Miscoding in the Nationwide Inpatient Sample Database Raises Questions About Validity for Arthroplasty Research

Area of Science:

• Arthroplasty Research
• Health Services Research
• Data Science in Medicine

Background:

• Administrative databases are increasingly used in arthroplasty research.
• The Nationwide Inpatient Sample (NIS) is a frequently utilized dataset.
• A lack of standardization in NIS data processing can lead to biased research findings.

Purpose of the Study:

• To assess data discordance in clinically relevant variables within the NIS for hip and knee arthroplasty.
• To propose a standardized data curation approach for NIS research on primary total hip arthroplasty (THA) and total knee arthroplasty (TKA).

Main Methods:

• Identified patients undergoing THA and TKA from the NIS (2016-2019) using ICD-10-PCS.
• Selected 63 relevant variables for comparison, including procedure codes and diagnosis codes.
• Analyzed 3,562,819 patient records using descriptive statistics.

Main Results:

• Significant discordance was found: 5.3% revision surgeries, 4.7% not elective, 2.3% non-primary THA/TKA, and 2.9-9.7% bilateral procedures.
• 6.2% of surgeries occurred outside the admission day, and 10.8% of THA/6.3% of TKA lacked orthopedic admission diagnoses.
• Overall miscoding reached 23.3% for THA and TKA studies.

Conclusions:

• Using the NIS without standardized data processing introduces substantial bias in studies of elective, unilateral, primary THA and TKA.
• A proposed stepwise data curation method can enhance the quality of hip and knee arthroplasty research using the NIS.