Regular Article
Evaluation of the predictive value of ICD-9-CM coded administrative data for venous thromboembolism in the United States

https://doi.org/10.1016/j.thromres.2010.03.009Get rights and content

Abstract

Objective

To determine the positive predictive value of International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM) discharge codes for acute deep vein thrombosis or pulmonary embolism.

Materials and Methods

Retrospective review of 3456 cases hospitalized between 2005 and 2007 that had a discharge code for venous thromboembolism, using 3 sample populations: a single academic hospital, 33 University HealthSystem Consortium hospitals, and 35 community hospitals in a national Joint Commission study. Analysis was stratified by position of the code in the principal versus a secondary position.

Results

Among 1096 cases that had a thromboembolism code in the principal position the positive predictive value for any acute venous thrombosis was 95% (95%CI:93-97), whereas among 2360 cases that had a thromboembolism code in a secondary position the predictive value was lower, 75% (95%CI:71-80). The corresponding positive predictive values for lower extremity deep-vein thrombosis or pulmonary embolism were 91% (95%CI:86-95) and 50% (95%CI:41-58), respectively. More highly defined codes had higher predictive value. Among codes in a secondary position that were false positive, 22% (95%CI:16-27) had chronic/prior venous thrombosis, 15% (95%CI:10-19) had an upper extremity thrombosis, 6% (95%CI:4-8) had a superficial vein thrombosis, and 7% (95%CI:4-13) had no mention of any thrombosis.

Conclusions

ICD-9-CM codes for venous thromboembolism had high predictive value when present in the principal position, and lower predictive value when in a secondary position. New thromboembolism codes that were added in 2009 that specify chronic thrombosis, upper extremity thrombosis and superficial venous thrombosis should reduce the frequency of false-positive thromboembolism codes.

Introduction

Increasingly, epidemiologic studies are using “administrative” hospital discharge data to identify patients with important vascular outcome events, such as deep-vein thrombosis or pulmonary embolism, which together comprise venous thromboembolism (VTE). Administrative data are computerized records that are gathered for some administrative purpose, but contain information that can be used for other purposes as well. In the United States (US), the Uniform Claim and Billing Form 04 (UB-04) requires International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) coding. The Joint Commission (TJC- the regulatory agency that oversees licensing of hospitals), the Agency for Healthcare Research and Quality (AHRQ - the US agency charged with improving the quality, safety, efficiency, and effectiveness of health care), and the Centers for Medicare & Medicaid Services (CMS) have launched quality measurement, quality improvement, and pay-for-performance initiatives that require identification of patients with acute VTE using a set of specific ICD-9-CM codes in hospital discharge records [1]. Researchers interested in vascular outcomes who use administrative data rely on ICD-9-CM codes to define presence or absence of acute VTE [2], [3], [4], [5], [6].

Prior to October 2009, there were twenty ICD-9-CM codes for VTE in non-pregnant patients, including 3 codes for pulmonary embolism, 10 codes for ‘thrombophlebitis’ (451 series) and 7 codes for ‘other venous thrombosis or embolism’ (453 series) [7]. In addition, there are several codes for pregnancy-related VTE (671 series) [8]. ICD-9-CM coding rules are evolving, as evidenced by the creation in October 2004 of 3 new codes that specify deep-vein thrombosis in the leg, thigh, or calf, and the creation in October 2009 of 21 new codes that specify additional locations and the acuity of VTE [9]. In addition, an ‘indicator’ to specify if a condition was present-on-admission (POA) was introduced in New York and California in the 1990s [10], and since October 2007 is required by CMS. This indicator should theoretically aid in the identification of hospital-acquired acute VTE events (i.e., not present-on-admission) [3], [4].

Only a few studies have retrospectively analyzed the positive predictive value of VTE codes. An early study showed that the most commonly used codes, when present in the principal position (specific for the condition that occasioned hospital admission), had very high positive predictive value for acute VTE (i.e., approximately 95%) [11]. However, other studies have reported that VTE codes in any position (principal or secondary) have much lower predictive value for acute VTE, in the range of 70-75%. In a multisite prospective cohort study, Cushman et al reported that the predictive value of any VTE code was 68% (95%CI: 63-74) [12]. More recently, Arnason et al [13] at a single hospital, and Heckbert et al [14], from the Women's Health Initiative, reported positive predictive values of 74% (95%CI: 64-82) and 78% (95%CI: 70-85), respectively. In another study using data from CMS, the predictive value of coding for deep vein thrombosis was 72% [15]. Finally, Zhan and coworkers analyzed the predictive value of ICD-9-CM coding for post-operative VTE using Medicare data and found a predictive value of only 29% for VTE [16]. These published studies have generally analyzed the predictive value of a group of VTE codes for any acute VTE event, not strictly lower extremity deep-vein thrombosis OR pulmonary embolism.

The aim of this study was to analyze a large sample of records from a wide array of hospitals throughout the United States to determine the predictive value of individual ICD-9-CM codes located in the principal position versus a secondary position for either any acute VTE or acute lower extremity deep-vein thrombosis or pulmonary embolism. Our objectives were to provide the ICD-9-CM Coordination and Maintenance Committee with the information necessary to enable them to restructure VTE codes in a more logical can comprehensive fashion, and to inform researchers and quality improvement professionals who use VTE codes for surveillance purposes about the predictive value of individual ICD-9-CM VTE codes [17].

Section snippets

Methods

The present study represents the compilation of three independent chart abstraction efforts by the University of California, Davis Medical Center (UCDMC), the University HealthSystem Consortium (UHC), and The Joint Commission (TJC). Because the research goals and abstraction methods were similar across projects, the first and last authors, who were involved in all three projects, decided to pool the data in order to enhance reliability. The UHC project also aimed to estimate the false negative

Results

The demographic characteristics of the samples from UCDMC, TJC and the two UHC subgroups (medical and surgical) that had a VTE code are shown in Table 1. The Joint Commission (N = 2052) sample had a median age that was 10 years older than both the UCDMC (N = 413) and UHC samples (N = 991), and the percentage of women in the TJC sample was also higher.

Discussion

All hospitals in the United States collect coded patient discharge data, which are used for a multitude of purposes [22] and are submitted to state health organizations and to various vendors, including The Joint Commission. The American Health Information Management Association recognizes that “the collection of accurate and complete coded data is critical to healthcare delivery, research, public reporting, reimbursement, and policy-making. The integrity of coded data and the ability to turn

Conflict of interest statement

The authors have no actual or potential conflict of interest specifically they have no financial, personal or other relationships with other people or organizations within three years of beginning the work submitted that could inappropriately influence their work.

Acknowledgment

This study was supported, in part, by a contract (290-04-0020, Support for Quality Indicators) with the Agency for Healthcare Research and Quality and, in part, by the Hibbard E Williams Endowment for General Medicine at UCDMC.

References (28)

  • Center for Disease Control. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM),...
  • Center for Disease Control. ICD-9-CM Official Guidelines for Coding and Reporting. Journal [serial on the Internet]....
  • J.M. Naessens et al.

    Impact of diagnosis-timing indicators on measures of safety, comorbidity, and case mix groupings from administrative data sources

    Med Care

    (Aug 2007)
  • R.H. White et al.

    Incidence and time course of thromboembolic outcomes following total hip or knee arthroplasty

    Arch Intern Med

    (1998)
  • Cited by (0)

    Presented at the International Society of Thrombosis and Haemostasis Meeting, July 16, 2009, Boston, MA.

    View full text