Abstract
Objective:
Relapse is a common measure of disease activity in relapsing-remitting multiple sclerosis (MS). The objective of this study was to test the content validity of an operational algorithm for detecting relapse in claims data.
Methods:
A claims-based relapse detection algorithm was tested by comparing its detection rate over a 1-year period with relapses identified based on medical chart review. According to the algorithm, MS patients in a US healthcare claims database who had either (1) a primary claim for MS during hospitalization or (2) a corticosteroid claim following a MS-related outpatient visit were designated as having a relapse. Patient charts were examined for explicit indication of relapse or care suggestive of relapse. Positive and negative predictive values were calculated.
Results:
Medical charts were reviewed for 300 MS patients, half of whom had a relapse according to the algorithm. The claims-based criteria correctly classified 67.3% of patients with relapses (positive predictive value) and 70.0% of patients without relapses (negative predictive value; kappa 0.373: p < 0.001). Alternative algorithms did not improve on the predictive value of the operational algorithm. Limitations of the algorithm include lack of differentiation between relapsing-remitting MS and other types, and that it does not incorporate measures of function and disability.
Conclusions:
The claims-based algorithm appeared to successfully detect moderate-to-severe MS relapse. This validated definition can be applied to future claims-based MS studies.
Introduction
Most patients with multiple sclerosis (MS) begin their life-long experience of the disease with an acute attack, involving most commonly the spinal cord or optic nervesCitation1. The presentation of MS then follows different patterns, and relapsing-remitting is the most common type. Relapsing-remitting MS (RRMS) is characterized by intermittent disease exacerbations with intervening periods of remission. Although residual disability might be evident during remission periods, progressive worsening is not observed during this time.
Relapses provide a gauge of disease activity in RRMS. In clinical practice and in trials, relapse is determined by assessment of neurological symptoms. Relapse is characterized by onset or worsening of neurological symptoms persisting for at least a day, preceded by a period of disease stability in patients with clinicopathologically-diagnosed MSCitation2. Symptoms secondary to a metabolic change, such as fever, are not classified as relapse and must be ruled out. Most relapses evolve over one to 7 days and remit within 1–3 monthsCitation3. Corticosteroids are typically used to treat acute exacerbationsCitation4 but are not generally part of the chronic therapy regimen.
Reducing the rate of relapse is a primary goal of disease-modifying therapy for RRMS, and relapse rate reduction is a widely used measure of treatment efficacyCitation5–9 and real-world effectivenessCitation10. Measuring relapse incidence is also important in studies of disease progression and exacerbating factors. Such studies have shown that relapses may influence the progression of disabilityCitation11–14 and have explored factors, such as infectionCitation15,Citation16 and stressCitation17,Citation18, that place MS patients at risk for relapseCitation19. Like research on precipitating factors, studies of relapse occurrence patterns and their relationships with baseline characteristics also rely on relapse identificationCitation20. Determination of relapse based on neurological assessment is a commonly used outcome measure in prospective clinical studies, however, its use in retrospective studies is currently limited.
Health insurance claims can provide a comprehensive set of postmarketing and epidemiological data which can be used to supplement the information provided by long-term follow-up studiesCitation21–23. Retrospective claims research relies on healthcare codes (International Classification of Diseases [ICD-9]) indicative of diagnoses, pharmaceutical prescription fills, medical procedures, and inpatient and outpatient care settings to track disease occurrence and healthcare utilization. Insurance claims related to the amount and types of care provided to MS patients will carry a diagnosis code for MS, but there is no diagnosis code specific to relapse and currently no validated claims-based method of detecting relapse. Applying this otherwise readily-available dataset to MS research requires defining criteria for relapse based on existing codes. Once defined, such an algorithm could be used for claims-based economic studies of MS.
The objective of this study was to test the content validity of an operational definition of MS relapse applied to healthcare claimsCitation24. To accomplish this, the identification of relapse based on a proposed claims-based algorithm was compared with relapse determined from medical chart review.
Methods
Study design and data source
Data from a healthcare claims database were retrospectively obtained for this validation study. The database includes claims from geographically diverse regions of the US, with coverage heaviest in the South and lightest in the West. Data were available regarding 18.1 million covered lives during the initial identification period. Medical, pharmacy, and enrolment data from commercial health plan members between January 1 2005 and February 29 2008 were used in the study. Healthcare claims data were used to identify MS patients for inclusion in the study, and the number of patients experiencing a relapse during a 1-year period was determined based on a claims-based algorithm as well as an examination of corresponding medical charts. Relapses identified using the algorithm are expected to be moderate or severe due to the need for a physician visit or prescription fill to generate a healthcare claim. The algorithm would not be sensitive to mild or short-duration exacerbations that do not lead to healthcare system utilization.
Claims and chart data were accessed in compliance with the Health Insurance Portability and Accountability Act (HIPAA)Citation25. A limited amount of patient-identifiable information was used only to obtain the medical charts. Patient-identifiable information was not included in the data set analyses. The study was approved by the participating health plan, a privacy board, and an institutional review board.
Patient selection criteria
The MS population was identified from among commercial health plan members based on claims dated during the identification period January 1 2005 through February 28 2007. To be eligible for inclusion, patients were required to have at least two claims with a primary diagnosis of MS (ICD-9-CM code 340.xx) at least 30 days apart during the identification period. Patients were required to be 18 years of age or older as of the date of the first claim with a MS diagnosis code during the identification period; that date was defined as the index date. Continuous enrolment in the health plan with medical and pharmacy benefits during the year after the index date, defined as the follow-up period, was also required.
Claims-based identification of relapse
The operational claims-based algorithm for relapse identification, which classifies MS patients according to care-seeking behavior associated with probable relapse status, was previously described by Ollendorf et al.Citation24. The algorithm required fulfillment of one of two criteria: (1) a claim with a MS diagnosis code in the primary position at any time during an inpatient hospitalization or (2) a claim with a MS diagnosis code in the primary or secondary position in an outpatient setting in addition to a pharmacy or medical claim for a qualifying corticosteroid (dexamethasone, methylprednisolone, prednisolone, prednisone, or adrenocorticotropic hormone [ACTH]) on the day of or within 7 days after the visit. Relapse and non-relapse cohorts were defined based on the operational algorithm. The relapse date was defined as the hospital admission date (criterion 1), or the service date of the first qualifying outpatient visit (criterion 2). Alternative claims-based relapse definitions were also created post hoc and tested against the chart-based detection method.
Medical chart review
Medical charts for a subset of MS patients, composed of equal numbers of patients from the algorithm-defined relapse cohort and from among those not meeting the algorithm-based definition for relapse (non-relapse cohort), were abstracted. Initially, 900 patients (450 with relapse and 450 without relapse) were randomly selected and all necessary approvals were obtained in order to contact physicians and access the medical records of patients in that sample. For each patient, the neurologist associated with the most MS claims for that patient was asked by the chart abstraction vendor to participate in arranging medical record review of the selected patient. If no neurologist was indicated, then the family practice physician associated with the most MS claims was contacted. The first 150 charts for each cohort available from participating physicians were used in the validation. The initial patient pool was thus randomly selected, but the final convenience sample included only patients of voluntary physician participants.
An abstraction tool was developed to extract relevant data from the medical records covering the follow-up period. Relapse was identified in the medical charts based on explicit statement of relapse, the presence of other key words (e.g., exacerbation, attack, flare, worsening or new neurological symptoms), or other descriptions consistent with relapse. Information about medications (e.g., new corticosteroid treatment), test results (e.g., new lesion), and signs and symptoms (e.g., weakness, numbness, sensory loss, diplopia, visual loss, gait disturbance, tremor, optic neuritis) was abstracted and used to infer the occurrence of relapse. Medical charts were abstracted by an independent firm.
Study variables
Data pertaining to patient characteristics and MS-related variables were determined both from administrative claims and medical charts when available. Patient age, sex, and geographic location were determined from health plan enrolment data. Prescribed medications, including corticosteroids (dexamethasone, methylprednisolone, prednisolone, prednisone, ACTH) and disease-modifying therapies (glatiramer acetate, interferon β-1a or β-1b, natalizumab) were determined from administrative claims. Adherence to disease-modifying therapy, a measure of the accuracy to which a prescribed medical regimen is followed, was calculated in terms of a medication possession ratio (MPR; total days supply divided by 365 days) using claims data. MPR greater than or equal to 80% was considered adherent. Information about procedures associated with MS (magnetic resonance imaging [MRI], lumbar puncture, evoked potential testing, plasma exchange) was also extracted from the medical charts. These data were applied in determining relapse based on the operational definition and to describe the MS-related healthcare use of the study sample. The alternative claims-based relapse definitions were based on various combinations of these variables.
Analyses
The data were analyzed descriptively. The relapses identified in administrative claims were compared with relapses abstracted from medical charts to determine true and false positives. The positive predictive value and negative predictive value of the claims-based algorithms were calculated based on the following formulas:
Results
Sample characteristics
Descriptive characteristics of the total MS population in the claims database and the subsample that had medical charts reviewed are shown in and . The majority of patients in the overall MS population as well as among patients with and without relapse were female. Consistent with greater representation of the South in the claims database as a whole, this region was the most heavily represented geographic area in the patient samples (). A total of 182 unique providers participated in the chart review, 97% of whom were neurologists.
Table 1. Demographic characteristics of the sample.
Table 2. MS-related medication in the follow-up period (claims).
Demographics of the subsample of patients with a relapse who had their medical records reviewed were similar to the remainder of patients with a relapse who did not have their charts reviewed; only geographic distribution differed slightly (). Comparisons of nonrelapsers versus relapsers among patients who had their medical records reviewed also suggested that this subsample was representative of the MS population. For example, with regard to age, those without an identified relapse were older than those meeting the definition for relapse in both the overall population (46.4 [SD 10.4] years vs. 43.9 [SD 10.8] years; p < 0.001) and the subsample of patients that had their charts reviewed (47.0 [SD 9.8] years vs. 43.2 [SD 9.6] years; p < 0.001).
Prescribing patterns for the disease-modifying therapies were not unexpected (). Claims for disease-modifying therapy in the subsamples with chart review were generally similar to those of the patients who did not have their charts reviewed. One statistical difference was found among patients without a relapse: a higher percentage used interferon therapy in the chart-review subsample than among those who did not have their records reviewed.
Methylprednisolone was the corticosteroid prescribed to the most patients (). Consistent with the operational definition, patients with relapses would be expected to have more corticosteroid use, although a few patients who did not meet the definition for relapse also had evidence of corticosteroid use. As shown in , both for patients with a relapse and those with no relapse, the percentage of the group that did not receive a corticosteroid was similar between those with and those without chart review. Thus, the distribution between patients who received a corticosteroid and those who did not was similar in the chart-review sample versus the cohort that did not have their records reviewed, although the proportions of patients with claims for specific corticosteroids differed slightly.
When stratified by presence and type of disease-modifying therapy, approximately one-third of patients were identified as having a relapse in each drug group; with the exception of the relatively small subgroup of patients receiving natalizumab.
Algorithm predictive value
The positive predictive value and negative predictive value of the operational claims-based definition of relapse are shown in . Based on the comparison with medical chart data, the claims-based criteria correctly classified 67.3% of patients with relapses (positive predictive value) and 70.0% of patients without relapses (negative predictive value) during the follow-up year. Of the 101 true-positive patients, 94 fulfilled the outpatient plus corticosteroid criterion and 17 had relapse based on inpatient hospitalization (10 patients fulfilled both criteria).
Table 3. Positive and negative predictive value of the operational claims-based definition of relapse.
The positive and negative predictive values of 13 alternative claims-based definitions are presented in . Two of these alternatives applied the operational criteria independently (, definitions 1 and 2). Each component provided a positive predictive value of approximately 68%, but the negative predictive value and the total number of correctly identified relapsed patients decreased relative to the combined algorithm.
Table 4. Positive and negative predictive value of alternative definitions of relapse.
None of the alternative definitions of relapse provided a greater positive predictive value than the operational definition. Two alternative definitions provided equivalent positive predictive values that were slightly lower than that provided by the operational criteria (, definitions 3 and 4). Both of these definitions used narrower criteria related to corticosteroid use during the follow-up period, but did not constrain the time period that the corticosteroid was filled to correspond with a physician visit. The alternative definition providing the highest negative predictive value also had the lowest positive predictive value (, definition 13). This definition incorporated several MS-related care-seeking behaviors, but these broad criteria did not provide relapse-associated sensitivity compared with the other definitions. The original operational definition provided the greatest positive predictive value and acceptable negative predictive value compared with the alternatives tested.
The operational definition was found to perform better among patients under age 45 than for the entire study sample. In this subset of patients (n = 140), the positive predictive value increased to 76.5%, while the negative predictive value decreased to 65.5%.
Agreement between claims and chart data sources
To further examine the validity of using claims information as a proxy for chart-based relapse, we compared the available data regarding relapse dates from claims and medical records. Among 93 patients with a relapse identified in both claims and charts and complete date information, the average number of days between the date of chart-based relapse and claims-based relapse was 28.3. Half of these patients had perfect agreement (same day identified in the claims and chart), and 75% of the sample had no more than 1 week between the date identified in the claims and the date identified in the chart. This comparison supports the validity of the data used in the claims-based relapse identification algorithm.
Claims and medical records were also compared with regard to MS diagnosis dates, disease-modifying therapy, and MS-related procedures. The mean difference between claims-based and chart-based diagnoses was 42.6 days. At least half of the patients had perfect agreement between the dates and at least 75% had no more than 51 days between the date in the claims and the medical chart.
The data source agreement was high with regard to disease-modifying therapy. Among patients with claims for glatiramer acetate, 95.6% also had it indicated in their charts, and 91.8% of claims-based intramuscular interferon β-1a users also had evidence of this prescription in the chart. Claims agreed with charts for 81.8% of patients with a claim for subcutaneous interferon β-1a and 87.5% of patients with a claim for interferon β-1b. Natalizumab use was uncommon in the study sample and none of the patients included in the medical chart review had a claim for natalizumab.
Likewise, evidence of MS-related procedures was in agreement between medical records and claims in most cases. Among patients with claims-based evidence of a MRI, 71.0% also had it noted in their charts. Among those with lumbar puncture indicated in the claims, 81.0% also had evidence of it in their chart, and 66.7% of patients with evidence of evoked potential testing in claims also had it indicated in their chart.
Discussion
The operational claims-based algorithm appeared to be successful in correctly identifying moderate-to-severe MS relapse. The algorithm performed better than all alternatives tested, and was applicable to MS patients receiving any disease-modifying therapy.
A high proportion of relapses were identified based on the outpatient criteria, and these would have been missed if only inpatient visits were considered. This observation suggests that the algorithm detects not only severe relapse requiring hospitalization, but also moderate exacerbations treated in an outpatient setting. The algorithm might not detect mild relapses or relapses of short duration that are not affecting activities of daily living and, therefore, are not treated with a corticosteroid.
Many MS patients in the total population of the claims database who did not meet the algorithm definition for MS relapse received prescriptions for corticosteroids. While corticosteroids that exert strong short-term anti-inflammatory and immunosuppressive actions remain the principal therapy for acute, symptomatic MS exacerbations to decrease the severity and duration of symptoms, corticosteroids are also used to treat many inflammatory conditions such as asthma, inflammatory bowel disease and rheumatoid arthritis and do not represent a unique therapy for MS relapseCitation26,Citation27. The relapse algorithm includes treatment with corticosteroids, but the other aspects of the algorithm contribute to the identification of MS relapse specifically.
The positive predictive value of the algorithm for MS relapse was shown to improve among patients less than 45 years of age. This finding may be the result of the natural history of relapse occurrence in MS. The most common age of onset for MS is in the second and third decade of life. Approximately 85% of all people diagnosed with MS begin with RRMS, but 5–15 years later symptoms begin to worsen. By 10 years 50% and by 20–25 years at least 80% of untreated relapsing patients will move into a secondary progressive state with fewer relapsesCitation28,Citation29. The claims-based algorithm may be more sensitive to identifying the relapses occurring while patients are still in the relapsing-remitting stage of their disease.
This validation of a claims-based algorithm for detecting relapse among MS patients opens up an available data set for research. The claims data accessed in this study had a high degree of agreement with chart data, supporting the reliability of claims data in MS research. Possible applications of the validated algorithm include retrospective claims-based analyses of MS drug effectiveness measured in terms of real-world relapse occurrence, similar to the study conducted by Ollendorf et al.Citation30. Claims-based research using relapse as a study variable can supplement the effectiveness data provided by long-term follow-up studiesCitation31. Retrospective claims data analyses provide shorter-term effectiveness data, but for larger samples. These approaches provide unique results that complement each other. Claims-based identification of relapse could also be used to improve on postmarketing data in studies of MS-related costs, healthcare utilization, and outcomesCitation24,Citation32,Citation33. The ability to identify relapse also broadens the applicability of claims data for epidemiological tracking of MS patients. Claims-based analyses could be used to investigate risk factors for relapse such as sex, age, and infection, although some previously-identified risk factors, such as stress, are not easily monitored with claims data.
Application of the algorithm is subject to certain limitations. The algorithm does not incorporate measures of function and disability, which, like relapse rate, are useful indicators of MS patient status. It does not provide information about the duration of relapse, which could be valuable to know in economic studies. The MS diagnosis code does not differentiate between patients with RRMS and those with other types of MS. Thus, certain healthcare utilization by patients with primary-progressive MS, for example, could be captured as ‘relapse’ based on the algorithm, but would not be defined clinically as an exacerbation. Inclusion of these patients may at least partially explain why the positive predictive value was not greater. Although a few statistical differences were noted regarding characteristics of patients who had their charts reviewed versus those who did not, these are not expected to affect the validation calculations or the applicability of the relapse identification algorithm. Approximately one-third of MS patients who had a clinical relapse did not have evidence of it in their healthcare claims. Attempts to identify this remaining third by adding treatments other than steroids and other putative relapse indicators improved the negative predictive value in a few cases and increased the number of correctly identified relapses in a few cases, but did not improve the positive predictive value of the algorithm. Some of these patients might have had a mild attack that did not warrant corticosteroid treatment or hospitalization, and thus did not meet the algorithm criteria. These observations support the idea that the algorithm detects moderate-to-severe relapse, and that identification of mild attacks might not be practical in claims research. Finally, this validation is limited by discrepancies between claims and chart data that arise due to the constraints of each data source. For example, claims and chart prescription information would differ if patients were prescribed medications which they never filled, or if the prescription came from a general practitioner rather than the neurologist who provided the medical record.
Conclusion
The proposed claims-based algorithm for identifying MS relapses appears to effectively capture moderate-to-severe relapses in claims data. This validation demonstrates the relevance of the claims-based measure as a proxy for chart data, and provides opportunity to incorporate existing data sources into MS research.
Transparency
Declaration of funding
This study was sponsored by Teva Neuroscience, Kansas City, MO, USA.
Declaration of financial/other relationships
M.O.-B. has disclosed that she is an employee of Teva Neuroscience. B.J.C. and M.V.L.-B. have disclosed that they are employees of i3 Innovus, a company that was contracted by the study sponsor to conduct the study.
Both reviewers have disclosed that they have no relevant financial relationships.
Acknowledgments
The authors thank Jonathan Johnson, i3 Innovus, Eden Prairie, MN for performing the statistical analysis, Jessica Wegner, i3 Innovus, and Health Information Solutions, Tampa, FL, for organizing and performing the chart abstractions, Howard Zwibel, MD, Neuroscience Consultants, Comprehensive Multiple Sclerosis Center, Coral Gables, FL, for assistance with developing the chart abstraction form, and Elizabeth J. Davis, PhD, i3 Innovus, for medical writing assistance.
Preliminary results of this study were presented at CMSC 24th Annual Conference, June 2–5, 2010, San Antonio TX, USA; ISPOR 15th Annual International Meeting, May 15–19, 2010, Atlanta GA, USA; and ECTRIMS 2009: 25th Congress of the European Committee for Treatment and Research in Multiple Sclerosis, September 9–12, 2009, Düsseldorf, Germany.
References
- Compston A, Coles A. Multiple sclerosis. Lancet 2002;359:1221-31
- McDonald WI, Compston A, Edan G, et al. Recommended diagnostic criteria for multiple sclerosis: guidelines from the International Panel on the diagnosis of multiple sclerosis. Ann Neurol 2001;50:121-7
- Johnson KP. Therapy of relapsing forms. In: Burks JS, Johnson KP, editors. Multiple sclerosis: diagnosis, medical management, and rehabilitation. New York: Demos, 2000
- Filippini G, Brusaferri F, Sibley WA, et al. Corticosteroids or ACTH for acute exacerbations in multiple sclerosis. Cochrane Database Syst Rev 2000;CD001331
- Polman CH, O'Connor PW, Havrdova E, et al. A randomized, placebo-controlled trial of natalizumab for relapsing multiple sclerosis. N Engl J Med 2006;354:899-910
- The IFNB Multiple Sclerosis Study Group. Interferon beta-1b is effective in relapsing-remitting multiple sclerosis. I. Clinical results of a multicenter, randomized, double-blind, placebo-controlled trial. Neurology 1993;43:655-61
- Johnson KP, Brooks BR, Cohen JA, et al. Copolymer 1 reduces relapse rate and improves disability in relapsing-remitting multiple sclerosis: results of a phase III multicenter, double-blind placebo-controlled trial. The Copolymer 1 Multiple Sclerosis Study Group. Neurology 1995;45:1268-76
- Jacobs LD, Cookfair DL, Rudick RA, et al. Intramuscular interferon beta-1a for disease progression in relapsing multiple sclerosis. The Multiple Sclerosis Collaborative Research Group (MSCRG). Ann Neurol 1996;39:285-94
- Randomised double-blind placebo-controlled study of interferon beta-1a in relapsing/remitting multiple sclerosis. PRISMS (Prevention of Relapses and Disability by Interferon beta-1a Subcutaneously in Multiple Sclerosis) Study Group. Lancet 1998;352:1498-504
- Khan OA, Tselis AC, Kamholz JA, et al. A prospective, open-label treatment trial to compare the effect of IFNbeta-1a (Avonex), IFNbeta-1b (Betaseron), and glatiramer acetate (Copaxone) on the relapse rate in relapsing–remitting multiple sclerosis: results after 18 months of therapy. Mult Scler 2001;7:349-53
- Young PJ, Lederer C, Eder K, et al. Relapses and subsequent worsening of disability in relapsing-remitting multiple sclerosis. Neurology 2006;67:804-8
- Binquet C, Quantin C, Le TG, et al. The prognostic value of initial relapses on the evolution of disability in patients with relapsing-remitting multiple sclerosis. Neuroepidemiology 2006;27:45-54
- Gorman MP, Healy BC, Polgar-Turcsanyi M, et al. Increased relapse rate in pediatric-onset compared with adult-onset multiple sclerosis. Arch Neurol 2009;66:54-9
- Hirst C, Ingram G, Pearson O, et al. Contribution of relapses to disability in multiple sclerosis. J Neurol 2008;255:280-87
- Tremlett H, van dM, I, Pittas F, et al. Monthly ambient sunlight, infections and relapse rates in multiple sclerosis. Neuroepidemiology 2008;31:271-9
- Correale J, Fiol M, Gilmore W. The risk of relapses in multiple sclerosis during systemic infections. Neurology 2006;67:652-59
- Potagas C, Mitsonis C, Watier L, et al. Influence of anxiety and reported stressful life events on relapses in multiple sclerosis: a prospective study. Mult Scler 2008;14:1262-68
- Brown RF, Tennant CC, Sharrock M, et al. Relationship between stress and relapse in multiple sclerosis: Part I. Important features. Mult Scler 2006;12:453-64
- Held U, Heigenhauser L, Shang C, et al. Predictors of relapse rate in MS clinical trials. Neurology 2005;65:1769-73
- Tremlett H, Zhao Y, Joseph J, et al. Relapses in multiple sclerosis are age- and time-dependent. J Neurol Neurosurg Psychiatry 2008;79:1368-74
- Birnbaum HG, Cremieux PY, Greenberg PE, et al. Using healthcare claims data for outcomes research and pharmacoeconomic analyses. Pharmacoeconomics 1999;16:1-8
- Berger ML, Mamdani M, Atkins D, et al. Good Research Practices for Comparative Effectiveness Research: Defining, Reporting and Interpreting Nonrandomized Studies of Treatment Effects Using Secondary Data Sources: The ISPOR Good Research Practices for Retrospective Database Analysis Task Force Report-Part I. Value Health 2009;12:1044-52
- Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol 2005;58:323-7
- Ollendorf DA, Jilinskaia E, Oleen-Burkey M. Clinical and economic impact of glatiramer acetate versus beta interferon therapy among patients with multiple sclerosis in a managed care population. J Manag Care Pharm 2002;8:469-76
- Health insurance portability and accountability act of 1996. Public law 104-191, 104th Congress. U S Department of Health & Human Services, 1996. Available at: http://www.cms.hhs.gov/HIPAAGenInfo/Downloads/HIPAALaw.pdf [Last accessed 12 January 2010]
- Frohman EM, Shah A, Eggenberger E, et al. Corticosteroids for multiple sclerosis: I. Application for treating exacerbations. Neurotherapeutics 2007;4:618-26
- Thrower BW. Relapse management in multiple sclerosis. Neurologist 2009;15:1-5
- Weinshenker BG, Bass B, Rice GP, et al. The natural history of multiple sclerosis: a geographically based study. I. Clinical course and disability. Brain 1989;112 (Pt 1):133-46
- Runmarker B, Andersen O. Prognostic factors in a multiple sclerosis incidence cohort with twenty-five years of follow-up. Brain 1993;116 (Pt 1):117-34
- Ollendorf DA, Castelli-Haley J, Oleen-Burkey M. Impact of co-prescribed glatiramer acetate and antihistamine therapy on the likelihood of relapse among patients with multiple sclerosis. J Neurosci Nurs 2008;40:281-90
- Ford CC, Johnson KP, Lisak RP, et al. A prospective open-label study of glatiramer acetate: over a decade of continuous use in multiple sclerosis patients. Mult Scler 2006;12:309-20
- Goldberg LD, Edwards NC, Fincher C, et al. Comparing the cost-effectiveness of disease-modifying drugs for the first-line treatment of relapsing-remitting multiple sclerosis. J Manag Care Pharm 2009;15:543-55
- O'Brien JA, Ward AJ, Patrick AR, et al. Cost of managing an episode of relapse in multiple sclerosis in the United States. BMC Health Serv Res 2003;3:17