UBC Faculty Research and Publications

Mental comorbidity and multiple sclerosis: validating administrative data to support population-based… Marrie, Ruth A; Fisk, John D; Yu, Bo N; Leung, Stella; Elliott, Lawrence; Caetano, Patricia; Warren, Sharon; Evans, Charity; Wolfson, Christina; Svenson, Lawrence W; Tremlett, Helen; Blanchard, James F; Patten, Scott B Feb 6, 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12883_2012_Article_745.pdf [ 498.36kB ]
JSON: 52383-1.0223764.json
JSON-LD: 52383-1.0223764-ld.json
RDF/XML (Pretty): 52383-1.0223764-rdf.xml
RDF/JSON: 52383-1.0223764-rdf.json
Turtle: 52383-1.0223764-turtle.txt
N-Triples: 52383-1.0223764-rdf-ntriples.txt
Original Record: 52383-1.0223764-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessMental comorbidity and multiple sclerosis:validating administrative data to supportpopulation-based surveillanceRuth Ann Marrie1,2,12*, John D Fisk3, Bo Nancy Yu2, Stella Leung2, Lawrence Elliott2, Patricia Caetano2,Sharon Warren4, Charity Evans5, Christina Wolfson6,7, Lawrence W Svenson8,9,10, Helen Tremlett11,James F Blanchard2, Scott B Patten8, for the CIHR Team in the Epidemiology and Impact of Comorbidity onMultiple SclerosisAbstractBackground: While mental comorbidity is considered common in multiple sclerosis (MS), its impact is poorlydefined; methods are needed to support studies of mental comorbidity. We validated and applied administrativecase definitions for any mental comorbidities in MS.Methods: Using administrative health data we identified persons with MS and a matched general populationcohort. Administrative case definitions for any mental comorbidity, any mood disorder, depression, anxiety, bipolardisorder and schizophrenia were developed and validated against medical records using a a kappa statistic (k).Using these definitions we estimated the prevalence of these comorbidities in the study populations.Results: Compared to medical records, administrative definitions showed moderate agreement for any mentalcomorbidity, mood disorders and depression (all k ≥ 0.49), fair agreement for anxiety (k = 0.23) and bipolar disorder(k = 0.30), and near perfect agreement for schizophrenia (k = 1.0). The age-standardized prevalence of all mentalcomorbidities was higher in the MS than in the general populations: depression (31.7% vs. 20.5%), anxiety (35.6% vs.29.6%), and bipolar disorder (5.83% vs. 3.45%), except for schizophrenia (0.93% vs. 0.93%).Conclusions: Administrative data are a valid means of surveillance of mental comorbidity in MS. The prevalence ofmental comorbidities, except schizophrenia, is increased in MS compared to the general population.Keywords: Multiple sclerosis, Administrative data, Validation, Prevalence, Depression, Anxiety, Bipolar disorder,SchizophreniaBackgroundAlthough depression and anxiety are considered com-mon in MS [1,2], population-based prevalence estimatesfor these conditions are rare. Even fewer prevalence esti-mates exist for bipolar disorder and schizophrenia in theMS population, and they vary widely [3,4]. The paucityof population-based studies of mental comorbidity mayreflect the challenges of conducting such studies. How-ever, such studies are needed given the impact of mentalcomorbidity in MS, including lower quality of life andreduced adherence to treatment [5,6]; and to minimizethe biases from using clinic-based samples.Studies of mental comorbidity could potentially use oneof several data sources including medical records review,self-report, interviews, or administrative data. Administra-tive data are population-based in publicly funded healthsystems such as Canada and are cost-effective and access-ible [7]. Such data are useful for assessing the burden ofdisease at the population level, including health servicesuse and costs [8]. Mental comorbidities can be assessed inclinical samples using structured diagnostic interviewssuch as the Composite International Diagnostic Interview(CIDI) although these are time consuming and depend* Correspondence: rmarrie@hsc.mb.ca1Department of Internal Medicine, University of Manitoba, Winnipeg, Canada2Department of Community Health Sciences, University of Manitoba,Winnipeg, CanadaFull list of author information is available at the end of the article© 2013 Marrie et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.Marrie et al. BMC Neurology 2013, 13:16http://www.biomedcentral.com/1471-2377/13/16heavily on recall of past episodes [9]. Administrative datahave the advantage that they are recorded during an epi-sode and need not be recalled later. Administrative data,however, are collected for health system management andare often inadequately validated [7,10]. Indeed, few pub-lished case definitions for mental comorbidity have beenvalidated, and efforts to develop and validate case defini-tions for depression have identified poor concordancewith the CIDI Short Form [11], and difficulties distin-guishing depression from anxiety [12].We aimed to validate administrative case definitions forseveral mental comorbidities in MS, and to describe theirprevalence among persons with MS versus a matched co-hort from the general population. We hypothesized thatthe prevalence of depression, anxiety, bipolar disorder andschizophrenia would be higher in the MS population thanin the general population.MethodsAdministrative dataWe conducted this study in Manitoba, Canada, usinganonymized administrative data obtained from ManitobaHealth (MH) which provides health care services for morethan 98% of Manitoba residents [13]. Since 1984, MH hasmaintained computerized records of health services claims,which can be linked using a unique personal health identifi-cation number (PHIN) identifying the person who receivedthe service. Physician claims include the PHIN, servicedate, and three-digit International Classification of Disease(ICD)-9-CM code for one physician-assigned diagnosis.Hospitalization records include the PHIN, admission anddischarge dates, and up to 16 discharge diagnoses. Before2004, diagnoses were listed using five-digit ICD-9-CMcodes and since 2004 they have been listed using ICD-10-CA codes. Since 1996, the Drug Programs InformationNetwork captures outpatient prescription drug dispensa-tions including date, drug name, and drug identificationnumber for Manitoba residents, regardless of payer. Thepopulation registry is updated when an individual migratesinto or out of Manitoba, or dies.Study populations and validation cohortUsing data from 1984 to 2006, we identified all Manitobanswith MS using a previously validated administrative casedefinition [14]. We identified up to 5 controls from thegeneral population for each MS case, matched on sex, yearof birth and region of residence (postal code), and exclud-ing anyone with an ICD9/10-code for any demyelinatingdisease as previously described [14]. As described in detailpreviously, the medical records of 430 persons with MSwere reviewed by a trained abstractor using a standardizeddata collection form [14,15]. Using each participant’sPHIN, these clinical data were linked with the administra-tive databases.Administrative case definitionsWe aimed to develop case definitions for depression,anxiety, bipolar disorder and schizophrenia using estab-lished approaches [16]. Developing case definitions formental comorbidity raised challenges. Although hospitalclaims provide 5-digit ICD codes, physician claims inManitoba have only three digits, reducing the specificityof coding. For example, at the 3-digit level, the samecode (296) describes bipolar I disorder, most recent epi-sodic manic (296.4) and major depressive disorder recur-rent episode (296.3). Therefore, we initially created an‘omnibus’ definition for mental comorbidity to capturepersons with any of the mental comorbidities of interest,followed by an ‘any mood or anxiety disorders’ definitionwhich captured depressive disorders, anxiety disordersand bipolar disorder. Finally, we developed case defini-tions for individual mental comorbidities. We identifiedICD-9/10 codes for the individual comorbidities and thecombination definitions (Additional file 1: Table S1).While incorporating prescription claims might improvespecificity of the case definitions, many medications usedfor mental comorbidities are used off-label for other pur-poses, particularly in MS [17]. To determine which medi-cations to include in our definitions we compiled a listof antidepressants (N06A), anti-anxiolytics (N05B), anti-psychotics (N05A), combination agents (N06C), andmood-stabilizing agents including anticonvulsants(N05AN01, N03AG01, N03AX09, N03AX12) avail-able in Canada based on the Anatomic TherapeuticChemical Classification System [18]. A multidisciplinarypanel comprised of two pharmacists (CE, MM), a psych-iatrist (SBP), a neuropsychologist (JDF), an epidemio-logist (SW), and a neurologist (RAM) independentlyreviewed this list and indicated whether each medica-tion was used (i) for each of the mental comorbidities ofinterest; (ii) other on-label uses including the specificcondition; and (iii) off-label uses, especially for MS. Tomeet our goal of enhancing the specificity of case defini-tions with prescription claims, we selected medicationsconsidered to be moderately specific for the comorbid-ities of interest (Additional file 1: Table S1), meaningthat the medication could not be used off-label for MS,and could not be used on or off-label for conditionsother than mental comorbidities unless an ICD codecould easily exclude the condition (e.g. epilepsy).We developed several case definitions for each comor-bidity by varying the number of physician, hospital andprescription claims required and the years of data requiredto classify a person as affected. Using our validation co-hort, we compared the classification of mental comorbid-ity according to the administrative case definitions versusdiagnoses based on medical records review for the 1 to5 year periods ending in fiscal year 2005/06. We report akappa (κ) statistic for agreement between administrativeMarrie et al. BMC Neurology 2013, 13:16 Page 2 of 8http://www.biomedcentral.com/1471-2377/13/16and medical records data [19], and the 95% confidenceinterval (CI) based on the normal approximation to thebinomial distribution. We interpreted κ as follows: slight(0-0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial(0.61-0.80), and almost perfect agreement (0.81-1.0) [19].Kappa indicates the proportion of agreement beyondchance and is calculated as (observed agreement – chanceagreement) ÷ (1 – chance agreement). Kappa is affectedby the prevalence of the condition of interest, however,such that if prevalence is very high or very low, chanceagreement is high and kappa is reduced with a maximumvalue of less than one [20]. Bias refers to the extent towhich the raters (i.e. administrative data versus medicalrecords data) disagree on the proportion of positive(affected) cases; greater bias, meaning a greater differ-ence in the proportion (prevalence) of positive rates, isparadoxically associated with higher kappas. Becauseboth prevalence and bias influence the magnitude ofkappa, we also calculated the prevalence index, biasindex, and the prevalence and bias-adjusted kappa forour preferred case definitions [20]. We estimated that asample of 400 persons can detect a k of ≥0.60 (substan-tial agreement) for comorbidities with ≥4% prevalence ifthe null hypothesis is k = 0.41, α = 0.05, and β = 0.20.Given the anticipated higher k (≥ 0.70) for bipolar dis-order and schizophrenia we estimated that our samplewould provide adequate precision for estimates of kwith prevalences ≥ 3% for these disorders.We computed sensitivity, specificity, positive predict-ive value (PPV) and negative predictive value (NPV) foradministrative definitions versus the “gold standard” ofmedical records review to identify whether an algorithmwould be vulnerable to over- or under-estimating theprevalence of the comorbidity. Further, we explored theimpact of these misclassifications on epidemiologic esti-mates by generating a range of ‘true’ prevalence esti-mates, and calculating the expected value of observedprevalence based on the sensitivity and specificity for thecase definition of interest [7,21].PrevalenceFor each comorbidity, we report the prevalence in theMS and matched cohorts. Once a person met the casedefinition, he or she was considered affected in all subse-quent years while alive and resident in Manitoba. Weestimated the point prevalence of the comorbidity onOctober 1, 2005 using mid-year population figures fordenominators and also calculated prevalence ratios (PR)by dividing the prevalence in those with MS by that ofthe control group. To enhance comparability with otherstudy populations, we age-standardized the findings tothe 2001 Canadian population, and calculated 95% CIsusing the exact binomial distribution. Using Poisson re-gression we calculated PRs and 95% CIs comparing theMS and general populations adjusting for age group(20-44, 45-59, ≥ 60 years) and sex. Cell sizes ≤ 5 weresuppressed.The University of Manitoba Health Research EthicsBoard and the Manitoba Health Information PrivacyCommittee approved the study and data access. Partici-pants in the validation cohort provided written informedconsent. Statistical analyses were performed using SASV9.2 (SAS Institute Inc., Cary NC).ResultsThe MS population included 4192 persons and thematched cohort included 20,940 persons (71.7% female).In the validation cohort most participants were White(91.6%), women (77.0%), with a mean (standard deviation)age at MS symptom onset of 33.2 (11.1) years [16]. Mentalcomorbidity was common in the validation cohort, with29.7% having any mental comorbidity, 29.2% having amood or anxiety disorder, 27.5% having a depressive dis-order, 6.5% having an anxiety disorder, 0.98% having bipo-lar disorder and 0.49% having schizophrenia.Omnibus definition: any mental comorbidityAgreement between the administrative case definitions(labeled A to Z) and medical records ranged from slight tomoderate (k = 0.11 to 0.51, Additional file 2: Table S2).The definition with the highest level of agreementrequired ≥ 1 hospital or ≥ 5 physician claims or ≥ 1physician claim and ≥ 4 prescription claims in 2 years (def-inition ‘Q’, k = 0.51); it had a sensitivity of 63% and a specifi-city of 86.8%. Using definition ‘Q’, the age-standardizedprevalence of any mental comorbidity in 2005 was 33.9%(95% CI 32.0-35.9%) in the MS population and 21.9% (95%CI: 21.2-22.6%) in the general population (PR 1.55; 95% CI:1.36-1.76). In both populations, the peak prevalence oc-curred in persons aged 45-59 years (Figure 1).Any mood or anxiety disorderAgreement between the case definitions (labeled A to Y)and medical records ranged from slight to moderate(k = 0.10 to 0.50, Additional file 3: Table S3). Thehighest level of agreement (k = 0.50) was achieved by sev-eral similar definitions, including definition ‘O’ (see below);all used prescription claims. Using definition ‘O’, (≥ 1 hos-pital or ≥ 5 physician or [≥ 1 physician AND ≥ 4 prescrip-tion] claims) the age-standardized prevalence of any moodor anxiety disorder in 2005 was 34.8% (95% CI 32.8-36.8%)in the MS population and 22.0% (95% CI: 21.3-22.7%) inthe general population (PR 1.58; 95% CI: 1.39-1.80). Thesimilarity of the estimates of any mood or anxiety disorderto those for any mental comorbidity reflects the predomin-ance of mood and anxiety disorders. In both populations,the peak prevalence occurred in persons aged 45-59 years(Figure 1).Marrie et al. BMC Neurology 2013, 13:16 Page 3 of 8http://www.biomedcentral.com/1471-2377/13/16Depressive disordersAgreement between the case definitions (labeled A to Y,Additional file 4: Table S4) and medical records rangedfrom slight to moderate (k = 0.11 to 0.49). The highest levelof agreement for a definition which did not use prescriptionclaims was moderate (k = 0.44), and used ≥ 1 hospitalor ≥4 physician claims in 5 years. Among all case definitionsthe highest level of agreement (k = 0.49) was achieved bytwo similar definitions (‘G’ and ‘P’); both used prescriptionclaims. Using definition ‘P’, which required (≥ 1 hospitalor ≥ 5 physician claims) or (≥ 1 physician claim and ≥ 7prescription claims) in 2 years, the age-standardized preva-lence of depression in 2005 was 31.7% (95% CI 29.8-33.5%)in the MS population and 20.5% (95% CI: 19.8-21.2%) inthe general population (PR 1.60; 95% CI: 1.41-1.82). Inboth populations, the peak prevalence occurred in personsaged 45-59 years (Figure 1).Anxiety disordersAgreement between the administrative case definitions andmedical records ranged from slight to fair (k = 0.02 to 0.23,Additional file 5: Table S5). The highest level of agreementfor any definition was fair (definition ‘N’, k = 0.23), andrequired (≥ 1 hospital or ≥ 2 physician claims) or (≥ 1physician and ≥ 2 prescription claims) in 2 years. Usingdefinition ‘N’, the age-standardized prevalence of anxiety in2005 was 35.6% (95% CI 33.7-37.7%) in the MS populationand 29.6% (95% CI: 28.8-30.5%) in the general populationA.051015202530354045OmnibusMood DisorderDepressionAnxietyComorbidityPrevalence per 100 personsB.051015202530354045OmnibusMood DisorderDepressionAnxietyComorbidityPrevalence per 100 personsBipolar DisorderSchizophrenia20-4445-5960+Bipolar DisorderSchizophrenia20-4445-5960+Figure 1 Age-specific prevalence of mental comorbidity in the MS (A) and general populations (B). Administrative case definitions used:Omnibus ≥ 1 Hospital or ≥ 5 Physician OR (≥ 1 Physician AND ≥ 4 Prescription claims) in 2 years. Mood disorder ≥ 1 Hospital or ≥ 5 PhysicianOR (≥ 1 Physician AND ≥ 4 Prescription) in 2 years. Depression ≥ 1 Hospital or ≥ 5 Physician OR (≥ 1 Physician AND ≥ 7 Prescription) in 2 years.Anxiety ≥ 1 Hospital or ≥ 2 Physician OR (≥ 1 Physician AND ≥ 2 Prescription) in 2 years. Bipolar disorder ≥ 1 Hospital or ≥ 3 Physician OR(≥ 1 Physician AND ≥ 3 Prescription) in 5 years. Schizophrenia ≥ 1 Hospital or ≥ 2 Physician in 2 years.Marrie et al. BMC Neurology 2013, 13:16 Page 4 of 8http://www.biomedcentral.com/1471-2377/13/16(PR 1.24; 95% CI: 1.12-1.38). In both populations, thepeak prevalence occurred in persons aged 45-59 years(Figure 1).Bipolar disorderAgreement between the administrative case definitions(labeled A to U) and medical records ranged from slightto moderate (k = 0.20 to 0.42, Additional file 6: Table S6).Several definitions had the highest sensitivity of 75%with specificities of 97% or higher. Agreement for thesedefinitions varied slightly, but confidence intervals over-lapped. Using definition ‘W’, which required (≥ 1 hos-pital or ≥ 3 physician claims) or (≥ 1 physician and ≥ 3prescription claims) in 5 years (k = 0.34), the age-standardized prevalence of bipolar disorder in 2005 was5.83% (95% CI: 5.01-6.65%) in the MS population and3.45% (95% CI: 3.17-3.73%) in the general population(PR 1.70; 95% CI: 1.55-1.87). Although the affectednumber of individuals was small, the prevalence of bipo-lar disorder was similar across age groups (Figure 1).SchizophreniaAgreement between all of the case definitions (labeled A toO) and medical records ranged from substantial to perfect(k = 0.67 to 1.0, Additional file 7: Table S7). Among casedefinitions with perfect agreement, the simplest definitionwith the highest sensitivity (100%) and specificity (99%)required ≥ 1 hospital or ≥ 2 physician claims in 2 years(definition ‘G’). Applying definition ‘G’, the age-standardizedprevalence of schizophrenia in 2005 was 0.93% (95% CI:0.61-1.26%) in the MS population and 0.93% (0.78-1.09%)in the general population (PR 0.95; 95% CI: 0.60-1.51). Al-though the small number of individuals affected requirescautious interpretation, the prevalence of schizophreniawas similar across age groups (Figure 1).Misclassification biasTable 1 shows the sensitivity, specificity, kappa, prevalenceindex and bias index for case definitions for which we pre-sented prevalence estimates. The definitions for bipolardisorder, anxiety and schizophrenia have high values forthe prevalence index indicating that kappa values will bereduced as compared to populations in which these condi-tions are more prevalent. Except for anxiety, the bias indexwas minimal. After adjustment for prevalence and bias, allκ increased except for schizophrenia which was already1.0. Graphical analysis of misclassification bias suggestedthat the case definitions perform reasonably well in theexpected range of prevalence for mental comorbidity inMS (Additional file 8: Figure S1) [1,2,22-28].DiscussionFew population-based studies have evaluated the preva-lence of mental comorbidity in MS [4]. To facilitate suchstudies, we validated case definitions for mental comorbid-ities based on hospital, physician and prescription claims.Our case definitions showed almost perfect agreementversus medical records for schizophrenia, and moderateagreement for any mental comorbidity, any mood or anx-iety disorder, and depression. The case definition for bipo-lar disorder had lower agreement, but acceptable sensitivityof 75% and high specificity of 97.5%. The case definitionfor anxiety had the lowest agreement but a specificity of82%. Further, persons with MS were at increased risk ofdepression, anxiety and bipolar disorder, but not schizo-phrenia when compared to the general population.Previous validation studies of administrative case defini-tions for mental comorbidity were often disappointing, andhave highlighted the challenges of distinguishing depres-sion from anxiety when using 3-digit ICD codes [11,12]. Ina Manitoba study, agreement was only fair (k = 0.26) be-tween surveys and administrative definitions for depressionwhich used hospital, physician and prescription claims [11]although this lower agreement may reflected their useof survey data and a broader range of prescription claimsthan in our study. Among persons newly treated with anti-depressants in Saskatchewan, Canada, agreement betweendepression identified on physician claims and medicalrecords was moderate (k = 0.54), similar to our findings[29]. We could not identify any published, validated casedefinitions for anxiety. Thus these validated case defini-tions augment the ability to conduct population-level sur-veillance of depression and anxiety. Despite challenges indeveloping case definitions for depressive and anxiety dis-orders sensitive and specific definitions were available forbipolar disorder (sensitivity 75%, specificity 97%). Among225 Americans, inpatient diagnoses of bipolar disorder,outpatient diagnoses of bipolar disorder by mental healthproviders, and outpatient diagnoses of bipolar disorder bynon-mental health providers that were accompanied by aprescription for lithium, carbamazepine or valproate, hadfalse positive rates below 10% [30]. However, we found thatbipolar disorder could be identified without such claims.Consistently, administrative case definitions for schizo-phrenia have performed well, with agreement betweenhospital claims for schizophrenia and medical recordsof 93.9-100% [31,32]. In American Medicaid data, thecase definition that we validated of either one hospitalor two physician claims for schizophrenia in two yearsidentified only 6% false positives (k = 0.76) [33].Collect-ively, this suggests that administrative data can accur-ately identify bipolar disorder and schizophrenia in theMS and general populations.Our approach is informative for researchers wishingto study mental comorbidity in other chronic neurologicdiseases, which share the potential problem of under-reporting of comorbidities due to coding biases [34].While prescription claims may add sensitivity, their useMarrie et al. BMC Neurology 2013, 13:16 Page 5 of 8http://www.biomedcentral.com/1471-2377/13/16for mental comorbidity is challenging in chronic neuro-logic diseases because of the frequent off-label use oftherapies. By restricting the breadth of prescriptionsused, and using them in combination with a physicianclaim for mental comorbidity we successfully createdvalid case definitions.Prior studies suggest that the annual prevalence of de-pression in MS is up to 14% with a lifetime prevalence ofup to 50% [1], and that anxiety disorders affect more than30% of persons with MS [2,22]. Our crude prevalence esti-mates of 33% for depression and 37% for anxiety based ontwo years of administrative data are consistent with thoseobservations. The age-standardized prevalence of bipolardisorder in the MS population was 5.83% (crude preva-lence 6.3%), 70% higher than in the general population.Studies in hospital or clinic populations suggested that bi-polar disorder affects 0.30% to 13% of the MS population[24-28]. The only one of these studies that used a truegeneral population control group reported that hospita-lized persons with MS had bipolar disorder twice as oftenas hospitalized controls (1.97% vs. 0.92%) [25]. Since thatstudy was limited to hospitalized persons, the prevalenceof bipolar disorder may have been underestimated, al-though the increased risk of bipolar disorder in MS wassimilar to our findings.The prevalence of schizophrenia was the same in theMS and general populations (0.93%). Two population-based studies, both using administrative data, evaluatedthe prevalence of psychosis, not limited to schizophrenia.In Taiwan, psychosis affected 7.5% of the MS populationand 2.0% of the general population (odds ratio 4.0) [35].In Alberta, Canada only 0.8% of MS patients had non-organic pyschoses including schizophrenia-spectrum dis-orders, and other non-organic psychoses, but this wasmore than in the general population [4]. Our findings ofan absence of an increased prevalence of schizophrenia inour MS population suggest a lack of increased risk whichmay reflect differences in the psychotic disorders studied(all versus schizophrenia alone), as well as the small num-ber of persons with schizophrenia.Medical records review for the validation cohort didnot involve all records of all health care providers overthe lifetime of study participants. For practical reasonswe also compared medical records to administrative datafor the 1 to 5 year period ending in fiscal year 2005/06,rather than from 1984 onward. Both factors may havereduced agreement between the data sources. Like med-ical records, administrative data only allow us to identifymental comorbidities for which the patient has beentreated; undiagnosed mental comorbidity cannot becaptured without a direct patient assessment. This studyhad several strengths, however. We validated the casedefinitions in a population similar to the one in which itwas applied, the design was population-based, we usedmatched general population controls, and we used mul-tiple types of administrative data.ConclusionsOur findings suggest that administrative data can beused for surveillance for mental comorbidities in MS,and should facilitate studies of the impact of mentalcomorbidity on health outcomes captured by administrativedata such as health care utilization. Our findings also pro-vide population-based data emphasizing the increasedprevalence of a range of mood and anxiety disorders in MS.Additional filesAdditional file 1: Table S1. Diagnosis and medication codes used toidentify comorbidities.Additional file 2: Table S2. Omnibus Definition: Administrative ClaimsCase Definitions as Compared to Medical Records Review.Additional file 3: Table S3. Mood and Anxiety Disorders: AdministrativeClaims Case Definitions as Compared to Medical Records Review.Table 1 Impact of prevalence and bias on agreement (kappa) between administrative case definitions and medicalrecordsComorbidity Case definition Comparison to medical records PrevalenceindexBiasindexAdjustedkappaNo. yearsof dataNo. and type ofclaimsaSens (95% CI) Spec (95% CI) Observed kappa(95% CI)Omnibus(Any Mental)2 ≥ 1 H or ≥ 5 P OR(≥ 1 P AND ≥ 4 Rx)63.5 (54.0, 72.2) 86.8 (82.1, 90.6) 0.51 (0.41, 0.60) 0.42 0.01 0.60Any Mood oranxiety disorder2 ≥ 1 H or ≥ 5 P OR(≥ 1 P AND ≥ 4 Rx)62.8 (53.2, 71.7) 86.9 (82.2, 90.6) 0.50 (0.41, 0.60) 0.43 0.01 0.60Depression 2 ≥ 1 H or ≥ 5 P OR(≥ 1 P AND ≥ 7 Rx)62.2 (52.4, 71.2) 86.7 (82.2, 90.4) 0.49 (0.40, 0.59) 0.45 0 0.60Bipolar disorder 5 ≥ 1 H or ≥ 3P OR(≥ 1 P AND ≥ 3 Rx)75.0 (19.4, 99.4) 97.5 (95.5, 99.4) 0.30 (0.036, 0.57) 0.95 0.02 0.94Anxiety 2 ≥ 1 H or ≥ 2 P OR(≥ 1 P AND ≥ 2 Rx)42.3 (23.3, 63.1) 82.2 (78.0, 85.9) 0.23 (0.022, 0.23) 0.74 0.12 0.69Schizophrenia 2 ≥ 1 H or ≥ 2 P 1.0 (15.8, 100) 0.99 (98.6, 100) 1.0 (0, 1.0) 0.99 0 1.0Marrie et al. BMC Neurology 2013, 13:16 Page 6 of 8http://www.biomedcentral.com/1471-2377/13/16Additional file 4: Table S4. Depression: Administrative Claims CaseDefinitions as Compared to Medical Records Review.Additional file 5: Table S5. Anxiety Disorders: Administrative ClaimsCase Definitions as Compared to Medical Records Review.Additional file 6: Table S6. Bipolar Disorder: Administrative Claims CaseDefinitions as Compared to Medical Records Review.Additional file 7: Table S7. Schizophrenia: Administrative Claims CaseDefinitions as Compared to Medical Records Review.Additional file 8: Figure S1. Assessment of misclassification bias foradministrative case definitions for mental comorbidities.AbbreviationsCIDI: Composite International Diagnostic Interview; CI: Confidence interval;ICD: International Classification of Disease; K: Kappa; MH: Manitoba Health;MS: Multiple sclerosis; NPV: Negative predictive value; PHIN: Personal HealthIdentification Number; PPV: Positive predictive value; PR: Prevalence ratio.CIHR Team in the Epidemiology and Impact of Comorbidity on MultipleSclerosis includes. Ruth Ann Marrie (University of Manitoba), Bo Nancy Yu(University of Manitoba), Stella Leung (University of Manitoba), LawrenceElliott (University of Manitoba), Patricia Caetano (University of Manitoba),James F Blanchard (University of Manitoba), Lawrence W. Svenson (Universityof Alberta), Joanne Profetto-McGrath (University of Alberta), Sharon Warren(University of Alberta), Christina Wolfson (McGill University), Nathalie Jette(University of Calgary), Scott B Patten (University of Calgary), Charity Evans(University of Saskatchewan), Helen Tremlett (University of British Columbia),John Fisk (Dalhousie University), Virender Bhan (Dalhousie University),Michelle Ploughman (Memorial University)Competing interestsRuth Ann Marrie receives research funding from: Canadian Institutes ofHealth Research, Public Health Agency of Canada, Manitoba Health ResearchCouncil, Health Sciences Centre Foundation, Multiple Sclerosis Society ofCanada, Multiple Sclerosis Scientific Foundation, Rx & D Health ResearchFoundation, and has conducted clinical trials funded by Bayer Inc. andSanofi-Aventis.John Fisk is the Director of the endMS Atlantic Regional Research andTraining Centre which is funded by the Multiple Sclerosis Society of Canada.He receives research funding from the Canadian Institutes of HealthResearch (CIHR) and in the past has received grants, honoraria andconsultation fees from AstraZeneca, Bayer, Biogen-Idec Canada, HeronEvidence Development Limited, Hoffmann-La Roche, MAPI Research Trust,Novartis, Sanofi-Aventis, Serono Canada, and QualityMetric Incorporated.Nancy Yu receives research support from the Canadian InternationalDevelopment Agency, the Multiple Sclerosis Society of Canada, CIHR, andManitoba Health and Healthy Living.Stella Leung reports no disclosures.Lawrence Elliott receives research support from the Canadian Institutes ofHealth Research, Health Sciences Centre Foundation, Public Health Agencyof Canada, and the Multiple Sclerosis Society of Canada.Patricia Caetano has worked on a research project funded by Amgen.Charity Evans reports no disclosures.Sharon Warren receives research funding from the CIHR, the CanadianHealth Services Research Foundation, Alberta Health Services and SSHRC.Christina Wolfson receives research funding from the Multiple SclerosisSociety of Canada, Canadian Institutes of Health Research, CanadaFoundation for Innovation, and Public Health Agency of Canada.Larry Svenson reports no disclosures.Helen Tremlett currently receives funding from: the Multiple Sclerosis Societyof Canada [Don Paty Career Development Award]; US National MS Society[#RG 4202-A-2 (PI)]; Canadian Institutes of Health Research [MOP: #190898(PI) and MOP-93646 (PI)]; Michael Smith Foundation for Health Research(Scholar award) and the Canada Research Chair program. She has receivedspeaker honoraria and/or travel expenses to attend conferences from: theConsortium of MS Centres, US National MS Society, Swiss Multiple SclerosisSociety, the University of British Columbia Multiple Sclerosis ResearchProgram, Teva Pharmaceuticals and Bayer Pharmaceutical (honorariadeclined) and ECTRIMS. Unless otherwise stated, all speaker honoraria areeither donated to an MS charity or to an unrestricted grant for use by herresearch group.James Blanchard receives research support from the Multiple SclerosisSociety of Canada, CIHR, Bill & Melinda Gates Foundation, CanadianInternational Development Agency and the United States Agency forInternational Development.Scott Patten was a member of an advisory board for Servier, Canada. He hasreceived honoraria for reviewing investigator-initiated grant applicationssubmitted to Lundbeck and Pfizer and has received speaking honoraria fromTeva and Lundbeck. He is an Associate Editor for the Canadian Journal ofPsychiatry and a member of the editorial board of Chronic Diseases andInjuries in Canada. He is the recipient of a salary support award (SeniorHealth Scholar) from Alberta Innovates, Health Solutions and receivesresearch funding from the Canadian Institutes for Health Research, theInstitute of Health Economics and the Alberta Collaborative Research GrantsInitiative.Authors’ contributionsRAM, JDF, SW, SBP, and HT conceived of and designed the study initially.RAM, JDF, LE, PC, CE, SW and SBP reviewed and selected diagnostic codesand pharmacotherapies for algorithm development. RAM, NY and SLanalyzed the data. All authors assisted in the interpretation of the data. RAMdrafted the manuscript. All authors revised the manuscript and approved thefinal version for publication.AcknowledgementThe results and conclusions presented are those of the authors. No officialendorsement by Manitoba Health is intended or should be inferred. Theauthors thank Melanie McLeod, PharmD for her assistance in reviewingpharmacotherapies for this project.FundingThis study was funded by operating grants and a Don Paty CareerDevelopment Award from the Multiple Sclerosis Society of Canada, theManitoba Health Research Council, the Canadian Institutes for HealthResearch, and the Rx & D Health Research Foundation. The sponsors had norole in the design, in the collection, analysis or interpretation of the data; inthe writing of the manuscript; or in the decision to submit the manuscriptfor publication.Author details1Department of Internal Medicine, University of Manitoba, Winnipeg, Canada.2Department of Community Health Sciences, University of Manitoba,Winnipeg, Canada. 3Health Sciences Centre, GF-543. 820 Sherbrook Street,Winnipeg, MB, Canada. 4Departments of Psychiatry and Medicine, DalhousieUniversity, Halifax, Canada. 5Faculty of Rehabilitation Medicine, University ofAlberta, Edmonton, Canada. 6College of Pharmacy and Nutrition, Universityof Saskatchewan, Saskatoon, Canada. 7Department of Epidemiology andBiostatistics and Occupational Health, McGill University, Montreal, Canada.8Research Institute of the McGill University Health Centre, Montreal, Canada.9Department of Community Health Sciences, University of Calgary, Calgary,Canada. 10School of Public Health, University of Alberta, Edmonton, Canada.11Surveillance and Assessment, Alberta Health, Edmonton, Canada.12Department of Medicine (Neurology), University of British Columbia,Vancouver, Canada.Received: 3 December 2012 Accepted: 4 February 2013Published: 6 February 2013References1. Goldman Consensus Group: The Goldman Consensus statement ondepression in multiple sclerosis. Mult Scler 2005, 11:328–337.2. Korostil M, Feinstein A: Anxiety disorders and their clinical correlates inmultiple sclerosis patients. Mult Scler 2007, 13:67–72.3. Rodgers J, Bland R: Psychiatric manifestations of multiple sclerosis:a review. Can J Psychiatry 1996, 41:441–445.4. Patten SB, Svenson LW, Metz LM: Psychotic disorders in MS: Population-basedevidence of an association. Neurology 2005, 65(7):1123–1125.5. Marrie RA, Horwitz R, Cutter G, Tyry T: Cumulative impact of comorbidityon quality of life in MS. Acta Neurol Scand 2012, 125(3):180–186.6. Mohr DC, Goodkin DE, Likosky W, Gatto N, Baumann KA, Rudick RA:Treatment of depression improves adherence to interferon beta-1btherapy for multiple sclerosis. Arch Neurol 1997, 54:531–533.Marrie et al. BMC Neurology 2013, 13:16 Page 7 of 8http://www.biomedcentral.com/1471-2377/13/167. Patten S: Integrating data from clinical and administrative databases inpharmacoepidemiological research. Can J Clin Pharmacol 1998, 5(2):92–97.8. Asche CV, Singer ME, Jhaveri M, Chung H, Miller A: All-cause health careutilization and costs associated with newly diagnosed multiple sclerosisin the United States. J Manag Care Pharm 2010, 16(9):703–712.9. Haro JM, Arbabzadeh-Bouchez S, Brugha TS, De Girolamo G, Guyer ME, JinR, Lepine JP, Mazzi F, Reneses B, Vilagut G, et al: Concordance of thecomposite international diagnostic interview version 3.0 (CIDI 3.0) Withstandardized clinical assessments in the WHO world mental healthsurveys. Int J Methods Psychiatr Res 2006, 15(4):167–180.10. Tricco AC, Pham B, Rawson NSB: Manitoba and Saskatchewanadministrative health care utilization databases are used differently toanswer epidemiologic research questions. J Clin Epidemiol 2008,61(2):192–197. e112.11. Martens P, Fransoo R, McKeen N, Burland E, Jebamani L, Burchill C, CosterCD, Ekuma O, Prior H, Chateau D, et al: Patterns of regional mental illnessdisorder diagnoses and service use in Manitoba: a population-based study.Winnipeg: Manitoba Centre for Health Policy, University of Manitoba; 2004.12. Kisely S, Lin E, Gilbert C, Smith M, Campbell L-A, Vasiliadis H-M: Use ofadministrative data for the surveillance of mood and anxiety disorders.Aust N Z J Psychiatry 2009, 43(12):1118–1125.13. Branch HIM: Population report. Winnipeg, Manitoba: Manitoba Health andHealthy Living; 2008.14. Marrie RA, Yu N, Blanchard JF, Leung S, Elliott L: The rising prevalence andchanging age distribution of multiple sclerosis in Manitoba. Neurology2010, 74(6):465–471.15. Horton M, Rudick RA, Hara-Cleaver C, Marrie RA: Validation of a self-reportcomorbidity questionnaire for multiple sclerosis. Neuroepidemiology 2010,35(2):83–90.16. Marrie R, Yu B, Leung S, Elliott L, Caetano P, Warren S, Wolfson C, Patten S,Svenson L, Tremlett H, et al: Rising prevalence of vascular comorbidities inMS: validation of administrative definitions for diabetes, hypertension,hyperlipidemia. Multiple Sclerosis Journal 2012, doi:10.1177/1352458512437814.17. Samkoff LM, Goodman AD: Symptomatic management in multiplesclerosis. Neurol Clin 2011, 29(2):449–463.18. Anatomic therapeutic chemical classification system. http://www.whocc.no/atc/structure_and_principles/.19. Landis JR, Koch GG: The measurement of observer agreement forcategorical data. Biometrics 1977, 33:159–174.20. Sim J, Wright CC: The kappa statistic in reliability studies: use,interpretation, and sample size requirements. Phys Ther 2005, 85:257–268.21. Kleinbaum D, Kupper L, Morgenstern H: Information bias. In Epidemiologicresearch: principles and quantitative methods. New York: Van NostrandReinhold; 1982:220–241.22. Dahl O-P, Stordal E, Lydersen S, Midgard R: Anxiety and depression inmultiple sclerosis. A comparative population-based study in nord-trondelag county, norway. Mult Scler 2009, 15(12):1495–1501.23. Wood B, van der Mei I, Ponsonby A-L, Pittas F, Quinn S, Dwyer T, Lucas R,Taylor B: Prevalence and concurrence of anxiety, depression and fatigueover time in multiple sclerosis. Multiple Sclerosis Journal 2012, doi:10.1177/1352458512450351. Epub ahead of print.24. Edwards LJ, Constantinescu CS: A prospective study of conditions associatedwith multiple sclerosis in a cohort of 658 consecutive outpatientsattending a multiple sclerosis clinic. Mult Scler 2004, 10:575–581.25. Fisk J, Morehouse SA, Brown MG, Skedgel C, Murray TJ: Hospital-basedpsychiatric service utilization and morbidity in multiple sclerosis. Can JNeurol Sci 1998, 25:230–235.26. Joffe RT, Lippert GP, Gray TA, Sawa G, Horvath Z: Mood disorder andmultiple sclerosis. Arch Neurol 1987, 44:376–378.27. Jacobs LD, Wende KE, Brownscheidle CM, Apatoff B, Coyle PK, Goodman A,Gottesman MH, Granger CV, Greenberg SJ, Herbert J, et al: A profile ofmultiple sclerosis: the New york state multiple sclerosis consortium. MultScler 1999, 5(5):369–376.28. Marrie RA, Horwitz RI, Cutter G, Tyry T, Campagnolo D, Vollmer T: The burdenof mental comorbidity in multiple sclerosis: Frequent, underdiagnosed, andunder-treated. Mult Scler 2009, 15(3):385–392.29. West SL, Richter A, Melfi CA, McNutt M, Nennstiel ME, Mauskopf JA:Assessing the Saskatchewan database for outcomes research studies ofdepression and its treatment. J Clin Epidemiol 2000, 53(8):823–831.30. Unützer J, Simon G, Pabiniak C, Bond K, Katon W: The use of administrativedata to assess quality of care for bipolar disorder in a large staff modelHMO. Gen Hosp Psychiatry 2000, 22(1):1–10.31. Rawson NS, Malcolm E, D'Arcy C: Reliability of the recording ofschizophrenia and depressive disorder in the Saskatchewan health caredatafiles. Soc Psychiatry Psychiatr Epidemiol 1997, 32(4):191–199.32. Kashner TM: Agreement between administrative files and written medicalrecords: a case of the Department of Veterans Affairs. Med Care 1998,36(9):1324–1336.33. Lurie N, Popkin M, Dysken M, Moscovice I, Finch M: Accuracy of diagnosesof schizophrenia in Medicaid claims. Hosp Community Psychiatry 1992,43(1):69–71.34. Elixhauser A, Steiner C, Harris DR, Coffey RM: Comorbidity measures for usewith administrative data. Med Care 1998, 36:8–27.35. Kang J-H, Chen Y-H, Lin H-C: Comorbidities amongst patients withmultiple sclerosis: a population-based controlled study. Eur J Neurol 2010,17(9):1215–1219.doi:10.1186/1471-2377-13-16Cite this article as: Marrie et al.: Mental comorbidity and multiplesclerosis: validating administrative data to support population-basedsurveillance. BMC Neurology 2013 13:16.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistributionSubmit your manuscript at www.biomedcentral.com/submitMarrie et al. BMC Neurology 2013, 13:16 Page 8 of 8http://www.biomedcentral.com/1471-2377/13/16


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items