SEARCH

SEARCH BY CITATION

Keywords:

  • Validity;
  • ventricular arrhythmias;
  • international classification of diseases;
  • medical record

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Background

Drug-induced pro-arrhythmia is a serious and unexpected event. Large administrative and claims databases can potentially identify drugs or interactions leading to cardiac arrhythmias. The purpose of this study is to evaluate the evidence supporting the validity of algorithms or codes to identify ventricular arrhythmias using administrative and claims data.

Methods

A search of MEDLINE database is supplemented by manual searches of bibliographies of key relevant articles. We selected all studies in which an administrative and claims data algorithm or code was validated against a medical record. We report the positive predictive value (PPV) for ICD-9 codes compared to medical records.

Results

Our search strategy yielded 664 studies, of which only seven met our eligibility criteria. Two additional studies were identified by peer reviewers. The most commonly included databases were Medicare and Medicaid, and the most commonly evaluated ICD-9 codes were 426.x and 427.x. The individual use of ICD-9 codes 427.x yielded a high PPV (78%–100%). The highest PPV was seen when both ICD-9 codes 427.x and 798.x were used (92%). The same codes yielded the highest PPV when found in the principal diagnosis position (100%).

Conclusions

The use of ICD-9 codes 427.x, alone or in combination with code 798.x, in the principal position is appropriate for the identification of ventricular arrhythmias in administrative and claims databases. Copyright © 2012 John Wiley & Sons, Ltd.


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Cardiac arrhythmias are abnormal heart rhythms that can be the result from structural heart disease or external factors like medications. Drug-induced cardiac arrhythmia is a problem faced by all clinicians who prescribe drugs and extends far beyond the realm of the cardiologist and cardiac electrophysiologist.[1] As the number of available drugs expands, the potential for patients to be prescribed with drugs with proarrhythmic potential, either alone or in combination with other drugs, continues to increase.

Administrative and claims databases provide the opportunity to conduct active surveillance to detect safety signals that emerge from drugs newly released into the market.[2] However, these data were not designed for research purposes and may be subject to information bias.[3, 4] Therefore, a critical first step in the process of accurately identifying safety signals is to assure the validity of algorithms or codes used to identify health outcomes.

The goal of this project is to identify algorithms that can be used to detect selected health outcomes of interest using claims data sources and describe the performance of these algorithms as reported by the studies in which they were used. This report summarizes the process and findings of the ventricular arrhythmias algorithm review.

METHODS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Search strategy

Details of the methods for these systematic reviews can be found in the accompanying manuscript by Carnahan and Moores.[5] In brief, the base PubMed search included terms to identify studies of administrative and claims databases with Medical Subject Headings (Mesh) that represented the health outcome of interest: “cardiac arrhythmias”[Mesh], and specific types of arrhythmias (ventricular fibrillation, tachycardia, premature complexes, torsades de pointes). Searches of the citation database of the Iowa Drug Information Service was also conducted. The details of these searches can be found in the full report on the Mini-Sentinel website: http://www.minisentinel.org/foundational_activities/related_projects/default.aspx. All searches were conducted in May 2010. All searches were restricted to articles published in 1990 or after. The rationale for only including articles published after 1990 was that older data on algorithm validity may not accurately reflect current coding practices. Mini-Sentinel collaborators were also asked to help identify any relevant validation studies.

Selection criteria

The abstract of each citation identified was reviewed by two investigators. If an investigator selected an article for full-text review, the full text was reviewed by both of the investigators. Agreement on whether to review the full text or include the article in the evidence table was calculated using a Cohen's kappa statistic.

Articles were considered for inclusion if they reported on original data where claims codes or algorithms that identified subjects with cardiac arrhythmias were validated against a medical record. This report presents data on ventricular arrhythmias, as a separate report will discuss atrial fibrillation.

Data abstraction

One investigator was responsible for completing the evidence table (LT), and the second confirmed the accuracy of the data abstracted (TH). Differences between the two reviewers were resolved by consensus. The key exposure variable was the claims code or algorithm.

The outcome variable of interest was the validation of the claims code against the documentation of cardiac arrhythmias in the medical record. Documentation of cardiac arrhythmias was defined as a clinical encounter documenting the occurrence of a cardiac arrhythmia or the objective documentation of the arrhythmia by an electrocardiogram. We also captured the setting and clinical context where the event occurred. Where possible, the positive predictive value (PPV) of the administrative code when compared to the medical record validation was reported.

RESULTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Literature search

The search strategy identified 664 abstracts. We selected 29 for full-text review. We excluded 378 because they did not study cardiac arrhythmias validated against a medical record, 219 because they were not administrative and claims database studies, and 37 because the data source was not from the United States or Canada. The Cohen's kappa for agreement between reviewers on inclusion versus exclusion of abstracts was 0.54. The reviewers disagreed mainly on the classification of categories for the excluded articles and not on the inclusion or exclusion of articles.

Of the 29 full-text articles reviewed, seven were included in the final evidence tables; seven were excluded because the HOI identification algorithm was poorly defined, and 15 were excluded because they included no validation of the outcome definition or reporting of validity statistics. Cohen's kappa for agreement between reviewers on inclusion versus exclusion of full-text articles reviewed was 1.0.

Mini-Sentinel investigators provided no published and no unpublished reports of validation studies that had been completed by their teams. Peer reviewers identified two additional articles that were included in the final evidence table, for a total of nine studies reviewed.

Setting and population of the included studies

Table 1 reports nine studies included in the systematic review. Of the nine studies, only one evaluated claims data from a national health insurance company.[6] Two evaluated private state/local health plans,[7, 8] while the remaining evaluated Medicare and Medicaid databases.

Table 1. Studies validating cardiac arrhythmias administrative codes against medical records
CitationStudy population, time period and inclusion criteriaDescription of outcome studiedAdministartive codes and algorithmValidation statistics
Enger et al. (2002)[6]United Healthcare administrative claims database from 1993 to 1998. 28,078 subjects who used cisapride and were not using antiarrhythmic drugs.Composite primary outcome to identify incident serious arrhythmic event defined as electrocardiographic documentation of ventricular fibrillation or torsade de pointes.426.xx14 cases were confirmed as serious ventricular arrhythmias of the 146 cases identified using codes (PPV = 10%).
427.xx
Hanrahan et al. (1995)[7]Harvard Community Health Plan in New England from 1988 to 1990. Study included those who had filled prescriptions for antihistamines in pharmacy files (N = 26,320).Ambulatory records, hospitalizations and emergency room visits for sudden death, torsades de pointes, complex ectopy, ventricular arrhythmia, syncope, and simple ectopy defined as clinically diagnosed with electrocardiographic documentation were evaluated.426, 426.0, 426.1, 426.10, 426.11, 426.12, 426.13, 426.3, 426.4, 426.5, 426.50, 426.51, 426.52, 426.53, 426.54, 426.6, 426.7, 426.8, 426.81, 426.89, 426.9, 427, 427.1, 427.4, 427.41, 427.42, 427.5, 427.6, 427.60, 427.61, 427.69, 427.9, 429, 429.2, 429.9, 780, 780.2, 780.3, 780.4, 785, 785.0, 785.1, 785.5, 785.50, 785.5161 cases of confirmed arrhythmias in the 1290 patients who were identified using codes (PPV = 5%). The majority of events were documented in the emergency room followed by inpatient hospitalizations.
Hennessy et al. (2008)[13]Medicaid programs of California, Florida, New York, Ohio, and Pennsylvania from 1999–2000. Study included users of cisapride, metoclopramide and proton pump inhibitor users.Hospitalization with a discharge diagnosis for ventricular arrhythmia or sudden cardiac death.427.1 paroxysmal ventricular tachycardia118 cases of confirmed in 126 patients (PPV = 92% (95% CI 86%–96%)) when the codes were used in any position.
427.4 ventricular fibrillation and flutterThe occurrence of an outpatient-occurring sudden cardiac death/ventricular arrhythmia (as is likely to be caused by prescription drugs taken in the ambulatory setting) had a PPV = 23%. When using a principal discharge diagnosis, the PPV = 100%.
427.41 ventricular fibrillation
427.42 ventricular flutter
427.5 cardiac arrest
798 sudden death, cause unknown
798.1 instantaneous death
798.2 death occurring in less than 24 h from onset of symptoms, not otherwise explained
McDonald et al. (2002)[11]Healthcare financing administration's Medicare Provider Analysis and Review inpatient hospitalization file from 1984 to 1995 of 4073 patients aged 65 years and older with a principal discharge diagnosis of arrhythmia.Hospital discharge with a principal discharge diagnosis of ventricular tachycardia or ventricular fibrillation/cardiac arrest.427.1 paroxysmal ventricular tachycardiaThe Medicare ventricular arrhythmia cohort was compared to the Seattle-area Myocardial Infarction and Triage Intervention registry.
427.4 ventricular fibrillation and flutterSensitivity 77%
427.5 cardiac arrestSpecificity 94%
Ray et al. (2001)[10]Tennessee Medicaid enrollees between 1998 and 1993. Study included patients who were aged 15–84 years, were not in a long-term facility and did not have life-threatening illnessesProbable sudden cardiac death defined as witnessed sudden collapse with no pulse or respiration, an unwitnessed collapse in a person known to be alive within the previous hour, ventricular fibrillation or tachycardia before the start of cardiopulmonary resuscitation, or autopsy findings consistent with a ventricular tachyarrhythmia.Not Reported701 cases were confirmed as probable sudden death among 1487 deaths (PPV 47%).
Staffa et al. (1998)[9]Ohio Medicaid population from 1986 to 1992. Individual paid medical claims submitted for reimbursement under Medicaid. Examined any claim for either astemizole or sedating antihistamines.Ventricular arrhythmias (paroxysmal ventricular tachycardia, fibrillation and flutter) and sudden cardiac death.Not reported8 cases of arrhythmias were confirmed out of 11 cases identified using codes (PPV = 73%).
Walker et al. (1999)[8]Saskatchewan Health administrative claims from 1990 to 1995. Included subjects were beneficiaries who received cisaprideArrhythmic events were defined as subjects with serious ventricular arrhythmia (ventricular fibrillation, sustained ventricular tachycardia and torsade de pointes).Not reported27 cases of arrhythmic events were confirmed among 199 patients identified using codes (PPV = 14%).
Chung et al. (2010)[14]Tennessee Medicaid enrollees between 1988 and 1993 aged 18 to 84 years who had at least 365 days of enrollment and had no evidence of life-threatening diseaseIncident sudden death associated with ventricular arrhythmias427.1 – paroxysmal ventricular tachycardia926 cases were evaluated for confirmation of the codes.
427.4 – ventricular fibrillation and flutterThe reported PPV (confirmed/reported in claims)
427.5 – cardiac arrest427.1 = 100%
427.4 = 100%
427.5 = 78%
427.9 – cardiac dysrhythmias, unspecified427.9 = 89%
Hennessy et al. (2010)[12]Medicaid programs in California, Florida, New York, Ohio, and Pennsylvania from 1999 to 2002.Incident sudden cardiac death as principal/first listed diagnosis as evidenced in the medical record of inpatient and emergency department claims combined.427.1 – paroxysmal ventricular tachycardiaThe reported PPV for principal diagnosis when compared to different medical record review strategies were:
85% when sudden death or ventricular arrhythmia was used
427.4 - ventricular fibrillation and flutter74% when ventricular arrhythmia was used
92% when sudden death was used
427.41 - ventricular fibrillation and flutterWhen a verbatim statement was found of sudden death:
66% when sudden death or ventricular arrhythmia was used
427.42 - ventricular flutter16% when ventricular arrhythmia was used
427.5 cardiac arrest97% when sudden death was used
Codes for sudden death:When a verbatim statement was found of ventricular arrhythmias:
798 – sudden death, cause unknown59% when sudden death or ventricular arrhythmia was used
91% when ventricular arrhythmia was used
798.1 – Instantaneous death32% when sudden death was used
When a ECG diagnosis was used:
48% when sudden death or ventricular arrhythmia was used
798.2 – Death occurring in less than 24 h from onset of symptoms84% when ventricular arrhythmia was used
19% when sudden death was used

The population reported in the studies included users of medications that can potentially cause drug-induced arrhythmia in five studies, and four studies included insured patients who had no life-threatening conditions.

Six of the nine studies listed in Table 1 reported the algorithms of international classification of diseases (ICD-9) codes to identify patients with cardiac arrhythmias; the remaining studies did not report the codes necessary for replication.[8-10] All of the reported codes were ICD-9 codes. No ICD-10, or diagnosis related group (DRG) codes were utilized in the algorithms. Procedure codes for Holter monitoring and electrocardiograms were used in one study to identify the population.

In all studies that reported ICD-9 codes included in the evidence table, at least two ICD-9 codes were used for identification of ventricular arrhythmias. The most commonly used codes for ventricular arrhythmias were 426.xx and 427.xx.

Validation criteria and definition and setting of cardiac arrhythmia

All but one study[11] included in Table 1 validated claims data through abstraction of medical charts. Documentation of cardiac arrhythmias in the medical records was generally based on clinical encounters; only two studies based their definition solely on electrocardiographic diagnosis.[6, 12] Two studies relied on electrocardiographic validation of ventricular fibrillation/torsade de pointes codes and reported a PPV of 10% and 84%.

Four studies validated administrative codes using clinical encounters (outpatient, inpatient, and emergency room documentation) and reported PPV of 5%, 47%, 73%, and 74%.[7, 9, 10, 13] One study linked a Medicare ventricular arrhythmia dataset with a detailed clinical registry and reported a sensitivity of 77%.[11]

Sudden death was used as part of the definition of cardiac arrhythmias in six of nine studies.[7, 9, 10, 12-14] In general, the PPV was highest in studies that used sudden death when compared to those not including sudden death as their definition of cardiac arrhythmia. Only one study that included sudden death and ventricular arrhythmias evaluated clinical encounters originating from the outpatient setting. The study included those events in the principal position for hospitalization or first listed in the emergency room and found an 85% PPV.[12] A second study identified patients who had sudden cardiac death and determined if it was arrhythmic in nature and found a PPV of 47%.[10]

Validation algorithms

One study[6] used all of the subcodes within the ICD-9 code classification for 426 (conduction disorders) and 427 (cardiac dysrhythmias) and confirmed the diagnosis in only 14 out of the 146 cases identified (PPV 10%). Another study that used specific subcodes of the ICD-9 code 427.x (427.1, 427.4, 427.5,427.9) reported a high PPV (89%–100%).[14]

One study used a limited and specific set of codes for ventricular arrhythmias (427.1, 427.4, 427.5) compared to a registry and found a sensitivity of 77% and specificity of 94%.[11] When an inclusive set of codes within the 427.xx and 798.xx (sudden death, cause unknown) codes was used, the PPV was 92%.[13]

When ICD-9 codes 426.xx, 427.xx, 780.x (general symptoms) and 785.xx (symptoms involving cardiovascular symptoms) were used, confirmation of cardiac arrhythmia only occurred in 5% of the population (PPV 5%).[7]

Four studies[8, 11-13] reported on the performance of the ICD-9 codes in the principal position as a discharge diagnosis of cardiac arrhythmias or in any position. One study used Medicaid databases (CA, FL, NY, OH, PA) and reported a PPV for ICD-9 codes 427 and 798 of 92% when used in any position.[12] A study using the Medicare database reported that the sensitivity of ICD9 code 427 as a principal diagnosis was 77%.[11] Another Medicaid (CA, FL, NY, OH, PA) study validated ICD-9 codes as principal diagnosis using electrocardiographic criteria and reported a PPV of 92% for sudden death codes, 74% for ventricular arrhythmia codes, and 85% when either code was used; however, the highest PPV was 100% when those codes were found in the principal position.[13] A study using claims data for principal discharge diagnosis from Saskatchewan Health confirmed 27 cases among 199 chart reviews (PPV 14%).[8]

Algorithms for defining drug induced arrhythmia

Five studies included subjects who were taking medications with potential pro-arrhythmic effect.[6-9, 13] The medications included were cisapride, metochlopramide, proton pump inhibitors, and antihistamines. For studies evaluating medications, the definition of cardiac arrhythmias was based on claims submitted from hospitalized patients with serious cardiac arrhythmias, and two studies included subjects who had sudden cardiac death. The studies with the highest PPV were those that included ICD-9 codes 427.1, 427.4x, 798.x.

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

This systematic review found that the highest PPV were seen when ICD-9 codes 427.x and 798.x were used. The use of specific subcodes of the 427 ICD-9 code that define specific ventricular arrhythmias like ventricular tachycardia (427.1) or fibrillation (427.4) yielded the highest results when the codes were used individually. When ICD-9 codes 426.x were added to the algorithm, the reported PPV lowered when compared to algorithms that did not use the code and therefore decreased the ability to identify only patients with cardiac arrhythmias. The PPV was higher when the codes were used in the principal position.

Drug-induced arrhythmia is becoming an important public health concern as the number of medications that can prolong the QT interval continues to grow. Therefore, in our report, we evaluated studies to determine the setting where the arrhythmic event occurred, with particular attention to studies reporting on outpatient events where most drug-induced arrhythmia could occur. One study reported that that the PPV for identifying potentially outpatient originating sudden cardiac death and ventricular arrhythmia precipitating hospitalization and emergency room treatment had a PPV of 85%.

Our report has several limitations that deserve mention. First, a few of the included studies did not report the specific ICD-9 codes used to identify cardiac arrhythmias, decreasing the number of available studies for the analysis. Second, there were no reports of other claims data used like pharmacy and DRG codes. Third, there was a lack of an evaluation of algorithms in higher risk patients, particularly those with heart failure. However, it would be expected that the PPV in high-risk patients would be higher than in unselected populations since the PPV is dependent upon disease prevalence in the source population. Fourth, current data provide little information on the number of cases of sudden death that might be missed by administrative and claims data. It is likely that some number of cases of sudden death will not reach medical attention and, therefore, will not be identified in these data. Future work should evaluate the sensitivity of administrative and claims data compared to death certificates. Last, our search was based on PubMed, and we did not search EMBASE and other databases that could have identified potential articles not found in PubMed.

Ventricular ectopy and nonsustained ventricular tachycardia are common in patients who have cardiomyopathies and heart failure. It has long been known that the frequency of ventricular ectopy is a risk factor for sudden cardiac death. This is a particularly important group of patients where there is a need for surveillance of new drugs and medication interaction.

There is a critical need to identify safety signals in claims databases that can potentially be used in large post-marketing surveillance to address the theoretical risk of excess mortality associated with modest degrees of drug-induced QTc prolongation or to identify new interactions between medications from large epidemiological observational studies.[15, 16]

In conclusion, there is a need for active surveillance of potentially fatal complications in high-risk patients, and this study provides a set of codes that can be used for active surveillance. We recommend the use of ICD-9 codes 427.x and 798.x in the principal position and generally do not recommend the use of ICD-9 codes 426.x, to identify patients with ventricular arrhythmias or cases of sudden death. We also recommend that future work be conducted to determine the sensitivity of claims data for identifying cases of sudden death, using death certificates as the criterion standard.

CONFLICT OF INTEREST

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

The authors have declared that there is no conflict of interest.

KEY POINTS

  • Few studies have evaluated the validity of ICD-9 codes in claims databases
  • The most common databases used are Medicare and Medicaid
  • ICD-9 427 alone or in combination with code 798 can identify ventricular arrhythmias with PPVs ranging from 78% to 100%.

ACKNOWLEDGEMENTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

This work was supported by the Food and Drug Administration through the Department of Health and Human Services (HHS) Contract Number HHSF223200910006I. The views expressed in this document do not necessarily reflect the official policies of the Department of HHS, nor does mention of trade names, commercial practices, or organizations imply endorsement by the U.S. government.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONFLICT OF INTEREST
  8. ACKNOWLEDGEMENTS
  9. REFERENCES