Abstract
The use of administrative databases in arthroplasty research has increased over the past decade. The Nationwide Inpatient Sample (NIS) is one of the first and most frequently used. Despite many published articles using this dataset, there exists no standardization resource accounting for the potential of biased results. The purpose of our study was to assess the amount of discordant data between clinically relevant variables and propose a standard for using this database in primary total hip (THA) and knee (TKA) arthroplasty.
An initial set of patients undergoing TJA were identified from the NIS between 2016 and 2019 using the International Classification of Diseases, 10th Revision, Procedure Coding System (ICD-10-PCS). All records with THA and TKA in any of the procedure variables (PR1-PR20) were included. A total of 63 relevant and consistent variables were selected for individual comparison including diagnosis-related group (DRG) (1), elective (1), procedure codes (20), day of main procedure (1), and diagnosis codes (40). Descriptive statistics were used. A total of 3,562,819 patients were included in the initial set.
Using DRGs, 5.3% were revision surgeries, 4.7% were not elective, 2.3% did not have THA or TKA as their primary procedure for hospitalization, 2.9% of THA and 9.7% of TKA were bilateral. A total of 6.2% of the surgeries were done day(s) before or after admission, 10.8% of THA and 6.3% of TKA were missing an orthopaedic diagnosis for admission. Many had multiple orthopaedic diagnoses for admission, 3.2% of THA and 0.7% of TKA. Overall miscoding was 23.3%.
Using the NIS without standardized data processing to study elective, unilateral, primary THA and TKA introduces major bias. A logical and stepwise approach to curate the data before analysis is proposed to improve research quality when using this database in hip and knee arthroplasty studies.