (Table 2).
Discontinuation occurred rapidly, with approximately 40% of the sample discontinuing their PMM by one month post-index; a subsequent steady decline in persistence was also observed over the remainder of the follow up period (Figure 1). Rates of adherence and discontinuation were similar between the PMM experienced and naïve cohorts. Adherence and persistence rates were also similar among experienced patients with one or more than one PMM class during the baseline period.  (Table 2). Acute migraine medication use temporarily decreased over the 30 days following index PMM discontinuation for all acute medication classes in both the experienced and naïve cohorts. The greatest declines were observed in triptans, barbiturates, opioids, and other analgesic medications. Rates of acute migraine medication use returned to baseline levels in the 90 days following index PMM discontinuation, potentially indicating a nonabatement or resurgence in migraine symptoms following PMM discontinuation.

Healthcare Costs and Service Utilization
Trends observed for all-cause and migraine-specific service use were similar ( proportional trends in costs were similar between experienced and naïve patients.
Pharmacy costs accounted for approximately one-quarter of total healthcare costs, while medical costs accounted for the remaining three-quarters of total healthcare costs.
Outpatient services had the greatest contribution to all-cause medical costs, accounting for more than one-half of all medical costs. Emergency room costs had the smallest contribution to medical costs.
There was also an approximate 1.6-fold difference in migraine-specific healthcare costs between the PMM experienced and naïve cohorts, with the PMM experienced patients having higher costs over the follow up period. Among the experienced cohort, migrainespecific costs were 2.3 times higher in patients with two or more PMMs ($3,992) compared to those of patients with only one PMM ($2,766) during the baseline period (Table 3b).
Again, proportional trends costs were largely similar between the naïve and experienced cohorts for migraine specific costs. Outpatient services again accounted for the majority of migraine-specific medical costs. Inpatient costs had the smallest contribution to migraine specific medical costs, although, PMM experienced members had increased inpatient costs compared to PMM naïve members.
Average costs for individual healthcare encounters were also described in the full sample of both experienced and naïve patients. Although the mean ± SD costs associated with an inpatient admission ($12,050 ± $10,133) or emergency room visit ($1,003 ± $1,028) far exceeded that of office ($124 ± $68) and neurologist visits ($135 ± $76), these latter services were more often utilized, leading to them accounting for a larger proportion of all-cause and migraine-specific healthcare costs.

Discussion
This retrospective study utilized administrative healthcare claims to examine treatment patterns, service utilization, and healthcare costs associated with migraine management 13 within a population of patients newly initiating or switching PMM treatment. Consistent with previous studies on migraine and PMM utilization, our population was found to consist primarily of middle-aged females. [2] Chronic migraine were also relatively rare within the sample, with only 3.1% of the PMM naïve sample and 7.2% of the PMM experienced sample having a specific diagnosis code indicating chronic migraine. These proportions largely replicate those found previously; the lower proportion of chronic migraine within the PMM naïve sample may be due to reduced diagnosis of chronic migraine or utilization of chronic migraine codes within newly treated migraine patients. [4,21] The reduced rate of chronic migraine in the PMM naïve sample may also point to transformation from episodic to chronic migraine over the course of disease within the sample. [4,21] Finally, it may be that chronic migraine patients have "given up" in terms of seeking effective preventive medications.
Migraine specific and all-cause healthcare costs were increased within the PMM experienced cohort, and within the experienced cohort, were higher for those who used two or more PMMs during baseline compared to those who used only one PMM. Although the increased comorbidity burden within the PMM experienced patients could account for a portion of this increase, the similar increase in migraine-specific costs suggests that comorbidities alone do not account for the increased cost of care observed within the PMM experienced cohort. It is likely that at least a portion of these costs emanate from the increased duration or severity of disease. [21] From a treatment perspective, the results of this study confirm findings from other analyses that adherence and persistence with PMMs is poor. [17][18][19] Low rates of both adherence and persistence were observed in both PMM experienced and naïve patients suggesting that lack of patient education or migraine experience within the PMM naïve cohort is not the only reason for poor adherence and discontinuation. Treatment patterns also indicate that currently available PMMs may leave patients with unmet needs regarding migraine management. In this study, high rates of PMM discontinuation followed by PMM switching, poly-therapy, or reliance on acute migraine medications were observed.
These patterns of treatment suggest that currently available PMMs may not be fully effective or may be associated with side effects, causing patients to switch to different PMMs or supplement their treatment regimen with either acute medications or polytherapy regimens. Additionally, cyclical patterns of PMM treatment indicated by rates of discontinuation, switching, and PMM re-initiation, especially within the PMM experienced sample, indicate that many patients are unable to manage their migraine symptoms, even with revisions to their treatment regimen. Acute medications intended to treat symptoms after onset, as opposed to preventing symptoms, were found to account for well over half of all migraine-specific pharmacy costs. This finding indicates that most patients are not relying on preventive medications to manage their disease, but instead are treating breakthrough pain in a reactive fashion.
The overall picture of the PMM experienced cohort was one of multiple medication switches, increased use of acute medications, thus increasing risk of transformation to chronic migraine, and increased costs compared with PMM naïve patients. Optimization of migraine treatment regimens to take a more proactive approach to symptom management is needed not only to help patients manage their migraine symptoms, but also to help reduce the cost of care associated with migraine management.
Limitations of this study include those that pertain to any administrative claims-based study, as these data are collected for the purpose of facilitating payment for medical services and lack clinical specificity found in medical records and physician notes. Further, this study sample was comprised of patients covered by commercial insurance; therefore, findings may not be generalizable to populations with other forms of insurance or to the uninsured. PMMs examined as part of this analysis are not migraine-specific therapies, although patients were required to have a diagnosis for migraine. Despite the requirement for a migraine diagnosis in close proximity to the initiation of PMM therapy, the possibility that PMMs may have been prescribed for conditions other than migraine prevention cannot be ruled out. Finally, patients were required to maintain 24 months of enrollment in the database prior to study index and an additional twelve months of enrollment following the index date, therefore study outcomes were not captured for those who discontinued enrollment during follow-up.
Overall this study found that despite the utilization of PMMs and other acute agents, many patients are unable to adequately manage symptoms of migraine and evidenced by ongoing use of acute migraine medications and migraine healthcare utilization. The more experienced the patients in terms of PMMs, the greater the acute medication use and the greater the cost. Since adherence was poor, the strategy of multiple sequential switches with current medications seems counterproductive. The results suggest that new treatments with greater efficacy or increased tolerability are needed in order to improve management of migraine symptoms and disease and overall disease progression. In conclusion, the low rates of adherence and persistence observed within this sample almost certainly contribute to poor migraine management; they likely also contribute to escalated healthcare costs, some of which are associated with inefficient or ineffective management of migraine symptoms. Further study to elucidate the underlying factors that contribute to poor adherence and persistence are certainly warranted.

Declarations
Ethics approval and consent to participate This study used anonymized, de-identified retrospective claims data from the MarketScan databases, and no patient identifiable data were used. Data were analyzed and reported 16 on a group level, and Institutional Review Board approval was not required.

Consent for publication
Not applicable.

Availability of data and materials
The data that support the findings of this study are available from IBM, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of IBM.