Validation of Serious Ventricular Arrhythmias in an Administrative Database

Cristina Varas-Lorenzo,1 Catherine Johannes,2 Lisa McQuay,3 Kirk Midkiff,3 Mary Rose Stang,4 Daniel Fife5 1RTI Health Solutions, Barcelona, Spain; 2RTI Health Solutions, Waltham, MA, United States; 3RTI Health Solutions, Research Triangle Park, NC, United States; 4Saskatchewan Health, Regina, ; 5Johnson & Johnson PRD, Titusville, NJ, United States

ABSTRACT BACKGROUND CONCLUSIONS

BACKGROUND: Pharmacoepidemiology • Evaluation of infrequent adverse events, such as drug-induced SVA, requires Table 1. Specifi c Codes for SVA or CA Table 2. Case Defi nition • Overall, 34 cases were confi rmed as idiopathic SVA according to • Information from electronic administrative studies examining serious ventricular large populations of exposed patients. Serious Ventricular Arrhytmia the prespecifi ed criteria (Figure 2). health databases, even when using hospital Diagnosis ICD-9 ICD-10-CA arrhythmias (SVA) may ascertain this First diagnosis of a fatal or nonfatal VA with hospitalization or confi rmed discharge diagnosis codes specifi c to SVA and • Population-based administrative databases are a useful resource for Defi nition • An additional 14 cases were confi rmed as SVA but did not meet the outcome from data on hospitalizations, by a specialist during the study period after the review of detailed patient’s postmarketing safety studies, but information in administrative claims alone SVA 427.1, 427.4 I47.0, I47.2, I49.00, I49.01 study criteria for idiopathic SVA. In three of these nonstudy cases, but there is limited information on the SVA as defi ned above and no prespecifi ed alternative cause was chronological information, may be insuffi cient is often not suffi cient to accurately identify outcomes such as drug-induced Idiopathic the SVA occurred in the hospital. documented immediately before or concurrently with the SVA occurrence validity of these diagnoses in administra- CA 427.5 I46.0, I46.1, I46.9 SVA for establishing an accurate diagnosis of SVA. SVA. in the same hospitalization episode in the hospital medical record • Among suspected CA cases, seven were confi rmed as SVA leading tive databases. An ongoing study of SVA Figure 1. Hospital Facilities in • Possible SVA cases can be identifi ed ICD-9 = International Classifi cation of Diseases, 9th Revision; ICD-10-CA = International Classifi cation of Diagnoses TdP, VT, VF to the arrest and without prespecifi ed alternative cause. • For drug safety studies with SVA as an and sudden cardiac death (SCD) offered Diseases, 10th Revision, Canadian Enhancement. endpoint, case validation is useful for the opportunity to perform such a valida- through hospital discharge diagnoses Athabasca Health Authority • Acute coronary • Cardiomyopathy • Aortic aneurysm • In 14 cases the CA had an alternative cause or was due to other codes; however, there is limited H? syndrome or (dilated, hypertrophic, dissection identifying SVA cases and appears essential for tion. myocardial infarction arrhythmogenic type of arrhythmia. Of note is that 36% of suspected CA cases identifying idiopathic SVA. information on the validity of these Legend • Valvular heart Hospital Designations Case Verifi cation (within 7 days) right ventricular) or occurred in the hospital. Stony Rapids OBJECTIVE: To describe the validity of a as of September 30, 2008 disease diagnoses, and additional sources of congenital heart disease • Validation of CA hospitalizations is essential for H Provincial Hospital • Pulmonary arterial two-stage process for identifying idio- H Regional Hospital Abstraction Process • Cardioverter District Hospital H Prespecifi ed information, such as hospital records, Community Hospital hypertension or acute • Chronic coronary heart Figure 2. Results of Case Validation identifying SVA as the leading cause and Keewatin Yatthé H Northern Hospital defi brillator pathic SVA among hospital discharges in H Mamawetan Churchill River H alternative are usually required to identify valid Highway • An abstraction form was developed and submitted with the study a pulmonary embolism disease with presence implanted specifi cally when the outcome defi nition the Canadian Saskatchewan Health (SH) causes of clinical heart failure SVA cases. La Loche protocol to the University of Saskatchewan Biomedical Research • Chronic coronary excludes in-hospital events. ? • Terminal illness, Validation of suspected cases of VT or VF database. H heart disease with leading to VA ? Ile a la Crosse H Ethics Board for review and approval. as manifested • A validation study was performed as presence of clinical • Acute stroke or 60 • Profi le review before chart abstraction can METHODS: The study population com- by, for example, 52% H? heart failure leading cerebrovascular disease part of an ongoing case-control study H? • SH contacted the regional health authorities for permission to coma or shock increase the effi ciency of the validation Loon Lake H? 50 PrairieMeadow North Lake Kelsey Trail prised over 86,000 users of domperidone Prince Albert Parkland to VA Big River ? ? H ? on the risk of SVA or SCD nested in a LloydminsterH H H? abstract charts and executed contract agreements with 10 regional process. Shellbrook? ? H ? H H Maidstone or a proton pump inhibitor (PPI) by pre- Prince Albert Hudson Bay ? ? ? H ? H Cardiac Arrest cohort of users of domperidone or NorthH H Tisdale health authorities and 6 affi liated facilities for access to hospital 40 ? ? Melfort H ? H H Unity H? scription (1990-2005). Hospitalizations Humboldt Kelvington Biggar Saskatoon ? Wadena? ? H H Unexpected circulatory arrest, usually due to a cardiac arrhythmia, PPI medications identifi ed from the H? H Preeceville records. Lanigan H? H? H? WatrousH? Wynyard Canora KindersleyHeartlandRosetown H? 30 ? ? Defi nition occurring within 1 hour of the onset of symptoms, in which medical with ICD-9 and ICD-10-CA codes specifi c H H ? 26% ? Outlook ? H H H ? Davidson Canadian SH database. H Lestock Sunrise Leader ? Balcarres H? H ? Melville H Esterhazy • The abstractions were performed by SH health workers after Fort Qu'Appelle intervention (e.g., defi brillation) reversed the event Herbert Indian Head to ventricular tachycardia (VT) or ventric- H? Regina H? Moose? Jaw ? Broadview H H? H Wolseley 20 17% Maple Creek H Moosomin ? • We present the results of the H Kipling completing a full day of in-person training conducted by RTI Health N = 65 (100%) REFERENCES ? ? ? • VT, TdP, or VF leading to the CA hospitalization, and H GravelbourgH? H H H? ular fi brillation (VF) as the principal diag- ? Regina Shaunavon H Arcola ? Five Hills Qu'Appelle ? H Assiniboia H Redvers H? Cypress H? Solutions project staff. validation process of SVA cases • CA not precipitated by another mechanism, such as complete validated Percentage ? 10 1. Zipes DP, Camm AJ, Borggrefe M, Buxton AE, Chaitman B, ? ? nosis were identifi ed. A cardiologist, H ? H ? H H H Diagnoses Sun Country 5% Kilometres atrioventricular block, other type of arrhythmia, severe bradycardia, or Fromer M, et al. ACC/AHA/ESC 2006 guidelines for identifi ed by specifi c hospital ? blinded to the exposure of interest, re- H 010020050 • Records were abstracted on-site at facilities, and supporting terminal illness 0 management of patients with ventricular arrhythmias and discharge codes. Saskatchewan Hospital Designations HISC: GIS Unit, RA, 01/05/2009, HospDesig_09.mxd viewed electronic patient profi les; cases as of September 30, 2008 Produced by Saskatchewan Health utilizing licensedgeospatial data from Information Services Corporation of Saskatchewan. materials (e.g., discharge summaries, electrocardiograms) were Idiopathic Non-idiopathic In-hospital No SVA the prevention of sudden cardiac death: a report of the Prespecifi ed • Same as above plus serious respiratory compromise occurring a b c with terminal illness, cancer, or other deidentifi ed and copied. SVA cases SVA cases SVA confirmed American College of Cardiology/American Heart Association alternative immediately before or concurrently with a CA Task Force and the European Society of Cardiology prespecifi ed known immediate causes of causesa • Materials were sent to RTI-HS after all information on the drug Committee for Practice Guidelines (Writing Committee to arrhythmia were excluded. Suspected OBJECTIVE TdP = torsades de pointes; VA = ventricular arrhythmia. Develop Guidelines for Management of Patients with exposures of interest were masked. a The presence of potential triggering factors for SVA, such serum electrolytic disturbances, dialysis, abuse Ventricular Arrhythmias and the Prevention of Sudden cases were validated by hospital medical Validation of suspected SVA leading To describe the validity of a two-stage process for identifying idiopathic SVA of recreational drugs, or pacemaker inserted, were not considered as alternative causes. Cardiac Death). Circulation 2006;114:e385-e484. record abstraction. The positive predictive Case Validation 60 to cardiac arrest cases among hospital discharges in the SH database. value (PPV) was calculated for the SVA • Before reviewing suspected cases, written criteria were established 50 screening algorithm. Validation of non- for case validation, following recommended criteria summarized in PPV cases was not performed. international clinical guidelines issued jointly by the American 40 METHODS • The PPV was calculated for the SVA screening algorithm as the 36% College of Cardiology, the American Heart Association, and the 32% CONFLICT OF INTEREST STATEMENT RESULTS: A total of 117 patients met the number of confi rmed cases among the total number of hospital 30 27% screening defi nition of SVA; after profi le Study Population European Society of Cardiology for management of patients with Johnson & Johnson PRD provided funding for the project under 1 charts abstracted with specifi c primary discharge codes for VF or VT. review, 74 were judged as suspected ventricular arrhythmias and the prevention of SCD. 20 a research contract with RTI Health Solutions. The authors are • A cohort of over 86,000 persons with at least one dispensing of domperidone N = 22 (100%) • The PPV for confi rmation of study cases for the CA screening also employees of RTI Health Solutions, Saskatchewan Health, and cases for abstraction. Of these, 65 records Johnson & Johnson PRD. or a PPI from 1990 to 2005 was identifi ed using the SH database. • Based on all the available information, the cardiologist classifi ed validated Percentage was calculated. 10 (88%) were abstracted and 34 were vali- 5% • After excluding persons with a diagnosis of cancer, the fi nal study cohort each reviewed case into one of the following categories for SVA: Disclaimer: This study is based, in part, on deidentifi ed data dated as idiopathic SVA (PPV = 52%; 95% 0 provided by the Saskatchewan Ministry of Health. The comprised 83,212 persons. − Defi nite interpretation and conclusions contained herein do not CI: 40%-65%). An additional 14 cases SVA leading to SVA with alternative In-hospital No CA RESULTS CA, no prespeci- prespecified cause CA confirmed necessarily represent those of the government of Saskatchewan were confi rmed as SVA but did not meet • Follow-up time for case ascertainment began with the date of cohort entry − Probable fied alternative or other arrhythmia or the Saskatchewan Ministry of Health. cause leading to CA the study criteria for idiopathic SVA. Thus, and extended until the date of an event, death, disenrollment from the SH − Possible • A total of 117 patients met the screening defi nition of SVA (Table 3). the PPV of the screening algorithm for all database, or the study end date of 12/31/2005. − Noncase. • After the review of the electronic patient profi les by a cardiologist SVA (idiopathic or not) among the vali- a • All study data were deidentifi ed and provided by SH after review and • The data evaluated for this classifi cation included: and after excluding those patients with malignant neoplasm, Confi rmed case of idiopathic SVA: includes cases validated as defi nite, probable, and possible. dated cases was 74% (95% CI: 63%-85%). approval of the study protocol by SH’s Data Access Review Committee. b Nonidiopathic SVA cases: SVA was confi rmed but did not meet the study case defi nition because there − Clinical symptoms at presentation 74 patients were judged as suspected cases for abstraction. was clear evidence of a prespecifi ed alternative cause. CONTACT INFORMATION CONCLUSIONS: Even when using hospi- Case Ascertainment c − Copies of ambulance emergency services • Of these, nine charts were not abstracted because the chart was in an In-hospital cases: Diagnosis was confi rmed as SVA or CA, but the onset of the event was during a tal discharge diagnoses, case validation is hospitalization episode (in-hospital events). Cristina Varas-Lorenzo, MD, MSc, PhD • Suspected SVA cases were identifi ed from hospitalizations with specifi c out-of-reach facility or had been destroyed. useful for identifying SVA and appears − Copies of electrocardiograms Director, Epidemiology codes for VT or VF as the principal diagnosis (Table 1). essential for identifying idiopathic SVA. − Copies of discharge reports • Completed abstractions were achieved for 65 suspected cases (88%). RTI Health Solutions • Hospitalizations with cardiac arrest (CA) codes were screened to identify Trav. Gracia 56, Atico 1 Profi le review before chart abstraction − Abstracted clinical information, including laboratory values • In addition, 22 suspected CA cases were abstracted. Table 4. PPV of Screening Criteria for the Confi rmation of Cases of SVA and CA can increase the effi ciency of the valida- potential SVA leading to CA (Table 1). Fulfi lling the Study Case Defi nition 08006 Barcelona, Spain − Other tests results, as available. tion process. Screening by Phone: +34.93.241.7761 • Chronologically recorded information in the databases was used to create Table 3. Subjects Selected for Electronic Patient Profi le Review and Recommended for PPV PPV of Confi rmed • This validation process was performed without knowledge of the Primary Hospital Diagnoses 95% CI Fax: +34.93.414.2610 electronic patient profi les. Medical Record Abstraction Formula Idiopathic SVA exposure status of interest. No validation was performed on non- Discharge Codes E-mail: [email protected] Algorithm Recommended • A cardiologist, blinded to the exposure of interest, reviewed electronic SVA suspected cases. Hospital Primary Discharge Code Idiopathic SVA 34 of 65 52% 40%-65% Presented at: 25th International Conference patient profi les; cases with terminal illness, cancer, or other prespecifi ed Diagnoses SVA Suspected for Abstraction VT, VF All SVA 48 of 65 74% 63%-85% on Pharmacoepidemiology & known immediate causes of arrhythmia were excluded. ICD-9 ICD-10-CA N N Therapeutic Risk Management 427.1, 427.4 I47.0, I47.2, I49.0 VT, VF 117 (100%) 74 (63%) SVA leading to CA, • “Suspected” SVA cases were selected for hospital record review and medical CA without prespecifi ed 7 of 22 32% 15%-53% August 16–19, 2009 427.5 I46.0, I46.1, I46.9 CA 138 (100%) 30 (22%) record abstraction. alternative cause Providence, RI, United States