International Statistical Classification of Diseases, Tenth Revision and the Definition of Laryngectomy

Abstract

This cross-sectional study uses data from the 2001 to 2011 Nationwide Inpatient Sample to compare International Statistical Classification of Diseases, Tenth Revision (ICD-10) with Ninth Revision coding systems for laryngectomy procedures.

The International Classification of Diseases, Ninth Revision (ICD-9) coding system has been used in the US since 1979 to track health care statistics, quality outcomes, and mortality.¹ With use of ICD-9 coding terminology, a well-established volume-outcome association has been demonstrated for many low-volume, high-risk surgical procedures, including laryngectomies.^2,3 In 2015, the US replaced ICD-9 with International Statistical Classification of Diseases, Tenth Revision (ICD-10) aiming to improve coding accuracy with greater detail in defining diseases and expanding surgical codes by grouping similar procedures. In pediatric surgery, this transition affected identification of perforated appendix readmissions,⁴ a key quality measure.⁴ We investigated whether use of ICD-10 coding altered identification of laryngectomy cases and the volume-outcome association for laryngectomy.

Methods

This cross-sectional analysis of patients with laryngeal or hypopharyngeal cancer was performed as previously described,² using discharge data from the Nationwide Inpatient Sample (NIS) 2001-2011. Race and ethnicity were collected in the NIS data set.² We used ICD-9 codes to identify patients who underwent partial or total laryngectomy, then converted ICD-9 laryngectomy codes to equivalent ICD-10 procedure coding system (ICD-10-PCS) codes (Figure 1). These ICD-10-PCS codes were then mapped to ICD-9 definitions to compare the 2 cohorts. This study was approved as exempt by the Johns Hopkins Medical Institutions Institutional Review Board. Using Stata, version 16 (StataCorp LP), we analyzed associations between variables using cross-tabulations, multivariable logistic regression, and generalized linear regression modeling.²

Figure 1. International Statistical Classification of Diseases, Tenth Revision (ICD-10) Mapping of International Classification of Diseases, Ninth Revision (ICD-9) Codes for Larynx Procedures.

Results

Among patients with laryngeal or hypopharyngeal cancer, ICD-9 codes for laryngectomy identified 45 156 cases between 2001 and 2011.² Patients included 36 215 men (80.2%) and 8941 women (19.8%) (mean age, 62.6 [range, 20-96] years); 0.9% were Asian individuals, 10.6% were Black, 4.6% were Hispanic, 81.8% were White, and 2.1% were listed as being of other race or ethnicity.² In the cohort of 51 512 cases identified using ICD-10-PCS definitions (mean age, 62.9 [range, 20-99] years), 40 924 (79.6%) were men and 10 491 (20.4%) were women (of those for whom sex was reported); 0.9% were Asian individuals, 10.6% were Black, 4.6% were Hispanic, 81.7% were White, and 2.2% were listed as being of other race or ethnicity. The proportion of laryngectomy cases classified as partial laryngectomy increased from 17.8% to 28.0% using ICD-10-PCS (Figure 2). We attribute this increase to inclusion of cases classified by ICD-10 codes previously classified by ICD-9 as “30.09, excision or destruction of larynx lesion.” Based on ICD-10-PCS definitions of laryngectomy, no association between hospital volume and in-hospital mortality, surgical complications, or costs was found.

Figure 2. Differences in Laryngectomy Case Volume Identified Using International Classification of Diseases, Ninth Revision (ICD-9) vs International Statistical Classification of Diseases, Tenth Revision (ICD-10) Definitions for Laryngectomy.

Discussion

This cross-sectional study reveals meaningful differences between ICD-9 and ICD-10 procedure coding definitions, with a 14.1% increase in procedures classified as laryngectomy by ICD-10-PCS that are defined as excision/destruction procedures only under ICD-9. The reduced accuracy in procedure definition resulting from expansion of ICD-10 codes has implications for data analysis based on administrative data sets, such as the Leapfrog Hospital Survey, which publicly evaluates performance based on minimum volume standards using ICD-10-PCS.⁵ Gourin et al² previously reported a significant volume-outcome association for laryngectomy, with minimum hospital volumes associated with significantly reduced morbidity, mortality, length of hospitalization, and costs based on ICD-9 codes. This association is abrogated when ICD-10-PCS codes for laryngectomy are used, likely because the ICD-10 definition includes nonlaryngectomy procedures.

The aim of ICD-10-PCS is to increase descriptive precision by expanding the number of available codes. There is a 19-fold increase in the number of procedure codes in ICD-10 vs ICD-9, with procedures grouped for more precise identification and tracking of specific diseases.¹ Our data reveal that use of ICD-10-PCS codes artifactually increases the number of laryngectomy cases by adding cases that fall under the same ICD-10-PCS grouping as laryngectomy, reducing the positive predictive value and accuracy of procedural coding through dilution of detail. Our data also reveal that this approach reduces the accuracy of laryngectomy cohort identification, increasing risk of misclassification. In epidemiologic terms, ICD-10-PCS increases the number of false-positives in identifying laryngectomy cases. This observation has implications for research and quality measurement, including publicly reported volume-outcomes work.^5,6 There are limitations to the use of administrative hospital discharge data, which are limited to discharge diagnosis codes and do not contain information about stage, prior treatment, or process measures related to perioperative care and decision-making, which could impact our findings. Nevertheless, the NIS is the largest US inpatient surgical database and represents the spectrum of US medical system practice. Based on these data, we suggest a pressing need exists for more granular detail than ICD-10-PCS provides to accurately identify procedural details beyond laterality in quality and safety research.